GPLi'_11LR OF TRE COOP.

G LLIïGUAGE

1.. ;1. G. Garman

Ph.D. Thesis

University of Edinburgh
1 973
 (i)

ST'r.T'i1.EIï T OF RESEARCH

Research for this thesis was started in September 1967, financed

by a grant from the Department of Education and Science, London.

This grant extended up to 1969, and included travel to and from India

in 1963: within the grant period the first three chapters were

completed (save for minor subsequent additions and updating of

references), and a preliminary version of Chapter 4 appeared as a

paper in the Dravlingpex series (Vol. 1, ro. 7). In December 1970

a revised version of this paper was written up (Garman, forthcoming)

while I was working on a project - Cross -Linguistic Studies of

Language acquisition - supported by a grant from the Social Science

Research Council to Dr. R.J.ales at the Cognition Project, Department

of Psycholo y, University of Edinburgh. The remaining work was done

after joining the Department of Linguistic Science, University of

Reading in October 1971. Part of Chapter 5 was presented as a paper

to the Second All -India Conference of Dravidian Linguists, 1972

(Garman, to appear). No other part of the thesis has been submitted

for publication. The thesis has been composed by myself.

L.i.G.Garman

Department of Linguistic Science,

University of heading.

Larch, 1973.
 ACKNOTaZi;DGEMEidTS

I want to thank all my informants : Dr. I .] .Iiuthanna (London,

September 1967); Sri N.]i.Vasu (i=ercara, hay-June 1968); Sri N.N.

Belliappa and Smt. N.G.Belliappa (Iiysore, July August 1968);

Dr. C.A.Pemmiah (Virajpet, September-November 1968); Sri E.K.

Appachu and Smt. M.N.Appachu (Ammathi, September 1968); Sri N.C.

Chinnappa (Virajpet, September 1968); Sri N.C.Subbayya and Smt.

N.P.Subbiah (Karada, SeptemberN ovember 1968); and Kum. N.M.

Belliappa (Edinburgh, November 1970 -September 1971). I owe my

particular thanks for hospitality as well as linguistic data to

Sri Id .Iú .Belliappa, Smt. N.G.Belliappa, Sri 1i.K.Appachu and Smt.

I..K.Appachu, Dr. C .A.Pemmiah, Sri N.C.Subbayya and Smt. N.P.

Subbiah.

I:y thanks go also to Dr. P.J.,;ales for encouraging me in every

way possible to continue this research while I was at the Cognition

Project (August 1969- September 1971), and to my colleague there,

P.D.Griffiths, for his valuable advice on a number of specific

points and for the benefit I gained generally from working with him

for more than a year. Dr. E.K.Brown kindly read a preliminary

version of the section on 'volitive' verbs in Chapter 3; Professor

L.B.Emeneau, Dr. G.Brown and Dr. E.C.Fudge provided helpful comments

on the Dravlinppex paper (the first version of Chapter 4); and Dr.

P.J.Trudgill has read the later drafts of the third and fourth

chapters: I am most grateful to them all.

It will be evident how much of this work has been influenced by

the approach to phonological description presented in Fudge (1967,

1969a and 1969b). It will be equally clear that Professor J.Lyons

has had a quite incalculable influence in the course of this

research, even though he did not supervise it; perhaps this

statement alone suffices to show the thanks that are due to him

from one of the former postgraduate students in the Edinburgh

Linguistics Department: it is a debt of gratitude common to us all.

Above all, I want to thank my supervisor, Dr. R.E.Asher; he

has earned my gratitude not merely for being my supervisor but for

constantly maintaining a judicious blend of encouragement and

criticism, and being endlessly patient, in the face of many long and

embarrassing delays. iithout his support in ways that are outside

the normal brief of a supervisor, the work might never have been

finally written up.

The typing was done by Ers. I1. uinlan, of the University of

Reading; she had a great deal of difficult material to work through

in very little time, and carried it out in a most helpful and

expert way.
 SULi-;rY

The purpose of this work is to present the main features of

the phonology, morphology and syntax of the Coorg language of

southern India.

Chapter 1 briefly describes the area where the language is

spoken and the most important points regarding the community of

the Coorgs (noting areas of further research interest, principally

in regional and social dialects). It also introduces and comments

on the rather sparse 19th- century literature on the language, and

records the important work done in this century by Professor

I. .B .Emeneau.

Chapter 2 is principally concerned with the sound system of the

language, describing it in some phonetic detail and demonstrating

alternative analyses in respect of a number of problem areas (in

particular, the stop consonant system). It is suggested that an

analysis in terms of phonetic contrast alone is inadequate, and that

frequency of contrast has to be taken into account when setting up

the phonological system. This shows the necessity of working with

underlying representations, together with a set of process rules,

in order to account for the surface contrasts.

Chapter 3 deals with syntax, illustrating with a particular

type of 'surface' transcription discussed in the preceding chapter.

The basic sentence patterns of the language are introduced, and then

attention is concentrated on those sentences involving the copula,

and on the transitive causative constructions in the language. It is

suggested that the facts of Coorg require a 'lexicalist' approach to

the statement of certain voice -related verb forms (which are therefore
 (v)

not to be treated in the syntactic component at all).

The conclusions of Chapters 2 and 3 constitute the basis for

the main thesis of the work. It is argued in Chapter 4 that the

lexicon has to include a generative subcomponent, working in terms

of 'Root' and 'Affix', in order to facilitate the statement of extra -

syntactic lexical relationships (such as hold between certain voice -

related verbs as described in Chapter 3); furthermore, that the

appropriate phonological system for the representation of these

lexical relationships is the one suggested at the end of Chapter 2.

Chapter 4 concentrates on the structure of verbs; Chapter 5 notes

some evidence in favour of a lexicalist approach to the description

of certain lexical relationships between nouns, and shows how the

analysis allows for a straightforward account of a particular problem

in derivational morphology (contrast of consonant length immediately

following a short root vowel). Finally, Chapter 6 suggests ways in

which the analysis may throw light on the structure of complex

lexical items.
 COIT TEIJTS

Statement of research

icknowl edements

Summary

Chapter 1: Introductory 1 -27

1.0 Background 2

1.1 The land 2

1.2 Recent history 4

1.3 The Coorg community 5

1.4 Population 7

1.5 Coorg literature 9

2.0 1-revious studies 10

2.1 Early general and linguistic studies, to the

end of the 19th century 10

2.2 20th century studies 18

2.3 The Coorg vowel system 20

3.0 Fieldwork for the present study 25

Chapter 2: Transcription 28 -109

1.0 Introduction 29

1.1 Inventory of phonemes: vowels 29

1.2 Inventory of phonemes: consonants 30

1.3 The word 32

1.4 The syllable 32

2.0 Contrasts 33

2.1 Vowels 34
 2.1.1 Short vowels, position and manner 34

2.1.2 Long vowels, position and manner 36

2.1.3 Vowels, short versus long 38

2.1 .4 /ae/ versus /e:/ versus /a:/ 39

2.1.5 "'asal versus oral vowels 39

2.2 Consonants 41

2.2.1 Contrasts of place 41

2.2.2 Contrasts of manner 44

2.2.3 Contrasts of length 46

3.0 Allophonic statement 48

3.1 Vowels 48

3.1.1 Short vowels 50

3.1.2 Relative positions of the short allophones 56

3.1 .3 Long vowels 57

3.1.4 Relative position of the long allophones 60

3.2 Consonants 61

4.0 Distribution 63

4.1 Vowels 63

4.1.1 Clusters 63

4.1.2 Long vowels 65

4.1.3 Back unrounded vowels 65

4.2 Consonants 73

4.2.1 Clusters 73

4.2.2 Distribution within the word 85

4.2.3 Further details on intervocalic distribution 93

5.0 Conclusion 100

5.1 Dravidian orthographies 101

5.2 The sound pattern of Coors 104

 (viii)

Chapter 3: Syntax 110 -195

1.0 Introduction 111

1.1 The basic syntactic patterns 112

1.1.1 Sentences negated by Lille] 114

1.1.2 Sentences negated by Lalla] ('Relational') 126

2.0 The copula 128

2.1 The copula forms La:gj- and [ull]-/[ir]- 129

2.1.1 The copula in existential /locational function 131

2.1.2 The copula in possessive function 132

2.1.3 The copula in relational sentences 132

2.1.4 The copula in interrogative and negative

sentences 133

2.2 The status of the copula 137

2.2.1 Kandiah's arguments 137

2.2.2 The possessive copula 147

2.2.3 The adverbialisation transformation 149

2.3 The analysis of the copula 152

2.3.1 Summary 152

2.3.2 The analysis 154

3.0 Voice 164

3.1 'Intransitive', 'transitive' and 'causative' 166

3.1.1 The data 166

3.1.2 Discussion 167

3.1.3 The analysis of causative verbs 173

3.1.4 The problem of 'volitive' verbs 175

3.2 Two proposals regarding 'volitive' verbs 178

3.2.1 Kandiah's proposal 178

3.2.2 Lyons' proposal 184

3.2.3 Syntactic relatedness 189

 (ix)

3.3 Conclusion 192

3.3.1 The role of syntactic deep structure 192

3.3.2 The role of the lexicon 193

Cha)ter 4: The structure of lexical items: verbs 196 -290

1.0 Introduction 197

1 .1 Verb classes 198

1.1.1 The non-past affix 198

1.1.2 The past affix 201

1.2 The verbal lexeme 204

1.2.1 The basic lexemes 204

1.2.2 Derived lexemes 205

1.2.3 Irre;;ular and defective lexemes 208

2.0 Preliminaries to the analysis of lexemic structure 209

2.1 Lexemic relations 209

2.2 The form of lexical entries 214

3.0 The generative component 221

3.1 Onset 226

3.2 Peak 229

3.3 Coda 234

3.3.1 Preliminaries 234

3.3.2 The system of elements 249

3.4 Cons Suff 264

3.5 Voc Suff 271

3.6 Suffix 2 275

3.7 I and II 276

3.8 Volit1 270

3.9 Volit2 282

 (x)

4.0 The rules cited in this chapter 282

4.1 The lexeme structure rules 282

4.2 The phonological mutation rules 284

4.3 The realisation rules 285

4.4 The articulatory sequence rules 287

Chapter 5: The structure of lexical items: nouns 291 -371

1.0 Introduction 292

1.1 Lexeme -final elements 292

1.2 Contrast of consonant length after a short root

vowel 293

1.3 Canonical forms 294

2.0 Preliminaries to the analysis of nominal lexemes 295

2.1 The nominal lexeme 296

2.2 Lumber 296

2.3 Case 299

2.4 Gender 301

3.0 The generative component 313

3.1 Onset 323

3.2 Leak 327

3.3 Coda 334

3.4 Cons Suff 350

3.5 Voc 3uff 359

3.6 Suffix2 364

3.7 Nom 364

3.8 Noun 366

4.0 The rules cited in this chapter 368

4.1 The lexeme structure rules 368

4.2 The phonological mutation rules 369

 (xi)

4.3 The realisation rules 370

4.4 The articulatory sequence rules 370

Chapter 6: The structure of complex lexical items 372 -430

1.0 Introduction 373

1 .1 Preliminaries to the analysis of coEle ::

nominal structure 373

1.1.1 Set 1 373

11.2 Set 2 376

1.1.3 Set 3 379

1.1.4 Set 4 380

1.2 The lexeme 383

2.0 The structure of complex nominals 386

2.1 Set 1: Lexeme Lexeme 386

2.1 .1 Lexeme = Base" Classifier 386

2.1 .2 Lexeme # Base" Classifier 396

n
2.2 Set 2: Base Postbase 399

2.3 Set 3: Base "Lexeme 412

2.4 Set 4: Lexeme "Classifier 417

2.5 A further case of linear derivation 420

3.0 The rules cited in this chapter 427

3.1 The lexeme structure rules 427

3.2 The phonological mutation rules 428

3.3 The realisation rules 428

Appendix I: Inventory of verbal lexemes, Classes 1 -5 431 -448

Appendix II: Inventory of derived verbal lexemes 419 -455

Appendix III: Inventory of irregular and defective

verbal lexemes 456 -458
 (xii)

Appendix_ IV: Inventory of nominal lexemes 459 -499

Bibliozraphy 500
 1

CHAPTER ONE

Introduc tory
 2

1.0 Background

1.1 The land

Coorg (an anglicisation of the Kanarese word /kodao /) is one of

nineteen Districts that make up the modern (i.e. post -1956) Mysore

State in southern India (see the map in Fig. 1 on p. 3). It lies in

the south-western corner of the state, defined by a district

boundary on three sides (west to north; north to east; and east to

south) and by the border with Kerala State on the fourth side (west

to south). Its greatest extent from north to south is about sixty

miles, and from east to west about forty miles, enclosing an area of

just over one and a half thousand square miles. It consists in the

main of mountainous country, extending across the peaks of the Western

Ghats, somewhat to the north of the Niligiris. The district town,

Mercara, lies at an altitude of almost four thousand feet, and the

highest peak in Coorg, about twenty miles away from Kercara in a

south-westerly direction, reaches five thousand seven hundred and

twenty nine feet. The average rainfall in Mercara is of the order of

a hundred and twenty inches a year, most of this resulting from the

south -west monsoon (July to September), which supports average to

good rice cultivation in the valleys. Forest reserve areas are all

that now remain of the once extensive forests (containing teak and

sandalwood) that have largely been cleared for other types of land

use; apart from rice fields, numerous coffee estates make a

considerable demand on land area. The district is famous in India

for its coffee, and also for oranges and bee -keeping; marmalade and

honey are marketed by co- operative societies.

 H A S S A N D I S T R I C T
"..
.Kodlipet,

S OUT H

1.
ti
C A N A R A
; . .
,
'. , ...
D I S T R I C T
Somwarpet.

.,.
'. M Y S O R E

COORG DISTRICT
.Mercara
hagamondla

.Siddapur

D I S T R I C T
9

Vi rajpet

. Ponnampet

12°N

.Srimangala

6
o .Cannanore

0 5 10 15 20 25 30 35 40
miles

Figure 1: Coors; District in Mysore State (post -1956)

 4

1.2 Recent history

In recent history, Coorg has undergone a number of political

changes which have affected its social structure. As an independent

territory, ruled by a dynasty of Kanarese -speaking Hindu Rajahs of

the Lingayat sect, it was placed under British protection in 1792

when Tipu Sultan, the Muslim ruler of Mysore, was holding much of

South India against the British and simultaneously laying claim to

Coorg. After the defeat and death of Tipu at the hands of the

British in 1799 (in which the Coorgs also played a notable part), a

Hindu dynasty was installed on the Mysore throne. But an uprising

occurred in Mysore in 1831, and was followed by a period of British

administration; and in 1834 Coorg was annexed to hysore, the last

Coorg Rajah being deposed by the British, who accused him of cruelty

towards his people. In 1858, when the British Government took over

the administration of India from the East India Company, Coorg became

a separately administered province, and remained so (even after

India's independence) until 1956; in that year the states of the Union

were redefined on a linguistic basis, and Coorg was merged with

Kanarese -speaking Mysore. This was a direct result of the influence

of the Kanarese- speaking Lingayat dynasty in Coorg, from the

beginning of the seventeenth century up to 1834. During this period

of more than two hundred years, the contacts between Coorg and the

Malayalam culture of the Malabar coast must have declined, in favour

of the Kanarese language and culture of the court. Srinivas (1952)

notes that 'Educated Coorgs are usually trilingual, knowing Kodagi,

Kannada and English. Kodagi is used in the home, Kannada in talking

to most non -Coorgs excepting halayalis, and English in official

matters, and occasionally in conversation with strangers' (1952: 9).

 5

It should be noted that English succeeded in establishing itself as

the language of officialdom after 1834 in much the same way as

Kanarese must have done under the Coorg Rajahs. Since 1956, Kanarese

has increasingly become once more the language which supplants Coorg

outside the home or circle of fellow -Coorgs.

What Srinivas (1952: 7) described as 'Coorg Proper', the 'core

and centre of the culture of the Coorgs', lying in the area bounded by

Mercara, Siddapur, Srimangala and Bhagamandala, is apparently a much

smaller area today. It is generally said to lie south and west of

I+iercara, north and west of Virajpet; perhaps as more of a concept than

a geographical area, it is of importance and is held in great and

sentimental respect by many Coorgs who feel that they have been cut

off from their culture and the roots of their language by 'Westernis-

ation'. It is fair to say that most of these Coorgs would not give up

their present occupations (in law, medicine, the armed services, etc.)

to go and live in the old conditions. However, it is still not

difficult to find people of middle age and upwards who are relatively

undisturbed by the encroaching world; but the existence of monolingual

Coorg speakers must be seriously doubted.

1.3 The Coorg community

'Coorg' is an ambiguous term, referring to the land, the largest

single community there, and the language spoken by that community.

It will be seen in what follows that it is also a complex term, in that

the notion of a 'Coorg community' is complex.

In Emeneau (1938) the Coorgs are described as follows: 'The

Coorgs are an endogamous community of about 40,000 people who form

approximately a quarter of the population of the province of Coorg....

They were originally the warrior owners of this tract and formed, and
 6

still form, the highest community in the social scale (barring the

few Brahmans who live in the province). It is doubtful how far they

can be considered a caste in any strict Indian sense of the word, for

they resolutely refuse to accept the ministrations of the Brahmans in

any of their rites' (1938: 123). Richter (1870), basing much of his

account on Iioegling (1855) together with his own long experience in

the province, had earlier noted that a number of the tribes speaking

the Coorg language are not accepted as part of the Coorg community;

they provide drummers at feasts, or are ironsmiths, or carpenters,

barbers, toddy-tappers, washermen, etc. A further group, known as

Amma Coorgs, preserve a very different kind of distinctiveness, although

numbering very few people (only 281 according to Richter, 1870; and

probably still not more than five hundred): of these, Srinivas (1952)

says that they 'exemplify a tendency which has always been present in

the caste system: a small group of people break off from a larger whole

of which they are a part, Sanskritize their customs and ritual, and

achieve a higher status than their parent body in the course of a few

decades' (1952: 35). J.S.F.Mackenzie notes in a preface to Lindsay

(1874) that 'The Amma Coorgs, a small subdivision numbering in all

300 souls, resemble their countrymen in language and costume but

neither eat animal food nor drink spiritous liquors, and only intermarry

with their own class'. The main body of Coorgs, however, relish any

meat except beef (pork and venison are delicacies) and are, in general,

fairly satisfied with their reputation for being able to drink

alcohol.

In contrast to these distinctions, however, Srinivas (1952) also

stresses the point that fairly high caste non-Coorgs are accepted into

the Coorg community with much greater ease than is found with the caste
 7

system generally (1952: 37), and he cites Moegling (1855) and Richter

(1870) to the same effect. One of the commoner points that one hears

raised in conversation among Coorgs these days is the influx of large

numbers of Muslims from north Kerala; apparently it is these people

against whom Coorg social solidarity is currently defined (in part),

and the distinctions marking off the relatively recent admissions into

the Coorg fold (e.g. Tulu or Kanarese Gowdas, Kanarese Okkaligas, etc.)

are correspondingly blurred.

The Coorg community as a social unit is therefore complex, and not

completely isomorphic with the corresponding speech community. It

would clearly be worth while to investigate what promises to be a rich

field of social varieties within the speech community. In addition,

Coorgs themselves generally recognise a number of regional dialects:

the north -eastern, the north -central, the south-central, and the

southern. Mercara is often cited as having its own dialect. All this

remains uninvestigated as yet.

1.4 Population

Connor (1870), reporting on the survey of Coorg that he carried

out between 1815 and 1817, estimated the total population of the area

to be about 40,000. The first census, in 1871 (cf. Lindsay, 1874),

reported a total population of 168,312, of which 26,389 were Coorgs

(there is no indication of whether this figure includes non -Coorg

communities speaking the Coorg language). Thus Coorgs represented

about 15, of the total population at that time; this figure increased

only slightly up to the 1921 census, when it jumped to 27.. Prior to

the 1871 census, there is the estimate of Moegling (1855), which

reckons 25,000 to 26,000 Coorgs; if Coorgs at this time also

 8

constituted about 15;x- of the total population, this would give a

figure of around 165,000 for the area, which, in view of the 1871

figure, seems only slightly too high. But it is difficult at first

sight to square this figure with that in Connor (1870) for the first

quarter of the century. One might reject Connor's figure as unreliable

except for the fact that he shows a generally high level of competence

in his report, and an estimate of the population clearly falls within

the area of his competence as a survey officer; probably his figure

needs only to be modified somewhat. In this connection, it is worth

noting what Moegling (1855) has to say regarding the Coorg population

prior to his day: 'They have much increased in number during the last

20 years. They are no more killed ad libitum by their Rajahs, nor

destroyed by harassing warfare. In former days they seem scarcely to

have mustered more than 4,000 or 5,000 fighting men' (1855: 28).

Later on, he points out that the old social order was in decay at the

time of writing as a result of the sudden increase in numbers in the

Coorg community: 'I am told, there are some houses in the country

containing sixty, seventy, eighty souls and upwards. Very fine and

patriarchal, if there be peace in the house: But, what fearful misery,

when such a house is rent by discord: In these days, the ancient

system is breaking up...' (1855: 32). M oeglin g's point about the

homicidal attitude of the Rajahs towards the Coorgs is highly contro-

versial, to say the least; as a missionary, he clearly favoured the

policies of the British administration in Coorg as opposed to those of

the earlier Hindu dynasty and must therefore not be taken as an

objective historical source: but his other point, regarding the

decimation of the Coorgs in warfare, fits with the historical fact

that Coorg was struggling for its existence against Muslim Mysore from

Ryder Ali's conquest of Bednur in 1763 to the death of his son Tipu in
 9

1799, and with the account in Rice (1878) that Coorgs formed the main

body of the Rajahs' armies (1878: 327). We may perhaps think of a

community of 10,000 Coorgs at the time of Connor's survey, in a total

population of probably rather more than 40,000. After the death of

Tipu, Coorg was considerably easier of access from the outside; roads

were reopened, trade contacts increased with the Kanarese- speaking

area formerly controlled by Tipu, and we probably have to recognise

an influx of Kanarese communities on quite a large scale. During this

period also, of course, the Coorgs were increasing in number, remaining

the largest single group in the territory.

Returning to the present century, the census for 1931 contains the

most detailed information prior to the reorganisation of the states;

Coorgs are estimated at 41,026, while the number of Coorg speakers is

given as 44,585 (all of whom are recorded as speaking some other

language also - undoubtedly Kanarese and/or English). At this time,

Coorgs formed about 25 of the total population in Coorg (163,327):

other speech communities in the area were listed as Kanarese (62,769,

but made up of various social communities), Malayalam (14,914), Tulu

(14,275), Yerava (10,026), Hindusthani (4,378), and Tamil (3,007).

Since 1956, figures have been available only for the new Mysore State

or for India as a whole; in 1961 there were 78,202 Coorg speakers in

Mysore State, while the figure for India (including Mysore) was 79,172.

1.5 Coorg literature

As a final point in this section, it may be noted that Richter

(1870) and Caldwell (1875) both correctly state the situation in their

day, that Coorg had no literature (in the sense of the term that

excludes the oral tradition of songs). However, Appacha (1906) is a

bold and conscious attempt to initiate a Coorg dramatic tradition, and

 10

was followed by Appacha (1908), and two other works in the same pattern

(for which references are not to hand). In these works, common Hindu

myths are worked into the form of a short drama, where characters tend

to speak either a rather stilted language (Kings, court officers, etc.),

or a colloquial variety (the common people); the text is broken into

at intervals with short songs, usually based on folk idiom, and

showing interesting features of rhythm and alliteration. On the whole,

the songs have outlived the plays, although productions of the complete

plays were put on in a number of places in Coorg during 1968, in

honour of the birth centenary of the poet. Appacha (1929), a second

edition of Appacha (1906), is particularly interesting, in that it

contains a prose introduction by the author which deserves consider-

ation as the first attempt to create a formal style in the language,

outside the framework of the dramatic form. In addition to Appacha's

work, mention should also be made of Chinnappa (1929), a translation

into the Coorg language of the Bhagavad alta. The same author had

earlier (Chinnappa, 1924) published a collection of, and commentary on,

the songs, sayings and customs of the C oorgs, in the Kanarese and

Coorg languages (unfortunately, the only copy of this work that came

to hand had the title page mutilated, so that the publisher could not

be discovered). At the Cauvery Ashram in Virajpet, Tukkatira S.

Poovayya is currently writing devotional lyrics and short poems in the

Coorg language (Poovayya 1966, 1968).

2.0 Previous studies

2.1 Early general and linguistic studies, to the end of the

19th century

Ellis (1816: 3) mentions 'C odugu' as one of the languages of south

 11

India related to Tamil and Malayalam; but it was not until Lt. P.E.

Connor carried out his survey of Coorg, 1815 -17, that there was much

information available on the people and the land, and it was only

much later than this that reliable language data began to be collected.

Connor's report remained unpublished for fifty two years (Connor,

1870), although presumably much of the information it contained

circulated prior to that date among interested people who were

connected with the East India Company in some way. The language was

of only incidental interest to Connor, of course; his professional

concern was with the nature of the terrain, and the social, economic,

and demographic particulars of the area. On these topics he shows a

sober and perceptive judgement that one could hardly expect in his

description of the speech of the Coorgs, but it is nevertheless

worth repeating his impressions of it here: 'That this mountainous

tract should have a dialect peculiar to itself is a phenomenon

deserving some remark: the attainments of a philologist would be

necessary to enter into those enquiries which so curious a subject

deserves; no attempt will be made here to investigate its Etymology

but perhaps a few brief remarks may be desirable.

'The dialect spoken amongst these hills is known only to the

Codugus, its use consequently is limited to a few people and it cannot

be considered as the channel of intercourse with any but themselves;

there is no data by which to establish its antiquity, nor can

scarcely a well founded conjecture be formed as to the origin of it,

unless from the apparent affinity it bears to the Mallialum, we are

entitled to conclude that it is most probably derived from it, and

that the alterations it has undergone have been caused by the detached

and ignorant state of the society amongst whom it prevails.

 12

'The language of the Codugus has much of the harsh and disagree-

able sound of the Malabar spoken on the Eastern Coast, it is unwritten

and like most oral dialects scanty and meagre; there are but few

abstract terms, and even the names of objects of sense are borrowed

(somewhat modified in the use) from other tongues. In short it appears

evidently that of a rude people, and has no claims to be considered as

a primary language.

'Having no alphabet, the Canarese character is always used, and

their own or that language written indifferently; the Native Officers

of Government, whose attainments alone reach thus far, comprehend

both perfectly...' (1870: II, 53-4).

We may probably rely on the information here regarding the social

status of the Coorg language, and the use of the Kanarese script; it

conforms to what is known of the history of the area, and - save in

the matter of literacy, in which present-day Coorg is better than

average - corresponds generally to the situation today. Even the

suggestion that Coorg might be derived from Malayalam is a more reason-

able guess in many ways than certain later attempts to see it as a

dialect of Kanarese (e.g. Caldwell, 1856). More tran thirty years

after Connor's survey, B.H.Hodgson included some Coorg material in a

comparative word list of some southern Indian languages (Hodgson, 1849).

He noted his indebtedness 'for these vocabularies to Mr. Walter Elliot

of Madras, whose name is a sufficient warrant of their perfect

accuracy' (1849: 2); but the fifty seven Coorg items are on the whole

so badly transcribed as to be nearly unusable. Long and short vowels

are confused, and a number of what can only be retroflex consonants

are represented as dentals.

In 1856, R.Caldwell published the first edition of his comparative

 13

grammar of Dravidian languages; however, this work (Caldwell, 1856) is

less useful as far as Coorg is concerned than the second edition

(Caldwell 1875), a discussion of which will be postponed until a brief

survey has been made of the other publications which appeared after

Hodgson (1849). Among these should be mentioned Koegling (1855);

this records the impressions of the land and the people set down by a

German missionary, the Rev. H. Moegling (mentioned above, p. 8 ), and

deals in particular with the last Lingayat Rajah, whom Moegling

regarded not only as 'a fool and a coward' (1855: 201), but also as an

insane despot, justly deposed by the British in 1834. This view

prompted a reply (Anon, 1857) from someone describing himself as 'An

officer formerly in the service of his highness Veer Rajunder Wadeer,

Rajah of Coorg', who dedicated his book to the Rajah and attempted to

rehabilitate the Rajah's reputation: 'The Rev. Mr. Moegling, an

American (sic) missionary, is the other person who has recently revived

and given increased circulation to these calumnious reports, in his

work entitled Coorg Memoirs...' (1857: 133). Undeterred however,

Moegling in the following year published the Kanarese autobiography of

the last Rajah, together with an English version which had been made by

Lt. R.Abercromby, previously published in 1808; he took the opportunity

of again stating his interpretation of the Rajah's personality

(Moegling, 1858).

Capt. R.A.Cole was appointed Superintendent of Coorg in 1865; two

years later, he published a grammar of the language (Cole, 1867), which

deserves assessment as the first serious description to appear in print.

In the preface to this work, he records that shortly after his appoint-

ment he travelled on duty to one of the west- central taluks, 'the very

stronghold of the C oorgs' , and 'found that many of them could not speak
 1#

Canarese, the official language, but only a dialect, which I could not

understand. The officials also, when desirous of making remarks to

each other, used to employ this dialect...' (1867: Preface, 3). It

should be noted that this account is consistent with that given in

Connor (1870) for the situation fifty years earlier; one has to assume

a gradual spread of Kanarese -Coorg bilingualism from the eastern parts

of the province, and one has evidence here that it had not at this

time penetrated to the western taluks. Cole's grammar is, unfortun-

ately, a very disappointing effort, even for its time. It consists in

the main of a collection of paradigms (of nouns, pronouns, and verbs),

followed by a list of 'irregular' verbs (most of which are perfectly

regular, given certain straightforward morphological statements),

more lists of 'adverbs', postpositions and conjunctions, a vocabulary

section, and some sample sentences. Cole, who clearly knew Kanarese

to some extent at least, used both the Kanarese script and a Roman

transliteration of this in citing the Coorg material, and as a result

there is a generally reliable representation of vowel length, retro-

flex consonants, etc. However, the transcription is defective in

other respects (see below, p. 22 ), the morphological statements are

not trustworthy, the classification of the nouns and verbs is arbitrary,

and the vocabulary frequently does not distinguish between Coorg items

and Kanarese loans. Two years later, another German missionary, the

Rev. A.Graeter, published a brief account of the land and the people

(Graeter, 1869). It was written in Kanarese (language and script),

and included some Coorg songs also (in Kanarese script). The same

author in the following year published a more extensive collection of

Coorg songs (Graeter, 1870), to which was appended a sketch of Coorg

grammar and a vocabulary (1870: 26 -48). Of the grammatical sketch it

 15

is fair to say that it seems to be largely independent of Cole (1867),

but represents no advance over that description, save in respect of

the transcription of the vowels (see below, p.23 ). It contains twelve

nominal and pronominal paradigms, one verb paradigm, and a list of

nearly fifty 'irregular' verbs - far larger than that given in Cole

(1867).

It was in this year that Connor (1870) appeared, reporting on the

survey of 1815 -17; and so did Richter (1870), a valuable manual or

gazetteer of the province by one of its most respected missionaries

and educationists, which included a section on the language (1870:

193-214). Richter records his indebtedness to Noegling (1855) as far

as non-linguistic matters are concerned; he says of that work that it

'furnishes extensive and correct information on social and religious

topics and has been largely made use of in this volume'(1870: Preface).

He makes no acknowledgment in respect of his language material, however,

although it is clearly based on the same data as is found in Graeter

(1870); the reason for this is that these two missionaries were

collaborating on the language at this time. Richter's transcription

is not systematically different from that used in Graeter (1870),

although it is employed a little more consistently.

The following year, C.E.Gover published a collection of south

Indian folk songs in translation (Cover, 1871), including a number from

Coorg (1871: 101 -46). He states that he is indebted 'to two earnest

and capable German Missionaries, the Revs. W. (sic) Graeter and George

Richter, for literal translations of the originals. The first named

gentleman collected and published them in the Kodagu vernacular and

also rendered most of them into English' (1871: 103 -4). It is note-

worthy, in view of the relatively consistent transcription used in

 16

Richter (1870), that Gover specifically says of Richter that 'Almost

single -handed he reduced the language to writing...' (1871: 104). In

this year too, Capt. R.A.Cole published his second work (Cole, 1871),

an account of the civil code prevailing among the Coorgs.

The final publication containing Coorg material prior to

Caldwell (1875) is Burnell (1873), a Coorg translation of St. Matthew's

gospel (XIII, 1 -34). Burnell notes that 'this specimen of the

Kodagu language is due to the kindness of the Rev. F.Kittel of the

Basel Mission; he has submitted it to the criticism of a number of the

most intelligent natives of Coorg, and has added the valuable notes

under the text' (1873: iii). He also refers to Richter (1870) and

Graeter (1970), and goes on to say that 'There is also a Grammar by

Capt. Cole, but it is to philologists of less use than pr. Graeter's'

(1873: v). The text is written in the Kanarese script, incorporating a

sufficient number of diacritical devices (undoubtedly Kittel's) as to

allow for an unambiguous representation of the vowels of Coorg (see

below, p.24 ) .

Thus, by the time Caldwell (1875) appeared, a fair amount of Coorg

material had been published, and a reliable system of transcription was

within reach; but many uncertainties remained in the morphology of the

language, and the syntax had received virtually no attention (although

the basis for beginning such a study was available in Burnell, 1873).

Caldwell has this to say of Coorg: 'Last in the list of cultivated

Dravidian languages is the language of Coorg; but although I have

thought it best to give this language a place amongst the cultivated

members of the family, the propriety of doing so seems to me still

more doubtful than that of placing Tulu in this list... The native

spelling of Coorg is usually Kodagu, properly Kudagu, from kuda, west,

 17

a meaning of the word which is usual in Ancient Tamil. In the first

edition of this work, this language had not assigned to it a place

of its own, but was included under the head of Canarese. It had been

generally considered rather as an uncultivated dialect of Canarese,

modified by Tulu, than as a distinct language. I mentioned then,

however, that Dr. Moegling, a German missionary, who had resided for

some time amongst the Coorgs, was of the opinion that their language

was more closely allied to Tamil and Malayalam than to Canarese. It

is not quite clear to me yet to which of the Dravidian dialects it

is most closely allied. On the whole, however, it seems safest to

regard it as standing about midway between Old Canarese and Tulu.

Like Tulu, it has the reputation of puzzling strangers by the

peculiarities of its pronunciation. A grammar of the Coorg language

has been published by Major (sic) Cole, Superintendent of Coorg, and

some specimens of Coorg songs, with an epitome of the grammar by the

Rev. B. (sic) Graeter of Mangalore' (1875: 32-3) . Caldwell's Coorg

material consists of (i) a paradigm of a neuter noun, (ii) a paradigm

of the first and second person pronouns, and (iii) a paradigm of a

verb. On the whole, it is cited in a usable form, although there

are some errors: thus, the nominative forms of the first and second

pronouns (1875: 416-7) show a short vowel where a long vowel must be

expected. It is possible that certain forms cited by Caldwell are

no longer current, of course. He apparently used Cole (1867) and

Graeter (1870) /Richter (1870) as sources, but some forms are from

none of these works. Relatively little work was published on Coorg in

the next few decades. Rice (1878), in a section on the language (1878:

279 -96), presents a slightly revised version of the description in

Richter (1870), placed at his disposal by Richter; none of the changes

 18

is of importance. Hodgson (1880) contains an unchanged version of

the Coorg material in Hodgson (1849). More important than either of

these, therefore, are the publications of Anon (1882) and Veil (1886);

the first is a selection of bible sentences in Coorg, and the second

a collection of Coorg proverbs, written in unmodified Kanarese

script in each case (the typographical evidence allowing for the

possibility that Veil was the author of the first as well as the

second). Appiah (1887) is an interesting but hardly important work

containing a proposal for an entirely new Coorg script, some Coorg

material, and a few paradigms; for comments on the script described

in the work, see below, p. 24.

2.2 20th century studies

The first important publication in the twentieth century to

incorporate Coorg material was the Linguistic Survey of India (LSI);

the relevant parts are Vol. I, parts 1 and 2 (1927 -8) . Of this work,

however, Emeneau (1955) notes that it 'did not extend its operations

into Madras, Hyderabad, and Mysore States, and consequently failed to

map the dialects of the greater part of the Dravidian- speaking

territory. Moreover, its mesh was so wide and its methods so hit -or-

miss that several of the Dravidian languages in central India were

missed altogether...' (1955: 153). This general criticism regarding

the methods used in LSI is borne out for Coorg also; the Coorg

material it used was mainly drawn from one of the worst available

sources, the data in Hodgson (1680), and was not even consistently

checked against the better material in the other source quoted,

Richter (1870).

For some time after this, nothing of importance appeared in print.

 19

However, during the period 1935 -8, Emeneau collected material in the

field on Toda, Kota, Kolami, and Coorg, and a number of his public-

ations from that time contain some Coorg material even if they are

concerned principally with other topics than a descriptive analysis

of the Coorg language. Emeneau (1938) is worthy of special note, as

it contains an extensive, if ritualised, discourse in the language;

Emeneau (1945) presents some additional material in the framework of

a comparative study.

Goda Varma (1946), as a result of independent fieldwork done in

1939, attempts 'to show that the Coorg language agrees more with Tamil

and I%ialayalam than with Kannada in point of phonology, structure and

vocabulary, although it has latterly to some extent been influenced,

in these respects, by Kannada, the language now employed for literary

and court purposes in the country' (1946: 418). Goda Varma goes on

to make a well -documented statement of certain aspects of Coorg

phonology (in respect of its vowel system) from a comparative and

historical point of view (although his statement has been superseded

by that worked out independently by Emeneau, presented in Emeneau,

1970a).

The most important milestone in Coorg linguistic studies was, of

course, the publication of Emeneau's excellent material in DED, DBIÂ

and DEDS (1961 -8). Emeneau has also clarified the comparative position

of the language (within the South Dravidian subgroup), in Emeneau (1967),

and has made further important statements (Emeneau, 1970a, 1970b,

1971 ). Kothandaraman (19E9) contains a comparative statement with

regard to the same general area of interest that is covered in Goda

Varma (1946) and Emeneau (1970a). Garman (1969) presents essentially

the same analysis of Coorg verbs as that in Emeneau (1967), but

 20

attempts a generative statement, and supplies some additional data;

a revised version is presented in Garman (forthcoming) where the

model used is not substantially different from that described in

Chapter 4 of the present work. Some of the additional material

appears in Burrow and Emeneau (1972).

The current situation with respect to Coorg linguistics is such

that two recent comparative studies, Shanmugam (1971) and Subrahman-

yam (1971), are able to use Coorg data with confidence. But it is

nevertheless frustrating that a comprehensive grammar has yet to

appear; in this respect, therefore, Doraiswamy (forthcoming) may

prove to be most important. Meanwhile, the following chapters of the

present work represent an attempt to provide a grammatical model in

terms of which a coherent and comprehensive description may eventually

be framed.

2.3 The Coorg vowel system

One of the outstanding features of the sound pattern of Coorg,

and one which has attracted comment from the earlier investigators, is

the nature of its vowel system; this recognises short and long
n it tr

unrounded vowels in the high and mid back positions ( /i, I :, e, e: /)

as well as the corresponding rounded vowels (lu, u:, o, o: /).

In respect of the high back unrounded vowel /i /, Coorg may be

interpreted as sharing in what is apparently an areal feature,

evidenced in Tamil and Malayalam varieties, Tulu, and the (Coorg)

Gowda variety of Kanarese (see, for example, Asher, 1966; Andrew-

skutty, 1971; Bhat, 1967; and Kushalappa Gowda, 1970). All these

languages show derounding of word -final /u /; and Coorg regularly has

it!
// where Mysore Kanarese (for example) has /u/, as in items such as

/maddi/ (Coorg), /maddu/ (Mysore Kanarese), 'medicine'. But, as is

 21

implied in the transcription here, there is a contrast, however

sporadic, between Coorg /i/ and /u/ in this position; furthermore,

all eight back vowels occur and contrast in other positions. Hence

the situation in Coorg is to a considerable degree distinct from that

in the other languages mentioned above, and its vowel system is

unique to it. Further discussion of the phonological status of the

elements involved must be left until the following chapter; what is

relevant here is the extent to which the early descriptions of the

language paid attention to this feature of its sound structure.

There can be little doubt that the back unrounded vowels (perhaps

together with the fact that certain lexical items are nasalised,

showing nasal vowels), were mainly responsible for the reputation

that Coorg had in the last century for being a 'difficult' language.

Thus Connor (1870) mentioned that it 'has much of the harsh and dis-

agreeable sound of the Malabar spoken on the Eastern coast' (1870: 53)

(presumably referring to the areal feature of final derounding); Cole

(1867) records his surprise on discovering that 'Dr. Moegling, the

celebrated German missionary, ... had given up the attempt to learn

the language, owing to the difficulty of conveying the exact pronunc-

iation of the words' (1867: Preface, 3); and Richter rather fancifully

observed that 'The Kodagu language is shorter, more simple, but less

refined than the Canarese and a convenient medium for conversation;

by its contracted, rounded forms with abrupt terminations in half

vowels, it does not require a great exertion of the organs of speech

and admits of chewing betel and retaining the precious juice, whilst

the flow of conversation is uninterruptedly carried on. Indeed, a

beginner should practise the pronunciation with his mouth half full of

water, till he can speak without spilling any' (1870: 194).

 22

It is convenient to begin a consideration of the transcriptions

that were proposed in respect of these back unrounded vowels with Cole

(1867). He notes there (obvious misprints have been corrected in what

follows) 'the peculiar sound given to the vowel u in most of the words

in which that vowel is to be found. The pronunciation is as near as

possible to that given to u in the word creature, and I have designated

it as a half or silent u, and had wished to have it caused to be

written in the Canarese character as follows j ru, and to have

denoted it in the Roman character by placing a dot under the vowel;

but the Press was not able to do so. This vowel has also another

peculiar pronunciation, chiefly occurring in the Dative case in the

affix of that case. This pronunciation is almost like that given by

the French to eu in the word queue' (1867: Preface, 13). It is fairly

clear from this that Cole thought in terms of just one extra vowel,

'silent u', of which the typical form was heard by him as being close

to the value of English / a/; its other,'peculiar pronunciation,

chiefly occurring in the Dative case', is undoubtedly the high back

unrounded quality. It is interesting that, although he recognised

only one 'extra' vowel, he wanted to relate it in terms of symbolis-

ation and terminology to the high back rounded vowel. Thus (taking

his example), the Kanarese syllabary can only represent the consonant

r by means of a (vowel deletion) diacritic on the basic form d ,

ra (i.e. the included vowel a is present in the unmarked symbol),

yielding OF ; if the sequence ru is intended, then the u- diacritic

is added to the basic symbol, yielding ZiS : Cole's symbolisation of

'silent u' is achieved by the expedient of using the vowel deletion

diacritic with the u- diacritic (not normally allowed, of course, in the

Kanarese script) .
 23

Graeter (1870), employing just the Kanarese script, uses the

consonant symbols with the vowel deletion diacritic most commonly to

represent the mid back unrounded vowel, as in '`' for /ra:jë /,

'king'; he uses a combination of this diacritic with the u- diacritic

(i.e. the same as the symbol in Cole, 1867) to represent the high back

unrounded vowel, as in M)210715

s 17 for /ra: jañg/, 'to the king'.

However, he is not consistent in his application of this distinction,

frequently indicating a mid back unrounded vowel where the high back

unrounded must be expected (e.g.TID IO 0 2' for /ra:jara gond /,

'by the kings', 1870: 26).

Richter (1870) uses only a Roman script, where o and u represent

the mid and high back unrounded vowels, respectively. Of them, he

notes that 'the former (ö) is pronounced like e in the French relative

pronoun "que ", or the ö in the German "Götter ", the latter (ü) sounds

like the French u in "vertu" or like the ü in the German "Mutter "'

(1870: 195). However, while Richter's script is adequate in respect of

the / o distinction, it also contains the symbol a which is nowhere

described or explained and which occurs in some forms where one must

expect either ö or a; hence, some uncertainty as to the precise nature

of the Coorg vowel system is still evidenced in this work. This is the

more surprising in view of the fact that at one point Richter seems to

recognise that the contrast between ö and a is neutralised under

certain conditions; thus, he gives the forms rájö, 'the king' and

rájangiz, 'to the king', adding in a footnote that 'Though corrupted at

the end of a word the a takes again its full sound as soon as new

syllables are added' (1870: 197). Possibly ä is an attempt to illustrate

this neutralisation, since it occurs only in those forms where neutral-

isation of the ö á contrast is found; but not in all such forms, as the

above example shows.

 24

Kittel's version of the Kanarese script, used for the Coorg

material in Burnell (1873), agrees with Graeter's in using the vowel

deletion diacritic to represent the high back unrounded vowel; he

notes that 'All Half Vowels have the Tulu sound' (1873: 1). The mid

back unrounded vowel he indicates by placing a dot above the symbol

for short or long e; in this way he implies a relationship between

the mid front and mid back unrounded vowels, as against the J'a

relationship implied in Richter (1870). It will be shown in Chapter

2 that each of these relationships is valid, depending upon the

position and environment of the vowels concerned.

Caldwell's Coorg material (in Caldwell, 1875) is cited in Roman

script, where the single apostrophe ' after a consonant indicates the

high back unrounded vowel; thus marat'l', for /maratl /, 'in the tree'

(1875: 319). In the one instance where the mid back unrounded vowel

is expected, the transcription shows am (in kejjam, first person

singular, affirmative mood, past tense, of the verb 'to do', 1875:

554 -5; cp. /kejje/, '(I) did (work), worked'). Appiah (1887), in

spite of the freedom consequent upon devising a wholly new script for

the language, nevertheless does not provide distinct symbolisations

for the back unrounded vowels at the same level as for the other

vowels; the high back quality is represented by a superscript dash,

, and the mid back quality by the superscript symbol ' ; in each case,

these diacritics are in combination with the basic symbol for a

consonant with the included vowel a.

The first systematically implemented Romanisation is found in

Goda Varma (1946), where the symbols i and e are used for the back

unrounded qualities; Emeneau uses the same symbols in his works,

marking length with a single dot after the vowel. The symbolisation
 25

in the present work is the same, except that the more usual length

marker /:/ is used.

It should perhaps be noted that contemporary practice concerning

the use of the Kanarese script for Coorg is to use the vowel deletion

diacritic on the basic consonant symbol for representing the high

back unrounded vowel; thus is ¡ra/ and V is /ri /. This means,

of course, that cr has a different value depending on whether it is

being used to transcribe Coorg ( /ri/) or Kanarese ( /r/), and that no

unambiguous representation of the consonant alone is possible when

this script is being used for Coorg. Regarding the mid back

unrounded vowel, the practice is to use the form for the consonant

followed by the mid front vowel (i.e. d , re, stands for Coorg

/re/ and /rë/) in all positions save word-finally, where e , ra

(hence standing for Coorg /ra/ and /rë/) is used. In this way,

current practice recognises both the relationship implicit in the

transcription proposed in Richter (1870) and that in Burnell (1873).

It is a consideration not to be lightly dismissed that this implement-

ation of the Kanarese script, while not 'phonemic', has gained wide

acceptance in Coorg, and apparently causes no difficulties in every-

day use. It will be seen in the next chapter that a solution of this

sort captures a number of regularities that are not primarily

phonetic.

3.0 Fieldwork for the present study

The main part of the fieldwork was carried out in Coorg between

May and November 1968. Prior to that, about twelve hours of informant

work had been completed in London, during September 1967, and this

provided a useful basis for preliminary analysis and subsequent work

 26

in the field. However, the data from these London sessions was later

seen as being quite heavily influenced by Kanarese, and has therefore

not been used as a primary source for the description of the language

in the following chapters.

host of the data which is represented in this study was collected

in Karada village, about ten miles west of Virajpet, and well within

the area which Coorgs generally consider to be the home of the Coorg

language and culture. But the initial fieldwork in Coorg was done in

hercara: here, the informant was N.Ii.Vasu; his parents have lived in

Mercara for some time, and his childhood was wholly spent in the

town, except for brief visits to the family home, which is in Karada.

His speech in many respects seems to be typical of his generation in

Mercara; it shows, for example, a fair degree of Kanarese influence

(chiefly at the lexical level), in connection with which the role of

the Government Arts College in Mercara is probably important (most of

the teaching there is carried out in Kanarese). However, Vasu and

his parents are all part of the larger family, Nadikerianda

( /nadike:riyanda /), whose ancestral home and lands are at Karada; the

fieldwork done in Karada subsequently revealed many respects in

which Vasu's speech was typical of that area, and which therefore

seemed to be characteristic of the Nadikerianda family. The inform-

ants in Karada (N.C.Subbayya and N.P.Subbiah) were a retired, well -

educated couple who had both been 'out' (in other parts of India) for

much of their lives; they had returned to Karada only within the

previous five years. They each spoke essentially the same variety of

the language, and this showed slight generation differences with that

of Vasu; these differences have been ignored in the preparation of this

work. Their educational attainments included a thorough knowledge of

 27

Kanarese and English, and may therefore be argued to be a disturbing

factor with regard to their Coorg speech; against this, however, it

should be noted that all the Coorgs in the Karada area apparently

know Kanarese fluently and English to a varying extent. Hence, in

respect of their knowledge of Kanarese (the language most likely to

interfere with Coorg), my informants were not untypical of the

people in the village generally. In practice, they showed themselves

to be ideal informants in many ways; they quickly became interested

in the purpose of the research and were extremely helpful in

pointing out respects in which they felt that their speech might

diverge from that of their contemporaries in the village; in all such

cases, the information was checked against the speech of whoever was

available. N.N.Belliappa, N.G.Belliappa, and N.N.Belliappa, who also

provided language data which is represented in this work, are all of

the Nadikerianda family too; although not maintaining frequent and

close contacts with Karada any longer, they show essentially the same

variety of speech as is found in that area.

 28

CHAPTER TWO

Transcription
 29

1.0 Introduction

This chapter is mainly concerned with a monosystemic analysis of

the sound pattern of Coorg, in terms of general phonemic theory. As

such, one of its purposes is to fit into the general format of the

monographs in the Linguistic Survey of India series currently being

produced by the Centre for Advanced Study in Linguistics at the

Deccan College, Poona (1967 - ) However, this analysis also provides

a convenient basis on which to compare some rather different proposals,

which are considered in the context of a discussion of phonological

distribution (pp. 63 -109).

1 .1 Inventory of phonemes: vowels

Coorg vowel phonemes have to be classified in terms of the

following phonological parameters: position (front /back; high/mid/

low/); length (short/long); and manner (rounded /unrounded). They are

set out in Figure 1.

Front Back

Unrounded Rounded

Short Long Short Long Short Long

11 ü
High i i: i i: u u:

n n
Mid e e: e e: o o:

Low (ae ) a a:

Figure 1. The Coorg Vowel Phonemes

(The parenthesised element // is of restricted occurrence.)

 30

In addition to these, however, a number of nasalised vowels

occur, as follows:
N N N N N n/
pp N
/v e, e:, e, a, a:, u,
/1, o: /,

and we therefore recognise the nasalisation marker /-1. /i / and

/ë:/ also occur, as the result of lengthening of /i, e/, respectively,

before the quotative particle (see p. 41). Emeneau (1938: 123, fn. 1)

notes that 'nasalisation is best evaluated phonemically as a feature

of the whole word and not of any separate vocalic or consonantal

phoneme'. Accordingly, he represents it in his transcription by

superscript n immediately after the form in question. Very few

lexical items show this nasalisation, and it is therefore not

convenient to set up a whole series of additional, nasal vowel

phonemes for Coorg: Emeneau's solution is probably the best, but in

this work the more conventional use of the tilde on the vowels will

be followed.

1.2 Inventory of phonemes: consonants

The relevant parameters for the consonant phonemes are: place

( labial /dental/retroflex/palatal/ velar/(glottal) ); length (single

geminate); and manner ( stop/nasal / lateral/trill /fricative/approx-

imant; voiceless/voiced). The consonant phonemes are illustrated in

Figure 2:
 31

Retro-
Labial Dental Palatal Velar (Glottal)
flex

voiceless p t t c k

Stop
voiced b d d j g

Nasal m n n ñ ri

Lateral

Trill r

voiceless (f) s s (h)

Fricative
voiced (z)

Approx-
v Y
iman t

Figure 2. The Coorg Consonant Phonemes

(The parenthesised elements /f/, /h/ and ¡z/ are of restricted

occurrence.)

The category 'Labial' in Figure 2 covers bilabial and labio-

dental articulations, as indicated by the symbols in that column;

similarly, 'Dental' covers dental to post -dental articulations, and

'Palatal' post -alveolar to palatal articulations. /c/ and /j/ are

post -alveolar affricates, but are conveniently classed with the true

stops in most phonological statements. It will be seen that the

categories 'Glottal' and 'Fricative (voiced)' are of marginal status.

Full details of the allophonic realisations of all these elements are

given in section 3.0 below.

 32

The issue of consonant length (which is not illustrated in Figure

2) represents a problem in analysis which will be considered further

in section 4.2.3; for the moment, the traditional categories 'single'

and 'geminate' will be made use of.

1.3 The word

In what follows, a unit called the word will be appealed to

fairly frequently. In all cases, 'word -initial' may be taken to mean

'potentially post- pausal (utterance -initial)', while 'word -final'

should be interpreted as 'potentially pre- pausal (utterance-final)'.

The uninflected forms of nouns (i.e. the 'subjective singlar') and

of verbs (i.e. the second person singular imperative) are, in each

case, a single word; but those inflected forms which show just one
;

main stress are also, in each case, a single word.

1.4 Stress

There is no contrastive stress in Coorg in the sense shown in

English import (verb/noun). However, it is not entirely predictable

from the word alone, as defined above. In order to state the occurrence

of stress a phonological unit called the syllable has to be set up,

compatible with the larger phonological unit, the woid. (i.e. every

word is wholly made up of syllables). Then, every word- initial

syllable has full stress; every word-final syllable has secondary

stress. Word- medially, a syllable has full stress if it contains a

long vowel nucleus, or a short vowel nucleus followed by a geminate

consonant cluster (the first consonant of the cluster being separated

from the second by the syllable boundary). there a syllable has

neither a long vowel nucleus nor a final consonant which is part of a

geminate cluster in the sense just described, it carries no stress.

 33

However, although this statement covers hundreds of examples, it

does not account for a large number of forms such as /ku:kudike/,

'large cooking vessel for rice', which shows full stress on the first

two syllables; in this case, and all the others of this type, we set

up an internal word boundary /4, and write /ku: +kudike/. Not

surprisingly, the occurrence of f +4 is frequently a guide to

morphemic analysis (e.g. /ku:1i /, 'rice'; /kudike/, 'cooking pot').

However, in cases such as /nu:putti/, 'vermicelli rice' (cp. /nu:li /,

'thread', and /putti/, 'rice preparation'), the internal boundary is

not marked, since the stress pattern is predictable with reference

to the phonemic sequence as it stands.

All syllables in word-initial position are of the form (C)V(C)

or (C)V; medially, they are CV(C) or CV, and a single intervocalic

consonant is to be taken as (the onset) of the same syllable as the

following vowel; and since no consonant occurs word -finally, except

in a few loans, such as /tapa:l/, 'post, mail' (from Hindi), word -

final syllables are open. Geminate consonants do not occur after a

long vowel; this is interpretable as a general condition on the

overall length of the syllable unit.

2.0 Contrasts

In what follows, the abbreviations '1, 2, 3, sg, pl' stand for

'first, second, third person singular, plural', respectively; 'imp'

is for 'imperative', 'caus' for 'causative', 'comp' for 'completive',

'opt' for 'optative', and '(in)tri for '(in)transitive': 'masc' and

'fem' stand for 'masculine' and 'feminine', and 'prox' and 'rem' for

'proximate' and 'remote'.

 34

2.1 Vowels

Wherever possible, these are illustrated in two positions:

(a) in the first syllable of the word, occurring with stress, and

(b) word -finally. In each case, the relevant element is underlined.

2.1.1 Short vowels, position and manner

/o/ does not occur finally, except in a few loan words such

as /foto /, 'photograph', /re:diyo /, 'radio', etc.

(i) High front versus mid front:

(a) /nilli/ stand (2 sg imp)

/nelli/ paddy

(b) /kani/
d narrow passage

/kande/ bulb of plant

(ii) High back unrounded versus mid back unrounded versus low:
n u
(a) /kittici/ was possessed (3 sg, 1, 3 pl)
n n
/'_ketti/ bundle

/katt / cot, bedstead

(b) /takkj/ language

/takke/ headman (of village)

/takk./ sufficient

(iii) High back rounded versus mid back rounded:

(a) /Pull/ orange, sourness

/poli/ break (tr; 2 sg imp)

(iv) High front versus high back unrounded versus high back
rounded:

(a) /kiri/ lip

/kirke/ small

/kurkë/ fox, jackal

 35

(b) /ma:di/ upper storey

fl
/na:djj jungle clearing

/ma:du/ may do

(v) Mid front versus mid back unrounded versus mid back
rounded:

(a) /etti/ arrive (2 sg imp)

/etti/ bull, bullock

/otti/ press (2 sg imp)

(b) /jalle/ sugar cane

/mallëf cockerel

( vi) Mid front versus low versus mid back rounded:

(a) /ketti/ cut, blow

/kátti/ wash (plate, hands) (2 sg imp)

/kotti/ tip, nipple

(b) /ba:ne/ open lands

/ba:n / sky

(vii) High front versus mid back unrounded versus high back
rounded:

(a) /?ppa/ will be (3 sg, 1, 3 pl)

/ëppa/ will raise (3 sg, 1, 3 pl)

/ppa/ will plough (3 sg, 1, 3 pl)

(b) /boll silver

/kullj/ dwarf (mast)

/boll2/ Bollu (man's personal name, pet

form)

(viii) Mid front versus high back unrounded versus mid back
rounded:

(a) /kemmi/ cough (2 sg imp)

/timmi/ sneeze (2 sg imp)

/kombi/ horn
 36

(b) /bille/ disk, badge

/billi/ bow (archery)

(ix) High front versus low versus high back rounded:

(a) /ikka/ now

/ákka/ then

/ukkate/ without boiling over

( b) /kari/ curry

/tare/ short, kind

/karE/ may be digested

2.1 .2 Long vowels, position and manner

/i: /, /i: /,
/9
ë:/ and // do not occur word-finally, except

where this is in the first syllable of the word, as in /ti: /, 'tea'.

(i) High front versus mid front:

(a) /cile/ bag

/ce:le/ sash

(ii) High back unrounded versus mid back unrounded versus

low back:

(a) ki:li lower, below

II
I
/ke:li/ ask (2 sg imp)

/ká:li/ grain

(iii) High back rounded versus mid back rounded:

(a) /kuta/
._ quarrel

/ko:ta/ coldness (not weather)

(b) /da:ruil and who ( /u:!, co- ordinative

marker)

/da rál/ : someone ( /o: /, indefinitive

marker)
 37

(iv) High front versus high back unrounded versus high

back rounded:

(a) /ki:ti/ torn piece

kida/ place below, down

/ku:ti/ shout, noise

(v) Mid front versus mid back unrounded versus mid back
rounded:

(a) /ne:ru/ may hang (intr)

/ne :ru/ may rise up

/to :ru/
: may leak

(b) /ayiiga121/ even they ( /e: /, emphatic marker)

/elliyJ somewhere ( /o: /, indefinitive

marker)

(vi) I;id front versus low back versus mid back rounded:

(a) /e:le/ which way?

/a:le/ that way

/o:le/ screwpine leaf

(b) /avalli/ even she (le: /, emphatic marker)

/aval.21/ she? ( /a: /, interrogative marker)

le :való1/ someone (fem) ( /o: /, indefinitive

marker)

(vii) High front versus mid back unrounded versus high back
rounded:

(a) /i:rici/ sawed (wood) (3 sg, pl)

fl

e:rici/ (liquor) rose to the head

/tri/ village
 38

(viii) Mid front versus high back unrounded versus mid back
rounded:
fl

(a) /teni/ honey

n ,
/tindici/ was finished (3 sg, pl)

/to:ndici/ leaked (3 sg, pl)

(b) see (v) above

(ix) High front versus low back versus high back rounded:

(a) /ni:ri/ water

/na:ri/ fibre of plant

/nu:ri/ one hundred

(b) see (iii) above

2.1.3 Vowels, short versus long

(i) High front:

(a) /kiri/ lip

/ki:ri/ mongoose

(ii) Mid front:

(a) /eri/ ragi (Eleusine corocana)

/e:ri/ wall of paddy field

(b) /a:nef elephant

/ta:n2_1/ even oneself ( /e: /, emphatic

marker)

(iii) High back unrounded:

(a) /ili/ get down (2 sg imp)

it It

/i:li/ drag (tr; 2 sg imp)

 39

(iv) Mid back unrounded:

t,
t,

(a) /eli/ leave (position) (2 sg imp)

/e li/: seven

(v) Low back:

(a) /madi/ purity, cleanliness

/madi/ upper storey

(b) /ku:vW shallow well

/pu :vá; ì flower? ( /a: /, interrogative

marker)

(vi) High back rounded:

(a) /nudi/ heel of foot

girl

(b) /kadu/ mustard seed

/paduJ and swamp ( /u: /, co- ordinative

marker)

(vii) Nid back rounded:

(a) /podi/ powder

/po: di/ fear

2.1 .4 /ae/ versus /e :/ versus /a:/

(a) /baeriki/ (financial) bank

/be : igici/ (rain) cleared

/a: rigi/ to that (thing), to it

2.1.5 Nasal versus oral vowels

(i) High front, short:

(a) /iva/ she (prox)

/ivë/ he ( prox )
 40

(b) /kavi/ poet

/kávl/ lie face down (2 sg imp)

(ii) Mid front, short:

(a) /bevari/ sweat

N w!
/keva/ wall

(b) /kanivei valley

N N
/kanuvW plains to the east of Coorg

(iii) Mid back unrounded, short:

(b) /cellavë scaly anteater

0+
/illZV/ relative (masc)

(iv) Low, short:

(a) see (i) (b) above

(v) High back rounded, short:

(a) /keyyadi/ let (it) be bored, dug

/keYyadi/ let (it) feel prickly

(vi) High front, long:

/kavi:ndi/ saying 'poet'

fl,
saying 'lie face down'

(vii) Mid front, long:

(a) /e:va/ which person? (fem)

"
/e:ve/ which person? (masc)

(viii) Mid back unrounded, long:

/ba:ve:ndi/ saying 'brother -in -law'

/má:ve:ndi/ saying 'father -in -law'

 41

(ix) Low back, long:

(a) /na:vu/ tongue

/mávi/ father's sister, mother-in -law

(x) Mid back rounded, long:

(a) /mo:va/ daughter

/mó:vë/ son

ftio
/P.I
/and /ë ; are only found in the quotative construction in

the data, i.e. followed by -/ndi /, before which all short vowels are

obligatorily lengthened. Such occurrences ( (vi) and (viii) above)

should perhaps be taken as word- final, but it will be seen that this

complicates the statement of what a word is.

2.2 Consonants

Retroflex consonants do not occur initially, save for stops

in a few loan words; e.g. (from English) /da:ktri /, 'doctor' (where

the English alveolar stops are interpreted in terms of the Coorg

system as being non-dental and hence retroflex), and (from Hindi)

/tapa:1/, 'post, mail' (where the initial stop is retroflex in the

source language). Where possible, contrasts are illustrated

(a) word -initially and (b) word -medially.

2.2.1 Contrasts of place

(i) /p/ vs. /t/ vs. /t/ vs. /c/ vs. /k /:

(a) /ratti/ space before house

/tattë/ goldsmith

/cattuva/ wooden spoon

/kattë/ dam in tank or pond

 42

(b) /ka:zi/ coffee

/pa:te/ cockroach

/ra:te/ pounding board

/a:ce/ day of the week

/ka:ke/ crow (n)

(ii) /b/ vs. /d/ vs. /d/ vs. /j/ vs. /g/:

(a) /bale/ 'bangle'

/dale/ noose

/,tale/ pliant stick

/gale/ long stick

(b) /a:di/ beginning, origin

/ca:di/ slander

/ja:di/ jasmine (Jasminum grandiflorum)

/ra:gi/ ragi (Eleusine coracana)

Instances of an ungeminated voiced bilabial stop are rare inter -

vocalically; /kurube/, 'shepherd', is one example. The form /kaybi/,

'sugar cane', should also be noted, showing another case where single

/b/ has to be set up outside word-initial position.

(iii) /m/ vs. /n/ vs. /n/ vs. /ñ/ vs. /A/:

(a) /ma: le/ necklace

/na:ni/ I

/ña:na/ shame

(b) /a:me/ tortoise

/a:ne/ elephant

/ba:ne/ open grazing land

/pu:ñe/ cat

/a riuñni/
: male child
 43

A further skewness of consonant distribution is evidenced here:

apart from the fact that initial retroflex nasals are not found,

it is also the case that /ñ/ is found initially in only a few words

(less than ten, even allowing compounds to count as distinct words),

and is found ungeminated intervocalically possibly only in the

item /pu:ñe/ given above. However, it occurs more commonly in inter -

vocalic position when geminated (as in /kuññi /, 'Coorg child').

Further, /ri/ does not occur initially; and it is rare intervocal-

ically, regardless of whether it is geminated or single: the last

form cited above is a compound of the elements /a:li /, 'man' and

/kuññi /, ' Coorg child', and /caE o:le /, 'chain' shows the stress

pattern of a compound also, although in this case the identification

of the elements is less certain (cf. /o:le/, 'ear ornament, earring').

(iv) /1/ vs. /1/:

( a) /a: 1i/ banyan tree

/a:li/ man, servant

( b) /lekka/ lesson

The only recorded instances of /1/ in word -initial position involve

loan words (such as the item given above, which is probably via

Kanarese), or else are sound -symbolic, as in the single Dry entry

under /1/ -, /lot/, /lotto: /, 'noise of a wooden cattle -bell'.

(v) /f/ vs. /s/ vs. /s/ vs. /h/:

(a) /fawn/ fan

/sa:la/ loan

/ sa:le/ school

/he:sige/ disgust (exclamatory)

 44

(b) /ka:fi/ coffee (^' /ka:pi/)

/mi :se/ moustache

/de :sa/ country, nation

/a:ha:ra/ food

/h/ is rare intervocalically; the only instances are found in

Sanskritic vocabulary items, and in consciously 'correct' styles of

speech (where, for example, /sa:hasa/ may occur instead of the more

usual /sa:sa /, 'effort').

(vi) /v% vs. /y/:

(a) /va:le/ 'ear ornament'0+/o:le/)

/va:la/ cardamom (.4/e:la/)

(b) /ka:va/ (crow) will caw

/ga:yai wound

The occurrence of ¡v/ and /y/ word- initially is discussed more

fully below (pp. 87 -92); a phonological alternation is involved

here, as indicated by the alternative forms noted above.

2.2.2 Contrasts of manner

(i) /p/ vs. /b/ vs. /m/ vs. /f/ vs. /v/:

(a) /va:ti/ song

/ba:de/ inner hall

/ma:le/ necklace

/fEe n/ fan

/va:le/ ear ornament

(b) /pa:la/ exclamation of pity

/kurub/ shepherd
 45

/a:me/ tortoise

/ka :fi/ coffee

/ka:va/ (crow) will caw

(ii) /t/ vs. /d/ vs. /n/ vs. /1/ vs. /r/ vs. /s/ vs. /z/:

(a) /ta:ti/ amulet

fl

/da :ri/ who

n
/na:ni/ I

/la:di/ tape

/ra:te/ pounding board

/sa:sa/ effort

(b) /ku:ti/ shout, noise

/cu:di/ needle

/u:ni/ gums

/nuli/ thread

/u:ri/ village

/mu:si/ smell (2 sg imp)

/dazan/ dozen

/z/ only occurs in the data in loanwords, and never initially.

(iii) /t/ vs. /d/ vs. /n/ vs. /1 /:

(b) /ka:ti/ bison

/ca:di/ slander

/a :ni/ nail (metal, wood)

/ga:li/ wind
 46

(iv) /c/ vs. /j/ vs. /N vs. /s/ vs. /y/:

(a) ¡ca:ne/ baldness

/j,a: ga/ place

/ña:na/ shame

/sa :le/ school

/la:lai cardamom

(b) /a :ce/ day of the week

/bo:l.a/ beauty

/pu:ñe/ cat

/desa/ country, nation

/ga:ya/ wound

(v) /k/ vs. /g/ vs. /ri/:

(a) /ke:ri/ rope

/Le:na/ thought

(b) /a:ka/ large field

/ja:Ea/ place

a:riuññi male child

2.2.3 Contrasts of length

The distinction between single and geminate consonants (of

equivalent place and manner) is found only in the context of a

preceding short vowel (see above, p.33).

(i) Stops, voiceless:

/kada.Re/ steps for getting over fence

/kapere/ frog

limit
fire
 47

/pata/ picture

/patta/ coronation

moon

jaggery cake

mean fellow

husked rice

(ii) Stops, voiced:

II

/ kurube/ shepherd

/ubba/ poles in slots forming a gate

ear (grain)

kite, hawk

place below

obstacle

holiday

Holeya girl

/noga/ yoke

/ogaadi/ let (child) wriggle on belly (opt)

(iii) iasals:

/keái/ ear

/emme/ female buffalo

/j ana/ people

/ c enea/ small quantity

/mana/ sand
- 11

/panni/ mud, land property

 48

/ñ/ and /ri/ never occur singly after a short vowel.

(iv) Laterals:

/kale/ scar

/jalle/ cane of sugar

play

thief (fem)

(v) Approgimants:

/ava/ she (rem)

/avvë/ mother

/days/ kindness

/bayYa/ place behind

The phonemes /r, f, s, s, h, z/ never occur geminate in the

language.

3.0 Allophonic statement

3.1 Vowels

It will be seen from Figure 1 (p. 29) that the rounded/

unrounded opposition only applies to mid and high back vowels. Front

vowels are lip- spread, in greater degree for the long vowels than

for the short, and for the high vowels than for the mid. Back

unrounded vowels in high and mid positions are slightly less lip-

spread than their corresponding front vowels. /a:/ has neutral lip

configuration, while /a/ has allophones that range from unrounded

to slightly rounded (see below, p. 54). The degree of rounding of the

back rounded vowels is greater for the long and the high vowels than

for the short and the mid.

 49

All vowel articulations are characterised by:

(i) relatively longer duration, with a correspondingly less

centralised quality, when followed by a single voiced consonant

allophone;

(ii) relatively shorter duration, with a more centralised

quality, when followed by a geminate voiceless consonant group;

(iii) a secondary articulation of retroflexion when followed by

a retroflex consonant;

(iv) voicelBs articulation when occurring without stress before

a voiceless consonant or, quite often, word -finally.

By (i) and (ii) three degrees of phonetic duration and quality

are established for the allophones of the short vowel phonemes and

similarly for those of the long, yielding six perceptible degrees in

all. Within one speech style (i.e. as might be found with a

particular informant on a particular occasion) the long degree for

the short vowels is shorter than the short degree for the long vowels.

For reasons of clarity of presentation, the phonetic transcription

used below does not mark these conditioned degrees of duration.

The secondary articulation of retroflexion ( (iii) above) is

more noticeable with the allophones of the long vowels than with those

of the short, and is confined to the latter part of their articulation.

For front vowels, retroflexion is accompanied by a centralised quality,

and for the high and mid back vowels (though with these the

distinction is harder to find) by a more truly back articulation,

together with slight lowering. In each case the effect is that of a

diphthong with a weak final element. /a/ and /a:/ show no

perceptible diphthongal glide, though retroflex 'colouring' is

evident. A subscript dot under the vowel symbol (or associated length
 50

marker) indicates secondary retroflexion in the phonetic trans-

cription below.

The phonemes /i /, /i: /, /e /, /e: /, /u/, /u: /, and /0/, /o:/

have allophones with relatively tense onset when they occur

initially. The result is a diphthongal type of articulation, with

a weak first element, as the tongue is momentarily displaced to the

greatest extent from its position of rest (the onset) and sub-

sequently reaches the slightly more centralised position for the main

component of the diphthong. The distance between the initial and

final positions is not so great for those vowels which have relatively

tense articulations throughout; /i: /, /e: /, /u:/ and /o:/ are tenser

than their short counterparts, and /i/, /u/ are tenser than /e/,

/o /: in each case the tense onset is correspondingly less prominent

where the following vowel articulation is tenser. The tense onset

is indicated by 0-1 (for the front vowels) and

vowels) in the phonetic transcription used below; the degree of

tenseness is not marked.

The remaining details concerning the allophonic realisations

of the vowel phonemes are illustrated below for each vowel phoneme

in turn.

3.1.1 Short vowels

(i) /i/ (short, high, front) has the following allophones:

PG] short, fairly lax, slightly retracted

from front position, occurring always with stress;

/ikka/ [I'tk'k now

/indi/ [tndut* today

/kiri/ [l}ttk lip

 51

[4] extra short, lax, fully centralised,

occurring medially without stress;

M
/kudike/ [Okwr, ; cooking pot

/ki riñj i/ L K tL-ÉñLj from the lip

[ti'j half long, fairly tense, slightly raised

and fronted from the position of [t],

occurring utterance-finally:

/kar / L1kEc -L i
curry

(ii) /e/ (short,mid, front) has the following allophones:

[64 ] short, lax, slightly retracted, occurring

with stress;

/bedi/ !I*144 L ' heat of the sun

/kela/ k1Co{. stomach

/rekke/ AVAE] wing

[ eT ] short, tenser, less retracted, occurring

with stress before /y/;

/ney/ [1nerj ] butter

[ £I. extra short, lax, retracted, occurring

medially without stress;

/pi reke/
: WtE.F(«] small cucumber

[E j short, lax, slightly retracted, occurring

utterance -finally;

/man lmen ] house

 52

(iii) /i/ (short, high, back, unrounded) has the following

allophones:

[ uu,Yj short, fairly lax, fairly advanced from

back position, occurring always with stress;

n ,e

/kirke/ (kuokn; small

I
/tripti/
,r

[ uz is ] satisfaction

[iUir{: ] the whole

[.j extra short, lax, fully centralised,

occurring medially without stress;

/kudire/ klpdwitJ horse

/adiñ j i/ [ldF from it (reni )

This allophone of /i/ is generally articulated further back than

[+t ] of /i /; however, for most speakers the distinction is lost before

a palatal consonant. Thus /adinji/, 'from it (rem)' (cp. /adi /, 'it

(rem)') and /alliñji /, 'from there' (cp. /alli/, 'there'), both show

a medial vowel articulation which is in the region Nt j to [iktj,

without contrast.

[w.] short, fairly lax and advanced, with a

tenser, raised offglide, occurring only in

utterance-final position;

/mull {'Mnt« thorn

(iv) /e/ (short, mid, back, unrounded) has the following

allophones:

[/lj short, lax, fairly advanced, occurring

always with stress;

n u
/embadi/ [AL(rba ut i eighty
,+ ,+

/ll/
,.( -

L9A iuL- win (2 sg imp)

 53

/gët.i gë/ [gnjt ] shrewd fellow

[ A ] short, lax, fairly advanced, occurring in

utterance-final position;

/mañsë/ [Ing45 ] man

(v) (short, low) has the following allophones:

[a.t] short, fairly lax, raised and retracted

from cardinal position, occurring with stress

in utterance- initial position before phonetically

short consonants;
, .J -
/ádi/ [d1-)11LT&; it (rem)

/age/ [kijE ] paddy seedling

/adi/ place below

[ shorter, lax, just lower and further back

than mid -open central, occurring with stress (a)

utterance -initially before a phonetically long

consonant articulation, and (b) following a

consonant in the initial syllable of a word;

/akka/ [ 6kaA] then

/árigadi/ [lenyri: ] shop

/máddi/ [[,i M«7t medicine

/madi/ [1tiC L- ] enough

[8 ] extra short, lax, mid -open, fully centralised,

occurring medially without stress;

1
/a:rane/ [ct:rane] sixth

/accaki/ f#Sakuc= i that much

/ma:diyandi/ ¡q; ,atr; doing

 54

For a number of speakers [8 j maintains its central quality

even when it occurs long with emphatic stress (only in the contin-

uative aspect forms of verbs); thus, beside the last item cited above

there is also
N NNI
:ai-IJA a :: rtm milk=L-1 he went on doing (it)

[
w J as for [ a ] but slightly lip- rounded,

occurring medially without stress before /1/

or /ndl followed by a back rounded vowel;

/ma dándu/ Irq. rnc.c j must do

: L

w
/ekkalu:/ 0i4&46.1."3 always

For the first of these forms, some speakers have /ma:dondu/,

showing phonemic merger of /a/ with /o/ in this environment. The

second item is related to the form /ekko:lu: /, 'always'; in this

regard, compare what was noted above for [a j under conditions of

emphatic stress. In this item too, a phonemic merger has apparently

taken place (based on the features of length and lip- rour_ding),of /a/

plus extra length to /o: /; little or none of the 'emphatic meaning'

remains.

[ a4] short, lax, advanced, occurring utterance-

finally;

/maril Lm YU4j tree

(vi) /u/ (short, high, back, rounded) has the following

allophones:

[Al short, fairly lax, fairly advanced,

occurring always with stress;

/kuru/ L QrcCJ seed

 55

/punde/ [11#114nc; wetness

.-.

/nucci/ broken rice grains

[j extra short, lax, fully centralised,

occurring medially without stress;

n
/kiduve/ rte4fil
hawk

/garuda/ [implutt,i] eagle

[W1 half long, fairly tense, slightly raised

and retracted from the position of [O],

occurring utterance -finally;

/pul [gyp u ] worm

/ca:vuJ [ a; biliJ death, corpse

(vii) /o/ (short, mid, back, rounded) has the following

allophones:

[DA. ] short, lax, fairly back, occurring with

stress before a short consonant or before a

consonant preceded by /r/ or homorganic nasal;

/kode/ umbrella
w
/porme/ [itch3`'rrve- outside

/nombala/ ae pain

[ O* ] short, lax, often fairly advanced, as

[OiT ] (except when followed by velar or retro-

flex consonants), occurring with stress before

phonetically long consonants;

/kokki/ W lvcuM
,l

beak

/kotti/
..
LL Orf*l,LTJ
w
shed

/toppi/ L (.43uLttti- leaves

I ut
/ gotti/ [aoh- .1 knowledge
Jw w
 56

[4ic] short, lax, advanced, with some deround-

ing, occurring with stress before /y/;

t
/2ymbadi/ `"c 4cmba1utr nine

/oy/ I'O j] nail (finger, toe)

[O] half long, fairly tense, close to cardinal

position, occurring utterance -finally (in just

a few loans);

/re:diyo/ hei1 qo-] radio

3.1.2 Relative positions of the short allophones

The diagrams in Figure 3 show the relative positions of

the short vowel allophones discussed above, and indicate the phonemic

groupings.

Figure 3: The short vowel phonemes

 57

3.1 3 Long vowels

(i) /i:/ (long, high, front) is realised everywhere as:

[ -'] long, tense, slightly lower than cardinal

position, occurring only with stress;

/i:/ [11-4:] this

/mi:se/ [Im< :SE] moustache

/i:ti/ [' :tt' 1 spear

(ii) /e:/ (long, mid, front) has the following allophones:

[e :] long, tense, close to cardinal position,

occurring with stress before a consonant which

is followed by a high vowel or before a word

boundary (indicated by a space in the phonemic

transcription) or utterance-finally;

/te:ni/ [ILe:n1uT; honey

/e: mane/ [
Je: mg ] which man?

/avai J Oat-w tie.:] she (rem) (le:/, emphatic

marker)

/e:/ Lie= which?

[Eat] long, tense, with a weakly retracted final

tongue position, occurring with stress before /r/

plus a non -high vowel;

/ne ra/
: [1nEA: L CO 1 time

[ET :] long, tense, occasionally closer to

cardinal position [a] than [E], occurring only

with stress before a consonant other than /r/

plus a non -high vowel;

/kemë/ Wrr1A6 barking deer

 58

/p°-te/ town, bazaar

/g°'na/ [14ET:114i thought

(iii) /i:/ (long, high, back, unrounded) is realised every-

where as:

[ut:] long, tense, close to cardinal position,

occurring only with stress;

[t : P¿] will finish (1 sg)

n
/ki:da/ Lkcu:ctt place below, down

/ni : la/ Unut:&t-i1 length, throughout

(iv) :/
e
/It
(long, mid, back, unrounded) has the following

allophones:

[ó*l] long, tense, occurring with stress before

a consonant plus high vowel;

/ke: pi/ [ikk rT: pi- 3 will ask, hear (1 sg)

n
y
li/
.
u
L
!

/P :
Itk
ÇuiT1
.
seven

[A :] long, tense, close to cardinal position,

occurring with stress before a consonant plus

non -high vowel;

/é:lane/ nlaV1£] seventh

(v) /a:/ (long, low, back) has the following allophones:

[a.( :] long, tense, strongly advanced, often

with a palatalised off -glide, occurring with

stress before /y/;

/ ri/ 1_1a [4 J carpenter

/ott ya/ ÌM41 j a4

nn
] urging (n)

[Q.=] long, tense, close to cardinal position,

occurring with stress elsewhere than before /y/;

 59

/a:ce/ [t Q£
a
day of the week
t

/na:vu/ [In:Wü; tongue

n
/a:di/ [ q:r=- goat

/a:/ [1Q:ß that (demonstrative)

(vi) ¡u:/ (long, high, back, rounded) is realised everywhere

as:

[i:] long, tense, slightly lower than cardinal

position, occurring only with stress;

L
/ u: ri/ village
ú:lrut*!

/mu : nd i/ [ * u: vichi
nn
t three

/ku: lin / [
I

k(1c f a(u_
cooked rice

/pu :/ [tP :; flower

(vii) /o:/ (long, mid, back, rounded) has the following

allophones:

[o:3 long, tense, close to cardinal position,

occurring with stress before a consonant plus

a high vowel, or before a word boundary or in

utterance -final position;

It
-1 l 4c,
/ko :li/
: Lk e417 ], stick

AA
/da:ro: mañsë/ is ar I

AvrnrIeN] some man or other

(/o:/, indefinitive marker)

/po:/ ['ft'O: go (2 sg imp)

[Jl CI long, tense, with a weakly advanced

final tongue position, occurring with stress

before /r/ plus a non -high vowel;

/bo:re/ [X71 :YE] difference

 60

[p:] long, tense, close to cardinal position,

occurring with stress before any consonant other

than /r/ plus a non -high vowel;

/bo:nda/ [1b3:n14-4! not wanted

/o:te/ [634E1 reed

/ro:ma/ [!r3:f1I-I hair (of body)

(viii) /are/ (long, low, front) is realised everywhere as:

[a04] long, tense, slightly raised from

cardinal position, occurring only with stress;

/faen(i)/ [if :1(ú4 fan

/.bae ik(i ) /:Yi(Culr) (financial) bank

3.1.4 Relative positions of the long allophones

The diagrams in Figure 4 indicate the positions of the

allophones of the long vowel phonemes, and mark the phonemic groupings.

Figure 4: The long- vowel phonemes

 61

3.2 Consonants

(i) /p, t, t, k/ are fairly tense and voiceless stops (bilabial,

dental, retroflex and velar, respectively). They are noticeably

but not very strongly aspirated when they occur prevocalically with

stress and are hardly aspirated at all elsewhere. They are unreleased

when they occur as the first element of a geminate cluster. /k/ is

pre -velar [+] utterance -initially when followed by a high front

vowel and intervocalically when preceded by a high front vowel; it is

post -velar [k] utterance -initially when a mid -open to low back vowel

allophone follows, and intervocalically when a mid -open to low back

vowel allophone elsewhere, it is mid -velar [ k] :

/ki: ti/ LKt :tw ] torn piece

nl

/ikka/ [1tkkQ-+ j now

/kela/ OkibtaA] stomach

/rekke/ LfElk(! wing

/ko:ta/ .C tot coldness (not climate)

a
/akki/ [144:11 husked rice

(ii) /b, d, d, g/ are the corresponding lax and voiced stops.

They have no feature of aspiration, and are fully voiced intervocal-

ically. They are unreleased when they occur as the first element

of a geminate cluster. /g/ has the same positional variants as noted

above for /k/. /d/ is a retroflex flap [r] when it occurs singly in

intervocalic position. /d/ in the same position usually has slight

but noticeable friction on release (not marked in the phonetic trans-

cription here). All single voiced stops are slightly shorter in

duration than their voiceless counterparts.

(iii) /c, j/ are in most positions blade -postalveolar

 62

affricates. /0/ is fairly tense and voiceless, /j/ is lax and fully

voiced intervocalically. They each share the aspiration character-

istics of the corresponding stop consonants. /c/ is [r], and /j/

is [d], when occurring as the first element of a geminate cluster.

(iv) /m, n, n, ñ, ri/ are voiced nasals. /n/ is dental when

followed by /t, d/, but post -dental elsewhere. For the rest, place

of articulation is as for the corresponding stop consonants.

(v) /1, 1/ are voiced laterals. /1/ is post -dental, and clear

in all positions; /1/ is retroflex and fairly dark (pharygealised

rather than velarised).

(vi) /r/ is a single, voiced post-dental flap [t] when

preceded or followed by a front vowel or voiced dental stop. When

it occurs intervocalically between other than front vowels, or with

a following voiced non -dental stop or affricate, it is an alveolar

flap or trill [r] (generally 1 to 3 taps; 1 before or after a voiced

stop or affricate, 1 or 2 after a short vowel, 2 or 3 after a long

vowel). It is voiceless [r] or [r,], when it is part of a cluster

with a voiceless dental stop, or non- dental stop or affricate. It

is a retroflex flap, followed by a 2 to 3 tap apico -alveolar trill,

when it follows a retroflex consonant (voiceless if that consonant is

voiceless, otherwise voiced):

/kari/ [ kpl1-J curry

/kere/ [WALE ] tank

/kare/ M] stain

/kuru/41n1 seed
ft
/tripti/ satisfaction

/bra: ma/ [rat: r n-'ï village

 63

/kirke/ [tkút jì small

/kutri/ LKlltilz pound, beat (2 sg imp)

w -óo

(vii) /f/ is a voiceless labiodental fricative.

/s, s/ are apico -postdental and blade -postalveolar

voiceless sibilants, respectively. /s/ has a retroflex allophone

[S] when followed by a retroflex consonant.

/z/ is the voiced counterpart of ¡s/

/h/ is a voiced glottal spirant [A] intervocalically;

initially it may have voiceless onset, or be voiceless [k],

(viii) /v,y/ are labiodental and palatal approximants,

respectively. /v/ has lip rounding except when it is preceded or

followed by a front vowel, in which case it has only lip compression

(see Ladefoged, 1971: 62 for this distinction), and is somewhat

closer; these two allophones are distinguished in the transcription

as [W] and [y], respectively.

/y/ is [jj in all positions.

4.0 Distribution

4.1 Vowels

4.1.1 Clusters

The Coorg material cited from Emeneau's fieldnotes in DED and

DEDS recognises four vowel clusters, of which the following are

examples (given here as they occur in DED and DEDS):

todia small garden

perie elder

podea sari
II

elee youth
 64

With regard to all these, it should first be emphasised that

such postulated clusters are of very low frequency in the sources

cited, and apparently contrast with forms such as (from DED):

n
odeye husband

However, none of my informants recognised such a contrast, and

all had a perceptible palatal approximant between the medial and

final vowels of all these items:

/todiya/ [tlrtja j small garden

/periye/ il'e4f4J/1-] elder

/podeya/ [IPIt4L1EE:14
sari
N /podiya/ [Pitrijai
/ëleyë/ £Jn ] youth

/ odeyë/ Ni¡aá /1 husband

It would of course be possible to eliminate /y/ from the trans-

cription in these contexts since it could be argued that the inter -

vocalic palatal approximant is predictable with reference to the

front quality of the preceding vowel; however, this would not affect

the issue regarding vowel clusters in the language, since such

'clusters' would have a purely technical status.

In addition to the TED and DDS material cited above, DBIA

records, from the same source, a further possible instance of a vowel

cluster:

go:due ( ?go:dve) wheat

For my informants, however, this item contains a consonant

cluster (-/dv / -) rather than a vowel cluster; the voiced dental stop

is followed by a voiced labiodental approximant with no inner rounding,

which is the appropriate form of the expected allophone of /v/ in

this context (see p.63 ) .

 65

4.1 .2 Long_ vowels

These generally do not occur in utterance -final position, except:

(i) in monosyllabic items:

/1 :/ this

/e:/ which?

/a:/ that

/pu :/ flower

(ii) as realisations of the emphatic marker (/e:/), the

interrogative marker ( /a /), the co- ordinative marker

: ( /u: /), and

the indefinitive (distinct from the indefinite form /ori /, 'a',

'one') marker ( /o: /):

/a: maneye:/ that very house, even that house

(cp. /mane /, 'house')

/ori mu:diya:/ a girl, is it? (cp. /mu:di/,

'girl')

/na:yu: pu:ñeyu:/ dog and cat (cp. /na:y, 'dog';

/pu:ñe/, 'cat')

/elliyo:/ somewhere or other (cp. /elli /,

'where ?')

(iii) in loanwords, such as /re:diyo/, 'radio'.

4.1.3 Back unrounded vowels

Thus far, the description has treated these as phonemic elements

in their own right, established on the basis of the contrasts

evidenced in section 2.1 above and tacitly following the criterion of

phonetic similarity in the assignment of allophones to phonemes:

hence, all long high back unrounded vowel articulations are assigned
 66

,fl
to i: /, all long mid -close to mid -open back unrounded vowel

articulations to /ë :/, and so on. It has been seen in particular

that front and back unrounded vowels of corresponding height and

length are in contrast. However, this analysis suffers from the

disadvantage of obscuring the nature of the distributional

characteristics of these vowels. First, we must define three distinct

positional classes of vowels: (i) utterance -final, (ii) utterance -

initial or medial, occurring with stress, and (iii) utterance -medial,

without stress. In position (i) there is full contrast between /i/

and /i /, and between /W and /e /; the only other short vowels that

occur in this position are /u/ and /a/, each of which is also in full

contrast with /i/ and /e /. But the contrast between /i/ and /u/ is

based upon just eleven items in the nominal system, and on a few verb

forms such as /ma:du/, 'may do' versus /ma:di /, 'do (2 sg imp)'.

In exemplification of the foregoing may

/akkj husked rice

/takkj speech, language

/kokk.J hook, crook

/akk./ elder sister

/kadi/ bite

/kad / side

/kad/ mustard seed

/kadW ocean

/pad4/ swamp

/ma:d1/ do (2 sg imp)

/ma:diyand.J doing

nad. middle

/ma d,/
:
may do
 67

/ma:dand
..
must do

ima:dalE/ might do

In addition, for a number of speakers the contrast between /ë/

and /a/ in this position is not on the same level as that between /e/

and /e /, or /a/ and /e/; free variation between the two is fairly

common, as is a merger of /ë/ with /a/.

The main point to be made about the vowels that may occur in this

position, however, is that the front and back unrounded vowels are

in full contrast.

In position (ii) the pattern is quite different. First, back

unrounded vowels most typically occur before a retroflex consonant,

and in this environment the possibility of a front vowel occurring

is very restricted (only ten cases in the data). Secondly, front and

back unrounded vowels occur and contrast more frequently before /r/

than any other single element; the occurrence of the front

vowel quality in this environment is unrestricted and there are more

than twelve cases where the back unrounded quality is also found.

Against this has to be set the fact that there are only five instances

of the back unrounded quality occurring elsewhere than before a

retroflex consonant or /r/.

For exemplification, it is convenient to order the data according

to the regular pattern, then to the exceptional pattern type (a) and

finally the exceptional pattern type (b), as follows:

The regular pattern; back unrounded vowels in a following retro-

flex environment, front and back unrounded vowels contrasting before

/r /, and front unrounded vowels occurring elsewhere:

/idi/ the whole

singing ceremony before harvest

festival
 68

/teli/ laughter

/keli/ esteem

/iri/ night
H n
/kirke/ small

/eraci/ flesh
n
/eraki/ eaves

/ne:ra/ time

/ne:ra/ bundle of paddy seedlings

/idi/ it (prox)

/ki:li/ fastening

/bedi/ heat of the sun

/te :ni/ honey

Exceptional pattern type (a); the front unrounded quality in a

following retroflex environment:

parrot

plant

spear

ladder

Exceptional pattern type (b); the back unrounded quality else-

where than before /r/ or a retroflex consonant:

/timmi/ sneeze (2 sg imp)

/etti/ bull, bullock

/ki:da/ place below, down

/e:vu/ will get up (1 sg)

Exceptional pattern type (a) is found in a number of loanwords

(such as li:ti /, 'spear') which may have more fully assimilated

 69

variants with the corresponding back unrounded vowel ( /i:ti /). It

is also found occasionally as the result of morphological relation-

ships within a paradigm, although here also the back unrounded

quality is found with many speakers; thus, /ni:ni /, 'you (sg)',

/ni:ki /, 'to you (sg)', but /ni:da /m /ni:da/, 'of you (sg)'. There

is also an environmental restriction, for certain speakers, against

back unrounded vowels occurring after a palatal consonant; in all

such cases, the corresponding front vowel is found instead: /cedi/

(for /cedi/ in the speech of some others), 'rage'; /celli/ (for

rr rr

/cells /), 'flea'; etc.

Exceptional pattern type (b) may arise through loss of the

relevant phonological environment, either through historical change

(as seems to be the case with /etti /, 'bull, bullock') or through

morphological alternation and constraints on resultant phoneme-

,H ,fl

sequences: cp. e:li /, 'get up (2 sg imp)' with e:vi /, 'will get

up (1 sg)'. Where morphological alternation is not'the reason,

variant forms with the corresponding front vowel are becoming

n n u
increasingly common, as in /etti /4 /etti /, 'bull, bullock'.

Finally for position (ii), there are no cases in the data of a

back unrounded vowel occurring after a bilabial or labiodental

consonant, and there are only a few (exceptional) instances of a front

vowel occurring after a bilabial or labiodental consonant and

followed by a retroflex (one such exception is the loanword /pe:te /,

'market, bazaar, town'). /u/ and /o /, on the other hand (as also

the low vowel /a/) occur and contrast fully in this environment (and

their long counterparts likewise). There is therefore a clear dis-

tributional relationship between most instances of front vowels in

this position and their corresponding back unrounded and rounded

 70

counterparts, as shown in Figure 5:

Bilabial
.back rounded/ Retroflex
Labiodental

Vowel.______ back unrounded /Non -labial Retroflex

front (elsewhere)

Figure 5: The distributional relationship between

front and back vowels in. position (ii)

In position (iii) (utterance -medial, without stress), all the

short vowels save /e/ occur and /i/ is in full contrast (except

immediately before a palatal consonant, as noted above, in 3.1.1

(iii) ) with /i/ and with /u /. DED records one item with a mid -back

unrounded vowel symbol in this position:

putteri harvest festival

However, since all my informants had [a] as the medial vowel artic-

ulation in this item, it is transcribed here as /puttari/.

Similarly, with nouns such as /appë /, 'father', the inflected forms

show /a/ rather than /ë /:

/ appana/ father (objective)

/appanda/ id. (possessive)

/apparigi/ id. (allative)

It was noted in the preceding chapter (p.23 ) that Richter (1670:197)

makes the same point.

,fl

Finally, i/ is generally not in contrast with /a/ in this

position, as long as the consonant that precedes is single: /i/

occurs when the preceding sequence is a high vowel followed by a

single consonant, and /a/ occurs after a sequence of non -high vowel
 71

plus single consonant. Further details of this distributional

pattern will be given in Chapter 4 (pp. 79 -9 ).

What emerges most clearly from these considerations is that

setting up a phonemic distinction between front and back unrounded

vowels for position (ii) (utterance- initial or medial, occurring with

stress) does not make clear the regular nature of the phonetic

patterning in this position. The distributional pattern can be

accounted for at the phonemic level only by appeal to the principle of

relative frequency of occurrence, as suggested in Fairbanks (1957);

in these terms, the four phonemes /i, i:, e, e:/ will be set up,

together with the statement that they are realised as back unrounded

vowels, of corresponding height and length, in a following retroflex

environment; and as back rounded vowels if in addition a bilabial or

labiodental consonant immediately precedes. Further, for certain

idiolects, the occurrence of a palatal consonant immediately before

the vowel in question will be stated as a condition preventing a back

unrounded articulation in a following retroflex environment. These

four vowel phonemes will therefore express the regular pattern of

distribution; in addition, it will be necessary to establish a

further set of four phonemes, /I, I:, E, E: /, in order to account

for the exceptional pattern type (a) (front quality occurring in a

n u it n

following retroflex environment), and another four, /i, i:, e, e:%,

for exceptional pattern type (b) (back unrounded quality elsewhere

than before a retroflex consonant). Finally, it will be stated that

/i, i:, e, e:/ most commonly occur, in contrast with /i, i:, e, e: /,

respectively, in the environment of a following /r/.

in place
Although such an analysis sets up twelve vowel phonemes

of eight, it has the virtue of being able to characterise

eight of
 72

the twelve as having very restricted distribution: the regular pattern

is stated in terms of a system of only four elements (setting aside

the ' /r/' environment, which remains a problem; this will be

explicable in terms of the system in Chapter 4, pp. 231 -234). It thus

also allows naturally for a trea(ment of loanwords (in which /I, I:,

E, L:/ play an important role) in terms of the approach outlined in

Fries and Pike (1949) .

Hence, a decision between the two analyses under discussion

(the one ignoring the frequency of the elements it establishes, the

other paying attention to this factor) might seem to be possible

simply on the basis of what price (in terms of an expanded phoneme

inventory) one is Prepared to pay for a revealing description of a

regular phonological pattern in the language: but this is only

apparently the case, since the issues involved are rarely so simple

in practice. Indeed, it is not difficult to show that there are

further difficulties in Coorg which stand in the way of even a

frequency- oriented phonemic theory; thus, on the one hand, there is

the morphological pattern evidenced in the set /ni:ni /, 'you (sg)';

/ni:da /a/ni:da/, 'of you (sg)', while on the other hand there is

that illustrated in the verb forms /ke:li /, 'ask (2 sg imp)':

/ke:pi /, 'will ask (1 sg)' (the transcription here is the one

described in section 2.1 above, based on phonetic similarity rather

than on frequency). It will be seen from Figure 6 that neither the

simple nor the modified (frequency -oriented) phonemic systems is able

to represent these forms in a wholly consistent way:

 73

Simple phonemic /ke:li/

transcription /ni:da /ni:da/ /ke:pi/

Modified phonemic /ni :ni/ /ke:li/

transcription /ni :da/

/ \ /nI :da/
I.
/ke :pi/

Figure 6: Comparison of phonemic analyses of two

morholo:ical patterns

From this it is clear that what is required for a satisfactory

statement of such relationships is a phonology which works in terms

of underlying forms and process rules (although exceptions will of

course remain); this will be the main concern of Chapters 4 and 5.

For a statement that takes into account the historical facts regarding

the production of back unrounded vowels in the language, see Emeneau

(1970a) and Kothandaraman (196 °).

4.2 Consonants

4.2.1 Clusters

(i) Below is given a list of two-consonant clusters which occur

(a) intervocalically and also in either (b) word -initial or (c) word -

final position, or both:

/pl/ (a) /e:rople:n(i)/ aeroplane

(b) /21e:t(1)/ plate

/pr/ (a) /tupri/ spit (2 sg imp comp)

(b) /ira: ni/ animal

/tr/ (a) /katri/ scissors

(b) /tripti/ satisfaction

/svetri/ sweater
/tr/ (a)

(b) /tre:n(i)/ train

 74

/gr/ (a) /bari/ bend down (2 sg imp comp)

(b) /'.tea :ma / village

/ha/ (a) /paksi/ bird ( -+ /paksi /)

(c) /saks(i)/ socks

/ks/ (a) /paksi/ bird (r+ /paksi/)

(b) /ksatriye/ warrior

/kr/ (a) /okri/ thresh (2 sg imp comp)

(b) /kraya/ price

/st/ (a) /a:sti/ property

(b) /stira/ permanence

/st/ (a) /kasta/ trouble

(b) /ste:san(i)/ station

(c) /po :st(i)/ post, mail

/sl/ (a) /mosle/ crocodile (..../mosle/)

(b) /slipar(i)/ slippers

/sr/ (a) /visra:nti/ calmness

(b) /gri/ Lord, Mister

All these items are loanwords, with the exception of /mosle/

.+ /mosle /, 'crocodile' (also /mosale/ro /mosale/) and the verb forms

showing the completive or intensifier element /r/ (marked 'comp');

these latter have alternative pronunciations without the consonant

n n n n
w
cluster, as /okkiri /, /baggiri /, etc. /paksi /, 'bird' is more common

in the data than the alternative form /paksi /; but each is apparently

the result of reborrowing or remodelling, since the fully assimilated

form /pakki/, which is recorded in DBIA with the meanings 'bird' and

'penis of immature boy' has now come to mean exclusively 'penis'. /kg/

is one of the most common types of cluster in Coorg (especially in

intervocalic position), corresponding in many cases to /ks/ in related

languages such as Kanarese.

 75

(ii) The inventory of two- consonant clusters which occur inter -

vocalically and in no other position is very large, although the size

depends partly on the rapidity of the speech style accepted for

analysis, and in particular on whether open transition between conson-

ants is taken as a type of clustering. The principle that has been

followed here is that consonant groups only qualify as a cluster when

there is close transition bet:reen them (no release of the first

consonant) or, in cases of looser transition, where there is some

allophonic feature of homorganicity holding across the consonant

boundary (as in /erci /m.s /eraci /, 'flesh', where the first form shows

/r/ realised as a voiceless tap in the environment of the following

voiceless /c /). The list given below contains all the clusters that

occur in the data; it is followed by a tabular presentation (Figure 7)

where an attempt is made to draw a distinction between systematic and

accidental gaps in the data.

/pp/ Al-322i/ bottle

/pt/ /tri tti/ satisfaction

/pt/ /tuptë/ (I) spat (comp)

/pc/ /nipciri/ cause to stop, stand (2 sg imp)

/pk/ /cei/ to the small metal bp (allative

case form of /ceppi /)

/pn/ /toma/ objective case form of /toppi /,

small leaves

/pl/ in the shell (locative case form

of /cippi/)

/pr/ /t4 / spit (2 sg imp comp)

/pv/ /talai/ (I) escape, shall escape

/py/ /ci:mandi/ combing (continuative aspect form)

 76

/tt/ /kutti/ tall narrow container

/tc/ /de :tciri/ cause to drive away (2 se imp)

/tk/ /po :tki/ to the male goat (allative case

form of /po:t /)

/tn/ /prayatna/ effort

/ti/ /mutli/ in the pearl (locative case form

of /mutti /)

/tr/ /ba:tri,/ take up (in hand)(2 sg imp comp)

/tv/ /otvi/ (I) squeeze, shall squeeze

/ty/ /keandi/ cutting (continuative aspect form)

/tt/ /potti/ box

/tk/ /ku :tki/ at the gatherin- (allative case

form of /ku:ti/)

/tn/ /ka:tne/ (I) showed

/tl/ /katli/ in the b»nJle (locative case form

of /katti/)

/tr/ /ku:tri/ sharpen (2 sg imp comp)

/tv/ /no:tvi/ (I) see, shall see

/ty/ /ketyandi/ building (continuative aspect form)

/cc/ /eccì/ scraps of food

/okte/ (I) threshed (paddy) (comp)

/ka:kciri/ cause to call (2 sg imp)

/kukke/ small basket

/nukne/ (I) pushed

/d -k1i/ in the direction (locative case

form of /dikq/

/kr/ /to:kra,/ of the gun( ossessive case form

of /to :ki/

/sa :ksi/ evidence

/kaksi/ litigation
 77

/kv/ /nikvi/ (I) comb, shall comb

/ky/ /nayand/ licking (continuative aspect form)

/bt/ /tabte/ (I) embraced (comp)

Il
/bc/ /tabcirì/ cause to embrace (2 sg imp)

/bb/ /obbe/ open drain

/bn/ /jibna/ objective case form of /jibbi /,

young areca nut

/bl/ /tabli/ orphan

/br/ /kabri/ seize with mouth (2 sg imp comp)

/bv/ /pabva/ (creeper) urines, will twine

/by/ /paandi/ (creeper) twining (continuative

aspect form)

/dp/ /modpa/ (thing) is, will be sweet

/dc/ /udcati/ (sun) rose

/dk/ /mudke/ old man

/dd/ /uddi/ common pulse

/dn/ /o:dne/ (I) read (past)

/dl/ /bedli/ in the heat of the sun (locative

case form of /bedi /)

/dr/ /ubri/ rub (2 sg imp comp)

/dv/ /tidvi/ (I) shall clean

tll

/dy/ /se:Ayandi/ drinking (deep draught) (continuative

aspect form)

/dk/ /na:dki/ to the disrict (allative case form

of /na:di /)

/dd/ /dadde/ stupid woman

/dn/ /o:dne/ (I) ran

/dl/ /ka:dli/ in the forest (locative case form

of /ka:di/)

/dr/ /mu:dra/ of the face (possessive case form

of /mu: di/)
 78

/ma:dvi/ (I) do, shall do

/ma:dyandi/ doing (continuative aspect form)

/gej je/ small bell

/nugte/ (I) entered by force (comp)

/bamiri/ cause to bend down (2 sg imp)

/pagde/ chess

/mur/ mould

/mune/ (I) swallowed

/bari/ bend down (intr, 2 sg imp comp)

/bai/ (I) bend down, shall bend down

(intr)

/ba ¿andi/ bending down (continuative aspect

form

/ram,pa/ hubbub

/kumte/ (I) churned (comp)

/kemciri/ cause to cough (2 sg imp)

/ turnbi/ flying beetle

/kummi/ mushroom

/ timase/ (I) sneezed

/gu :mra/ of the owl ( ossessive case form

of /gu:mi/)

/samsa:ra/ family

/samsaya/ doubt

/kemvi/ (I) cough, shall cough

/keRrandi/ coughing (continuative aspect form)

/nen2i/ (I) think, shall think

/ante/ of that sort

/kinte/ (I) tore into strips (comp)

 79

/nc/ /minciri/ cause to flicker (2 sg imp)

/nk/ / benni/ to the back (body part) (allative

case form of /benni /)

/nd/ /indi/ today

/nn/ /ponni/ gold

/nr/ /kinri/ tear into strips (2 sg imp comp)

II
/ns/ /mansi/ mind, conscience

/nv/ /minva/ (lightning) flashes, will flash

/ny/ /pauandi/ (rain) drizzling (continuative

' aspect form)

/kunte/ lame man

Air?/

/nc/ /enciri/ cause to say (2 sg imp)

/nk/ /ponki/ to the wife (allative case form

- of /pormi /)

/nd/ /kanda/ piece of meat

/nn/ /erne/ (I) said

/nn/ /ginni/ joint (wrist, fingers)

/r1/ /kanl i/ in the eye, (locative case form

. of /kann /)

/nr/ /ponra/ of the wifg (possessive case form

of /ponni /)

/nv/ /envi/ (I) say, shall say

by/ /enyandi/ sa/ ying (continuative aspect form)

tile

tacky secretion of jackfruit

Coorg child

man

/Ac/ /muñciri/ cause to dive (2 sg imp)

/ilk/ /briki/ (financial) bank

 80

/rig/ /tan_ge/ younger sister

bin/ /murine/ (I) dived

/riri/ /caririo:le/ chain

/rir/ /muiiri/ dive (2 sg imp comp)

/Ay/ /muñvi/ (I) dive, shall dive

/riy/ /muJiandi/ diving (continuative aspect form)

/lp/ /balpi/ (I) pull, shall pull

/it/ /balte/ right hand

/lt/ /celte/ (I) scattered

/lc/ /koiciri/ cause to kill (2 sg imp)

/lk/ /ka:lki/ to the leg (allative case form

of /ka:ll /)

hair (of head)

objective case form of /a:li /,

banyan tree

/11/ /kalli/ stone

/lr/ /ko:lra/ of the stick (possessive case

form of /ko:li /)

/ls/ /kelsa/ work

/lv/ /celvi/ (I) scatter, shall scatter

/ly/ /baize/ big

/lp/ /kalpi/ (I) play, shall play

/it/ /boite/ white

/it/ /tulte/ (I) made small hops, jumped

/lc/ /kulce/ (I) took a bath

/lk/ /a:lki/ to the map. (allative case form

of /a:li/

/palme/ talk, gossip

/a:lna/ objective case form of /a:li /, man

 81

/ko:lli/ with a st,ck (locative case form

of /ko:li/)

/11/ /kalli/ liquor

/1r/ /a:lra/ of the man (possessive case form

of /a:li /)

/ls/ /tolsi/ sacred basil

/1v/ /ku:lva:di/ village servant

/ly/ /a:lyandi/ ruling (continuative aspect form)

/rp/ /urne/ small bag for betel quid

/rt/ /parti/ cotton cloth

/rt/ / certe/ coconut shell

/rc/ /kurci/ chair

/rk/ /karki/ carbon

/rb/ /kurbe/ shepherd

/rd/ /kardi/ bear

/rj/ /korj,i/ marshy place

/rm/ /porme/ outside

/rn/ /u:rna/ objective case form of /u:ri /,

village

/rl/ /u:rli/ in the vi ?lage (locative case form

of /u:ri/)

/rv/ /ja:rvi/ (I) slip, shall slip

/ry/ /pa:rvandi/ leaping, flying (continuative

aspect form)

/st/ /pustaka/ book

/st/ /kasta/ trouble

/ss/ /here/ disgust (exclamatory)

/sn/ /mu:sne/ (I) smelled (tr)

/sl/ /mosle/
crocodile
/sl/ /mosle/
 82

¡sr/ /mu :sri/ smell (tr, 2 sg imp comp)

/sv/ /mu:svi/ (I) smell, shall smell (tr)

/sy/ /mu:Ayandi/ smelling (tr, continuative aspect

form)

/sc/ /pascima/ the West

/sr/ /visra:nti/ calmness

/sv/ /asvatta/ holy fig tree

/vt/ /cavte/ cucumber

/vt/ /cavti/ kick, step on

/vk/ /cavka/ white head -cloth

/vd/ /kavdi/ shell, cowrie

/vn/ /pavni/ gold £1 coin

/vv/ /novva/ (something) pains, will pain

ivy/ /kava/ the poets (plural form of the

noun /kavi /)

/ yp/ /kaype/ gall bladder

/yt/ /poLt,i/ blow, cuff

n
/yt/ /bayti/ evening
Il
/yc/ /a Se/ (I) sent

/yk/ /kaki/ to the hand (allative case form

of /kay /, hand)

/yn/ /lame/ egg-plant

/yl/ /masli/ peacock

/yr/ /ayri/ carpenter

/yv/ /posvi/ (I) beat, shall beat

/vy/ /payyu/ cow

 83

p t t ck b d d j g m n n ñ ri( 1 1 r'f s s h z v y

p x x x x x 1 x x ac

t a 6 x x 1 6 x x x x x
t 5 6 x 3 xi 6 1 3 5 x ,x x x x
c 2 2 2 x 2 2 2 2 1 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2

k x x x, 1 x x x : x x Ix x

bl1 x x x x x x x x
d x 1 6 x x x 6 x x x x x
d 5 6 1 3 x 5 6 x 3 5 x x x x ¡x x
j 2 2 2 1 2 .2 2 2 x 2 2 2 2 2 2 2 2 2 2 2 2 2 2!2 2

g x x 1 x x x x x x

m x x x x x x x x x x x

n x x x x x! x x x x x x
n x x x x x x x x x x
N
n x x x x
ri x x x x x x, x a

l x a x x xl x x x x x x x

l x x x x x . x x x x x x x x

r x x x x x x x x x x x 4 x x

f 4

s x x x x x x x 4 x x
N x 4 x
s x

h 4

z 4

v x x a x x x

y x x x x x % x x x x

Figure 7: Table of intervocalic consonant clusters

Each entry x indicates the occurrence of a consonant cluster of which

the first element is the consonant in that row and the second element is

the consonant in that column. .here no entry is made, the cluster

 84

concerned does not occur in the data; for the rest, the following

entries indicate the principal restrictions against certain consonant

sequences (systematic gaps):

1 = stops or affricates differing only in voice cannot form a

cluster

2 = an affricate is released except before another affricate

3 = retroflex stops are released before a nostalveolar affricate

4 = /r, f, s, s, h, z/ never occur geminate (always released)

5 = retroflex stops are released before labial stops and nasals

6 = stops differing only in the retroflex versus dental

opposition do not form a cluster.

(iii) Three- consonant clusters occur only intervocalically. In

general, they show the following structure:

voiced stop

nasal

lateral trill
+ stop +
trill lateral

approximant

/s/

The following are examples of this pattern:

/dkr/ /badkri/ live (2 sg imp comp)

/ndr/ /pundri/ squeeze (2 sg imp comp)

jlcr/ /balcri/ snatch (2 sg imp comp)

/rtr/ /artri/ cut (2 sg imp comp)

/vtr/ /cavtri/ kick, step on (2 sg imp comp)

/mu:scri/ cause to smell (tr, 2 sg imp comp)

/scr/

/dkl/ /badkli/ in the property (locative case form

of /badki /)
 85

/ndl/ /mandli/ in the village green_.(locative

case form of /mandi /)

/ltl/ /eltli/ in the writinc (locative case

form of felt /)

/rkl/ /erkli/ in the eaves (locative case form

of /ërki /)

In addition, there are the following instances:

/ktr/ /da:ktri/ doctor

/f5n/ /ofsnal/ optional

/nfl/ /markinflo:r/ American wheat

/ksm/ /laksmi/ Lakshmi (woman's name)

(iv) Four- consonant clusters are quite rare; they usuPlly consist

of an auproximant followed by a sequence of nasal plus stop plus trill:

/vndr/ /navndri/ squeeze (2 sg imp comp)

4.2.2 Distribution within the word

Retroflex consonants generally do not occur in word -initial

position, except in a few loan words:

/tre:n(i)/ train

/ti:/ tea

/tapa:l/ post, r il

/da:ktri/ doctor

In addition, /i /, /r/, /s /, /s/ and /h/ are relatively rare in

this position; DED, DEDS and DBIA record only the following items:

DED he:sige disgust (exclamatory)

lot, lotto: noise of wooden cattle -bell

rampa hubbub

rekke wing

sa:la loan

se:d- drink (deep draught)

 86

DED se:r- join (intr)

DEDS ratte upper arm

DBIÀ sadala looseness

sauna small

seri correct

To this list may be added:

/1 /- /lekka/ sum, account

/lo :ka/ world

/r/- /raja/ holiday

/rasa/ sweetness

/ra:gi/ ragi

/ ra:je/ king, Rajah

/ra:te/ plank for pounding rice

/ra:ni/ queen, Rani

/ro:ma/ body hair

/ra:ksase/ ogre, Rakshasa

/ ratte/ cultivator

/situe/ lion

/suru/ beginning

/sanne/ gesture

/sonne/ zero

/saddi/ noise

/sotti/ property

/sonta/ waist

/sa:li/ row, line

/se :di/ tired feeling in legs

/so :ki/ evil influence

 87

/sa:sa/ effort

/si :ta/ coldness

/samaya/ time

/sambala/ salary

/savkiya/ health

/sasa:ra/ presumption, cheekiness

/saha:ya/ help, assistance

/sa:ma:na/ thing, article

/so :ma :ri/ wastrel, idler (m, f)

/sa:pa/ curse

/ga:le/ school

/su :le/ trident

/sarira/ trunk of body

Most of these items are loans via Kanarese, and are acknowledged

(and used) as such; some have Coorg equivalents (e.g. /sarira/ =

/tadi /).

The situation regarding the approximants /v, y/ in word -initial

position (not just for Coorg but for related languages also) has

frequently been obscured by what will be argued here to be a separate

issue, namely the occurrence of tense onset to front, and back rounded

vowels occurring word- initially (see above, p. 50). The issue as to

whether or not the tense onset should be assigned to ;y/ (for front

vowels) and /v/ (for back rounded vowels) has proved to be a fairly

long -lived one in phonemic descriptions of the languages of the group;

and the same issue, although not stated in the same terns, was a

problem facing those grammarians and missionaries who, prior to the

present century, were concerned with transliterating Dravidian languages

into Roman script.

 88

Caldwell (1875) states the position succinctly in terms of the older

tradition: 'There is a tendency in all the Dravidian languages to

pronounce e as if it were ye, and o as if it were wo. In colloquial

Tamil this pronunciation, though often heard, is seldom represented in

writing; but in modern Canarese and Telugu, y before e, and v or w

before o, are often written as well as pronounced' (1875: 122). Caldwell

himself decides to regard this Phenomenon as a corrupt habit of

pronunciation, hence not to be recognised in the transcription. He has

not been consistently followed in this practice by later grammarians,

however; and two quite different and more recent works, Bright (1958)

and Hamanujan (1963) both phonemicise tense onsets as /v /- and /y / -.

Only Bright (1958: 12), however, presents arguments specifically

for this analysis, and these are taken up for consideration here.

(i) His first point has to do with the phonetic similarity between

the tense onset and the corresponding approximant (he illustrates with

reference to the palatal point of articulation, but his arguments

clearly extend to the bilabial as well) .

(ii) His second point is that in some idiolects of Kanarese certain

loans show initial [e:], thus establishing a contrast which has to be

reflected in the transcription.

(iii) His final point is that medial [e:] arising out of the

junction of two words, the first of which has its final vowel elided,

may occur with or without the tense onset, thus giving rise to

stylistically variant pronunciations.

The situation in Coorg is very similar to this, except that the

tense onset tends to be regular and strong for some speakers, and inter-

mittent and weak for others: thus, in setting up a statement for Coorg

as a whole, one is concerned not so much with the occurrence of the

tense onset as with its privilege of occurrence. This, however, does

 89
not affect the main issue.

It is clear that Bright's first point is consistent with the

phonetics- oriented approach of phonemics generally, but even within

that theory it may be argued that the concept 'phonetic similarity'

should be used as a heuristic device rather than a theoretical tenet

(as, for instance, even in such a work as Bloch, 1950). It is

doubtful whether Bright's appeal to the concept could be defended in

any context save that where a type of (broad phonetic) transcription

might be required for cross varieties comparative work. Fairbanks

(1957) and Chomsky (1964) present arguments against the 'biuniqueness'

principle that is at stake here (the first from a standpoint within

phonemic theory) .

Bright's second point apparently does not apply to Coorg; but

even within phonemic theory his position has been argued against as

being inadequate for handlin; situations where 'co- existent phonemic

systems' (Fries and Pike, 1949) seem to be operating. If such loans

as Bright refers to are exceptional in terms of the overall system,

then the optimal transcription will mark them as such by the use of

some additional symbol (perhaps /L: /), rather than mark the regular

cases (in this instance, by a preceding /y /-).

His final point is really answered by himself, since he notes that

'the difference could...be assigned to an occurrence of plus juncture'

in the representation of that pronunciation which preserves the tense

onset in medial position, 'but without any gain in simplification of

the phonemic system' (1958: 12). From the discussion of the second

point above, it will be seen that phonemicising the tense onset actually

complicates the system marginally, and so Bright's disinclination to

take up the use of the plus juncture will hardly be justifiable if his

second point is not conceded; furthermore, it has also been seen

 90

(pp. 32 -3) that an internal juncture element is required on

independent grounds.

Moreover, if the tense onset is assigned to the phonemes /v/ and

/y/, a very curious distribution pattern for the vowels results,

whereby only the low vowels /ae , a, a:/ and the unrounded vowels

/o, i:, e, e:/ may occur word -initially.

The position adopted here, then, is that Bright's alternative

analysis is correct, whereby (in agreement with the practice in

Caldwell, 1875) the vowels with tense onset are taken as 'allophonic

realisations of the simple vowels' (Bright, 1958: 12).

A phenomenon of the sound system of Coorg which is related to the

foregoing should be pointed out here, even though it throws light on

the status of the vowels /a, a:/ as much as on the distributional

patterning of consonants. It concerns the vowels /e:, o, o:/ (in the

data; but it may also be discovered to involve /e/) in initial

position. Sequences of tense onsets [3] or [w] plus vowel occurring

initially in such items as

/e:lakki/ cardamom
11,
/oraki/ sleep

/o :le/ ear ornament

are optionally replaced in the speech of many informants by the

sequences L; 4: j, [WV], L`N 4 '] , respectively, yielding the following

phonemicisation:

/ya:lakki/ cardamom

/ varalo/ sleep

/va :le/ ear ornament

It should perhaps be pointed out that an analysis which treats

the tense onset in the first three forms cited above as sub -phonemic

is in principle just as able to account for the alternative forms

(with initial jy/ or /v/ followed by a low vowel) as one that

 91

phonemicises the tense onset. The low central vowel quality is

systematically related, as maximally unmarked, to the mid front

quality (marked for front articulation, with a contingent feature of

lip -spreading) and to the mid back rounded quality (marked for back

and lip -rounded articulation. In those styles of speech where the

tense onset features of [front articulation, lip-spreading] or [back

articulation, lip -rounding] are particularly strongly implemented,

the phonological opposition may be entirely carried in this segment

of the diphthongal complex, and not marked in the following vocalic

component (yielding /a, a: /).

The fact that this phenomenon is not thorough -going in its

application in Coorg should not be a reason for underestimating its

importance in Dravidian phonology generally. The same sort of alter-

nation is attested for related languages (Bright, 1958: 15, mentions

the alternation between the forms given in his transcription as /vondu/

and /vandu /, 'one', for Kanarese) and in particular seems to be

reflected in the facts of Toda as recorded in DED and DEDS: e.g. Ta.

oti, To. war -, with the sense 'break in pieces' (DED 799). However,

the situation in Toda is at once more comprehensive and complex than

in the other related languages; a tentative statement by Emeneau for

Proto- Dravidian { o in Toda ( o is less certain) is put as follows:

*
'Otherwise than after p -, o > wa: when the next syllable (the last
*
in the word) has -ay (which is lost in To.); otherwise, o > wi:'

(Emeneau, 1970b: 38). In the Toda examples given there the phenomenon

is not restricted to initial position in the word, which suggests that

the formal relationship between, for example, Ta. kotai, 'west wind,

summer', and To. kwa:r, 'monsoon, year' (DED 1827) is part of the,same

alternation, and that this is a rather general feature of Dravidian

 92

phonology. There is no case of /va/ or /va:/ alternating with /o /,

/o:/ in Coorg in non -initial position, but a substitution of this

sort does occur in other languages; thus Upadhyaya (1968: 139)

records Kanarese items (from the Vakkaliga dialect spoken in the

Nanjangud area) which show /va/ or /va:/ after initial /k, g, t, n,

b, m, s/, where the standard language has /o/ or /o: /.

As a final point in this section, it may be noted that there are

in general no instances of word -final consonants, the only exceptions

being a few loan words such as:

/sa:ks(i)/ socks

/tre:n(i)/ train

/tapa:l/ post, mail,

where the parentheses indicate that final /i /is also a possibility in

most cases.

However, a number of nouns ending in /a/ show intervocalic /1/

after the /a/ before certain suffixes:

/ava/ she (rem)

/avale:/ even she (rem; /e: /, emphatic

marker)

/ayriga/ they (rem)

/ayrigala/ they? (rem; /a: /, interrogative

marker)

/mayma/ daughter -in -law

/maymalu/ daughter -in -law also ( /u: /, co-

ordinative marker)

Another set of nominals, with final /e/ or /ë /, show intervocalic

/n/ before certain suffixes; in these forms, the contrast between the

mid back unrounded and the low vowels is lost before the /n/:

/ra:je/ king, Rajah

/ra:janu: ra:niyu:/ king and queen, Rajah and Rani

 93

Finally, certain verb forms with final /e/ (1 sg past) show

/a:n/ in place of /e/when followed by certain suffixes:

/pa:rine/ I leaped

/pa:rina:na:/ did (I) leap? ( /a: /, interrogative

marker)

As far as the noun forms are concerned, it is convenient to treat

the intervocalic /1/and /n/ as belonging to the base form rather than

the suffix, and hence to write /ava(1) /, 'she (rem)' (thus distinguish-

ing these nouns from those such as /mara/, 'tree', which do not show

the intervocalic /1/ in any environment), and, analagously, /ra:je(n) /,

'king, Rajah'.

4.2.3 Further details on intervocalic distribution

This section is mainly concerned with the features of length and

voicing of intervocalic consonants, and in particular with the way in

which they relate to the length of the preceding vowel.

(i) Stop consonants preceded by a homorganic nasal.

Thus far all such sequences have been represented by /m/

before /p/ or /b/, /n/ before /t/ or /d /, etc., as follows:

/mp/ /ramna/ hubbub

/mb/ /jamba/ pride

/nt/ /ante/ of that sort

/nd/ /andi/ that day

/nt/ /kunte(r_)/ lame man

/pundé(n)/ quarrelsome fellow

/IA/
/ñc/ /añci/ tile

/ñj/ /kañi/ rice gruel

/Ilk/ /naiiki/ to us (N /narigaki /; allative case

form of /nai ga(1) /)

/tame/ younger sister

 94

Since a nasal consonant which is not homorganic with the following

stop is also found in all cases where the following stop is voiceless

(see pp. 78-9 above), it is not possible to make use of just one nasal

consonant symbol, say, ¡n/, in all these instances. However, it would

be possible to do so unambiguously in all cases where the following

stop is voiced; but this would lead to a highly idiosyncratic

distributional pattern for /n/ with respect to the other nasal

consonants, and must be rejected as an ad hoc solution. Another

solution would be to set up a distinct element /N/ before all stops,

which is realised as a nasal homorganic with the following stop in each

case; this would then contrast with other nasal consonants in the

environment of a following voiceless stop. Such a symbol would

correspond to the 'anusvara' in the Kanarese script. Further, the

Kanarese anusvara may also be used before a nasal consonant symbol,

thus providing an alternative way of representing geminate nasal

sequences, and /P?/ could clearly be allowed to pattern in an analogous

fashion.

A decision between these analyses does not seem very important,

since the problem is raised in the context of a phonemic approach, and

this is not what will be used in setting up the phonological system for

the language (Chapters 4 to 6). The transcription described in the

first sections of this chapter, allowing ¡m/ to occur immediately before

/p/ and /b/, ¡n% before /t/ and /d /, etc., will continue to be used in

the citing the data to be accounted for; but at the systematic

phonemic level a polysystemic approach will be adopted, and this will

recognise a nasal element that is unmarked for place of articulation

as the source for all nasal consonants that are homorganic with a

following stop.
 95

(ii) Voiced and voiceless stops after a homorganic nasal.

A more important point concerning the homorganic clusters

of nasal plus stop has to do with the voicing opposition of stops;

the symmetry of the presentation above, in which both voiced and

voiceless stops are illustrated following a homorganic nasal, is

misleading. Voiceless stops in this position are much less frequent

in the language than are voiced stops. In many cases, the occurrence

of a voiceless stop is a mark of a loanword (e.g. /pace /, 'dhoti')

n u
or of a contracted form (e.g. /nariki/ for /narigaki/, 'to us').

Among the simple nouns in the data, there are only 9 instances of a

voiceless stop in this position, as against more than 80 instances of

a voiced stop; while with the verbs (taking only the uninflected forms)

the ratio is 1 instance of a voiceless stop to 38 instances of a

voiced stop. Furthermore, with the single exception of /kavriki /,

'vital spot', all instances of a voiceless stop in this position

follow a short vowel in the word -initial syllable (e.g. rampa/,

'hubbub', /ante/, 'of that sort', etc.).

(iii) Voicing contrasts between stops in intervocalic position.

It has been seen that there is a contrast between single

and geminate stops, nasals, laterals and approximants in intervocalic

position where the preceding vowel is short (pp.46 -8 above). However,

the length distinction apparently evidenced in such forms as:

/kada,pe/ steps for getting over fence

/kame/ frog

/nelaçi/ moon

/açci/ jaggery cake

 96

mean fellow

husked rice

is susceptible of treatment as a conditioned alternation; the geminate

consonant clusters are found after a short vowel in a stressed

syllable (the first in the word), and the single consonants after a

short vowel occurring in an unstressed syllable. The counter -examples

to this statement, exemplified in the contrasts for dental and retro-

fle:: voiceless stops:

fmitif limit

f tittif fire

/pata/ picture

/patta/ coronation

are extremely rare, occurring in only the forms just cited and the

following:

fkatef story

f pataf kite

f tiket(i) f ticket

f a2aru: paf rarity

fuuava:saf fasting

/kace:ri/ office

/kutumba/ family

all of which save the last are clear instances of loan words.

In all other cases,single stops occurring in this position are

voiced. Furthermore, although contrasts can be found easily enough for

single and geminate voiced stops in the data, the following observ-

ations have to be made: first, geminate voiced stops, like their

voiceless counterparts, are not found other than immediately after a

 97

short vowel in a stressed syllable; secondly, even in this position

they are relatively infrequent (only 12 instances in the uninflected

forms of verbs, as opposed to 47 instances of geminate voiceless

stops; and only 28 instances in the uninflected forms of nouns, as

opposed to 120 instances of geminate voiceless stops). The

distinction between voiced and voiceless stops in intervocalic

position after a short vowel is thus intimately bound up with the

single versus geminate opposition, and with the occurrence of stress.

After a long vowel (always occurring with stress), the contrast

between single voiced and voiceless stops is quite regular (see

pp44 -6 above), except that /b, does not occur; geminate consonants,

whether voiced or voiceless stops, or nasals, laterals or approximants,

do not occur in this position.

Regarding the distinction between single and geminate nasals,

laterals and approximants in intervocalic position, the situation is

fairly straightforward, since with these consonants there is no cross-

cutting distinction of voice; they contrast as single versus gemjrate

only after a short vowel in a stressed syllable.

The pattern described here, and in the section above on homorganic

clusters of nasal plus stop, may be summarised conveniently in the

table of Figure 8 (the use of the symbols P and B is from Krishna-

murthi, 1961):
 98

After a short vowel, After a long vowel After a short vowel,

in stressed syllable (always in a in unstressed syllable
stressed syllable)

(p) p P

B B B

pp

(BB)

(Np)

NB NB NB

CC C C

P = voiceless stop, single; PP = geminate voiceless stop cluster

B = voiced stop, single; BB = geminate voiced stop cluster

i. = nasal consonant homorganic with following stop

C = non -stop consonant, single; CC = non-stop geminate cluster

Parentheses enclose infrequent consonant types.

Figure 8: Intervocalic consonant types

(iv) A phonemic solution for intervocalic stops.

In respect of the situation regarding single and geminate

voiceless stops in intervocalic position, as just described, Bmeneau

(1967: fn.8) proposes the following solution:

'A phonemic statement for Kodagu will recognise that voiceless

stops and the affricate c have two major allophones each, based on

quantity. After a short unaccented vowel (i.e. in a word's first

syllable) a voiceless stop is long and heterosyllabic; elsewhere within

a word, i.e. after a long accented vowel or after an unaccented vowel,

it is shorter, and the syllabic division precedes it: e.g. mati

 99

[matti] "ax", ma:ti [ma:ti] "change:"

'This statement covers hundreds of examples. Exceptions to it,

viz. with a short voiceless stop after a short accented vowel, are so

rare that they may be covered by the use of special symbols; e.g.

[kate] "story" (a borrowing ultimately from Sanskrit) contrasts with

[ katte] "donkey ", and the two may be written kaTe and kate.'

Thus, in terms of this analysis, P and PP in the first column of

the table in Figure 8 may each be written as P, and brought into a

symmetrical opposition to P in the other two columns. This solution

therefore apparently regularises the data in an elegant way, and as

such was made use of in Garman (1969) in the context of a generative

phonological description; however, it will be argued below (pp.104 -

109)that it is not suitable for this purpose since it actually obscures

certain regularities in the patterning of phonological elements.

A final point remains to be made here. A number of items show

/v/ or /y/ after a short vowel in a stressed syllable, followed by a

single consonant or by a homorganic cluster of nasal plus stop:

/cavti/ step on, kick (2 sg imp)

/davde/ cheek

/pavni/ gold £1 coin

/gavli/ large lizard

/kavriki/ vital spot

/ñavndi/ squeeze (2 sg imp)

/bayti/ evening

a bi
/kaybi/ sugar
sug cane

/bayne/ egg-plant

/mayli/ peacock

/kuyndi/ feel prickly (2 sg imp)

 100

In such cases the sequence of short vowel plus approximant

patterns very much like a long vowel in respect of the possible

following consonants; thus, geminate consonants may not occur, and

there is a regular contrast in voicing among stops occurring singly,

but a voiceless stop preceded by a homorganic nasal is rare (only

found in the item /kaviki /). Hence it is possible to include these

cases under the second column of the table in Figure 8 above.

Concerning the contrast between B and BB, and between C and CC

in intervocalic position Emeneau (1967: fn. 8) remarks:

'Other consonants do not show the sane complementary distribution

with regard to quantity as do the voiceless stops...and are treated

phonemically as single and double respectively. For voiced stops and

the voiced affricate j this contrast has little functional yield,

but it must be recognised.'

The Coorg material in Emeneau (1938, 1967 and 1970a) is cited in

the phonemic transcription that treats length as an allophonic feature

of voiceless stops (here called the Emeneau type A transcription),

while that used for Coorg in DED, DEDS and DBIA marks the geminate

versus single distinction for all consonants (here called the Emeneau

type B transcription, corresponding to that presented in the earlier

sections of this chapter).

5.0 Conclusion

In this section the problem of voiced and voiceless stops in

intervocalic position is further considered, in the light of a phono-

logical solution which is based upon certain Dravidian orthographic

systems. The first part consists of an outline of the salient

features of the four main Dravidian orthographies, and of an

 101

examination of how the contrasts that these recognise as basic are

reflected in Coorg (5.1); the final part (5.2) looks ahead briefly

to the requirements of a phonology of underlying representations such

as will be described in Chapters 4 to 6.

5.1 Dravidian orthographies

The four main orthographies that are the concern of this

section are those used for Tamil (Ta), Malayalam (Ha), Kanarese (Ka)

and Telugu (Te). They fall into two broad categories, the Ta /Ma and

the Ka/Te.

(i) The Ta/Ma system

This is more clearly expressed in the Tamil script than

in the Malayalam, since the latter script has been fairly extensively

modified in order to take account of a large number of Sanskritic

loan words which show regular contrast between voiced and voiceless

and between aspirated and unaspirated stops. Thus, distinct

symbols which always represent voiced stops, and others for aspirated

stops (see, for example, Andrewskutty, 1971), but these are confined

to the Sanskritic component of the vocabulary. Native Malayalam

vocabulary items, on the other hand, are written in a way which corres-

ponds to the basic system of the Tamil script, employing stop symbols

which are unmarked for voice. The situation with regard to the Tamil

script is altogether more regular, as may be seen from the following

observation in Shanmugam (1973):

'There have been attempts to introduce new symbols, especially to

represent the voiced stop consonants, by some popular journals and

writers. Although this practice is not universally accepted, it is

more common now to represent the voiced ba as 013 (that is, by the

bold type of the letter used form). The representation of other

 102

voiced stop consonants has not gained currency.' (1973: 5).

According to this system, then, there is (ignoring the exception

just noted) only one stop symbol for each place of articulation; and

this symbol may occur either word- initially or word medially; when

medial, it may either be preceded by a symbol representing a homorganic

nasal, or be intervocalic; when intervocalic, it may be either single

or geminate. The possibilities, together with the (generalised)

corresponding phonetic values, are illustrated, for the velar place

of articulation, in Figure 9:

Initial Medial

with homorganic intervocalic, intervocalic,

nasal single geminate

ik k 'zk

[k] L09] [6] [kk]/[ k ]

Figure 9: The functions of the velar stop symbol

in the Ta /Ma system

The first row presents a conventional transliteration of the

Ta/Ma orthographic system, with N standing for the anusvara. The

stop symbol represents a fairly tense voiceless stop initially, a

fully tense voiceless stop when geminate intervocalically, a fairly

lax voiced stop after the nasal, and a fully lax, fricative

articulation, usually voiced, when it occurs singly in intervocalic

position. The fully tense stop is long after a short vowel in a

stressed syllable; otherwise, it is only half-long (represented here

by [k.]). This variation between long and half-long quantities is

wholly regular within the Ta/Ma system, as is pointed out in Firth

(1934: iii -iv) for Tamil:

 103

'It is only in certain cases that double stop consonants like

pp, tt, tt, kk are heard really doubled in the Italian sense....

They are, however, always voiceless and energetically articulated....

But they vary considerably in length.... If the tt of kattu "caw"

were represented by an index of two, the index of the tt of kaattu

"protect" would be one and a half.'

The orthography thus operates primarily in terms of an inter -

vocalic opposition between geminate (tense) and single (lax)

categories, with respect to the stops. In initial position, and

following a homorganic nasal, this opposition is neutralised. Voicing

is not a primitive category of the system. It is true that Fowler

(1954), expressly dealing with a heavily Sanskrit-influenced variety

of Tamil, sets up the lax stops /b, d, g/ in his phonemic system

(mainly on the basis of word -initial contrasts in certain items); but

Fairbanks (1957) successfully shows that these stops are best handled

outside the main system (occurring infrequently, and only in Sanskrit,

Portuguese and English loan words), and that the orthography

accurately reflects the central sound pattern of the language (though

this last point is not made explicit).

(ii) The Ka/Te system

By contrast, the Ka/te system recognises both a voiced and

a voiceless series of stop symbols; again, illustration will be made

with reference to the velar place of articulation, with broad

phonetic values indicated, as set out in Figure 10:

 1
04

Initial iledial

with homorganic intervocalic, intervocalic,

nasal single geminate

Figure 10: The functions of the velar stop symbols

in the Ka/Te system

Initially, there is a recognised voicing contrast, and this is

carried through to all positions. Unlike with the Ta/Ma system, kk

only occurs after short vowels in stressed syllables, and hence

always represents a fully long voiceless stop articulation; the

fairly tense half -long voiceless stop which occurs after long vowels

and after short vowels in unstressed syllables is represented by k.

Thus the two systems bear even less correspondence to each other than

is apparent on first sight.

At this point it is convenient to widen the discussion to include

Coorg.

5.2 The sound pattern of Coorg

The main features of the sound pattern of Coorg have been

presented and discussed in the preceding sections of this chapter,

and it is fairly clear where Coorg stands in relation to the two

orthographic systems which have just been introduced. It is important

first of all to see the Ta /Na and Ka/Te systems not as competing

phonological solutions to cover essentially the same basic data, but

 105

rather as complementary to each other, reflecting the rather

different surface sound patterns in the languages they are used for.

Coorg, with frequent contrast between voiced and voiceless stops in

initial position (even in native vocabulary items), and with

demonstrable, if relatively infrequent, contrast between voiced and

voiceless stops medially, certainly stands within the Ka/Te system

insofar as its surface sound structure is concerned.

However, it has been seen that a simple phonemic analysis which

takes the voicing contrast between stops to be a primitive feature

of the system (e.g. the Emeneau type B transcription) is unable to

reflect the distributional pattern of the elements concerned. The

representation of voiceless stops in the Emeneau type A trans-

cription constitutes one attempt to reflect this pattern more

faithfully, within a phonemic framework; however, this only high-

lights the nature of the patterning of voiceless stops, and says

nothing about voiced stops, or homorganic clusters of nasal plus

stop, although these also stand in need of explanation. The regular

pattern for stops in Coorg - i.e., ignoring the parenthesised

elements in the table of Figure 8 above (p. 98) - is illustrated here

in Figure 11:

Initial riedial

with homorganic intervocalic, intervocalic,

nasal single geminate

P, B NB P, B PP

Figure 11: The basic distributional pattern for

stops in Coorg
 106

It may be seen from this that, ignoring the contrast between P

and B in word -initial position, the patterning of stops in Coorg

closely corresponds to what is found in the Ta/Ma system. The only

difference concerns the occurrence of both P and B in the third column

of the table; and it is clear that this difference is only apparent,

since it has been seen (pp. 95 -8 above) that P and B are not

generally in contrast in this position, B occurring after a short vowel

in a stressed syllable, and P elsewhere, i.e. after a long vowel, and

after a short vowel in an unstressed syllable. Indeed, this only

serves to demonstrate more clearly theparallelism between the Coorg

pattern and that expressed in the Ta/Ma system. This fact, and the

patterning of medial stops generally in Coorg, is most conveniently

reflected by the use of a single stop symbol (with phonetic values

indicated for the velar place of articulation), as in Figure 12:

With homorg- Intervocalic, Intervocalic, Intervocalic,

anic nasal single geminate, after geminate,
short vowel in elsewhere
stressed syllable

NP P PP PP
r39 1 rkk1 rk
9

Ti;ure 12: The regular pattern of medial stops in Coorg

With regard to the entry PP in the rightmost column, it is

relevant to recall the description in Firth (1934) regarding the

phonetic quantity of the corresponding stops in Tamil (p.103 above),

and also to note what was recognised earlier (p. 61) regarding the

relatively longer duration of 'single' voiceless stops in Coorg

compared with their voiced counterparts. In other words, the phonetic

facts are not at variance with the distributional pattern that is being
 107

claimed to hold for Coorg equally as for the Ta/Ma system.

Clearly, then, in respect of its most regular stop contrasts in

medial positions, Coorg falls within the Ta/Ma system; but in order to

formalise this fact, the optimal phonological system has to fulfil the

following conditions:

(a) it has to be polysystemic; that is, it must allow for a

systematic distinction between sets of elements operating at different

positions within the word (at least, between word -initial and word -

medial positions)

(b) it must operate in terms of underlying representations, and

derive certain surface contrasts by rule.

In Chapters 4 to 6 an attempt is made in this direction.

In conclusion, it is worth making two connected observations.

First, the type of phonological system which has been suggested here

as appropriate for underlying forms in Coorg has been approached by

means of an examination of the frequency of the contrasts between

certain elements within a simple phonemic system; hence the degree of

the coincidence between the resulting system and that expressed in the

Tamil and Malayalam orthographies constitutes substantial evidence for

the view that this orthographic system is indeed (as has often been

claimed) a suitable representation of the most fundamental phono-

logical contrasts in at least a large number of Dravidian languages

(if not all; it is, for example, an empirical issue whether a

phonology of underlying representations for Kanarese and Telugu would

have to be set up in the same general way as for Coorg). This is not

altogether surprising, perhaps, in view of the historical claims that

have been made for the analysis that the Ta/Ma system provides for stop

consonants; Caldwell (1875) was the first to provide a name for the
 108

central characteristic of this analysis when he noted that the Tamil

and Malayalam orthographies exhibited what he called 'the peculiar

Dravidian law of the convertibility of surds and sonants' (1875: 137).

His view was that this principle had been a characteristic of the

proto -language and that Tamil preserved this state of affairs most

faithfully of all the Dravidian languages; on the other hand, 'in the

Northern Dravidian dialects' (by which Caldwell appears to refer to

the non -literary languages of North and Central India, but the situation

in Kanarese and Telugu is essentially similar), 'the difference between

surds and sonants is generally expressed by the use of different

characters for each sound, in imitation of the system of the

Devanagari' (1875: 137).

Caldwell's belief that the proto- language must have had a system

of stops built around the 'convertibility' principle was countered

later in Bloch (1919), but has since been supported with a great deal

of evidence, for example in Burrow (1938) and Krishnamurthi (1961: 24)

(also Emeneau, 1970b, where however some dissatisfaction is expressed

regarding the symbols -k-, -kk-, etc.; this is apparently Emeneau's

viewpoint subsequent to his proposal of the Eneneau type A

transcription for Coorg). It is fair to say that, at time of writing,

an assumption of the historical v alidity of the convertibility

principle is quite uncontroversial. A concluding remark in Burrow

(1938) may be cited here as typical: 'Telugu and Kanarese have, for

reasons that are obscure in most cases, introduced secondary voicing

into primitive Dravidian words' (1938: 722).

The second observation that has to be made is that there is

evidence on morphological grounds also (as will be shown in Chapters 4

to 6) for setting up a single series of stops in the underlying forms

 109

for Coorg lexical items. Thus, pairs of lexical items such as the

following (cited in Emeneau type B):

/ku:di/ join (intr; 2 sg imp)

/ku:ti/ id (tr; 2 sg imp)

/ka:di/ forest

/ka:ti/ bison (animal inhabiting forest)

/todi/ touch (2 sg imp)

/totte/ (I) touched

/ju rigi/
: stray (intr; 2 sg imp)

/ju:ki/ id (tr; 2 sg imp)

show a morphological relationship which may be generalised. as:

-(N) P + P- _) -PP -,

with the voicing distinction being a secondary correlate of the under-

lying contrast of consonant length (geminate versus single). In this

connection, the work of Kumaraswami Raja (Iumaraswami Raja 1969a,

1969b) on the regular alternation between -IdP- and -PP- (with -NPP-

forming an inadmissible sequence) is particularly important, and has

been incorporated in the analysis suggested in the later chapters of

this work.
 110

CILtLPTER T:1 EE

Syntax
 111

1 .0 Introduction

This chapter does not aim at a complete survey, however non -

detailed, of the field. This is partly for reasons of limited space

(in a study which is primarily morphological and phonological), and

partly because seven months' fieldwork has not yielded sufficient

material in depth - nor an adequate insight into the language - for

anything more than a superficial treatment, save in certain fairly

well defined areas. A number of general studies of the syntax of

Dravidian languages are available (e.g. Ramanujan, 1963 for Kanarese;

Agesthialingom, 1967 for Madurai Tamil; and Kandiah, 1967 for Ceylon

Tamil, all within transformational-generative theory); and while these

have certainly outlined an approach to the syntax of the Dravidian

group, very little detailed work has yet been done at all. The data on

Coorg syntax is in general only sufficient to indicate that Coorg to

a large extent fits into the general framework established thus far;

but (more interestingly) in certain areas it seems to provide evidence

that this framework requires modification, or at least further

discussion. What follows, therefore, is, first, a description of the

basic sentence patterns of Coorg, and secondly, a fairly detailed

study of (i) the copula, and (ii) the causative and transitive

constructions of the language. These topics have either been dealt

with very briefly in descriptions of other languages, or else

analysed in detail (e.g. Kandiah, 1967), but in a way that seems

inadequate in the light of what a description of Coorg must achieve.

Duch of what has been analysed for other languages of the group has,

inevitably, proved to be relevant to this study; it is hoped, and

believed, that the analysis of these two problem areas in Coorg

syntax will have relevance for the description of other South Dravidian

languages.
 112

The transcription used for presenting the data is Fmeneau type B

(see Chapter 2, p. 100). This should perhaps be thought of as a type

of useful broad transcription rather than a strictly phonemic one

(e.g. it does not treat quantity as an allophonic feature of inter-

vocalic voiceless stops); accordingly, it is enclosed in square

brackets, but these of course should not be interpreted in the same

way as the square brackets of the preceding chapter. In the

subsequent chapters also (Chapters 4 to 6), square brackets will

continue to be used to indicate the Btaeneau type B transcription, and

slashes will mark off elements of the phonology of underlying forms.

Theeneau type B transcription is more useful for syntactic and

morphological work than the Emeneau type A, since it provides a

wholly consistent representation of consonant gemination, which turns

out to be a very common morpho- syntactic phenomenon. )here necessary,

elements which are subject to deletion rules are enclosed in

parentheses in those contexts where they would be deleted; for

example, very commonly, word -final 411, before another word with an

initial vowel. In all cases, the forms cited are the 'full' forms

of slow and careful speech.

1.1 The Basic Syntactic Patterns

The following. sentences are superficially identical:

1. [adi ba:tij

it came

'It came'.
u_
2. [adi basti]

it duck

'It's a duck'.

and rests
The ambiguity of the phonological realisations of
1 2
 113

«_
on two factors: first, the homonymy of the noun [ba:tij, 'duck' with

the verb [ba :ti],'came'; and secondly, the surface identity of the

two distinct types of syntactic structure involved in these sentences.

The first factor is not relevant to this section; the second is, and

may be illustrated by consideration of the negative counterparts of

1 and 2. These are:

u

3. [adi bandile ]

it came -not

'It didn't come'

4. [adi ba:t(i) alla]

it duck not

'It's not a duck'.

(For the purpose of the present discussion, the differing surface

realisations of the basic form of the verbal element in sentences 1

and 3 are not relevant.)

Sentences 1 and 3 represent the positive and negative forms of

one type of syntactic pattern, and 2 and 4 those of another type. In

each case the negating element is underlined. [le] in 3 is

morphologically related to [ille], which is derived historically from

an old Dravidian negating verb; [alla] is similarly derived from

another old negating verb. DED 198 gives several paradigmatic forms

for alla in Tamil, and notes that in Modern Tamil that form is used

for all persons and numbers, although originally it was only the

neuter form; and the Old Tamil cognate of [ille] (DED 2106) was a

verb with a full negative paradigm. Coorg shows only the invariant

forms [alla] and [ille] and it is convenient here to regard them as

negating verbal particles. Traditionally, they are glossed along the

following lines: [alla], 'to be not so- and-so'; [ille], 'to be not,
 114

to exist not (in a place)'. [alla] negates only those syntactic

patterns of the Subject nominal +Predicate nominal type, while [ille]

negates patterns where the predicate is verbal. This statement will

be revised later, but serves for the moment.

1.1.1 Sentences negated by [ille]

The class of structures negated by [ille] covers a wider variety

of sub -types than that negated by [alla], and so will be dealt with

first here.

(i) Intransitive and transitive verbs.

1 illustrates a one -place verbal predicate structure (using the termin-

ology of Lyons, 1968: 350 -371): a fairly small class of verbs,

including [po:] -, 'to go', [ca:l] -, 'to die', [bu :1] -, 'to fall',

etc., occurs as predicate in this pattern. The sentences below show

other, but related patterns:

5. [ave na:yina kondati]

he dog killed

'He killed the dog'.

N 11 il It

6. [ave ku:l(i) undati]

he rice ate

'He ate the rice'.

N
N 11 11

7. [ave undati]

he ate (rice)

'He had a meal'.

In the foregoing, 5 shows a two -place structure, with the second

nominal showing the objective case marker (underlined). 6 also shows

the same pattern (although, as commonly with Dravidian inanimate

nouns, the second nominal shows no overt objective marker -see Asher,

1968: 91, where the same fact in Malayalam is accounted for in terms
 115

of a 'zero exponent of accusative case'), but the grammar requires

that it also be related to 7. As in English, it is possible to

think of contexts where the utterance which

H 11 tl
8. *Lave kondatij

'He killed'

underlies is perfectly acceptable; but there is an essential

difference between 7 and 8 in that one does not have to look for

contexts wherein the utterance that 7 underlies would be accepted.

That is, no anaphoric (i.e. sentence to sentence) or deictic (i.e.

sentence to situation of utterance) reference need be invoked in the

case of 7. In other words, 7 is grammatical, and 8 is not, although

it may be acceptable in a given context. (On the distinction

between 'utterance' /'sentence', 'acceptable' /'grammatical', and the

broader distinction between 'primitive' /'linguistic' terms, see

Lyons, 1968: 137 -42, 172.)

However, it may be argued that, while the derivation of 8 from 5

is not a requirement of the grammar, it is necessary that 7 be thus

derivable from 6. A fairly strong case for this can be made out for

the verb Lunnj -, 'to eat (rice)', since it is in contrast with

another verb, Ltinnj -, 'to eat (anything which is not rice)'. Thus:
II

7. Lave undatij

'He had a meal'

and
N II It

P. Lave tindatij

he ate (not rice)

'He had a snack'

may be argued to depend, for semantic analysis, on an appropriate

predicate nominal, which is subsequently deleted. In 9 this

 116

obviously may be any of a large class of nominals referring to non -

rice dishes, or other edibles; in the case of 7 the second nominal

is restricted to [ku:li] 'cooked rice', [tari gu:li

it yesterda s rice
'yesterday's

(served again)', etc. Hence the same syntactic relationship as

holds between 7 and 6 also holds between 9 and 10:

N fl It

10. [ave pandi kari tindatiJ

he pork curry ate

'He ate the pork curry'.

Hence both of these verbs presuppose certain predicate nominals and

the sentences in 7 and 9 are clear cases of 'pseudo- intransitive'

constructions. A number of other verbs, such as [o :d] -, 'to read',

II 11

[elid] -, 'to write', [pare] -, 'to utter, speak' etc., belong in the

same class, which might be called, following Lyons (1968: 360 -1), the

class of 'inherently transitive verbs'.

(ii) Impersonal constructions.

A superficially quite different pattern from any that has so far

been considered, but which is also negated by Lille], is illustrated

by the following:
n u
11. [duddi ni:ki bo :nduvad

money you -to needed - interrog.

'Do you want money?'

12. [naki po:ra3

me-to suffices -not

'I want more'.

13. [naki novva]

me -to pains

'I feel sore'.

 117

14. [na:ki kulipa]

me -to is -cold

'T feel cold'.

t,
15. [na:hii perta dura nadapakayya]

me -to much distance walk -able-not

'I can't walk very far'.

In each of the sentences 11 -15, as the literal and free trans-

lations indicate, the notional subject is not the grammatical subject:

the case marker (underlined) is the 'allative'. Further, in 12, 13,

and 14 no other nominal occurs: in 11 [duddi] occurs, and in 15 the

111,

nominal group [perta du:ra] occurs, but whereas [duddi] may be

interpreted as the grammatical subject of 11, [partadu:ra] is quite

clearly a modifier of the verb [nada] -, 'walk' in 15. For 13 one

might tentatively supply [nombala], 'pain' as a (deleted) grammatical

subject - but this suggestion is probably motivated by morphological

considerations rather than syntactic ones; namely, that the noun

[nombala] and the verb [novva] are formed on the same root [no] -;

[ga:ya] 'wound' has equal syntactic justification to be considered as

the grammatical subject. With 12, an even wider choice exists for

determining the putative grammatical subject. By contrast, in 14 it

is difficult to find anything to play this role. All these sentences,

therefore, differ from one another quite considerably, and yet all

are related in some sense. In traditional terms, they are all

examples of 'impersonal' constructions, and are kept distinct as such.

However, it seems to be the case in other Dravidian languages, and is

certainly the case in Coorg, that the verbs in these constructions are

comparatively rare, defective, or morphologically unique: thus

tl
[kuli] -, 'cold feeling is', [no] -, 'pain (intr)' and [kayy] -, 'be able'
 115

all haveIsradigms restricted to neuter forms, while [bo:nd] -, 'be

needed' is a morpholoically irregular formation from the root

[bo:d] -, 'to beg'; and [po :r]- is a uniaue verb form occurring only

with the negative verbal suffix -[a], '(something) is not enough'.

This suggests that the most economical analysis will asign the

irregular features of 11 -15 to the verbs occurring therein, and will

derive these sentences from regular one -place underlying structures:

this will require that [kayy]- and [kuli]- be marked in the lexicon

for obligatory subject deletion, and that all the others, except for

[bo:nd] -, be marked for optional subject deletion. It is possible

that the correct treatment of [no]- is to make two entries in the

lexicon, one with the meaning 'to pain (intr)' and the other with the

meaning 'pain is'; then the first will be marked for no subject

deletion, and the second for obligatory subject deletion: this is

tentative, however, as the data contains no conclusive evidence one

way or the other. [bo:nd]- will have to be marked for non -deletion

of subject: in the situation of utterance, something like

[bo:nduva:], 'wanted ?' may occur; but this is to be handled in terms

of utterance (where perhaps a rule of topic deletion is required) and

not in terms of sentence by a rule of subject deletion.

In all cases where there is obligatory deletion of subject, the

question arises as to what element should be posited for the subject

in the underlying structure. The operation of a concord rule between

the subject nominal and the personal ending of the verb places limits

on the choice of subject nominal for these sentences; that is, the

form of the personal ending definitely excludes the first and second

person singular, and the second person plural. In a language such as

Ceylon Tamil, which is described by Kandiah (1967), the system of

 119

personal endings does not show the fairly advanced syncretism of the

Coorg system, and it is possible to set the limits even closer on the

choice of subject nominal; Kandiah demonstrates that, in terms of the

concord rule, it is 'most convenient to set up the third person

neuter singular pronoun from among the various possibilities' (1967:

335). In Coorg, where all singular subjects save the first and second

person pronouns, and all plural subjects save the second person

pronoun are in concordial relation with the same personal ending of the

verb, the situation is not so clear. However, it is not irrelevant to

point out that this personal ending in Coorg is historically an old

third person singular inanimate form; furthermore, native speakers'

intuitions point to Ladi], 'it' (the third person singular inanimate

proform) as the 'supplied' subject, and this corresponds to Kandiah's

'third person neuter singular pronoun' subject for these - Kandiah

informally calls them 'subjectless' - sentence types.

(iii) 'Existential' constructions.

The following group of sentences is also negated by Lille]:

n
16. [miñña'ku: ikkaku.: eccaki bo:re]

before -to-and now-to -and what- extent -to difference

'Between now and formerly, how much difference there ist'

u n n,
17. [afrikatili sima uadi]

Africa -in lions are

'There are lions in Africa'.

18. [de:va undi]

God is

'God exists'.
 120

NN ir rl

19. [avada kunni avada arel i) undij

she-of child she-of waist -in is

'Her child is at her waist' (Method of carrying small children).

u n n
20. [moyra avada beratil(i) undi]
..

ring she -of finger-on is

'The ring is on her finger'.

II It
21 . [a: to :tatili ku:va unclad

that estate -in well is - interrog.

'Is there a well in that estate ?'

22. [a: to:taki ku:va unda:j

that estate -to well is - interrog.

'Is there a well for (the use of) that estate ?'

23. [i:
NN
kunnikii
' ikka po : di
u_
candi i

this child -to now fear is

child now afraid'.

r n
24. [na:ki pani undil

me -to work is

'I have work (to do)'.

n rr n
25. [a: na:yik i) or(i) udda ba:l(i) uadi]

that dog -to a long tail is

'That dog has a long tail'

rr

26. [na: da pakka cers.a dudd(

.
i ) und
..
]

me-of-prox. some money is

'I have some money'.

All the foregoing sentences consist of a subject nominal in the

'subjective' case, and a predicate containing a form of the verb

[u11] -, 'to be' and also, or_tionally, some sort of predicate

 121

expansion, typically consisting of a nominal phrase with the head

nominal in the 'locative' (marked by -Ili)) or the allative case

(- [ki]). 16 shows the result of optional deletion of [candi] (the

appropriate form of [u11] -, 'to be'), a stylistic variation which is

open to all the examples of this type, except sentences like 18 and

probably 17 as well. There is a scale of preference in the matter

of verb deletion in these sentences which is not strictly relevant

to the present discussion, and is moreover difficult to grade

precisely; but it seems to be the case that in maximally unmarked

instances, the verb is frequently deleted where a locative expansion

is within the predicate, and frequently retained where a locative

expansion is a sentential adjunct. Thus, [undi] could more naturally

be deleted from 19 and 20 than from 17. Another factor that has to

be taken into consideration is the predicate nominal itself: thus

[bo:re] frequently tolerates deletion of [ull]- in my data. By

'maximally unmarked' instances is meant instances where there are no

grammatical categories for the expression of which the occurrence of

the verb is obligatory. Thus, the category 'interrogative', one of

whose markers is the suffix [a:], is realised on the verb in 21 and

22 above, and under these conditions the verb is non-deletable. The

only possible form of the interrogative counterpart to 16 would thus

be:

27. [miññaku: ikkaku: accaki bo:re unda:]

before-to now-to that- extent difference is - interrog.

-and -and -to

'Between now and formerly, is there that much difference ?'

where the verb occurs. Tense markers ('past' and 'non -past') are also

realised on the verb, and under these conditions too the verb is

non -deletable.
 122

The similarities of sentences 16 -27 may therefore be brought out

by including them all in a class which may be labelled - prompted by

the gloss on the verb that is common to them all -'existential'.

From this point, it is necessary to note that a variety of notional

'functions' is served by this class, and that these have to be

clearly distinguished by the grammar in order to account for the

transformational relations that these sentences contract. The

principal functions, as indicated by the translations given, are

existence, location and possession.

(a) Existence and location

The relation between these functions is demonstrable with

reference to 17, which bears formal similarities to the expression of

the basic,existential function in 18 and to the expression of the

basic locational function in 19 -21; with 17 may be compared:

It
28, [sima undi]

'Lions exist'

and
11

29. [sima afrikatil(i) undi]

'The lion is in Africa' /'The lions are in Africa'.

The presence of existential function as opposed to locational

function in a given sentence would seem in some cases to depend upon

whether the locative expansion (where this exists) is considered a

sentential adjunct or part of the predicate group. In the first case,

the existential function is present, and in the second it is the

locational. This distinction is relevant to the issue, mentioned

above, of whether the verb [u11]- appears in the surface structure:

its absence is possible only with locational function. However, the

distinction between a sentential adjunct and a predicate expansion is

 123

not always easy to maintain, as Lyons points out (1968: 390). In the

cases just considered, it is true to say that where the locative

phrase precedes the subject nominal, the preferred interpretation is

that it is a sentential adjunct (existential function). However,

stylistic considerations, for example emphasis, may interfere with

this, and the distinction is wholly neutralised in adjectivalisation

transforms; thus:
rr tt

30. [afrikatil(i) ulle sima]

Africa -in being lions

'The lions in Africa' /'Lions in Africa...',

is interpretable in two gays, as shown by the translation, the first

(ignoring the difference in number) being derived from the structure

underlying 29 and the second from that underlying 17. (Concerning the

way the feature of definiteness in the subject nominal reflects this

distinction, see below, p. 154).

(b) Existence and possession.

The relation between the functions of existence and possession is

best approached via a consideration of certain sentences which

contain what is formally an allative expansion. Thus, 22 shows what

is basically an existential sentence with a sentential adjunct

expressing 'benefactive' sense by means of the allative case marker.

21 and 22 are distinct primally in that 22 makes no assertion that the

well is located inside the estate. The adjectivalised transform of

21 is:
u rr n
31. [a: to:tatil(i) ulle h.i:va.1

that estate -in being well

'The well in that estate...',

but
 124

32. [a: to:tatira ku:va.]

that estate-of well

'The well of that estate...'

is closer to the interpretation of 22 than that of 21 since, like 22,

it makes no assertion that the well is actually within the estate.

The case marker, underlined, in 32 is the 'possessive', and the

grammar has therefore to handle the functions of existence and

possession in such a way that the relation between 22 and 32 becomes

clear; as also the relation between 24 and 33, and between 25 and

34:

33. [na:da panic

me -of work

'My work...'

34. [a: na:yira udda ba:li]

that dog-of long tail

'That dog's long tail...'

(c) Location and possession

A further expression of the function of possession to some

extent overlaps all this, and is illustrated by 26. But this too is

formally linked with the expression of existence, and makes explicit

also the relation implied in the foregoing between possession and

location. The form [pak_ka] in 26 is glossed just as 'prox.'

(proximate), since it is not easy to isolate its meaning; but it is

clearly related by my informants (and conveniently so from the

descriptive point of view) to the [pakka] that occurs in other types

of construction, as:
H

35. [naigaki ka:di pakk(a) a:ci, na:di du:r(a) a:ci;

us -to forest near has become, town far has -

become

'For us, the forest is near, the town (lit. 'district') is far
away.'
 125

(A conventional statement of a life of .retirement, preceding death.)

The relation between location and possession found in the use of

[ pakka] depends crucially on the animate or inanimate status of the

nominal with nenitive inflection in the construction:

36. Nominal;Possessive marker + roakka]

Where the nominal is animate, the interpretation is possessive; where

it is inanimate, the interpretation is locative. Thus:

n n n n
37. [ maratira pakka on balye kall(i) undil

tree -of-prox. a big stone is

'Near the tree there is a big stone',

which is to be compared with 26 above.

(d) Alienable and inalienable possession

Finally on [ pakka], it is to be noted that, given an animate

nominal, the area where this construction does not overlap the

possessive function of the 'allative- existential' construction

exemplified by 24 is precisely where a distinction between alienable

and inalienable possession seems to be drawn; thus:

rr rr rr

38. * [a: na:yira pakka ori udda ba:l(i) undi]

that dog -of-prox. a long tail is

makes no sense, except in the case where there is a long tail lying

on the ground beside the dog. Body parts are central cases of

inalienable possession, but the grammar will have to recognise also

that:

39. * [na:da pakka pani undi]

me -of -prox. work is

is not a grammatical alternative to 24 above, which therefore is also

a case of 'inalienable' possession.

Concerning the use in this way of such labels as 'alienable' and

 1 26

'inalienable', the position adopted here is that suggested in Lyons

(1966) in the context of a discussion of traditional 'notional'

terminology for the parts of speech - namely, that, where a formal

description has set up the relevant categories (in this case, compare

38 and 25), the question of labelling the categories is a secondary

matter, and one that is best determined with reference to the

clearest, or nuclear, cases (which will also be the most numerous,

or at least the largest single class). Thus, in Coorg, the terms

'alienable' and 'inalienable' apply to the formal distinction between

sentences such as 25 and 38, and therefore may be applied, as labels

for a particular formal feature, to cases such as 24 and 39 (to

account for the ungrammaticality of 39).

Finally, an interesting distinction of meaning along the same

lines can be found by comparing 26 above with:

40. [na:ki cenna dudd(i) undij

'I have some money',

whereby [duddi] could most typically be translated in 26 as 'coins'

(as well as 'money'), and in 40 as 'money' (in preference to 'coins'),

the implication being that in 26 the money is on my person, whereas

in 40 I am referring to my wealth.

1.1.2 Sentences negated by [alla] ('Relational'

The class of sentence types negated by [alla] is, in the

broadest syntactic terms, homogeneous, since each example is of the

general pattern:

41. Subject nominal + Predicate nominal

However, the class includes sentences which express a number of

different notional functions, and these are indicated, in conventional

set -theoretical terms, in the examples given here:

 127

tl N tl

42. [ninna:ndi bandave bollu]

yesterday having- Bollu Subject Predicate

came-he
nominal nominal
'The man who came yesterday is Bollu'
n
43 [ave angadika:rej

he shopkeeper

'He's a shopkeeper'.

44. (a: manse rayte]

Subject Predicate
that man farmer E
nominal nominal
'That man is a farmer'.
tt tt tt u
45. [na:da tammane ori nalle ti:n ika:re]

me -of younger a good eater

brother

'icy younger brother is a good eater'.

p M
a
46. [kodave kinnariga perta a: ro: gya: raj
Subject Predicate
Coorg boys very healthy-ones
nominal nominal
'Coorg boys are very healthy types'.

42 exemplifies identity predication, while 43 -45 show class-membership

(with role -predicates in 43 and 44), and 46 shows class -inclusion.

Every kind of nominal group in the language may occur as the subject

nominal in this class of sentences, and the examples cited above have

been chosen to illustrate the range of possibilities (thus, a

pronominalised subject showing the adjectivalisation transformation

in 42; an untransformed pronoun in 43; demonstrative and possessive

nominal groups in 44 and 45; and a modified indefinite nominal group

in 46). Certain restrictions operate on the type of predicate

nominal that may occur, however; for a discussion of these, and the

implications they hold for the form of the underlying structures of

these sentences, see below, pp. 154-158. It should be emphasised that

 128

there are no formal grounds for distinguishing the different

functions noted above for 42, 43 -5 and 46, and that these sentences

therefore all express a single syntactic relation. Further, since

the symbols =, E and C all may be interpreted as representing types

of relation holding between things or classes of things, this

syntactic pattern will henceforth be called 'relational'.

2.0 The copula

The last two basic sentence patterns - the existential and the

relational - raise the issue of how Coorg handles the notion of 'being',

and therefore an analysis of the copula verb follows here. Tradition-

ally, the copula is treated as a verb in its own right; however, it

has been suggested in more recent theoretical work that it is not a

constituent of deep structure at all, but simply a 'vehicle' for

expressing categories that are normally marked in the verb, such as

tense, mood and aspect; the corollary of this being that in

maximally unmarked sentences it may not occur at all: for a represent-

ative statement of this position, see Lyons (1966, 1967 and 1968) .

Kandiah (1967), in a discussion of Ceylon Tamil, considers this

question and presents six arguments for according deep status to the

copula. All of his arguments are relevant to the discussion of the

copula in Coorg, but are not accepted here. Accordingly, in what

follows the Coorg verbs of being will be examined, further examples

given and a critique made of Kandiah's position - which, since it is

not published, will be set out fairly extensively. It is maintained

here that the arguments against Kandiah's position are also relevant

to the situation in Ceylon Tamil, as described by him.

In view of the fact that Kandiah's position is criticised below,

 1 29

it is worth making the point at this stage that his work contains

the first explicit treatment, in transformational-generative terms,

of the copula in Dravidian, and it is precisely because it is an

extremely detailed and valuable survey of Dravidian verbal syntax

that it is used as the framework for discussion here.

2.1 The copula forms [a:gj- and [uii]-/[ir]-

Two verbs enter into the discussion of the copula in Coorg,

one of which exhibits a mixed paradigm drawn from the forms of two

distinct verbs of being. The three historical verbal bases which

are thus involved have the following forms in Coorg:

(i) [a:g]-

(ii) [ir]-

(iii) [ul1]-

An examination of the cognate forms of these three items in other

Dravidian languages (listed in DED 282 for [a:g] -, DED 407 for [ir]-

and DED 599 for [ull] -), together with their meanings, is enough to

show that there is a fairly wide semantic distinction between the

entries at DED 282 on the one hand, and those at DED 407 and 599 on

the other. DED 282 typically records glosses such as 'become, come

into existence, happen, occur, prove to be, be what ought to be, be

right, be possible', etc. (in Coorg, [a:gj- means 'become, happen').

On the other hand, DED 407 is most commonly recorded as 'exist,

remain, sit down, be in a place, live, be stable', etc. (in Coorg,

[ir]- means 'be (in a place)'); while DED 599 contains such glosses

as 'to be, to be there, exist, live', etc. (in Coorg, [ull]- is

sm having merged).
identical in meaning with [ir] -, the two paradigms
mer g ed )

that
Given these semantic areas of coverage, it is not surprising

in any
the forms in DED 407 and 599 may be wholly interchangeable
 1 30

particular language - and this is the case at least for certain

varieties of Tamil, for example. But in Coorg this close relation-

ship has resulted in a merger of the two forms, where they are in

complementary distribution with respect to the syntactic environment

in which they occur. Thus, from [irk- are drawn those forms that are

marked for past or future tense, or mood, or aspect; while the

remaining forms are made on the base [u11]-. Prima facie, it would

seem to be appropriate to say that the [ull,- forms are marked for

present tense, but this is not so; in adjectivalisation transforms,

the forms -[iñje], and -[ippe] are possible (with past, and future

reference, respectively) but are quite rare,- [ulle3 generally

occurring. Thus constructions such as 31 (above) are truly tenseless.

It is only in contrast with the finite past and future tense forms

(on the base [ir] -) that the [ull]- forms may be glossed as

distinctively 'present tense'; in what follows also it will be seen

that it is convenient to regard just the past and future tense

forms as marked.

Thus there is no reason for setting up [ull]- and [ir]- as

distinct from each other in the syntactic base; they form a single

system, where there is a three -way tense distinction (past, present,

future). As such, this system is unique, since in the rest of the

Coorg verbs there is only the two-way distinction between past and

non -past.

Further, [u111- /[ir]- on the one hand and ta:g] on the other

express the aspectual distinction between unmarked stative

sentences (showing static aspect) versus those that are marked for

dynamic aspect, respectively (for the terms 'static' and

'dynamic', and the relation holding between them, see Lyons, 1968:397f.).
 131

The examples below illustrate the foregoing.

2.1 .1 The copula in existential/locational function

N
« 11 11 u
47. [ave manel(i) undi]

he house -in is

'He is in the house'.

11 /T 11
«.
4Q. [ave manel(i) inat; ,

he house -in was

'He was in the house'.

si
49. [ave manel(i) ipna_]

he house -in will -be

'He will be in the house'.

50. [a: varatili, dumba ma:Age undij

that tree -in many mangos are

'There are many mangos in that tree'.

tl II
r« 11

51. [indi ni:da kelas(a) a:ci inni ni:ni po:kalu]

today you -of work became now you can-go

'Your work ended today; now you can go'.

52. [ente bo:ndiye:ñgiyu a:na]

what needed -if -even will- become

'WhAtever you want will happen'.

53. [nariga santo:sat(i) ikka]

we happiness -in let -be

'Let's be joyful:'

47-49 illustrate the three -term distinction of tense; 50 on the

one hand and 51 and 52 on the other show the opposition between

static and dynamic, respectively; and, finally, 53 represents a

typical modal ('optative') formed on the ir- base (the relevant verb
 132

form is underlined in all these examples, and in those that follow).

2.1.2 The copula in possessive function

The distinctions of tense and aspect that are illustrated

above in sentences expressing existential and locational functions

also hold with regard to possessive constructions involving the

copula:

54. [ni:ki raja iñjata:]

you -to holiday was?

'Did you have a holiday?'

55. [avangi madi a:ci]

he -to enough became

'He had enough'.

2.1.3 The copula in relational sentences

Ay
N
11 tl 11

56. [ave on da :ktrij

he a doctor

'He's a doctor'.
II 11 11

57. [ave ori da:ktr(i) a:ci]

'He became a doctor'.

N 11

58. [ave ori da:ktr(i) a:pa]

'He will become a doctor'.

11 11

59. [kinnanda mu:di ceriyad(i) aci]

boy -of face small-it became

'The boy's face grew small' (Description of effect of being

scolded).

However, in the case of relational sentences it is not possible

to make distinctions of tense without also marking dynamic aspect in

the copula: that is, tense may only be marked on the base [a:g] -,
 133

'to become', and not on the base [irk -, 'to be, be in a place'.

(This consideration confirms the distinction drawn earlier between

the classes of relational and existential sentences on the grounds of

their different negating elements.) Thus, it is only on the basis

of this distinction that the grammar is able to account for the

fact that the perfective version of 57 (dynamic aspect):

eo
N Ii It 11 11

60. [ave on da:ktri a :yitundij

he a doctor become -is

'He has become a doctor/is a doctor'.

implies the unmarked static sentence 56, but that the counterpart of

the unmarked copula in the locative sub-type of existential sentence

shown in 47 is not formed on the base [a:g]- as in 60 but is one of

the directionally -differentiated pair [bard -, 'to come' /[po :] -, 'to

go'; thus:
N 11 II II

61. [ave maneki banditundifpo:yitundi]

he house -to come-is / gone-is

'He has come/gone to the house'.

implies the unmarked static 47 in the same way that 60 implies 56

(see also the discussion on pp. 161 -162, regarding the connection

between [bard- /[po:]- and the copula forms in existential sentences).

2.1.4 The copula in interrogative and negative sentences

The final examples relevant to this discussion of the copula

are concerned with interrogation and negation. (For convenience,

examples given earlier will be re- introduced here.) In the following

examples, the affirmative- declarative sentence is followed by the

negative, then the interrogative, and finally the negative- interrogat-

ive counterparts:
 134

(i) Existential sentences (unmarked)

(a) existential function

18. [de:va undi]

'God exists'

62. [de :va ille]

God not -be

'There is no God'

63. [de:va unda :1

'Does God exist ?'

64. [de:va illeya:]

'Is there no God ?'

(b) locational function

ti
h
11 11 II

47. [ave manel(i) candi]

'He is in the house'

65. [ave manel(i) ille]

'He is not in the house'

66. [ave manel(i) unclad

'Is he in the house ?'

67. [ave manel(i) illeya: )

'Is he not in the house ?'

(c) possessive function

11 T 11 It

25. [a: na :yik(i) or(i) udda ba:l(i) undi]

'That dog has a long tail'

68. [a: na:yik(i) or(i) udda ba:l(i) Ille]

'That dog doesn't have a long tail'

69. [a: na:yik(i) or(i) udda ba:l(i) unda :]

'Does that dog have a long tail ?'

 135
11 11
It

70. [a: na:yik(i) or(i) udda ba:1(i) illeya:]

'Does that dog not have a long tail ?'

(ii) Relational sentences (unmarked)

It 11

43. [ave angadika:re]

'He's a shopkeeper'

71. [ave axigadika:ranalla]

he shopkeeper- not -be

'He's not a shopkeeper'

4 tt
72. [ave arinadika:rana: ]

'Is he a shopkeeper ?'

73. [ave arigadika:ranallava: ]

'Is he not a shopkeeper ?'

(iii) Sentences marked for aspect /tense

In the next set of sentences, the negative counterparts to

examples of dynamic aspect are given:

n n n
74. [a: maratili, dumba ma:Age a:pile] (cp. 50)

'In that tree, there won't be many mangos'.

1I

75- [avángi madi a:yile] (cp. 55)

'He didn't have enough'.

N II 11 II II

76. [ave ori da:ktr(i) a:pile] (cp. 58)

'He will not become a doctor',

and in the final set, the negative counterparts are provided for

those static sentences which are marked for tense:

77. Lave manel(i) iñjile] (cp. 48)

'He was not in the house'

N
N tl II

78. [ave manel(i) ippile] (cp. 49)

'He won't be in the house'.

 136

The foregoing illustrate that in existential and relational

sentences which are marked for tense or aspect the copula behaves

like any other verb in the language by being supplemented by ille

in the negative (cp. 3 above): but where the copula is unmarked in

an existential sentence, it is, as pointed out in Asher (1968: 104)

with regard to the same problem in Malayalam, 'not supplemented,

but replaced by the negative form'. At the same place, Asher also

handles Malayalam relational sentences (which he calls 'equational';

Asher, 1968: 99), noting that the maximally unmarked forms of these

show aane in the positive which is replaced by alla in the negative.

As a consideration of the examples given so far will suggest, the

situation in Coorg is slightly different, in that no overt copula

form equivalent to Malayalam aane occurs, at least in unmarked

sentences. Under elicitation conditions my data shows one case of a

copula in this context:

79. [adi nalled(i) (und)a:j

it good -it (is) - interrog.

'Is it a good thing ?' /Is it good ?',

but there is every reason to believe that this is highly aberrant;

the informant on this occasion described the parenthesised copula as

'unnecessary', and never used it again on subsequent occasions:

furthermore, it was given under artificial conditions, where the

informant seemed to be trying to make his language as 'explicit' as

he could to an outsider.

Granting that this is so, then the situation for Coorg is that

the negative element [alla] replaces nothing, and relational

sentences are distinct in that the interrogative suffix -[a1 in

these sentences occurs on a nominal (see 72), without requiring a

 137

copula as 'vehicle'. However, the grammar has to allow for this

in

other cases as well, where emphasis is linked with interrogation;

compare the following:

y II II
80. [ave ninna:ndi bandávána:J

he yesterday having-come -he?

'Is he the one who came yesterday?' (non -emphatic).

N N It N It

81. [avana: ninna:ndi bandave]

'Is he the one who came yesterday?' (emphatic he).

2.2 The status of the copula

Examples 47 to 80 above, together with 16 to 46 given in

discussing the basic sentence patterns, provide sufficient material

for considering the six arguments for the deep status of the copula

which are given in Kandiah (1967: 71 -6). These arguments are

presented here and analysed.

2.2.1 Kandiah's arguments

(i) Kandiah's first argument (1967: 71) is 'a general theoret-

ical point', namely that 'the criterion of the frequency of occurrence

may not be legitimately invoked by the grammar to exclude the

Copulative Verb from treatment'.

In answer to this, it may first be noted that denying deep

status to the copula is not excluding it from treatment, but treating

it in a particular way. Secondly, the fact that a grammar of Coorg -

or of Ceylon Tamil - which denies deep status to the copula may have

numerically as many cases of 'writing in' the copula in the surface

structure as a grammar which includes the copula in deep structure

has of 'deleting' the copula from the surface structure in order to

fit the data is a relatively minor issue, since the theoretical

 138

principle here s not frequency of occurrence but predictability

of occurrence: or refining this, predictability of the privilege of

occurrence. That is, if the :ramnar can account for the distribution

of the copula in the corpus it is describing by reference to the con -

figuration of the relevant underlying syntactic ctructures, then

there is nc justification for including the copula in deep structure:

aná this holds true even where there is, as noted above (p. 121), a

scale of preference in the matter of occurrence cf the copula in.

existential sentences. Two possible ways of handling the scale of

preference would be:

(a) to 'write in' the copula in surface structure before the operation

of certain stylistic transformational rules optionally deleted it

in certain cases - e.g. when the predicate nominal is [bo :rej,

'difference' (see above, p. 121).

(b) to delay 'writing in' the copula until certain optional rules in

the transformational component of the language have specified

whether in any particular instance the copula will occur. These

rules will be rather similar to optional stylistic rules, but

since stylistic rules are usually formulated as deleting

elements rather than optionally writing them in, it may be

necessary to keep the two distinct.

The choice between (a) and (b) would be a purely empirical one,

however, and would not affect the status of the copula. All this is

a purely theoretical answer to a purely theoretical argument.

(ii) Kandiah's second argument (1967: 71 -2) is really a

continuation of his first, since it shows that the criterion of

frequency, if followed, would favour setting up the copula as an

element of deep structure, since there are cases where its 'retention'
 139

is preferred to its 'deletion'. Because the criterion of frequency

of occurrence has been denied relevance to the issue above, no

specific answer to this argument is given here.

(iii) The third argument Kandiah presents (1967: 72-3) rests

on the fact that, in Ceylon Tamil, the negative particle may

optionally show gender and number concord with the subject nominal.

He gives the example (numbered to fit in with this work):

82. kantan vayttiyar alla(r)

'Kandan isn't a doctor'

where the parenthesised r is the optional concordial suffix.

The argument itself is in two parts. The first states that it

would be exceptional to have this as the sole instance where concord

is (optionally) realised on an element which is not a verb (a

particle), so that a verb has to be posited here, even though it has

no surface realisation.

The second part of the argument is that it would be counter-

intuitive to take the negative particle as a verb; because this would

ignore the differences between particle and verbal behaviour for the

sake of the concord feature of resemblance, and because it would

create the 'curious situation' whereby the copula pattern has no verb

in the positive but a verb in the negative.

It should first be pointed out that number concord is marked

on predicate nominals in Coorg, and generally in languages of the

group (this is one of the peculiarities of the predicate -constituent

nominal as opposed to other types of nominal; see, for example, the

discussion in Bach, 1968: 102-3). Hence it can only be that Kandiah

is here concerned with the particular phonological form of the

concordial suffix he cites, which is identical with the type

 140

associated with verbs rather than nouns. In this situation it may

be held either that the particle is a peculiar type of verb or else

that verbal -type concord is realised on particles as well as on

verbs. But it is just as exceptional to find that with alla, and

only with this element, the verb -deletion rule may optionally erase

only the base form of the verb, and not the concordial marker: and

then that alla may have these non- erased elements attached to it

(as if it were a verb, which is still more exceptional if alla is

not taken to be a verb). The point is that, whatever analysis is

adopted, there is something unusual about it.

The second point is more crucial, since it depends in part on

what one is prepared to call a 'curious situation'. The theory

advanced in Lyons (1966, 1967 and 1968) which seeks to deny that the

copula is a constituent in deep structure is based in part on the

notion of markedness: in these terms, the copula is seen as a purely

superficial 'vehicle' for overt marking of categories such as tense,

mood, aspect, etc. It is not at all curious, from this standpoint,

that Coorg and Ceylon Tamil should agree in having no surface

structure copula in positive sentences but in showing a form that

might be interpreted as a copula in negative sentences. Therefore

the reference to the 'curious situation' is no valid argument against

taking alla as the negative copula form.

Hence Kandiah's only reason for not taking alla as a verbal

element is that it is 'counter- intuitive' to do so. Against this

sort of objection argument on the basis of a closed corpus is clearly

unreasonable. However, it is not without relevance to this issue to

recall that alla (and ille) are traditionally (i.e. in the native

tradition as well as the Nestern) treated as verbs: and that from the
 '141

historical point of view, to the extent that Ceylon Tamil gives

evidence of the form aliar in such cases as 82, it would seem to be

preserving what is recorded in DED 198 as aliar, the third person

plural masculine form of the Tamil verb al- 'to be not so- and-so',

and to the extent that it shows alla Ceylon Tamil would seem to be

in accord with what DID notes of modern Tamil, that alla has come to

be used for 'all persons and numbers'.

(iv) In his fourth point, Kandiah states (1967: 73 -4) that

because copula verb phrases take similar expansions (e.g. time -

phrase expansions) to those of other verb phrases, the copula has to

be posited in deep structure; for such expansions are developed out

of his node Pred.P., and 'if the Copula Verb were not posited, the

Time expansion would have to be developed differently for the two

different patterns. In Copula sentences it would have to be

developed as a non -adverbial element of equal status with the subject

NP and the Complement, while in other sentences it would have to be

developed as an adverbial expansion of the verb, of lower status than

the subject NP'.

Kandiah's proposal, in terms of his Base Component, and concerning

the deep structures required for an analysis which treats the copula

as a purely surface element may be set forth for Coorg as follows:

 142

83. Non -copula pattern

%S -

NP Pr¡d .P.

VP

Verb

Tune P. V

[bollu] [ikka] [undandundi]

Bollu now is- eating (rice)

'Bollu is taking his meal now'.

84. Copula pattern

NP Time P. Complement

NP

I n
[bollu] [ikka] [da::ktri ]

Bollu now doctor

'Bollu is now a doctor'.

However, this is not what is proposed here; it is not even

possible to accept the Non -copula analysis for the sake of argument,

for the reasons set out in what follows.

It would seem to be the case that the class of expansions that

may be taken equally by the two patterns is quite restricted - as is

discovered when looking for examples. On the other hand, a large

number of expansions are peculiar to either the one or the other

 143

pattern. Thus, adjectival expansions involving any member of the

class of 'true' adjectivais are restricted to an NP node: the same

point is made, with reference to Kalayalam, in Asher (1968: 95) -

' ... "adjectivais" in Nalayalan form a category whose occurrence will

be stated as being restricted to pre -nominal position'. Further,

adverbial expansions such as [melle], 'slowly, gently' (a 'true'

adverbial), or a large class of abstract nouns showing the 'locative'

case, such as [nida:natili, (lit. 'in the end'), 'eventually, in a

leisurely way' or [jo:rili, (lit. 'in force, speed'), 'forcefully',

are all by definition restricted to the verbal node. What is

suggested here - and Kandiah also suggests this in a footnote to his

argument - is that those expansions which are taken equally by both

copula and non-copula patterns are to be treated as sentential

adjuncts, or possibly as Fred. P. adjuncts. This would involve

revising the P- marker 83 so that Time P. stands as an immediate

constituent of S or of Pred. P.; and it would not be difficult then

to formalise the two patterns, 83 and 64, in such a way that their

Time P. expansions are of comparable status.

(v) Kandiah's fifth point (1967: 74 -5) concerns the formal

(phonological) identity between his copula aaka and the verb aaka,

'to become (cp. Coorg [a:g3- 'to become'). He states that, to show

the syntactic similarities between the two, it is necessary to set

both up as constituents in deep structure; that only in this way can

the grammar account for the similarity of:

85. murukan panakkaaran aanaan

'1 urugan became a rich man'

and
 144

86. kantan vayttiyar aavaar

'Kandan is a doctor'

Kandiah then says that because there are differences in these

verbs, despite their basic similarity, the deep structure must

recognise two verbs, aakal and aaka2. He calls them VCop and
Act
(referring to the forms double -underlined in 85 and 86,
VCop Idea
respectively.

It is clear that Kandiah's point here is consistent with his

decision to accord deep status to the copula; but it in no way

supports that decision. The situation illustrated for Ceylon Tamil

by 85 and 86 has been set out above for Coorg, in examples 56 to 59.

There, it was argued that the distinction between 'becoming' and

'being' sentences in Coorg (parallel with Kandiah's patterns with

and respectively) may be handled in terms of

VCop Act VCop Iden'
dynamic and static aspect: and that in sentences marked for dynamic

aspect the copula occurs in surface structure as the base [a:gj -,

whereon the appropriate aspectual marker may occur. The situation

is not materially altered by the fact that Coorg has no surface

structure copula form in aspectually unmarked sentences (i.e. a

corresponding form to Kandiah's VCop Iden That is, if there is no

'

objection in principle to an analysis which says that what is

realised as aavaar in 86 is represented by zero in Coorg surface

structure - and this sort of analysis is tacitly assumed in any cross -

language comparison - then this analysis at no point makes contact

with the issue of the status of the copula. Hence there would seem

to be no reason for proposing a radically different status for the

Coors copula from that for the Ceylon Tamil copula on the basis of
 145

this particular difference in the data. To summarise the reply to

Kandiah's fifth point, it may be stressed that the argument for two

copula verbs, to handle what are here called static and dynamic

sentences, is not an argument for the deep status of the copula;

since an analysis which recognises static and dynamic aspect, but no

copula, in deep structure is in principle able to account for the

same facts. If it can be shown that it is desirable on other

grounds to set up static and dynamic aspect, then this would be an

argument a rainst setting up the copula as a deep element (for a

discussion of static an,- d;,-namic aspect on other grounds, see pp. 161-

162 below): for the foment it is sufficient to note that there is

nothing in uandiah's fifth point in favour of such an analysis.

(vi) Kandiah's sixth and final point (1967: 75 -76) is related

to the foregoing. It is that, because the copula verb occurs in

'relativised' constructions, it must be a deep element. It is not

suggested here that this is put forward as an argument in the same

sense as the previous five, but more as a confirmatory piece of

evidence: but, under either interpretation, it is insufficient.

Thus far, examples of 'relativisation' (this term is from Kandiah,

1967; but &sher, 1968: 105 -6 is more satisfactory in calling this

'adjectivalisation', and it is this latter term which is used in

this chapter) have occurred in 30, 31, 42, 80 and 81 above: of

these only 30 and 31 involve the copula. These latter two are

repeated here, and further cases of copula -relativisation given:

It It

30. [afrikatil(i) ulle sia]

Africa -in being lions

'The lions in Africa' /'Lions in Africa...'

 146

It It
31. [a: to:tatil(i) ulle kuva]
that estate -in being well

'The well in that estate...'

87. [ikka ulladi po:yiruva]

now being-it will-completely-go

'What exists now will pass away'.

88. [su:rarigala:ne ratiariga]

heroes-being kings

'Heroic kings /Kings whogaerre heroes...'

w j
"
89. [santo:sava:ne kelasa]

harTiness-being work

'Joyful work...'

90. [santo :savulle kelasa]

happiness -being work

'Joyful work...'

The above examples cover the main types of copulative function

discussed so far (save for possessive function: this is reserved for

analysis immediately following this), but show no new facts that

would require a distinct treatment of the copula from that proposed

for verbal predicate constructions. Thus, 30, 31 and 87 show the

form [ull]- in constructions derived from existential predicates

(30 showing neutralisation of the existential /locative contrast, 31

showing a preferred locative interpretation and 87 an unambiguous

existential interpretation), whereas 88 shows [a:g]- in the

construction derived from the relational predicate type. Finally,

89 and 90 represent an overlap of the existential and relational

Predicate types, with no easily discernible difference in meaning.

The overlap is best treated as a characteristic of the syntactic

 147

behaviour of [santo :sa], which (like a number of other loanwords

denoting abstract concepts) apparently stands between the clear

instances of nominals and existential predicate elements (e.g. verbs).

In this connection it is important to notice that [santo :sa] (and the

other words of this type) may occur predicatively in a sentence

negated either by [alla] (the relational construction) or by [ille]

(the existential construction), again apparently with only a slight

difference of meaning. Providing a helpful gloss for such items is

rather difficult; perhaps 'happiness' and '(being) happy' capture the

distinction involved here.

Finally on Kandiah's sixth point, it may be noted that an analysis

which treats the copula as a superficial element will have to examine

the adjectivalisation construction illustrated here for the purpose

of deciding at what stage the copula is to be written into the

string: i.e. whether before the transformation or after it. In this

context, Kandiah's point is relevant, and would seem to point to the

preference of writing in the copula before the transformation so that

its base form may be determined with reference to the underlying

predicate construction, thus obviating the necessity for providing

distinct rules for determining the shape of the copula in both

predicate and adjectivalised structures. But it in no way argues for

the deep status of the copula.

2.2.2 The possessive copula

In conclusion to his discussion of VCop Iden and VCop Act'

Kandiah goes on to describe a pattern which he keeps distinct and

calls 'Possessive'. This is the counterpart in Ceylon Tamil to what

has been described above (p. 134) as the 'possessive function' of the

copula in Coorg. For this pattern, Kandiah sets up a third verb

 148

aaka, which is interchangeable with two other verbs in this pattern,

iru and ulla. This interchangeability is stated as holding at least

in the cases where there is an abstract subject nominal. iru and ulla

are cognates of Coorg [ir]- and [ull]- respectively. Kandiah says

(1967: 79) that the 'possessive' verb aaka is deletable from surface

structure, but that it is nonetheless necessary to 'posit' it; by this

he appears to mean that it must be accorded deep status: 'The reasons

for doing so are identical with some of the reasons above for

establishing that there is a VCop and will not be repeated

Iden'
here'. He thus takes it as a sort of copula (by implication) and

this agrees with the analysis of Coorg presented above, in examples

16-26 and 31 -34. The situation in Coorg is somewhat different in

that [irk- and [ull]- have merged in the language in a way that they

have not in Ceylon Tamil; but whether this materially affects the

status of the copula may be doubted. As Kandiah suggests, the nature

of the subject nominal (in feature terms, [± abstract]) seems to be a

conditioning factor in the interchangeability of aaka and iru and

ulla in Ceylon Tamil 'possessives', and an analysis which excluded

the copula from deep structure could account for the surface alter-

nation of these three forms in this construction with reference to

the subject nominal. Certainly there is no justification for

handling 'possessive function' copula sentences in Coorg as distinct

from the other copula types. Asher (1968) gives pertinent examples

from Malayalam, where irikkuka and unta (cognate with [ir]- and

[ull]- respectively) have remained distinct verbs; it is noted there

that 'the use of irikkuka in a way that allows it tc be considered

(1968: 102)
without question a 'verb of being' is very restricted'

and this restriction would seem to involve setting

up a number cf modal
 149

categories for the deep structures of copulative sentences so as to

account for the variations within the locative type that are

illustrated there for Malayalam. Thus, in the frame

avan viittil , unta makes the unmarked assertion 'he is at

home' and nena implies that this is stated by 'someone who is

elsewhere than in the house'; while irikkuka in avar verute

viittil implies that "They are just at home' (i.e. they have
00

no work to go to.)' (196E: 103).

This concludes the discussion of Kanr?iah's position regarding the

status of the c o';ula .

2.2.; The adverbialisation transformation

It remains to ;oint out one further construction in which the

copula occurs, before summing up. This construction is described in

Asher (1968: 106-7) as the result of an adverbialisation transform-

ation. The following examples are given for illustration:

n v n
91. [ettira bayya ke:ra:yi ke:rira bayya a:la:yi ba:nati po:pakkuva:i

bullock behind rope- rope -of behind man- sky-in go- is -it-
-of become become not?

'Isn't there a man at the reins of a bullock flying through the

sky?' (From a story text, describing a miraculous vision).

n_
92. [male jo:ra:yiti pojjati.j

rain force -become rained

'The rain beat forcefully'

n u
K o:dici]
93. [ave po:diya:yit(i)

he fear- become ran

'He ran fearfully'.

 150

u u 11 II

94. [ammatti na:dili otte poytili vi:rana:vi.tu: na:di.ka:rada

Ammathi Nad -in, single blow-in famous- person people -of-the

-become -and Nad-of

'In Ammathi Nad, Kullachanda Chondu, the famous single- sticks

kodi kuññiya:yitu: kullacanda co:ndu bolandandiñjati

beloved child-become Kullachanda Chondu was- growing-up

-and

fighter and beloved child of his people, grew up'.

The underlined forms in 91_94 are from the base [a:g]-, 'to
n_
become'. 91 shows [a:yi] where the others have [a:yitij (this latter

suffixed by the conjunctive particle [u:3 in 94) and this divergent

form needs explaining. The story from which 91 is extracted is a

traditional one, and 91 appears to represent a conventional form of

the description. It is just possible that [a:yi] looks to Malayalam

aayi (as set out in Asher, 1968), but there is no need to look

further than Coorg for an explanation: quite commonly, and especially

in set phrases, the verbal form in -[iti.] (traditionally called the

past participle in European descriptions of Dravidian languages, and

expressing perfective aspect) shows a short form which consists just

of the past tense marker; thus:

ti 11 R

95. [ave catti po:cij,

which shows the past tense marker underlined (followed by a non-

ft

morphological element, or 'enunciative' vowel [ij) is said, under

conditions of elicitation, to be the short form of:

11 11

96. [ave cattiti po:cil

he having went
-died

'He died' (almost with the sense 'He went and died').

In the case of [a:g3- the past tense marker is -[i] suffixed to

 151

a form of the base which does not show final -[g], and therefore there

is no difficulty in taking [a:yi] as the short form of [a:yiti].

Granted, then that there is only a single underlying form to be

accounted for in these examples, it is clear that they all show a

pattern that is apparently related to copula constructions. However,

not all the examples are of the same sort. 92 and 93 show

nominal + [a:viti] as a type of manner adverbial, while

[vi:rana:yiti] and [kodi kuññiya:yiti] (the forms without the

conjunctive suffix) in 94 are clear cases of 'past participle'

function. The relationship between the two types of construction is

superficially clear; but there are differences that require a distinct

treatment for each. In Coorg, the intonation pattern is the clearest

marker of these differences, and Asher (1968) is able to draw on the

written tradition of Malayalam to point out the same sort of fact -

that the 'participle' is graphemically a free form while the adverb -

ialisation marker is 'inseparable from the noun which it follows'

(1968: 106). Exactly what analysis should be adopted for each of

these patterns is unclear (and 91 would appear to be a problem in

statement in any event since it shows forms which are on the border-

line between participial and adverbial function), but it seems

natural to assume for both patterns that they are derived from an

underlying copula predicate construction. Thus:

97. [ettira bayya ke:r(i) uadi]

'At the back of the bullock are reins'

98. [male jo:r(i) undi]

The rain is heavy'

If
v w "
99. [avangi po:di undi]
..

'He is afraid'
 152
N
If
100. [ave kodi kuññi]

'He is a beloved child'

illustrate the sort of constructions to be posited for 91 -94. The

adverbialisation transform is to some extent similar to the adjectiv-

alisation transform discussed above (p. 145), but it is to be noted

that it obligatorily involves that form of the copula ([a:g] -) which

indicates dynamic aspect: the sort of alternation noted in

adjectivalisation between 89 and 90 above is therefore not possible.

This would sug7est either that the operation of the transform be

restricted to embedded copula sentences which show dynamic aspect, or

else that its operation should obligatorily assign dynamic aspect to

the embedded sentence. A reasoned choice between the two cannot be

made at this point (but see p. 163 below).

2.3 The analysis of the copula

What follows here is a summary of the various types of copula

predication in Coorg and an outline of the overall analysis they

suggest.

2.3.1 Summary

There are two basic patterns involving the copula, the relat-

ional and the existential. The first of these shows no surface

copula in maximally unmarked sentences, and shows [alla] in the

negative in otherwise unmarked sentences: distinctions of tense, mood

and aspect are marked on the base [a:g]-, and the negative counter-

parts of these sentences show morphologically regular contractions of

the base [a:g-] + [fille;. The second pattern optionally shows the

copula verb [ull]- in maximally unmarked cases, and [ull]- is replaced

by [fille] in the negative forms of otherwise unmarked sentences: tense

 153

and mood are marked on [ir]- (for static aspect) and on [a:g]- (for

dynamic aspect), and the negative counterparts of these sentences

show regular forms with either [ir]- or [a :g]- + [ille]. The copula

system is unique in Coorg in showing what is superficially a three -

term distinction of tense: however, it is probably better to take

this rather as, first a distinction between marked and unmarked, and

then a two-term distinction of tense within the marked category. In

terms of this analysis, the following table may be set up to

illustrate the range of possibilities:

Unmarked ( for Marked

tense/asr^ef)
Past Future

Dynamic Dynamic

Pos // [ull]-
a:n]-
a: p]- [ipp]-
[

[a:c]-

Neg [alla]/[ille] [a:yile] [iñjile; [_a:pile ] [irpile]

Figure 1: The forms of the copula

In the first, 'unmarked', column the table shows the forms for

the copula in both relational (before the slash) and existential

(after the slash) sentences. All the remaining four 'marked' columns

are relevant to existential sentences, but the dynamic versus static

contrast does not apply to relational sentences that are marked for

tense: only the forms recorded in the 'Dynamic' column may occur in

these. For simplicity, the various modal categories have been

eliminated from the table, but they present no problem of statement

since they are formed on [ir]- or [a:g]- exactly as described for the

two categories of tense.

 154

2.3.2 The analysis

(i) The formalisation of the deep structure of copula

sentences raises many problems, and what is offered here is not to

be taken as a final statement. However, it is worth pointing out

that the formal distinction in Coorg between unmarked relational

sentences (negating with [alla]) and unmarked existential sentences

(negating with the distributionally far more common [ille]) agrees

with the proposals concerning noun phrases in Bach (1968: 103 -4).

There it is argued that the peculiar properties of predicate

nominals (such as concordial relationship with the subject nominal,

and restrictions on the kinds of determiners that may occur -

features that are attestable for Coorg also) lead one to regard

predicate nominals as not being noun phrases in underlying structure;

and that a treatment of them which introduces them into noun

phrases by way of embedded relative predications can account for

their peculiarities. Bach envisages 'a noun phrase consisting of a

pronominal-like item (of a nature yet to be precisely determined)

plus a further embedded sentence ' (196;: 104). Adapting Bach's

argument from the restrictions on determiners with predicate nominals,

it is possible to have:
If
N II 11

101 . [ave (ori) da:ktri;

he (a) doctor

'He is a doctor',

but the following is not natural without an expansion:

N II 11

102. [ave a: da:ktrij

he that doctor

'He is that doctor'

is derived from
This can be explained if [da:ktri] in both cases
 155

the subject NP node of an embedded sentence. In the first case the

indefiniteness of the predicate nominal is signalled by the fact that

the embedded sentence is an existential copula type, such as under-

lies:
n n
103. [da:ktri undi]

'There is a doctor'.

The indefiniteness of the subject NP in this pattern is a

feature of English also: Lyons (1968: 390) points this out in the

following way - 'Existential sentences typically have an indefinite,

rather than a definite, subject: this fact raises the possibility

that they should be treated, in a syntactic analysis of their deep

structure, as indefinite locatives...' On the other hand, 102 would

be derived from some other sentence pattern, i.e. which allowed a

definite subject nominal, and this would also account naturally for

the need in this case of an expansion (the predicate of the embedded

relative sentence).

(ii) In terms of this analysis, the grammar would recognise

[alla] as the negating particle of a particular class of expressions

of which the predicate constitutes a nominal derived from an embedded

relative sentence. This analysis is supported by what seems to be an

important fact; namely, that the only other use of [alla] in the data

is in cases where a (negative -interrogative) marking is required on a

whole sentence (as opposed to an element within it):

104. [ave bappa]

'He will come'

105. [ave bapp(a) alla:]

'He'll come, won't he ?'

 1 56

106. [idi na:da pustak(a) alla:]

it me -of book isn't it?

'It's my book, isn't it ?'

104 -6 show the similarity of function played by [alla:] in 105

(with an intransitive verb) and in 106 (relational sentence): in the

former case, the embedded relative may be most clearly illustrated by

another English version which, less idiomatically in English,

translates the Coorg more faithfully:

107. '(It is (the case) that) he will come, isn't it ?'

(iii) A possible formalisation of the foregoing would set up a

basic bipartite sentence pattern of NP + Pred., together with the

feature [+ relatj (relational) assigned to this configuration;

[+ relatj would automatically be assigned the further feature

[+ pS],indicating that a type of relative sentence embedding is

obligatorily to be erfcrmed:

108.

Pred.

+ relat.

= S

At this stage er_ obligatory 'segment tra_nsbrration' (as used in

Jacobs and Rosenbaum, 1968: 84) will introduce NP and S as

constituents into the deep structure:

/
109.
S

YP Pred
[+_relat2

NP
 157

This gives the basic configuraticn of the required deep

structure. The NP introduced by the segment transformation rule is

rewritten as a proform whose exact nature it is not possible to

state at this stage: however, it would seep, plausible to have

assigned to it automatically all the features of the subject NP of

the embedded sentence. During this process, further development may

be blocked at any stage where the features of the subject NP are not

suitable (for example, there would appear to be certain restrictions

on having pronouns as predicate nominals; thus:

p'
No" N 11

110. [a: manse aver

that man he,

is probably to be excluded as ungrammatical). The embedded S is

developed as an existential predicate sentence with existential

function; it would seem possible in principle to restrict its develop-

ment to this pattern by reference to the feature [4- relit] in the

matrix sentence, or else to leave it to a relational transformational

rule, which will have to erase the subject NP of the embedded sen-

tence, to block further development of the construction if the

embedded sentence is not an existential predicate type. Below is

given the proposed deep structure tree for sentence 43:

 1 58

NP Pred.
[+ relat.]
pro

ju_N

[ave.! NP

proform

hI

1?
..
C-
Pred.

+anim. AT
+loc.
+human -def.
I t

[a_*iqádika: re]
-def.
+anim.
+human
N
N tt tt

43. lave arigadika:re!

'He's a shopkeeper'

(For the feature composition of the embedded predicate, see the

section immediately following.)

(iv) A simpler deep structure configuration may be proposed for

the existential pattern. The principal differences here are between

the existential, the locative and the possessive functions, and the

examination of these that has been made suggests that these are all

linked. Following Lyons (1968: 390), it is proposed here that all

these be treated as types of 'locative' constructions. In these

terms, the existential function is filled by an indefinite locative

predication, and the locative and possessive functions are expressed

as(respectively) static and dynamic variants of a definite locative

predication. The Pred. node in this construction would be assigned

the feature [+ loc.], and then this would be further developed as

[+ def.], which would in turn be developed as [± dynamic]. Therefore,

the P- Marker for the purely existential sentence 18 would be as

follows:
 1 59

11 2. s

NP Pred.

+loc.
P_
-def.
-dyn.

At this point, a segment transformation rule will introduce a

Loc.P. constituent into the PP'Iarker, assigning it the features

[+ loc., - def.] that are specified for the Pred.:

113.

NP Fred.

Loc.P. -+loc.
N - +loc. -def.
-def. ]
[de:va] ^ I I

[undi]

1E. [de:va undi]

'God exists'.

Where the Loc.P. is [-def.], it is realised as zero. Where the

Pred is specified as [ +dyn.j, the copula will mark this aspect on the

base [a:gJ -.

Where [ +toc.] is developed as [ +def.], however, the Loc.P.

established by the segment transformation rule will be developed

further; the P -N arker for sentence 47 will be as follows:

 160

114a.

NP Fred.

Loc.P +loc.
L+loe. +def.
+def -dyn.
pro l

NP

N
tl n
Lave] [mane] -[li]
+Id. - +Affix
-anim +loc .

+loc.

N
N It 11 11

47. [ave maneli candi]

'He's in the house'.

The essential similarity, noted above, between this pattern

and that which expresses possessive function may be illustrated by

presenting the P- Farker for 40:

114b.

NP Pred.
-
Loc. 2.
Yodif.. C+loc . +def.
-dyn.
_
r

IiP

pro
n_ n

[ cenna] [duddi] [na:]- -[ki] [undi]

+Pro. [+Affix
+anim. +loc.
+hun.
+loc .

It It fl_

40. [na:ki cenna dudd(i) undi]

'I have some money'.

 161

The criterial difference between 114a and 114b is that in

114b the nominal (pronoun) under the node Loc.P. is specified as

[ +anim.], whereas in 114a it is [- anim.]; but this is actually a

simplification of the issue. The feature [ +hum.] also seems to be

relevant, inasmuch as it is possible to express in this way either

possessive or locational function, given a [ +anim, -hum.] noun such

as [na:yi] 'dog'; thus both the following occur:

fl_
115. [a: na:yili pulu undij

'There are worms in that dog' /'That dog has worms'

n n u n,
25. [a: na:yik(i) or(i) udda ba:l(i) undid

'That dog has a long tail'.

rr
At this stage it is not clear how far the fact that [a: na :yili]

in 115 is a sentential adjunct rather than a predicate expansion

affects this issue.

(v) The analysis suggested above for existential sentences,

whereby the expression of existential, locational and possessive

functions is linked to the notions 'locative', 'definiteness' and

'dyna*nicity', naturally suggests the possibility that a two -term

directional distinction within the dynamic category may also be

found within the same (i.e. copula) system. It is the case not only

that such a distinction is found, but also that it seems to provide

evidence of a close relationship between 'local' and 'grammatical'

functions (for these terms, see Lyons, 1968: 295 -304) involved. Thus

the (stative) locative [maneli] , 'in the house' is in opposition to

both the dynamic forms [maneki], 'to the house' (allative) and

[manefiji], 'from the house' (ablative). Using the location of the

house as a point of reference, the least marked verb forms collocating

with [maneki] and [maneñji] are [bar] -, 'come' and [po:] -, 'go',
 162

respectively. For the same sort of reason, it has been proposed

(Lyons, 1968: 397 -9) that 'come' and 'go' in English be treated as

directionally- marked forms of the dynamic copula. Further support for

such an analysis within Coorg is not hard to find, for [bar]- and

[po:]- are commonly used as auxiliary verbs just as are forms based

on [irÌ -, [ill]- and [a:g] -; therefore, an analysis which tries to

show a relationship between the copula and a,i7 _li ry verb systems

is able to point to their surface identity in Coorg. Examples of

the use of [bar] - and [po: 3- as copula forms are:

116. [annane po:ce:rigi

that go -if
manner

'If it turns out like that...'

n n
117. [ávárigi cedi ba:ti]

him -to anger came

'He got angry'.

Examples of their use as auxiliaries are:

118 [áve catti_ po:cij

he having went
-died

'He died'.

119. [áve geddandi ba:ti]

he winning came

'He was being successful'.

The relation between the local function of the allative marker

-[ki] and its grammatical function is exemplified by the possessive

As far as the
types of copula predication presented above (p.132 ).

ablative marker -[ñji] is concerned, it plays an instrumental role

with the
in certain contexts with animate and inanimate nominals;
 163

latter type, it is interesting to note that the basic

locative form

in -[i j is also a possible marker of instrumental

function. Thus:
it Il
120. [a: su:leñji na :da duddi pa:1(i) a:ci -j

that prostitute me -of money ruin became

-from

'I lost my money because of that prostitute:'

rl n n
1 21 . five kuññina ko:lili pojjati]

he child stick -with beat

'He beat the child with a stick'.

N11
ti MN II N 11

122. [ave 'nnina ko:linli pojjati]

he child stick -with beet

'He beat the child with a stick'.

(The last two sentences are fully synonymous, as far as could be

determined during collection of the data.)

(vi) Finally, it will be noticed that in the P- Harker 114b the

locative affix to the pronoun has the same form as the allative
n_
marker -[kij. This is dictated partly by the fact that the Loc.P.

nominal is [ +anim, +hum], since with this class of nominals the basic
n_
locative marker -[lij cannot occur. But this does not wholly account

for the fact that -[ki] occurs, for it has been noted above that it
n
seems best to regard -[ki] as the marker of one sort of dynamic

locative predication; if this is accepted, the problem arises as to

how to introduce the íbature [ +dyn.] into the P Harker 114b. Clearly

it cannot be specified for the Pred. node as this stands above, for

this would not have the desired effect (it would mean that a copula

formed on the base [a:gj- would occur in the surface structure, with

the sense of money coming into possession of the notional subject).

Possibly what one has here is further evidence to support the thesis
 1 64

in Bach (1968) that predicate nominals should be introduced into

sentences by way of relative clauses; for, if this approach is

adopted, it would be possible to account for -[kij on the Loc.P.

nominal by introducing that nominal into the underlying structure

in 114b via a relative clause with the feature [ +dyn.] specified in

its predicate. (It is worth noting that exactly this process may be

recognised in order to account for the facts of the adverbialisation

transforms (pp. 149 -152) which always show [a:gj- in surface structure.)

3.0 Voice

The category of voice is appealed to in discussion of the data

in this section: but since the morpho- syntactic phenomena that are to

be dealt with here hardly correspond to what is traditionally handled

under this term (i.e. 'active' and 'passive'), some explanation and

justification of this usage is in order.

It should first be noted that restricting the category of voice

to Indo- Buropean active and passive constructions (also the creek

'middle') hardly allows for revealing cross - linguistic statements to

be made. Thus, for example, it is pointed out in Lyons (1968: 378-

80) that one of the prime functions of the passive is to permit the

construction of 'agentless' sentences; in Coorg, on the other hand,

where there is no passive in the accepted sense, what is traditionally

called the 'causative' construction fills this role (see the

discussion below, pp. 172 -173). Hence, one may want to recognise in

this a typical function of what may in general terms be called

'voice', implemented in different constructions in different

languages. In much the same spirit, Hockett (1958) proposes a

characterisation of voice in terms of general grammatical theory as

 165

follows: 'Voice-distinctions apply to verbs, and have to do with the

relationship between the subject and the verb, the verb and its

object, or the verb and some other noun tied to it in an intimate

way' (195E: 236). In particular, Lyons (1968: 380-6) provides a

formalisation of certain intransitive, transitive and causative

constructions of English in the context of a discussion of voice;

and Palmer (1971) explicitly recognises 'active', 'passive',

'causative', 'reciprocal' and 'reciprocal- causative -adjutative'

within the category of voice in Tigrinya: 'clearly, if we do not

confine voice to the European patterns of active and passive they

(the five types just referred to) are simply different voices of the

Tigrinya verb' (1971: D5).

However, one danger in the use of the term voice in this section

must be specifically guarded against. It is traditionally a label

for a grammatical category, where 'grammatical' is generally under-

stood to refer to the syntax of the language in ouestion: in what

follows, however, certain intransitive, transitive and causative

constructions will be introduced as involving 'voice-related' verb

forme, without prejudicing the issue of whether the relationship in

_uestion is to be specified in the syntactic component, or in what

manner. This issue will receive special consideration later on

(pp. 178-92), where it will be shown that a number of voice-related

verbs in Coorg actually require a type of lexicalist analysis: the

formalisation of this is a separate issue again, which will be taken

up towards the end, and developed in the following three chapters.

As a starting point for the discussion below, the proposal in

Lyons (196e: 384) is taken up, whereby the relationship between certain

one -place, two- place and three -place constructions may be formalised by
 166

including S as a possible rewrite of Pred in a rule such as:

Verb

123. Pred Nom + Verb

3.1 'Intransitive', 'transitive' and 'causative'

The rule just cited formalises the distinction between

'basically transitive' nuclei (where Pred is expanded. as Nom + Verb)

on the one hand, and, on the other, 'embedded two -place nuclei'

(where ?red is rewritten as S, which is in turn expanded as a one -

place nucleus). In terms of this analysis, the grammar recognises

both one -place and tyro -place nuclei, of which all the first are

'intransitive' and of which some of the latter are 'basically

transitive' and the rest are 'derived transitive' nuclei. There is no

case in the data of a verb which would require the setting up of a

'basically causative' (i.e. a primitive three -place nucleus) pattern,

but there are a number of cases of what may be called 'derived

causatives' (where Pred is rewritten as S, which is then expanded as

a two -place nucleus).

3.1.1 The data

The following sentences illustrate the facts to be accounted for:

1 24 . [ payyu m e va 3

cow graze (intr)

'The cow grazes'

125. [attaka:re payyuna mera]

cattle -man cow grazes (tr)

'The cowherd grazes the cow'

 167

126. [paksiya maratinji pa:rici]

birds tree-from flew

'The birds flew from the tree'

,1 T,

127. [ponna:li nelli pa:tici]

woman paddy winnowed

'The woman winnowed the paddy'

1 28 . [kunnni pandi kari tindati ]

child pork curry ate

'The child ate the pork curry'

IT 11 11

1 29 . [awee pandi kari ku iki ti:ticij

mother pork curry child -to fed (by hand)

'The mother fed the child pork curry'

Also:

130. [rayte (attaka:rana s di ) payyuna

a me:çittati]
..

farmer, (cattle- man -by- cow caused -to -graze

means -of

'The farmer got the cow grazed (by the cowherd)'

n 11 n
131. [sika:ri paksiyala maratinji pa:ricittati]

hunter birds tree -from caused-to -fly

'The hunter made the birds fly from the tree'

11
MM 11 11

132. [ta:yi (avvana gundi) kunnina pandi kari timbiçittati]

grand- (mother-by- child pork curry caused-to -eat

mother means -of)

'The grandmother got (the mother to get) the child to eat the

pork curry'

(Parenthesised elements are optional.)

3.1 .2 Discussion

Sentences 124 -9 illustrate what is traditionally called an

 1 68

intransitive- transitive relationship. Sentences 124, 126 and 128

are, further, related to 130-2, respectively, as non-causative to

causative. The immediate concern of this discussion is to character-

ise the differences between 125 and 130, between 127 and 131 and between

129 and 132. However, it is first necessary to make explicit certain

features of the examples given.

(i) The principal difference between 124 and 125 (apart from

the extra nominal in 125) is the fact that the transitivised form of

the verb has -[p]- as the non-past tense marker where 113 shows

-[vii- (underlined). This is most conveniently analysed by setting up

the category 'bilabial place of articulation' as the marker of non-

past tense, and then recognising within this two sub -categories,

namely, 'lax' ( -[v] -) and 'tense' ( -[p] -) manner of articulation.

This is the analysis that lies behind the traditional approach to this

area Dravidian morphology and syntax; in these terms, the

transitivised forms all show 'tense' articulation of the non-past

affix. This outlines one of the most common forms of morphological

alternation between pairs of verbs in Coorg that are related to each

other as 'intransitive' to 'transitive' (for the most part; there are

some cases of 'transitive' verbs linked to what appear to be

'causatives' by similar processes, and these are discussed immediately

below). However, it is not the only type of alternation, as may be

understood from an examination of 126 and 127, where the 'lax' -

'tense' variable in the verb is not a marker of past or non -past

tense but is the final element of the verb base; and -[r]- and -[t]-

represent the lax and tense articulations of this, respectively. In

128 and 129 the alternation is different again, the opposition this

time being between the 'lax' -[nd]- and 'tense' -[t]- (accompanied in
 169

this instance by lengthening of the preceding vowel); the nasal in

the lax form represents the final element of the base, and the

following voiced stop represents the marker of past tense; while the

voiceless stop in the tense form represents the base -final element.

It will be shown in the next chapter that 'lax' and 'tense' are

useful phonological categories for treating a whole range of these

alternances, and that what is involved may be analysed as two

distinct but related types of morphological alternation, each

expressing the same function. In this section it is necessary to

consider the syntactic implications of this voice function. In this

respect, 124 and 125 are wholly regular, being straightforwardly

related as intransitive to transitive respectively. 126 and 127, on

the other hand, show a phenomenon that is very common among such

pairs of verbs; namely, a tendency towards specialisation of meaning

of the transitive counterpart. It should be emphasised here that the

distinction between the opposition -[v]- /-[p]- on the one hand and

-[r]- /-[t]- on the other has nothing to do with the semantics of the

verbs involved; there are numerous cases of semantically regular

transitives showing final -[t]- where their intransitive counter-

parts show -[r]- (e.g. [a:r] -, 'to dry (in tr)' /[a:t] -, 'to dry (tr)').

The semantic connection is not hard to supply between the pair

[pa:r] -, 'to fly, jump' and [pa:t] -, 'to winnow (= make chaff fly

off from grain)', and the formal features of the verbs involved

encourage the establishment of the connection. Informants would

typically volunteer these connections, where they felt perhaps that

some explanation was required to an outsider. The way in which the

of the
grammar formalises these intuitive connections is a part

argument of this chapter. Regarding the skewness of the meaning-

 170

opposition in the pair of verbs under discussion it should be

pointed out that Lyons (1968: 360) discussing English examples of

what is apparently a very similar set of verbs, such as lie /lay,,

fall/fell, etc., notes that the transitive verbs in these cases are

'pore specialised in their selectional possibilities than the

corresponding intransitives'. It is exactly in selectional

restrictions that the distinction between [pa :r]- and [pa:t]- is tc

be looked for (i.e. what sort of object nominal is made to 'fly' or

'jump'). (The fact that Lyons calls such verbs as /a, fell, etc.

'morphological causatives' is not a reason for supposing that he is

dealing with a different type of verb to what is being illustrated

here from Coorg - rather, his terminology is different, because his

analysis is in this respect different from what will be presented

here; see below, pp.184-188).

Finally, 128 and 129 are distinct from the foregoing in that 128

is itself a transitive construction, and it would seem odd to have a

description whereby a process called, say, 'transitivisation' could

be applied to such. Accordingly, it is decided here to adopt the

terminology of Kandiah (1967, 1968) wherein a similar set of verbs

in Ceylon Tamil is called the set of 'volitively related' verbs, the

'involitive' being the forms in 124, 126 and 128, and the 'volitive'

being those in 125, 127 and 129. (However, this does not imply that

Kandish's analysis of this type of verb relation is accepted; see

below, rp.178 -184) .

Only eight transitive verbs, out of a total of more than eighty

to the
intransitives, are found in the set of verbs that are open

volitive relation. Only in two cases ([bud] -, 'to let go' /[budIt] -,

'to make let go' and [pad j -, 'to suffer '; [nadit]
-, 'to make suffer')
 171

are the pairs related as transitive to causative; in two other cases,

there is simply a specialisation of meaning in the volitive fora

([mä:rJ -, 'to sell' /[ma:t] -, 'to change' and [biri_] -, 'to open (jacer-

fruit , dismantle (hcuse)' (with lax non -past and past affixes)/

[biri] -, 'to spread. (leaves, blanket)' (with tense non -past and past

affixes)); and in the remaining four cases, there is a fairly close

semantic correspondence but not (it will be argued below) such as

would justify calling the volitive form in each case the 'causative'

of the corresponding transitive ([ari] -, 'to find out' /[arip] -, 'to

inform, tell'; [tinn] -, 'to eat (not rice)' /[t5 :ti -, 'to feed (by

har_ä)'; [k?:n] -, 'to see, seem' /[k_a:t] -, 'to show'; [kada] -, 'to

cross /[kadat] -, 'to take across').

(ii) Sentences 130-2 all show verb forms with -[c]- the

'causative' marker (double underlined) followed by the element

-[itt] -. There are good reasons for keeping these elements

distinct; from a purely descriptive point of view the evidence is

as follows: -[c]- is followed by -[ i tt]-- in all past forms of caus-

ative verbs, and by -[ir]- ni -[id]- in all non-past forms. But

-[itt]- (past) and -[ir]- w -[id]- (non -past) are also the form of

the 'completive' suffix to verbal bases. Therefore it is decided

here to state that in Coorg morphotactics the causative suffix is

obligatorily followed by the completive suffix, since this will

avoid repeating the sane set of realisation rules for two distinct

suffixes. From a comparative point of view also it is more useful to

set up -[c]- as the causative suffix, rather than the forms -[cir] -ry

-[cid]- (non -past) and -[citt]- (past); on this point, it should be

noted that Emeneau (1967) sets up -ç- for the causative suffix in

Coorg.
 172

(iii) It remains to characterise the differences between 125,

127 and 129 on the one hand, and 130, 131 and 132, respectively, on

the other. There are certain superficial similarities - for example,

in each case concord operates between the subject nominal (the first

element in each sentence) and the personal ending of the verb.

Iioreover, [payyuj, the subject nominal of 124,becomes the object

nominal (showing the objective case marker -[na]) in both 125 and

130. The same statement cannot, of course, be made for [paksiya]

in 126, but only because of the selectional restrictions that operate

against having a nominal such as [paksiyala] (to quote its objective

case form in -[la]) as object of [pa :t] -; this does not affect the

fact that the basic sentence pattern relationship between 126, 127

and 131 is the same as that between 124, 125 and 130, respectively.

Thus, [nelli] in 127 is a further example of what has been noted

already for inanimate nominals - an 'objective' form with no overt

case marker. 129 and 132 show greater superficial differences;

[kuññina] is objective in 132 but [kuññikì] shows the allative case

affix -[ki] in 129. However, this is superficial, in the sense that

129 can most naturally be accounted for by specifying [ti:t]- in the

lexicon as taking 'food' nominals as direct object and [+anim.]

nominals as indirect object, and glossing it here as 'to feed to'

rather than as 'to feed'. r urthermore, some informants hesitate

between the choice represented by ['kuññina] and [kuññiki] in such

sentences as 129, with no determinable semantic differentiation; this

situation is presumably dependent upon the fact that [pandi kari], as

an inanimate nominal, may optionally have no overt objective

case

marker.

But there are basic differences between the volitive and causative
 173

patterns, and the greatest of these is that in 130, 131 and 132 the

notional subjects of the main action are identical with the

grammatical subjects of the sentences 124, 126 and 129, respectively.

(By main action is meant 'the grazing' in 124 and 130, 'the flying'

in 126 and 131 and 'the eating' in 12E and 132, rather than 'the

causing' of these actions.) By contrast, in 125, 127 and 129 the

grammatical and notional subjects are in coincidence. The simplest

test of this is the following: that, ontologically, it is left open

in 125 as to whether the cow actually chewed any grass, and in 129 as

to whether the child actually swallowed what was fed to it. The

corres-:ending question, as to whether the grain actually jumped or

flew in 127, is almost meaningless, since the divergence of meaning

is so great: but it should be noticed that this is only a special case

of the meaninj difference that is being pointed out here - thus, to

graze a cow does not 'mean' the same (necessarily) as to make it

graze and to feed a child pork curry is not necessarily to make it eat

it. In 130-2, however, what is being asserted is precisely that the

co;; grazed, that the birds flew from the tree, and that the child ate

the pork curry - together with the additional assertion that this

was caused by some ultimate agent (the farmer, the hunter, t ^V grand-

mother). The immediate agent (the subject of the corresponding

volitive construction) is tyrically not s'cified, being introduced

where necessary via some sort of embedding (the cowherd, in 130; the

mother, in 132).

3.1 .3 The analysis of causative verbs

For the sorts of reasons just discussed, sentences such as

130-2 have traditionally been related to 124, 127 and 129, respect-

ively, as 'causative' to 'non-causative', and there seems no reason to

 174

object to this analysis, or this terminology (cf. the remark in

Bally, 1932: 54 that 'logiquement, causer c'est faire qu'une chose

soit'). The same terms, for the corresponding morpho-syntactic

relationship, have also been adopted in transformational -generative

works such as Kandiah (1967), where arguments are presented for

taking a sentence such as 130 as the result of embedding the

structure underlying 124 in the Object, position in:

u
133. Lrayte Object ma:dicij

farmer O bject did

'The farmer did Object'.

Further, the intermediary (instrumental nominal) La;taka:ran_a

gundij of 130 is treated in the same analysis as the result of

another embedding into the matrix sentence before the embedding of

124; thus:

134. Lrayte attaka:rana gundi Object ma:dici],

where the underlined section is the (reduced) embedded sentence, and

n_
where Lgundi] is historically derived from the Perfective participial

form of the verb Lkoll] -, 'to take', but in much modified form. The

corresponding form in Ceylon Tamil (and in Malayalam; see, for

example, Asher, 1968: 95) is not a problem in statement as gundi]

is in Coorg, since it is not a diachronic peculiarity. Concerning the

Coorg form, it is worth pointing out that Cole (1867) only records it

as [gonduj, but that none of my informants had [o] as the first

vowel. In many cases, initial [g]- was not present, or else only

sporadically so, and the form was then identical with - and by some
n,
people identified with - the copula Lundij. (In some people's

speech, loss of [g]- was accompanied by another feature, namely the

use of the possessive case marker [daj- f+ -Lra]- instead of the

 1'(5

objective on the preceding nominal, thus: Lattaka:randa undij.)

Possibly this may be held to have influenced the quality of the vowel,

but the matter is extremely uncertain. [attaka:rana] is in the

objective case in the embedded sentence in 134, and the sense of

134 itself may be given literally as 'the farmer, having taken the

cowherd, did Obiect'.

This analysis is taken here to be substantially correct,

although Kandiah does not make it clear under what node the

instrumental sentence is to be embedded; Possibly a manner adverbial

node might be argued for. Further, it is not at all obvious that it

would be correct to take [ma:d] -, 'to do, make' as the causative verb

form for the matrix sentence in 133 and 134; it does not appear in

surface structure (in this construction), and it could be argued that

the choice of a particular lexical item for this purpose is artificial,

and that a better (because more abstract) formalisation would be to

embed the structure underlying 124 into a configuration which had

Lraytej as subject NP and whose Pred was assigned the feature

[ +caus.]. But, all such details aside, the method of analysing

causative sentences as embedded two -place or three -place nuclei, is

not in dispute. Embedding is the only way to formalise the fact that

the object NP of the causative sentence corresponds in each case to

the subject NP of the related involitive construction.

3.1.4 The jroblem of 'volitive' verbs

Now it is necessary to turn back to a consideration of the

volitive sentences 125, 127 and 129. To clear the way for a discussion

of these, it may be recalled here that it has been proposed to

recognise the following types of primitive sentence structure:

 176

135. Intransitive: one -place nucleus.

V.

Lpaktiya] L pa: r]-

birds fly

E paksiya pa:rici]

'The birds flew'

136. Transitive: two -place nucleus.

ìiP Singular) Pred (Past)

NP (Singular)

vtr

L na: yi] L_oYYJ-

he dog beat
N
Lave na:yina pojjatij

'He beat the dog',

and, in addition, the following derived sentence structures:

 177

137. Causative: embedded two -place nucleus.

caus.j

YP (Plural) Pred

Pro VP

intr
k ,f
L ave] Lpakgiya] Lpa:r]-

,_-Fcaus.j

he birds cause-to-fly (Past)

V.
w/ II II tt _

[ave pakgiyala pa:ricittati]

'He made the birds fly'.

138. Causative: embedded three -Place nucleus.

./
NP (Singular) Pred (Past)
[ +caus.]

NP (Singular) Fred

NP (Singular) VP
I.
I

Pro i% N N V,
tr

L t l: yi] L k- nni] [pandi] Licari] [ tinn]-

[ +caul.]

grandmother child pork curry cause -to -eat (Past)

It tt

Lta:yi kuññina pandi kari timbicittati]

'The grandmother got the child to eat the pork curry'.

 178

(These -I'iarkers are greatly simplified and given by way of

illustration only.)

It will be noticed that by this analysis there are two competing

methods of accounting for two -place nuclei (either as primitive or as

derived). The choice implies a borderline area between the two,

between 'basically transitive' verbs such as Lpoyy]- 'to beat' and

the causative two-place construction of 137. Any analysis of volitive

verbs will therefore have to come to a principled decision to treat

volitives either as derived forms (together with the causatives), or

as basically transitive, at least so far as the syntactic component

is concerned (i.e., it is left open for the moment whether, by this

analysis, volitive verbs are derived with respect to their involitive

counterparts in some other part of the grarnr r).

3.2 Two proposals regarding 'volitive' verbs

Two recent discussions are relevant to the issue just raised;

Lyons (1968: 359 -60, 383 -5) and Kandiah (1968); this last work presents

materially the same argument as is found in Kandiah (1967), but most

references are made here to the later version, since this is published.

However, neither of these analyses is accepted here. what follows is

a critique of their proposals as these apply to the situation in

Coorg, before an attempt at a third analysis, which will have implic-

ations for, and be further developed in, the following chapters.

3.2.1 Kandiah's proposal

The verbs that are the subject of discussion here, and in

Kandiah (196), form a small subset of the total stock of Dravidian

verbs and most of them have coLnate forms in more than one language

of the group: thus, many of Kandiah's volitively related pairs of verbs

 179

have closely corresponding forms and meanings in Coorg (cp. Coorg

Larij-, 'to find out' /Laripj-, 'to inform, tell' with Ceylon Tamil

arintaan, '(he) found out' /arivittaan '(he) informed (someone)',

where -vi- in the second form is the form of the volitive suffix

which this verb takes in Ceylon Tamil: even the form of the particular

volitive suffix in this verb would seem to be morphologically similar

to the final element in Coorg [aripj -, since Kandiah sets up

-vi-N -Pi- iv -ppi- as the variant forms in Ceylon Tamil). The once

productive process that has led to their creation as a distinct sub-

set is no longer operative. The issue has been complicated further in

the southern Dravidian languages, excluding Kanarese, by the fact that

the original morphological alternance marking the voice distinction

between involitive and volitive forms has at some stage been extended

analogically to include verbs that were once outside the subset.

Thus Coorg -Lp]- and Ceylon Tamil -vi- N -Di-Al -phi- in the

case just cited go back to an old causative suffix: see Kandiah

(1967: 124-6). On the original voice suffix -nk ( involitive)/ kk

(volitive) and the spread of the morphological relationship, see

Emeneau (1967). The use of 'analogically' at this point in the text

perhaps assumes too much in the diachronic situation, but may be

justified by the fact that all the types of morphological alternance

involved may be handled in terms of the 'lax' /'tense' opposition:

see above (p. 168) and the following chapter.

Typically, numerous types of base -final elements, the tense

affixes and an old causative marker all have to be taken into account

when discussing the morphology of these verbs in the languages of the

group.

(i) While the morphological aspect of the volitive relationship

 180

is wholly susceptible of statement,

Kandiah (1968) points out that

there are certain interesting differences in

syntactic behaviour
between the involitive and volitive forms of many verbs (in
Ceylon
Tamil: but his demonstration may be paralleled from Coorg also).

He shows that the involitive form of a given verb may not necessarily

take the same class of subject or object nominals as its volitive

counterpart; and this has been shown to be the case in Coorg also

(above, p. 170). He also shows that volitively related verbs may

take adjuncts - say, adverbial expansion - quite independently of each

other, and this too may be illustrated from Coorg; thus, in the

following examples:
n u n n
139. Lnelli nida:natil(i) onañgici]
.

paddy end -in dried (intr)

'The paddy eventually dried'.

n n n n n n
140. Lkinne nelli nida:natil(i) onakici]

boy paddy end -in dried (tr)

'The boy eventually dried the paddy',

the verbs tonarigi- 'to dry (intr)' and Lonak]- 'to dry (tr)' (showing

the original type of morphological marking of the voice distinction,

underlined) are independently modified by [nida:natili] (a locative

NP with adverbial function); that is, 140 does not assert that the

boy caused the paddy eventually to dry, but that after some period

of not drying the paddy he eventually did perform the action. (slowly

or Quickly). Thus the semantic evidence is against a transformational

relationship between 139 and 140, at least in terï.s of the thesis out

forward in Katz and Postal (1964) , Chomsky (1965) etc. Conc¢rning

the possibility that transformations may affect meaning- (Chozsky,

1 971 ; Partee, 1971; etc.) it should be noted that the range of meaning
 181

that may be so affected is apparently rather restricted (typically

involving neutralisation of semantic oppositions in certain syntactic

environments), and that the hypothesis 'has so far eluded both

demonstration and refutation' (Partee, 1971: 21). (Note that even

the sort of grammar proposed in Chomsky, 1971 can only handle

regular meaning- changes transformationally, and therefore cannot

account in that manner for the specialisation of meaning typical of

volitive verbs.)

A further problem which Kandiah (1968) points out for Ceylon

Tamil and which is also found in Coorg is that, with certain nominals

as subject or object, only the involitive, or only the volitive form

of a given verb may occur. But this, together with the independent

taking of adjuncts noted above, is only a special case of the

undisputed fact that the volitive verbs often show a distinct

semantic development when compared with their involitive counterparts.

Thus, besides the case of [pa:r] -, 'to fly, jump' /[pa:t] -, 'to

winnow', Coorg also has (a brief selection):

Involitive Volitive

[para] -, 'to crawl' [parat] -, 'to spread (grain)'

[olamb] -, 'to gargle' [olap] -, 'to wash (plates,

pots)'

[ka:n] -, 'to see, seem' [ka:t] -, 'to show'

etc.

From this, it is clear that any analysis which attempts to derive

a sentence containing, say, [ka:t] -, 'to show' from a sentence with

seem' will run into the difficulty of

having to
[ka:n] -, 'to see,

account for the fact that, where there is a [ +anim., +hum.] object

(in the sense of 'meet')

nominal, [ka:n] -, and not [ka:t] -, may occur
either [ka:n]- or [ka:t]-
and where the object nominal is [- anim.]
 182

may occur, but that in the former case no indirect object in the

allative case is possible, whereas in the latter case it is possible

and frequent. The important point here is not that the situation

for these two verbs is particularly difficult to state (it will be

seen that they are much closer in meaning, and therefore in syntactic

behaviour, than either of the other pairs listed above); it is rather

that the statement, once made, is not useful for such other pairs

as [pa:r]- /pa:t] -; [para]- /[parat] -; [olamb]Aolap] -; etc., and that

numerous statements will have to be made, some of great complexity,

to cover all the volitively related verb -pairs in the language.

(ii) In the face of such difficulties, Kandiah (1968) rightly

concludes, first, that it is impossible to specify the domain of

operation of a transformational rule, and its resultant string, such

that it can satisfactorily account for the derivation of all

sentences with volitive verb forms from corresponding sentences

with involitive forms; and, secondly, that it would be undesirable

to do so, since 'any attempt...to treat volitively related verbs as

a single base unit differentiated by a transformation would, even if

successful, fall short of descriptive adequacy for it would deny

deep status to the difference between them, and not allow to either

of them the prerogative it in fact enjoys of behaving differently

in the BASE from the other' (1968: 230 -231). To this it may be

added that, even if it were possible to specify all the necessary

information in the transformational component of the grammar, it

would be undesirable to do so because it would mean that an

exceedingly large number of transformational rules would have to be

set up, of which very many would be operative only with one verb in

the language; and, to the extent that there

were some general
 183

principals of operation underlying many of these

rules, the fact of

their existence could not easily be formalised by the grammar, since

each rule would be separate from each other.

(iii) There is, therefore, a fair amount of evidence to show

that sentences with volitively related verbs are not to be transform-

ationally related to each other. Kandiah draws the following

inference: 'if two verbs are to be related to each other in terms of

the deep structural relationship of volition, and if the base, while

making valid and necessary statements about them, can only differ-

entiate them, then it appears reasonable to assume that their relation-

ship has to be established BEFORr, they enter the base' (1968: 231).

This is basically correct, especially in the stress laid on an extra-

syntactic base relation between these verbs: but the unnecessary PTA

limiting assumption is made that the volitive relationship is a deep

syntactic one in some sense, and it is this which leads Kandiah to

set up a 'Pre Base Component' wherein the relationship can be

established. However, it is not clear from his work in what sense one

can talk of a 'deep structure' which is not identical with 'structure

generated by the syntactic base', where a syntactic base is postulated.

Of course, it is possible to set up a non-syntactic deep structure

by working in terms of a semantic, rather than a syntactic, base

(for example, the work of IicCawley, 1968, Lakoff, 1'71, and others)

but in this case there is no independent syntactic base as Kandiah

envisages; and his proposed PreBase Component is certainly not to be

interpreted as a semantic base. Thus, his Fre Base Component is of

uncertain status, being (in default of being anything else) a sort of

syntactic component to do the work which Kandiah has correctly argued

should not be done by the syntactic base itself. A further objection

 184

to his proposal is that the Pre -Base Component requires rules of

the fora (1968: 236):

141. w +X +Y Z

which has unfavourable implications for the meta -theory of phrase

structure rules, and, as a consequence, for the proposal that the

syntactic base of a natural language can be formalised as a type of

phrase structure grammar. This fact, taken on its own, is admittedly

not a counter-argument; the position adopted here is that any work

which is conducted within the framework of transformational -

generative grammar must substantiate such a claim as this in a

rigorous way however, and therefore to show that it is not a

necessary claim (in order to account for the facts) is to invalidate

it. Accordingly, kandiah's proposal must be held to be well -

motivated but finally ill-conceived; and the solution that is offered

here represents an attempt to show that a Pre -Base Component is not

necessary (or desirable) for a statement of the relation between

involitive and volitive verbs. But first it is necessary to consider

the discussion of 'morphological causatives' in Lyons (1968).

3.2.2 Lyons' proposal

It has been seen that Kandiah (1968) presents an argument for

excluding the volitive relation from the syntactic base proper, and

that the facts of Coorg likewise support this. The proposal in

Lyons (1968: 350 -71) however, is that such a relation has to be made

in the syntactic base (although it should be borne in mind that this

is not the same base that Kandiah,1968 assumes), and, moreover, that

it is to be formalised in the same way as the causative relation

(i.e. by sentence -embedding). Therefore his arguments will be

illustrated here from Coorg (although he presents them mainly in

 185

English, French and Turkish), and examined. The crucial point in

the discussion is that of 'syntactic relatedness', and it will be

suggested here that this is not a simple notion.

(i) Lyons first sets up what he calls an 'ergative' pattern

(1968: 352), where the same verbal form occurs in both one -place and

two -place nuclei, and where the grammatical subject nominal of the

one -place structure corresponds to the grammatical object nominal

of the corresponding two -place structure. In this pattern, the most

natural interpretation is that the object nominal is the notional

subject of the action of the verb in the two -place nucleus, the

grammatical subject nominal being taken as some sort of agent.

Thus, the following sentences:

142. Lkodi bi: jici]

flag waved

'The flag waved'

M
143. [ave kodi bi: jici

he flag waved

'He waved the flag',

illustrate the (very rare) pattern in Coorg of a transitive sentence

being derived from its corresponding intransitive by what Lyons

calls the 'ergative, or causative, transformation' (1968: 352).

(ii) The second pattern that he sets up he calls 'causative',

and distinguishes two sub -types within it, namely, the lexicalised

and the non- lexicalised types (Lyons, 1968: 352-3, 369). The latter

need not detain us here, since it has already been considered

(examples 130 -2); the former type may be exemplified in Coorg by:
1! 11 tt

144. Lmudike cattati]

old -man died

'The old man died'

 186

145. Lba:leka:re mudikana kondatij

youth -male old -man killed

'The youth killed the old man'.

Of such cases, Lyons says (1968: 352) that they are 'pairs of

different verbs between which the same syntactic and semantic

relationship holds in corresponding transitive and intransitive

sentences' (i.e. as has been illustrated in 142 and 143 above).

Therefore 145 is a derived two -place structure (the result of

embedding the underlying structure of 144 into the Pred node of a

deep structure matrix with Lba:leka:re] as subject nominal.

The analysis thus far requires setting up the following trees:

146. Ergative:

NP Singular) Pred (Past)

+ caas.]

Pred

UP

pro N V
erg
N
N II _ i

Laves l kodi] ]-

he flag wave
V
bi: jici]
tl

Lave kodi

'He waved the flag'.

 187

147. Lexicalised Causative:

FP (Singular) Fred (Past)

LTcaus.]

NP (Singular ) Pred

VP

V.
ntr
I I
II

Lba:leka:re] Lmudike] Lca:i]-

youth old man die

I1 _
Lba:leka:re mudikana kondatij

'The youth killed the old man'.

(The non -lexicalised causative pattern has already been illustrated

above; 137-8, p. 177).

(iii) Lyons then turns to 'morphological causatives' and says

that these 'fall between the two extremes of ' lexicalisation', on

the one hand..., and the use of the 'same' verb... in both transitive

and intransitive sentences, on the other' (1968: 360). He seems

here to be talking in terms of morphology, for he goes on to note

that ergative verbs in transitive sentences 'may be described as

being derived from the corresponding intransitive verbs...by means

of a morphological process of 'zero modification' ' (1968: 360).

He proposes to formalise the syntactic relation between such verbs

as have here been termed 'involitive' and 'volitive' by the same

method as in the other types of causative; namely, by deriving

sentences with the volitive verb from the corresponding sentence with
 188

the involitive verb. His reason for doing so is that the verbs thus

formed are 'syntactically related as 'causatives' to corresponding

intransitive verbs and (from a historical point of view) derived from

them by what were once more or less productive morphological

processes' (1968: 359 -360). It has been shown above, however, that

they are not adequately accounted for in Coorg as the "causatives'

of corresponding intransitives', and it is not hard to show that

the same state of affairs may be argued to exist in English; thus,

it is possible to point to a distinction in meaning between such

pairs as 148 -9 and 150-1:

148. John laid Bill down ('morphological causative')

149. John made Bill lie down ('true causative')

150. John felled something ('morphological causative')

151. John made something fall ('true causative')

in order to make a case for a distinct deep structure in each case

(e.g. 'to lay' does not mean 'to make to lie', but something more

like 'to make, by one's own agency, to lie'; and 'something' in 151

is commutable with a far wider set of nouns in English than is

'something' in 150 - setting aside metaphorical usage). If it can

also be shown - and it would seem possible, from the examples above -

that the distinction between the 'true' and the 'morphological'

causative in each case is not generalisable in the language, then a

further case has been made out for saying that the relation between

'morphological causatives' and their corresponding intransitives is

not one that can be handled satisfactorily by the syntactic base at

all; compare what was noted of the volitive relation above, (pp. 180-

183) .
 189

3.2.3 Syntactic relatedness

The discussion above notwithstanding, however, there clearly is

some sort of syntactic relation between sentences containing such

verbs, and this has to be explained. In this connection, it will be

recalled that Kandiah (1968), while presenting arguments expressly

against formalising the volitive relation in the syntactic base, at

the same time assumed it to be a 'deep structural relation'; and

(1968: 220) he quotes Robins in support as saying that inflectional

formations (such as mark the volitive relation) enter into and mark

syntactic constructions.

The suggestion is made here that at least two 'grades' of

syntactic relatedness have to be recognised, a 'strong' and a 'weak',

(where the strong grade implies the weak, but not vice versa). Then

a 'weak syntactic relationship' will be such as holds between all

intransitive structures of a language on the one hand, and all

transitive structures on the other; the 'intransitive -to- transitive'

relationship. Then the inflectional formations that mark volitively

related verbs in Coorg are the realisations of the morphological

expression of this weak syntactic relationship. b 'strong syntactic

relationship', by contrast, is the sort of relationship that has been

treated in transformational -generative grammar in terms of the

notion 'derivation'; that is, one sentence is derived from the

underlying structure of another when there is a strong syntactic

relationship between them. In these terms, there is a strong

syntactic relationship between the non- causative /causative sets

124/130, 126/131 and 129/132 above, and between the ergative examples

142/143 above. But there is only a weak syntactic relationship

between 124/125, 126/127 and 128/129, since in these cases the

syntactic base can only relate the verbs involved as intransitive to

 190

transitive (or one type of transitive to another type, in the case

of 128129), and beyond this can only differentiate them, as Kandiah

(1968) points out.

For these reasons, it is proposed here that volitive verbs be

introduced into the base as basically transitive verbs in primitive

two -place nuclei; since only in this way can the grammar account for

the fact that they behave as independent elements in the base. It

should be noted that the same argument holds, mutatis mutandis, for

what Lyons (1968) calls 'lexicalised causative' verbs, since the

difference betwo.en these and his 'morphological causatives' is purely

a feature of the lexicon of a given language, and not a syntactic

matter at all. Thus, in Coorg, Lka:t] -, 'to show' is the volitive

form of [ka:n] -, 'to see' (or its 'morphological causative', in

Lyons' terms), but also the 'lexicalised causative' of Lno:t] -, to

look at, see'; this case shows particularly clearly that the

distinction between the two types of relation here is wholly a matter

of relatively low-level rules. Thus the verbs which Lyons calls

'lexicalised causatives' must also be introduced into the base in

primitive two -place nuclei, and it must be left to the semantic

component to relate sentences which contain these verbs with those

that contain their intransitive and semantically related counterparts.

Indeed, the only reason for setting up such pairs as die /kill, eat/

feed, etc. is the semantic relationship that holds between them.

It is relevant at this point to recall that the crucial weakness

of Kandiah's position is the failure to distinguish between

'(in)volitive verbs' and '(in)volitive sentences' (1968: 226 -7):

cf. particularly his statement that 'both the relationship and the

differences between any pair of volitively related verbs are deep

 191

structural matters' (1968: 230), where 'deep structure' is

apparently taken as a level prior to lexical insertion. Concerning

the proposals in Lyons (1968), it would seem that the status of the

derived two -place and three -place structures is intermediate between

that of 'deep structure' in the sense of Chomsky (1965) and the

semantic configurations of generative semantics in the sense of

AicCawley (1968). For this reason it is not always clear whether

they are intended as formalisations of 'syntactic' or of 'semantic'

structure.

Finally, it may be pointed out here that the pair of verbs

most often quoted in support of a syntactic analysis for 'lexical

causatives' - die/kill (Coorg [ca:1]-/Lkoll] -) - is in many ways an

unhappy choice for exemplifying the group as a whole. This is

because it is extremely difficult, if not impossible, to find

contexts wherein the 'true' causative make to die could possibly be

in contrast with the lexicalised form kill. However, contrasts can

be found in many other cases, as will be evident from the following

English sentences with ' lexicalised causatives' (where in each case

the last part of the statement is impossible if the 'true' causative

form of the verb is substituted for the lexicalised form in the first

part):

152. I'm telling him the truth, but he won't listen.

153. I'm showing him the book, but he won't look at it.

154. I'm feeding the child, but it won't eat.

In view of the above examples, it would seem that the synonymy

of make to die and kill ( Coorg Lca:vicir]- and [koll] -) is an

ontological rather than a grammatical fact, and that it should

therefore not be marked by the syntactic base, but by the semantic

 192

component instead. If this is accepted, then any analysis which

posits a strong syntactic relationship between a sentence with die

and a corresponding sentence with kill (i.e. derives the latter from

the same structure as underlies the former) is open to the charge of

mixing levels.

3.3 Conclusion

3.3.1 The role of syntactic deep structure

Luch of the foregoing discussion of volitively related verbs

in Coorg might be open to interpretation as evidence against the

notion of a distinct level of 'deep syntax'. In particular,the fact

that Kandiah (1968) is able to show that a grammar incorporating the

syntactic base component proposed in Chomsky (1965: 120 -3) is

incapable of stating the volitive relationship might be taken as an

argument in favour of a generative semantics approach. It is true

that a grammar of the sort described in IIcCawley (1968), Lakoff

(1v71), and others, which has a semantic instead of a syntactic base

(working initially in terms of semantic deep structures, and then.

allowing for progressive lexical substitution of feature -specified

nodes within these deep structures as the derivation proceeds) will

not be open to the charge of failing in principle to characterise

the difference between such pairs as English lay /make to lie,

feed /make to eat, etc., and the corresponding Coorg forms. In such

a model of grammar, if a distinction in meaning between the members

of such pairs is recognised, it will be formalised in terms of

distinct feature -specified deep structures of the semantic base.

However, such a grammar is open to the complementary charge of

failing to account for the morphological similarities between the

members of such verb -pairs. This is because it will characterise the

 193

relationship between [pa:r] -, 'fly, jump'/[ pa ;tj-, 'winnow (grain)';

Lka:n] -, 'see, meet' /[ka:t] -, 'show'; [no:t] -, 'look at, see'/

[ka:t] -, 'show'; [cad] -, 'die' /[troll] -, 'kill', in essentially the

same manner; namely, in terms of (more or less) related semantic

configurations. It is true, as has been argued above, that one

difference between such pairs (lexicalised versus non-lexicalised) is

irrelevant from the point of view of syntactic relations (see p. 190).

But to talk simply in terns of 'syntax' and 'semantics' is not

sufficient, as far as the volitive verbs are concerned; for these

display morphological regularities as well (indeed, their morpho-

logical relationships are more regular than either their syntactic

or semantic relationships), which a generative semantics approach

treats as fortuitous matters of lexical realisation. In this respect,

such a description is unable to account for part of the explicit

knowledge that Coorg speakers have concerning their language (p. 169).

To this extent, therefore, it would seem a necessary (but not a

sufficient) requirement that a grammar of Coorg have a syntactic,

rather than a semantic, base.

3.3.2 The role of the lexicon

If the volitive relationship cannot be handled either by a

semantic base or by the sort of syntactic base described in Kandiah

(1968), then it clearly has to be formulated elsewhere, as is

correctly argued by Kandiah (1968: 231); see p. 183 above. The

proposal that is made here is that the appropriate place is the

lexicon, where this is understood to be outside the syntactic base

(as opposed to the organisation of the lexicon and the categorical

subcomponent in Chomsky, 1965: 141 -2). However, there are two

competing aims which the lexicon has to achieve in this case; first,
 194

it has to list the individual members of volitively related verb

pairs as independent lexical items, each with its own set of syntactic

and semantic features such that lexical insertion rules of the type

described in Chomsky (1965: 122) may operate; secondly, however, it

must make the morphological relationship between members of such verb

pairs explicit. For this reason, it must contain, in addition to

lists of entries as ordered pairs of phonological and syntactic -

semantic features, a generative subcomponent which will derive the

members of each volitively related verb pair from a common underlying

structure.

It is because it must contain a generative subcomponent (see the

following chapters for an attempt to formalise this, and Chapter 4

pp. 209-220 for further arguments for it) that this lexicon must be

located outside the syntactic base; such a model operates in terms of

the two independent and co- ordinate primitive notions 'sentence' and

'lexeme' (the latter specifying the domain of the categorial rules

of the generative base component of the lexicon).

Finally, it may be noted here that such a model of grammar is in

principle able first, to account for the morphological relatedness

but syntactic independence of volitively related verbs; secondly, to

handle all lexemic relationships (by which is meant morphological

relationships between lexemes as generated under the symbol 'lexeme',

whether verbal, nominal, or between these categories) which cannot

be accounted for in the syntactic base because they are only partly

regular; and thirdly, to distinguish between those morphological

processes that are productive (to be handled in the syntactic base)

and those that are non-productive (to be handled in the base component

of the lexicon). In respect of this last point, it should be noted

 195

that the model prol;osed here (which is a type of 'word- and -paradigm'

in the sense of Robins, 1959) is somewhat different from that

described in hatthews (1965a, 1965b, and 1967) .

 1 96

CHAPTER FOUR

The Structure of Lexical Items: Verbs

 197

1 .0 Introduction

In this chapter, the Coorg verbal system is examined, particularly

with respect to its verb classes (1.1) and the nature of the verbal

'lexeme' (1.2); the implications of the discussion of volitive verbs in

the preceding chapter are taken up, and considered in relation to

lexemic relations generally, and the form that lexical entries must

take (2.0 - 2.2); the generative component of the lexicon is then

described (3.0 - 3.9); and finally a list of the proposed rules is

provided (4.1 - 4.4) .

The data that is discussed is presented fully in Appendices

I -III (pp. 431 -58), where it is organised in terms of the verb classes

and canonical forms described below. Each item is numbered within its

class for ease of reference, and is generally cited in the text

together with its reference number in DED, DEDS, DBIA or DEN, if it

has one.

The canonical forms are stated in terms of the cover symbols C

(standing for all consonants, including intervocalic [v] and [y], but

excluding those covered by the symbols S and N), S (standing for Lv]

and Lys immediately before a consonant), N (for a nasal consonant

homorganic in place of articulation with a following stop), and V

(standing for all vowels). V is a long vowel; CC is a geminate

consonant cluster. The marginal contrast between nasal and oral vowels

is not marked at this level of representation, of course. The table

in Figure 1 below shows all the canonical forms that are represented

in the inventory of Appendix I; it should be noted that no dissyllabic

type shows a long vowel, and that short -vowelled monosyllabic types

show a greater variety of shapes than the long -vowelled types:

 198

I :onosyllabic Dissyllabic

Open (C)V, (C)V (C)VCV

(C)VC

(c)voC , (c)7C (C)vcvC

Closed (C)VSC

(c)vNc
(c)vuC (C)VCVNC
(c)vsNc }

Figure 1: The canonical forms of verbs in Appendix I

1 .1 Verb classes

Verb classes in Coorg are defined with reference to the forms

of the non-past and past affixes that may occur with different

verbs. Five such classes are recognised in the inventory of

Apendix I. The basis for them is examined in some detail here.

1.1.1 The non -past affix

The forms of the non -past affix may be characterised as either

'lax' or 'tense' . The lax forms are -Lvj- Ar -I:vvj - uvj -. The

first follows any vowel which is not a short vowel in the initial

syllable of the verb; i.e. it follows a long vowel in the initial

syllable, or a short vowel in a medial syllable. It also follows

certain (C)VC- forms (of Class 4 only, Appendix I), where the preceding

consonant is Llj except before the non-past and past affixes, where it

is deleted. Contrast the non -past form of the Class 1 item 141:

L a: luv j - rule (fron_

341

with those of Class 4 items 17 -19:

u

723 [e:vj- get up (from i e:i] -)

4402 Lba:vj- (child) lives, rule (from Lba:lj -)

Lbu:vj- fall (from Lbu:lj -)

4457
 199

follows a short vowel in the initial syllable of the only verb

of (C)V structure taking the lax non -past affix, the Class 2 item 1:

3143 [now]- pain (intr)

In all other cases the form of the affix is --[uv]- (i.e. when a

consonant precedes).

The tense forms of the non -past affix are [p] -e4 -[pp]- N -Lb] -.

The first two have a similar distribution to their lax counterparts

-[v]- and -[vv] -. Thus, -[pp]- is found after verb forms of the

structure (C)V -, as in the Class 5 items 1 -2:

781 [opj]- be suitable, consent

2904 [ topp]- salute,

while Lp]- is found after a long vowel in the initial syllable or

after a short vowel in a medial syllable. -Lpp]- also occurs in

certain (C)VC- and (C)VCC- verbs where the final -C or -CC is deleted

before the affix, and -[p]- also occurs after certain (C)VC- verbs

where the final -C is deleted similarly; the items affected are the

Class 3 items 2 -4:

407 [ipp]- be (in place) (from [ir] -)

2526 [ tapp] - give (to 1st /2nd person) (from [tar]-)

4211 [bapp]- come, (from [bar] -)

the Class 3 item 7 and the Class 5 items 12 -13:

3043 [nipp]- stand (from [nill] -)

3708 [popp]- fight (from [por] -)

[kapp]- steal, (from [ kal] -)

1 156

and the Class 5 items 14-19:

2354 [co:p]- become tired (from [co:l] -)

2936 [co :p]- be defeated (from co :l] -)

427 [i:p]- drag (from i:l] -)

904 [o :p]- have intercourse with (from [o :1] -)

 200

1677 [ke:p]- hear, ask (from [ke:l] -)

3588 [pu:p]- bury (fron: [pu:l] -)

In addition, -[p]- occurs after certain Class 5 verbs, all of which have

(C)VC- structure, and where the final -C is [y]; these are the Class
5

items 5 -11:

39 Laypj- send

1047 [kayp]- be bitter

1136 [kayp]- remove, undo (knot)

[cúympj- (fly, ant) bites

2894 [túymp]- (snake) hisses

4057 [muyp]- increase

4297 [bayp]- (stomach) hungers

3765 [PaYP]- id.

Note that where the verb has a nasal vowel, a homorganic nasal segment

precedes the affix; this is not treated here as part of the affix.

Finally, just three items show non-past forms with the affix

-[b]- preceded by the homorganic nasal. The unaffixed forms of these

verbs are of (C)VCC- or (C)VC- structure, where final -CC is [nn] or

[nn], and final -C is [n]. In each case, therefore, the nasal segment

[ m] preceding the affix -[b]- is explicable in terms of the final

consonants of the corresponding unaffixed verb forms. But the affix

-[b]- itself is to some extent problematic, in that it may be class-

ified as 'tense' (since it involves complete oral closure, like the

other tense forms) or as 'lax' (since it is voiced like the other lax

forms). The problem lies really in the definition of the terms

'tense' and 'lax', however, and the overall analysis is not substant-

ially affected by the choice of one classification over the other.

Arbitrarily, then, -'Lb]- is taken as 'tense'; the verbs concerned are

thus placed in Class 3 (rather than Class 2), being the items 5, 6 and
 201

8 within that class:

516 [umb]- eat (rice)

2670 [timb]- eat (not rice)

1203 [ka:mb]- see, seem

1.1.2 The past affix

The past affix is either vocalic (Class 1 verbs) or consonantal

(Classes 2 -5); the consonantal affix may, like the non-past affix,

be characterised as either lax or tense in form.

The vocalic affix is [i]. It does not occur in certain forms

of Class 1 verbs even though these forms are definitely to be called

'past': for example the verb [ma:d] -, 'do, make', has the forms
n m,
(I) did
n 11,
the having done (something)
n u
[ma:dine kelsa] the work done

without [i]. However, the occurrence of -[i]- or -[in]- is not

determined with reference to the form of the verb but rather to what

follows the suffix, and hence does not concern us here.

The consonantal past affix has the following lax forms: -FL nd]- /v

-[ñj]- ^+ -[nd] -; in each case the first element is a nasal, which is

homorganic with the following voiced stop that constitutes the second

element. The tense forms are of three series, corresponding to each

of the places of articulation found in the lax forms: -[t]-4 -[ tt]- 0v

add] -; -[ci-N -[cc] -^

If the dental place of articulation is taken as unmarked, it is

possible to account naturally for the palatal (lax or tense) forms as

occurring after a final -V which is [i], or final -C which is [y]: see

the Class 2 items 28-57 and the Class 5 items 34-62 (with final [i]),

and the Class 2 items 4-8 and 12 -17, the Class 4 items 10 -14, and the
 202

Class 5 items 5 -11 (with final -[y(y)]). It is also possible to

account for retroflex articulation of the affix (lax or tense) as

occurring after all final retroflex consonants except for certain

cases of -[1] in Classes 4 and 5; these instances can be regularised

by deriving the -[1] in these items from a peculiar phonological

element which does not constitute the necessary environment for the

retroflex form of the past affix, but is nevertheless realised as a

phonetically retroflex lateral. The items involved are the Class 4

items 17 -19:

723 [ edd] - get up

4402 [badd]- (child) lives, rule

4457 [budd]- fall

and the Class 5 items 16, 17, and 19:

427 [i:t]- drag

904 [o :t]- have intercourse with

3588 [pu:t]- bury

In addition, the tense forms of the affix are voiced when they follow

final -[y] or -[1] of (C)VC- verbs and final -[yy] or [ii] of

(C)VCC- verbs; in all these cases, the vowel of the past form of the

verb is short and is followed by geminate consonants. The items

concerned are the Class 4 items 10 -19 (note that this includes the

items 17 -19 cited above). This voicing is to be treated as a feature

of the class, predictable with reference to the preceding consonant.

The geminate voiceless forms of the affix occur after short

vowels in the first syllable of the verb. -[tt]- is found in two

instances with verbs of the form (C)V; they are the Class 5 items 1

and 2:

781 Lotti- be suitable, consent

2904 [ to ttl - salute

 203

It is also found in three cases where verbs of (C)VC -form show final

-[r] (deleted before the affix); these are the Class 4 items 8 and 9,

and the Class 5 item 12:

2833 [tett]- pay (penalty

3622 [pett]- bear (child)

3708 [peat]- fight

-[cc] occurs after two homophonous verbs of the form CVC- where the

final -C is [y] (deleted before the affix); these are the irregular

verbs

4540 [becc - cook (rice)

4565 [becc]- keep, appoint

-[ tt]- occurs after (C)VC- verbs where the final -C is [d], which is

deleted before the affix; these are the Class 4 items 1 -7. The table

in Figure 2 summarises the foregoing discussion.

Non-past Pas t

Class 1 -[ uv]- -Li]-

Class 2 -[uv] -0 4v]- -L vv -l.nd]- « ¡ nj]- "

Class 3 -l.P]- N -[PP]- a
-[bJ-
-[nji Lnd]-
Class 4 -[uv]- « -[v]- -[ tt]- »

. ..
Class 5 -[P]-^' -LPP]- -[tt]- -c]-
M +

-L cci- w w

Figure 2: The forms of the non-past and past affixes for

the five verb classes
 204

1.2 The verbal lexeme

We have thus far talked in terms of non -past and past affixes

being added to 'verbs' or 'verb forms' of varying canonical shapes.

We now refine the terminology for these basic forms of verbs. The

verbal lexeme is defined as the minimal element which is required to

account for that part of the whole verb complex, through its paradigm,

which cannot be analysed in terms of syntactic, closed-system

elements such as non-past, past, modal and person affixes, and which

is therefore to be inserted in the pre terminal string of the syntactic

component (in the sense of Chomsky, 1965; 84) independently of these

elements. The lexeme consists of a base plus a classifier;

specification of the base together with a specification of the forms

of the non -past and past affixes appropriate to a given verb (supplied

by the classifier) is sufficient to predict any form of a regular

verb. The base plus non -past or past affix constitutes the non -past

stem or past stem, respectively.

1.2.1 The basic lexemes

In the inventory of verbs in Appendix I there are four main

columns: in the first is given the inventory number of the items within

each class; in the next column is the DED, DEDS, DBIA or DEN reference

number (where this exists); the third column contains the base for each

verb, followed by the non-past and past stems (in that order); and

finally the fourth column provides a brief gloss. For the few items

that are not recorded in DED, DEDS, or DBIA, a tentative reference

number is supplied in column two in parentheses. Within each base

shape subgroup the entries are ordered with respect to their final

elements; all further ordering within the groups is by initial elements.

In both cases the order followed is that of Dravidian orthographic

 205

n n n rr

tradition; however, the extra vowels [ i] , [ i : , e[ e]] , [ : follow

]

[i], [i:], [e], [e:], respectively:

rr rr rr rr
N
[a, a:, i, i:, i, i :, u, u :, e, e:, e, e:, o, o :, k, g, n, c, j, n,

t, d, n, t, d, n, p, b, m, y, r, 1, V, s, s, 1].

The nasal -oral contrast in the case of a few vowels has been ignored

in the ordering of the items concerned.

Appendix I contains most of the lexemes that will be discussed

here; it was noted in the preceding chapter however, that there is a

regular morphological alternation, expressing the volitive relation-

ship between certain verbal lexemes, and by virtue of this relation-

ship it is possible to derive one lexeme from another. These

derived lexemes are not included in Appendix I, but are set out in

Appendix II; those lexemes in Appendix I which have a volitively related

lexeme in Appendix II are marked with an asterisk; those that have two

such forms are marked with two asterisks.

1.2.2 Derived lexemes

e now examine in more detail the morphological relationship just

noted and whose syntactic function was dealt with in the preceding

chapter. We have in fact to recognise two distinct but related

processes at the morphological level, which we shall call for now P1

and P2. P1 operates on certain lexemes of Classes 1, 2, 3, and 4 to

form volitively related lexemes of Class 1; P2 operates on certain

lexemes of Class 2, and on one of Class 4 to form volitively related

lexemes of Class 5. Note that some Class 2 items have a derived

lexeme in both Class 1 (by P1) and Class 5 (by P2). The field of

operation of P1 and P2 is summarised in Figure 3.

As has been noted already (Chapter 3, pp. 168 -171), it is usual

in Dravidian studies to characterise the sort of lexeme relationship

 206

Class 5

P2

Class 2 }
_'1

Class 3

r,

P1

Class 4
P1

Figure 3: The field of operation of P1 and P2

expressing the volitive relationship

involved here in terms of 'transitive' lexemes being derived from

corresponding 'intransitive' lexemes. However, 'transitive' and

'intransitive' are only terms of convenience (see Asher, 1966: 26)

and do not fit the syntactic and semantic facts in all cases. Part

of the reason for this is that P1 and P2 are no longer productive

processes in Coorg (and related languages). In what follows, there-

fore, the lexeme relationships mediated by P1 and P2 will continue to

be discussed in terms of the notion 'volitivity' (from Kandiah, 1968).

A brief survey of the data in Appendix II shows that Class 1

items 1-6, Class 2 items 30 -74, Class 3 items 4 and 13-19, and Class 4

items 2 -7 all exhibit a final increment, -(V)C in the lexemes that

are derived by Pl. The remainder of the lexemes derived by this process

show mutation of the lexeme -final consonants, as set out in Figure 4:

 207

Base -final consonants

Lax Tense

(Non -derived (Derived base)

base)

-kk

-rig -k
Class 1

Classes 1, 2

Class 3

Class 2

Classes 1, 2

Class 1

Figure 4: The operation of P1 at the phonological level

Three exceptions are found, in the derived lexemes related to the

Class 3 item 6 and the Class 4 items 17 and 19: the first two show

vowel length alternation, and the last shows an unexpected lexeme-

final consonant.

The operation of P2 is restricted to mutation of the past and

non-past affixes of Class 2 lexemes, such that they undergo class

change, to Class 5. Exceptions to this are the derived lexemes of the

Class 2 items 35 and 63, and the Class 4 item 17; the lexeme derived

from the Class 2 item 63 shows an unexpected lexeme -final vowel [i],

while the other two show exceptional mutations of lexeme structure.

All these six exceptions are excluded from treatment in what

follows. The phonological mutation effect of both P1 and P2 is

 208

conveniently handled in terms of tense and lax articulation of conson-

ants; the voice relationship that they express is one of the principal

concerns of this chapter.

1.2.3 Irregular and defective lexemes

The inventory of verbal lexemes is completed in Appendix III by

a list of those lexemes that are irregular in terms of the classific-

ation scheme presented above, or that are defective in the speech of

my informants. Six of them, marked with an asterisk, have derived,

voice -related lexemes, which are also given in Appendix III; of these,

five are derived in a perfectly regular way by P1, to form lexemes of

Class 1, and one is derived by P2, again in a regular way. However,

in this last case, the resulting lexeme is still irregular in respect

of its lexeme -final element, [y], which does not occur before the non-

past and past affixes; the irregularity is to be assigned to the

lexeme, and not to the operation of P2 (which affects only the forms

of the extra- lexemic non-past and past affixes that occur with the

lexeme).

Finally, the following three items:

333 La:valic]- yawn

1192 [ka:pa:d]- protect

DBIA 145 Ljams]- utter (incantation)

this
all show unusual lexeme composition, and are not dealt with in

chapter; however, the first two of these will be returned to in

Chapter 6, where it will be shown that their structure is explicable

in terms of the system as developed at that point.

 209

2.0 Preliminaries to the analysis of lexemic structure

We turn now to the matter of establishing a framework to generate

the greater part of the data that has been discussed in the preceding

section. The purpose is twofold: first, to specify the verbal

lexemes in the lexicon in the most economical and revealing manner;

and secondly, to provide for the development of the underlying

representation of the lexeme from the point where it is transposed

from the lexicon into a given syntactic string to the point where it

signals a set of instructions to the articulatory organs.

It is first necessary to state certain general assumptions

regarding the analysis that will be proposed, and, as far as possible,

to justify these.

2.1 Lexemic relations

It is assumed here, with Chomsky (1965) and others, that a lexical

entry must contain information that is necessary for:

(i) specification of its phonetic composition,

(ii) a. appropriate placement of the lexical item in a pre -

terminal string generated by the syntactic component,

b. specification of its behaviour with respect to subsequent

syntactic rules, and

c. interpretation of the lexical item by the rules of the

semantic component.

There are, however, grounds for suspecting that (i) above may be

inadequate as it stands, in the sense that certain important sub -

regularities in relationships between lexemes in Coorg can only be

stated if the internal morphological composition of lexemes is spec-

ified in the lexicon. Consider, for example, the following sets of

related items:
 210

2310 Lse:d]- drink (deep draught); Class 1 verb

fl

[se:di] tired, pulling feeling in legs; noun

2333 [cukk]- grow stout; Class 1 verb

[ cukki] wrinkles

2655 [tir]- turn round (intr); Class 1 verb

[tirik]- id., twist (tr); Class 1 verb

[tiri] wick; noun

3537 [puiig]- rot, fester; Class 1 verb

[puiigi] mould, mildew; noun

[pulu] worm; noun

4524 [boli]- become white; Class 5 verb

fl fl

[bolipi] whiteness; noun

[boli] light; noun

[bolica]
lamp; noun
[bolaki]

[bolli] silver

An adequate grammar must recognise the morphological and semantic

relationships that hold between items within such sets, and has to be

able to state the extent of the coincidence between the dorains of

these distinct relationships. However, where relatedness between

lexical items at both the morphological and semantic levels is the

result, diachronically, of once -productive processes in a language

which are now only semi -productive (or not productive at all), a

problem in analysis arises. Regarding this sort of problem, Chomsky

felt able, as late as 1965, to propose a transformational derivation

of 'destruction' in a construction such as 'their destruction of the

 211

property (1965: 184 -5); but the existence of such sets as 'horror',

'horrid', 'horrify'; 'terror, - 'terrid', 'terrify', etc. as the result

of 'Quasi productive processes' was recognised as tending to the 'very

unfortunate conclusion' that such items as these should be specified

in the lexicon directly (1965: 186).

The reason why this conclusion should have been felt to be

unfortunate is relevant to our purpose here. Chomsky saw, correctly,

that 'it is clear that from the point of view of both the semantic and

the phonological interpretation it is important to have internal

structure represented in these words' (1965: 186); but such a view is

of course incompatible with the proposal to specify these words

directly in the lexicon, as long as it is simultaneously assumed that

the phonological representation of a lexical entry be with respect to

phonetic composition alone (i.e. that morphological information be

ignored; see (i) above). However, this does not ex lain why morpho-

logical information should not be given in the lexicon, though it

does make clear certain assumptions regarding the respective roles of

the lexicon and the syntactic comyonent in Chomsky's 1965 model of

grammar. A tentative solution to the problem which Chomsky proposes

within his 1965 model (1965: 187) is the setting up of certain 'Stem'

elements (such as 'horn -') which can be inserted into a string formed

by prior insertion of certain other elements (such as ' -ify') into the

preterminal string. More recently, however, he has moved to a

'lexicalist' position (Chomsky, 1970), and has proposed a particular

formalisation (the X- convention) for representing the internal

structure of complex lexical items such as 'destruction'. The rules he

now proposes are still clearly part of 'syntax', however, and the

lexicon still has its entries in the form of simple linear concatenations
 212

of phonological elements, the only purpose of the concatenated

strings being to specify the phonetic composition of the item concerned.

All this is mentioned here simply to make the point that the

'lexicalist' position is logically distinguishable from particular

formalisations of it - for example, Chomsky's own, and the modified

version proposed in Stóckwéll et al. (1968: 5-8). The analysis to be

presented in this chapter also represents a lexicalist approach, but

one within a different model of grammar, especially insofar as the

role of the lexicon is concerned.

Since our main concern in this chapter is with Coorg verbs, and

since a number of these show a voice- relationship (see Appendix II, and

the preliminary discussion in the preceding section) which is a

particularly clear type of the lexemic relationships being considered

here, it is relevant to recall the conclusions at the end of Chapter 3

regarding volitively- related verbs; it was seen there that no syntactic

solution was acceptable, once the decision had been made to account for

such related forms by setting up a single underlying form in each case.

It was seen that the notion of a 'volitive sentence' is an invalid one,

and that one should speak instead of a 'volitive verb'. Finally, it

was proposed that this relationship be handled by setting up, in each

instance, a single underlying form in the lexicon. This is, of course,

an 'unfortunate' conclusion if the lexicon is of the type assumed in

Chomsky (1965). It is, however, possible to formulate a lexicalist

position (as the facts regarding volitively related verbs demand) by

modifying the lexicon suitably. It is worth noting that Stàckwell et

al. (1968) reach a similar conclusion, although their lexicon does not

in fact account for derivational processes: they note, 'Although in

principle we would like to have a single complex entry for items such
 213

as produce, productive, production, product, etc., and though we have

tentatively explored some possibilities in this direction, there are

so many complex problems that nothing has reached a formalizable

state' (1968: 937).

Cheng (1966) and especially Fudge (1969b) have indicated possible

ways in which to expand the lexicon so as to accommodate a generative

sub -component. Such proposals have been criticised (Brown 1970) for

their dependence on the syllable as a deep phonological unit, and for

the 'unnecessarily complicated theoretical framework' that a generative

component requires. However, while Brown's arguments for an inter-

pretive syllable as a distributional unit at the level of phonetic

realisation are persuasive, it is unclear exactly how a lexical entry

such as she assumes, which recognises linear concatenation as the only

relation holding between systematic phonemic elements, is able to

reflect the fact that many lexical items exhibit internal structure

which is essential to the stating of relationships between them and

other items. It is also not clear, though, that setting up a deep

syllable is an appropriate way to solve this problem in all languages.

It is a useful unit for Chinese, certainly (see Cheng, 1966; also

Tay, 1970): but the critical issue with regard to the description of

particular languages is whether the deep syllable is a suitable

generative unit, as opposed to being a useful distributional unit at

the deep phonological level for the purpose of stating certain rules

(e.g. the discussion of French adjectives in Fudge, 1969b: 255 -6).

The facts of Coorg apparently require that we work in terms of the

'morph' (in the sense of Bazell, 1953: 51 -62), which is only 'feebly

linked to syllabic pattern (Bazell, 1953: 62) in this language;

no evidence for a distinct deep syllable has been found.

 214

In the analysis proposed below, no presuppositions are made

concerning the syllabic status of the generative phonological unit;

instead, we work in terms of 'root' and 'affix', where the root is

set up on the basis of the morphological similarities that have to

be accounted for between related lexical items. It has to be

admitted that there is a circularity to this approach, since the

notion of what are 'related lexical items' is prior to the deter-

mination of what the form of the root actually is. But this circular-

ity is inevitable in a discovery procedure, and not vicious as long

as (i) semantic data, (ii) morphological data, and (iii) speakers'

intuitions are balanced against each other, and never relied upon

individually.

2.2 The form of lexical entries

We now consider the problems involved in an approach which

attempts to provide morphological information in that part of the

lexical entry which consists of a string of systematic phonemes.

A lexical entry for a complex such as

3537 L Puilg] - rot, fester

L Pgi J
mould, mildew

L PuluJ worm

must contain a structured string of systematic phonemes, representing

a ;'eneralised derivational statement of each of the three items based

on the same root; this will include specification of the affixes

that are involved in the morphological composition of each item. In

the above example, the root must be represented as a string of three

elements; the first realised as a voiceless bilabial stop; the

second as a short high back rounded vowel; and the third as a retro-

flex lateral when a vowel (an affixal element) follows, and having
 215

no phonetic realisation otherwise. There are some general points

raised in this sort of analysis which require comment, however.

(i) Linear versus parallel generation.

The most frequently occurring problem is deciding exactly

what relationship holds between items resulting from derivational

processes such as are illustrated here: thus, is the Ong]- of

[ puñgi] 'the same as' the [gig]- of [puñg] -, or only phonologically

n.
identical to it? If it is 'the same', is [ pungi] a deverbal nominal

or is [puñg]- a denominai verb? The general answer is that both types

of relationship between lexemes have to be allowed for. Thus it will

be a virtue'of the model to be able to derive forms syntagmatically

where this is appropriate and allow for paradigmatic derivation else-

where; i.e. the model should be operationally parallel (in the sense

of Neisser, 1967: 72) in those instances where it is not clear that it

should be linear. A clear case of linear generation (syntagmatic

derivation) will be found in the case of voice- related verbs; a further

case will be discussed in Chapter 6 (pp. 420 -427) concerning certain

gender- related nouns. Where the answer is not clear, as in the case

cited above, parallel generation allows for the uncertainty and does

not force an answer which cannot be supported from the data.

(ii) Semantic analysis.

Since the root is common to both the nouns and the verb in

the lexical items considered here, and since it is set up on the

basis of the analyst's assessment of what constitutes a reasonable,

and non-fortuitous coincidence of formal and semantic criteria, it

would seem necessary that the semantic component should be sensitive

to the internal structure represented in their complex lexical entry.

This would crucially involve a semantic reading for the root element,
 216

in advance of the category distinction between noun and verb and

neutral to the distinction between these; only in this way could the

economy achieved in having a single entry based on a uniquely -

specified root be reflected in the operation of the semantic compon-

ent. For these reasons, it would seem to be necessary as well as

advantageous to abolish the distinction between a lexical entry and

a dictionary entry; this is a contingent simplification of the other-

wise complicating step of removing the lexicon from the base

component of the grammar and expanding it so as to include its own

generative subcomponent as well as the lexical lists of the language.

The operation of the semantic component is thus sensitive not only

to the assigned structure of the preterminal string of the syntactic

base, but also directly to the structure of lexemes generated by

the lexicon base (just as the phonological component is sensitive

directly to their phonological specification). Therefore readings

must be provided for each of the affixes specified in the entry, in

such a way that they can be amalgamated with the reading for the

root by some explicit process. Then, a lexical item, such as Lpulu],

or [puñgi], or [puñg] -, has its form defined generatively by what is

a possible sequence of root plus affixal elements, and has its

semantic structure defined by a complex reading which is based upon

component readings for the root and each of the affixal elements

involved. In practice, affixes tend to supplement and modify the

reading of the root, and to provide information regarding the

syntactic category (noun, verb, etc.) and syntactic behaviour of the

item they help to constitute. They are frequently the source of what

Quine calls 'stage directions' (Quine, 1964: 474); thus, 839 [olap]-

will be entered under the root /01 / -, and will be specified as

containing (among other elements) a bilabial affix; this affix would

 217

have an associated semantic reading which would restrict and refine

the reading for the root, which would have to be something like

'cleanse, with water'. The result would be - in highly simplified

terms - ('cleanse, with water' ('said of plates, pots, etc.')).

This can be made a little clearer, perhaps, by adapting some

proposals in Bierwisch (1970). Working within a componential model

of semantic structure, Bierwisch envisages meaning structures

involving 'arguments' (symbolised as k) which receive grammatical

indices (denoted by etc., for 'grammatical subject' and

d'

'grammatical direct object', respectively) and which are related to

each other by certain 'logical constants' (such as 'and', 'not', etc.)

and by a number of 'relational components' (such as CAUSE, CHANGE TO,

HAVE, etc.) ( Bierwisch, 1970: 175 -7). If this model is used for the

semantic specification of a lexical item, and if a generative source

for complex lexical entries, working in terms of roots and affixes,

is set up, then the schema in Figure 5 shows the way in which the

semantic reading of a complex entry may be set out.

The items represented in the entry in Figure 5 are:

3255 [para]- crawl; Class 3 verb

[parat]- spread (grain); Class 1 verb

[pabb] -» (creeper) twines; Class 1 verb.

[para]- and [parat]- are volitively related items, by process Pi . We

are assuming here, for the purpose of exemplification only, that all

these forms can be accounted for by a root /par/- and affixes - /a /,

-/b /, and /t/. Square labelled brackets around these elements

indicate their derivational status (root, affix); braces indicate

that one and one only of the elements enclosed must be selected.

Unlabelled square brackets enclose information concerning the syntactic

 218

+ Verl,
1

7
Affix Class
Base

+ Verb

Affix Class 1

Base

+ Verb
i

Affix_ Class 1

Base

L e xe iii::

and and

X PERFORM ( ) Xs CAUSE (Xd PERFORM ( ) )

J

-.. and and

<7.r:::2A`i'I-,, Xs')
S
UMAN X. ¡VEGETATE xdllo

and. and and

< uCC`LED X > <COLLECTIVE Xdllb

ani and

< ;Ks7 not <GrROWING X'>

and and

Figure 5: A complex lexical entry with associated semantic readinEg

 219

category and class of the item concerned; angles,

< i> , enclose

information concerning selection restrictions; unfilled parentheses

indicate that part of a complex reading which is made up by the

semantic reading of the root (common to all the possible complex

readings). The semantic reading for [para] -, 'crawl' is arrived at

in the following way: the features assigned to the root /par/- and

the affix /a/ are combined, as indicated by the converging pathways

in Figure 5; at the node thus foamed, this combined reading is

supplemented by a further complex of features which are peculiar to

that node and which further develop the type of predication involved.

Thus PERFORE ( ) will have to specify an extension in a horizontal

plane for [parat] -, 'spread (grain)'. But presumably the distinction

between extension and dislocation with respect to KOVELEET, in the

items [pabò] -, t(creeper) twines' and [para] -, 'crawl', respectively,

need not be handled in the semantic specification of these items,

since it is part of the semantic specification of a creeper that it

is not a type of living thing that is bodily mobile.

Note that it is not the function of the pathway from affix to

node to contain such information as Xs PERFORM ( ) or Xs CAUSE

Xd ( PERFORE ( ) ). This is because not all lexical items which

have the affix -/a/, and which are verbs, and even which belong to

Class 3, would be appropriately specified as Xs PERFORE ( ); for

example, 3897 [mara] -, 'forget', 200 [ala] -, 'crave', etc., which

have to do with inner states or feelings. The pathway from affix to

node is rather to be seen as defining the node to which are attached

the idiosyncratic semantic features of the corresponding lexical item.

Note also that, while in the example given in Figure 5 the

morphological structure assigned by the lexeme generator is reflected

exactly in the semantic configuration, this is not a necessary state

 220

of affairs: all that is required of the match between the morpho-

logical and semantic levels is that it be exhaustive, and either one -

one or one -many.

Finally, it will be appreciated that the labelling of semantic

features applicable to a root in advance of a category distinction

between noun and verb presents difficulties if natural l anguage is

used for this purpose: for example, the use of the label hOVEKENT

in Figure 5 is not intended to imply that the 'basic' meaning of the

root /par/- is nominal rather than verbal.

(iii) Ordering of complex entries.

Thirdly, there is a problem in deciding whether a complex

lexical entry should be made by root (with the affixal possibilities

represented as subordinate to this), or by affix (followed, in each

case, by a statement of the roots that may occur). It has been

Figure 5 first of these alternatives is the

correct one, and the considerations in (ii) above concerning the sort

of information that affixes contribute to the semantic reading of a

lexical item support this.

(iv) Root and affix.

Finally, in a number of cases no univocal evidence for a

root -final element is available from an examination of sets of

apparently related items; thus, it is frequently impossible to tell

whether a consonant occurring immediately after a root vowel is the

realisation of a root -final element, or of an affixal element (with

the root -final element either not being present in underlying repres-

entation, or else having no realisation at the phonetic level). In

what follows in this chapter, and in the next two, it will become

clear that the analysis proposed is one which is in search of further

 221

data, in order to throw light on such points of obscurity. For this

reason, reference will occasionally be made to cognate forms in

related languages (especially Tamil, Malayalam, and Kanarese); the

justification for this lies in the fact that in a large number of

cases there are regular correspondences between the roots set up on

the basis of internal evidence from Coorg and those evidenced in the

related languages. Where internal evidence is scanty, therefore,

comparative data forms one possible source for framing an expectation

of what forms might also have to be accounted for in Coorg as a result

of further fieldwork. However, such data has been used as circum-

spectly as possible.

The rest of this chapter is concerned with formulating the

generative component of the lexicon in such a way as to account for the

data in Appendices I and II; particular attention will be paid to the

voice relation that holds between verbal lexemes, but occasionally

appeal will be made to non-verbal forms, where these throw light on

the root and affix structure of a verbal lexeme. The issue of noun -

verb and noun -noun relations will be taken up in the following

chapters.

3.0 The generative component

In what follows, a number of rules will be presented and

illustrated. In many cases, modifications or additions to rules will

be required as the examination of the data proceeds, and for this

reason they will simply be numbered serially as they are appealed to;

but in each case a reference will be given in parentheses, after the

rule in question, to its final form, according to the order of present-

ation in sections 4.1 - 4.4 below.

 222

(i) The lexeme.

The generative coEponent of the lexicon takes the notion

'lexeme' as primitive, and characterises the possible derivations

from it. The lexeme is not to be confused with the base, since it is

the function of the lexeme to specify not only morpho -phonological

form but also to provide such information as the fact that a certain

lexical item is a verb, of Class 1, etc. The following Lexeme

Structure (LS) rules are required:

1. Lexeme -- 11 Base I] Class(ifier) (LS 1)

2. Base -' Root (+rffix1) (±Affix2) (LS 2)

3. Class Noui (LS 3)

etc.

4. Verb (LS 4)

The elements I and II specify the two major categories of verbs,

those of Class 1, and those of Classes 2 -5, respectively. Their

further development is reserved for later on in this section.

(ii) The root.

This consists of an obligatory Peak (at which a system of

vowels operates), optionally preceded by Onset and optionally followed

by Coda (at each of which places distinct but related systems of

consonants operate). The sequence Peak Coda Patterns in identical

fashion with respect to all the phonological rules regardless of

whether Onset is present or not, and this fact is reflected in the

form of the rules which expand the root; they allow for a three -Mace

structure of which the last two places, Peak and Coda, are dominated
 223

by a single node, Nucleus:

5. Root -+ (Onset) Nucleus (LS 5)

6. Nucleus Peak (Coda) (LS 6)

(iii) The affixes.

These elements are expanded by rules 7 -9. They are dis-

cussed in some detail below. Volit and Volit2 are set up to account
1

for the operation of P1 and P2, respectively, as described briefly in

the first section.

Voc(alic) Suff(ix)
7. :=ffixl (Ls 7)
Cons(onantal) Suff(ix)

fSuffix2
8. Affix
2 } (LS 8)
Voice Suffix

Volit,
9. Voice Suffix - (Ls 9)
Volit

The lexemic structure as developed up to this point is illustrated in

Figure 6.

Over this configuration the following two conditions operate:

Suff

VOlit1
10. NOT: (LS l0)
Volit,

Voc Suff

11. NOT: Suffix2 + II (LS 11)

The first of these specifies that Coda must be present when Voc Suff

occurs, and that Suffix2, Volitl, and Volit2 only follow an element
 224

# Lexeme #

[t Base fl Classifier

Root (+ Affix ) (+ 1_ffix2) Verb

r
\

Nucleus i

Su ff . Suff2

(Onset) Peak (Coda)

OnS.Suff.
foc
Volit
L

Fire 6: generalised structure of the verbal lexeme

either at Coda or Cons Suff; the second ensures that Suffix2 cannot

occur in verbal lexemes of Class II.

(iv) The phonological system.

At this stage in the derivational process, systems of phono-

logical elements are to be made operative at the positions defined in

the lexemic structure. In spite of the polysystemic implications of

this statement, there is essentially one phonological system, comprising

consonants and vowels, within which are specified the subsystems at

different positions in the lexeme.

Three points have to be made in advance briefly concerning this

system.

(a) It has been seen (Chapter 2) that a phonemic statement

for Coorg has to recognise a contrast between voiced and voiceless

stop consonants for all places of articulation; but also that this

contrast exists only for stop consonants, and that even among these
 225

the contrast is in many cases marginal and is in general highly

asymmetric with respect to both distribution and frequency of the

elements involved. It was also seen Chapter 2 that a phonological

system of Coorg taking frequency and distribution of consonants into

account would bear a close resemblance to the system which is expressed

in the orthographies of Tamil and Malayalam, and which has also been

proposed for the proto- language, in respect of the voicing feature in

stops. In this type of system a contrast between tense and lax

articulation is seen as the prime distinction involving these

elements, the tense stop being indicated by gemination of the

appropriate symbol. The possibility of voicing (together with other

features of lax articulation, such as frica.tion) is then allowed for

with respect to the lax stop in certain positions, viz. intervocal-

ically, and after a homorganic nasal.

(b) The second point is that the phonological system is

defined here in terms of elements that are largely abstract and

'feature- sized' (Fudge, 1967: 3). The phonological segment is provided

by the lexemic structure; what is required of the system of phonology

is that it specify the features that fill out the given segments.

These features are denoted by letters and numerals (A, 3, I, II, b,

n, etc.) which combine with each other within the segment to form
A
complex symbols joined by . However, we use a more mnemonic trans-

cription for the discussion below, set off in the text by the slashes

/ /, and matrices are provided wherever necessary to state the

equivalences between the two systems: thus /p/ is AI, /b/ is Alb,

/m/ is Ain, etc. However, in the statement of the realisation rules,

the abstract system of representation is reverted to, since each

letter or numeral directly or indirectly has its effect at the

 226

systematic phonetic level, in terms of instructions to the articulatory

organs.

(c) Finally ~, the phonological system (however represented) is

assumed to be subject to two distinct types of rule: the Phonological

Mutation (p: ) rule, and the Realisation (R) rule. It is further

assumed that mutation rules at the phonological level, which rewrite

one phonological element into another, serve the purpose of stating

irregularities and subregularities in the language, and that they are

ordered; while realisation rules (applying only after the last PN rule)

rewrite phonological elements as systematic phonetic units according

to very general states of affairs in the language, and are unordered

(see the discussion in Fudge, 1969a). The extrinsic allophones

yielded by the realisation rules are subject to mutation as a result

of overlapping articulations; this is handled by the Articulatory

Sequence (AS) rules, which are mutation rules operating at the

physical phonetic level. Since they deal with phenomena which are

directly attributable to the properties of the speech organs, they

are in some sense 'natural', though of course it is not necessary that

they be universal; the notion 'natural' in this context is not incom-

patible with the notion 'language specific'.

3.1 Onset

The system operating at this point is illustrated in the matrix

in Figure 7. Parentheses (wholly without systematic status) indicate

those elements that are marginal to the system: for example, the

parenthesised elements of the second row (from the top) each occur not

more than five times in the data. /ñ/ occurs only in the Class 1 item

157, and /s/ only in the Class 1 items 119 and 139:

2386 [ñavndj- squeeze

 IV

Figure 7: The system at Onset

2310 [se :d]- drink a deep draught

2312 [se:r]- join (intr)

No retroflex consonants occur. The second and third columns (from

the left) are unmarked and marked, respectively, under B; and also

the first, second and third rows are unmarked, marked and marked,

respectively under I. This is intended to formalise the distinction

between marginal and central features in the phonology. B, in the

second column, defines the unmarked place of articulation which is

neither A (corresponding to labial articulation in a general way)

nor C (velar place of articulation); Bj, however, defines a marked

variety of B in the third column (marked for postalveolar to palatal

articulation). Similarly, Row I represents the unmarked type of

stop consonant; Ib and In define those instances where complete oral

closure is marked for the features of voice and lowering of the

velum, respectively. /c/ is unmarked with respect to I but marked

with respect to B, whereas /ñf is marked with respect to both these

 228

features; in this way ¡c¡ is stated as being less marginal than /ñ /,

and this is reflected in their relative frequency of occurrence

( /c/ is the least frequent element of the first row, but much more

frequent than the other two elements of the third column) . The

feature notation recognises the peculiar status of marked types of

articulation by providing for them a second -order symbolisation -

in this case, one that looks to the phonetic facts more obviously

than do the first -order symbols (j = postalveolar to palatal, b = the

voiced counterpart of p for 'plosive', n = nasal) .

The realisation rules which operate on the features defining the

matrix in Figure 7 are as follows:

12. A --. [bilabial p(lace) of a(rticulation_) j (R1)

[postalveolar p. of a.] / j

13. B [ postdental p. of a. ] / II (R2)

[dental p. of a.] elsewhere

14. C --* [velar p. of a.] (?3)

[contact, vocal cords vibratin^, / b

15. I - [contact, velum. lowered ¡ n (M)

[contact elsewhere

16. IV -+ [occlusion, groove friction] (27)

The terms 'contact' and 'occlusion' are from Halle (1ß58).

Certain points concerning the formulation of the rules have to

be noted here. First, it is assumed that systematic phonemic elements

are spelled out always with the features on the horizontal axis of the

relevant matrix preceding those on the vertical axis (thus /p/ is AI,

not I.!; and /ñ/ is BjIn, not InBj). Secondly, the usual phonetic

assumptions are made concerning the movement of articulators: thus

 229

[postalveolar p. of a.] specifies the tongue blade as the active

articulator, and [postdental p. of a.] specifies the tongue tip as the

active articulator, etc. In addition it is taken that, for Coors,

lowering of the velum implies simultaneous vibration of the vocal

cords (there are no voiceless nasal consonants in the language). From

what has been said already regarding the role of stop consonants in

Coorg phonology, it will be clear that the feature [contact] specifies

no vibration of the vocal cords; this feature is either specified

directly as an additional phonetic feature [vocal cords vibrating] in

certain environments, or indirectly (for example, by the presence of

the feature [velum lowered]) .

Finally, note that the second -order phonological features j, b and

n never occur on the left hand side of the arrow in the realisation

rules; they are employed only in stating the environment for the realis-

ation of first -order features. First -order features, on the other

hand (ri, B, C, I, IV), may occur in either of these positions. The

distinction in function which this positional difference marks will be

referred to as that between 'contextual function' (of second -order or

first -order features) and 'articulatory function' (of first -order

features only), respectively.

3.2 Peak

At this point in the lexemic structure two interrelatable sys-

tens operate. The first accounts for vocalic quality at the

systematic phonetic level, while the second, consisting of just the

optional element /LI, specifies long phonetic duration: see Figure

8. /L/ does not occur in the data with / and a/; this is taken

to be an accidental gap (nasal vowels are of very low frequency). In

terms of the system here, /a/ is the maximally unmarked vowel, with
 2 o

Fi;--jure 8: The systems at Peak

/i/ and /e/ marked for front articulation (this is the function of

the second -order symbol j in this context), and /u/ and /o/ marked

for back, lip- rounded articulation (the function of w). /ú/ and

/á/ are marked for lowering of the velum (n performs the same

function here as in the system at Onset). The nasality feature thus

marked on these Peak realisations is extended throughout the lexeme

by a phonetic mutation rule, AS1; the rule AS2 also provides for the

homorganic nasal segment occurring before the nonpast and past affixes

of Class 5; see the Class 5 items 8, 9, and 72:

[cuy]- (LcuymP]- (fly, ant) bites

2894 Ltúyj- (LtXympj- Ltúyñc] -) (snake) hisses

1927 cáyéj- (Lcáympj- Lcáygntj -)

chew
Lcávéj- (Lcávémpj- Lcávéntj -)

The main realisation rules which operate on this system are given
n n

here (excluding for the moment the elements /i/ and /e /, which are

considered in more detail below):

 231

[front of tongue highest, lips spread] j

lips rounded/ w (R9)

aback of tongue highest,
lips unrounded] else.

velum lowered]/ n

18. 1 [constriction, (Rio)

-] else.

velum lowered] /w n
1r[midopen tract,
/
_] {w
19. 2 -, (R11)
fvelum lowered]X n
open tract,
-]
/k

20. L -+ [vocal cords continue to vibrate, unaltered (R12)

vocal tract]

Notice the distinction here between the realisation of X which

is marked for lip- spreading, and the unmarked lip position for this

feature (unrounded, for the elements /a/ and /á/).

Furthermore, while the feature 1 is always realised as

Lconstriction], the feature 2 ranges from [midopen tract] to [open

tract], depending on whether X is marked for lip- rounding /lip-

spreading, or is unmarked, respectively. The term 'constriction' is

from Halle (1958). It is assumed here that [constriction], [midopen

tract] and [open tract] automatically specify vibration of the vocal

cords.
n n
There is, however, a problem with the elements /i/ and /e /. An

examination of the data shows that for the great majority of cases it

is possible to state that the high and mid front unrounded vowels at

Peak have retracted allophones (still unrounded) of corresponding

height before a retroflex consonant (see Chapter 2, pp. 67 -70).

 232

Locically, exceptions to this can be of two kinds; either the back

unrounded quality occurrin!- without a followin- retroflex consonant,

or the front unrounded quality occurring within this environment.

Since the latter type of exception does not occur in the verbal system,

we leave it on one side here. Concerning the other type, most of the

instances involve a followinrc postdental flan or apico- alevolar trill

(LT.], in terns of the transcription used here). Before this element,

both back and front unrounded vowel qualities occur and contrast; see,

for example, the Class 2 items 22 and 23:

2373 Lne:rj- hanE (intr)

23S0 [ne:rj- rise up

The situation may be regularised at the cost of setting up two elements

/r/ and /r'/ instead of just /r/ at Coda, both of which are realised

in identical articulatory terras, but which represent distinct deep

phonoloical categories. /r'/ is related phonologically to the class

of elements that underlie retroflex consonants at the systematic

Phonetic level. After this is done, the only exceptions left are the

Class 1 items 17, 75 (alternative form), and 116 (alternative form),

and the Class 4 item 15:

2634 Ltikk]- strain (at delivery of child)

2740 [timm]- sneeze

2857 Lde:t]- drive away

1 641 [ gell J- win.

ìs1e accordingly establish a marked column for these cases, under what

is a third -order feature, r, under j. This formalises the fact that,

n

so far as the system at Peak is concerned, /i/ and /e/ are to be seen

as types of /i/ and /e /, respectively (it will be shown below that a

different solution is required for /i/ at Voc Suff).

 233

We therefore supplement the realisation rules R9 and R11 as

follows:

lips rounded]/ w
[back of tongue highest,
lips unrounded]/ jr

21. (R9)
2

[front of tongue highest, lips spread]/j

velum lowered] /w n
[midopen tract,
417. -J/ j(r)

22. 2 -> w (R11)

velum lowered] /X n
[open tract,

ir:
-]/x

Notice that the two main subparts of R9 have here had to be

re-ordered, since the environment 'j' (formerly in the first sub-

part, as in 17 above) includes the environment ' +jr' unless it is

preceded by this environment in the statement of the rule.

The derivation of the back unrounded vowels at the systematic

phonetic level is handled by the realisation rule R9, therefore.

But this rule, by a simple extension of its operation, also accounts

for the fact that, in the environment of a preceding labial consonant

and a following retroflex consonant (including /r'/ - the alternative

forms of the Class 2 item 26 are possibly revealing in this regard),

only the back rounded quality of high and mid vowels may occur

Ca] and L a: J may also occur, a fact that tenus to confirm the

unmarked status of /a /; see Chapter 2, pp. 90-92). R9 accounts for

this restriction quite naturally by associating the retraction of

front unrounded vowels with the following retroflex environment (the

tongue root tends to retract with retroflection of the tip), and the
 234

rounding of the back unrounded vowels thus formed with the preceding

labial environment. For a detailed discussion, see Emeneau (1970a).

Therefore, R9 has to be supplemented further by what may be

called for convenience the 'vowel retraction' rule:

lips rounded]/A04 j ()* (L)'Br

23. X -i [back of tongue highest,
lips
unrounded] /{ I( )4 j() (L)ABr
C

(R9)

'Br' defines the phonologically retroflex consonants at Coda and

Cons Suff. The specification 'A(r and {r;} OA ' signifies that

it is immaterial to the operation of this rule whether it is AI, r'sb

or AIn which forms the preceding consonant in the first case, or

B /CI, B /CIb or B /CIn which forms the preceding consonant in the second

case. Likewise, ' j()' signifies that it is immaterial whether the

phonological segment in question is Xjl or Xj2. Note that this

convention of empty parentheses is quite different from the use of

parentheses in '(L)', where they indicate that the element enclosed

may or may not be present.

3.3 Coda

3.3.1 Preliminaries

The nature of the Phonological elements operating at this

position represents a crucial point in the analysis. As noted

earlier, in a number of cases there is apparently no good reason for

deciding that a consonant imrediately after a vowel at Peak is the

realisation of an element at Coda rather than Cons Suff, or vice -

versa; if no other consonant or consonant cluster follows within the

same lexeme, and if no evidence from related lexical iters is to hard,

 235

and if the consonant in question has in any case to be allowed for at

Coda and Cons Suff, there is no principled way of deciding the issue.

The Class 1 items 31, 32, 43, 69, 93, 96, 99, 101, etc. are all

cases in point. Where no principled decision is possible, the grammar

has to be able to state the equivalence of the possible derivations

at a lower level, and this is allowed for here; thus, the phono-

logical rule PM6 (which erases all +boundaries) states, in effect,

that whether a given consonant derives from Coda or from Cons Suff

is immaterial to the following rules.

In other cases, three kinds of evidence regarding the two

consonantal systems are available. The first is from an examination

of constantly recurring elements in base -final position as candidates

for treatment as suffixes. Language -internally, this can be a hit -

and -miss procedure, although the form of the base -final element is

often a guide; thus, it is the case in Coorg that a homorganic

cluster of nasal plus stop, of whatever Place of articulation,

frequently occurs in base -final position and never base -medially, and

is thus almost certain to be amenable to treatment as a suffixal

element. Once the suffixal system can be determined with some degree

of confidence, one is in a position to determine the elements at

Coda. More direct and revealing evidence is sometimes available in

the existence of related lexical items in the language (see the

earlier discussion concerning 3537 [purig] -, 'rot, fester', p. 214 ).

Finally, we may work with the simplifying assumption that all base -

medial consonants are representations of Coda; this follows directly

from the postulation of a root of (C)V(C) structure, together with

the condition LS 10 above. From this evidence, we have to account for

at least the following systematic phonetic elements at this position:

dental and retroflex laterals, nasals, and voiced stops; the

 236

apico- alveolar tap; and the labial and palatal semivowels. However,

we have already seen that the apico -alveolar tap must have two

sources at the systematic phonemic level to account for certain Peak

realisations, and consideration of sets of related lexical items

leads to the establishment of still other elements at Coda. The

complete system is shown in Figure 9.

Figure 9: The system at Coda

Before discussing these elements individually, some general points

have to be made here concerning base structure.

At the phonetic level of contrast, the following canonical base

shapes are found:

Class 1 (C)VC
(C)VNC
(C)VCC

(C)VC (C)VNC

(C)VSC (C)VSNC

(C)VCVC (C)VCVNC
 237

Classes (c)v
2 -5
(c)v
(c)vc
(C)VCC

(c)VC

(c)VCV

(c)vcvc

(i) Base vowels.

The first point to be noted is that long and short root

vowels contrast only before the base -final elements -C and -NC in

Class 1, and only before base -final -C or base -finally themselves in

the pattern exhibited in the remaining classes. In all other cases,

the root vowel is short, viz. before base -final -CC, and when

followed by the sequences -S(K)C and -CV(i)C in Class 1 and -CV and

-CVC in Classes 2 -5. The problem of the lack of length contrast

before base -final -CC is taken up below (pp. 240 -245); first, however,

we examine the other sequences which occur only after a short root

vowel. Concerrin: these, the following points have to be made:

(a) In Class 1, -S- and -C- after the short root vowel are

in minimal contrast at the transcriptional level in .9npendix I: -S-

is either [v] or [y]; and -C- is never [y], is [v] just in the Class

1 item 172:

4516 [ bevar] - s-:: ea t ,

and is otherwise [d, d, n, 1, 1, r]. It will be shown below, more-

over, that the single instance of apparent contrast here (in the item

just cited) is the result of the non -systematic status of the

Appendix I transcription in respect of the symbol [v].

(b) The second base -vowel -V-, before -(N)C -, only occurs
 238

after base -medial -C -; and it is in all cases either [a] or [ij.

Further, it is possible generally to predict which of these two vowels

will occur, given the root vowel; [i] occurs as the second base -vowel

when [i] or Lu] are the realisations of Peak, while [a] occurs after

all other root vowels ([i] does not occur at Peak in these base

types in the data). Only the Class 1 items 165 -6, 169-70 and 185

fall outside the scope of this statement:

3623 [porik]- pick up small objects

4402 [badi;k]- live (happily)

.11

725 elid]- write

3729 [porid]- undertake an office

[moliiig]- be pressed.

Each of these shows Lij where the pattern noted above would require

Lai; we shall set these items aside here, as not relevant to our

immediate purpose. It will be seen in what follows that they cannot

be generated by the model that is described in this chapter and the

next, but require a more powerful generative component which will

be formulated (principally on the basis of complex nominals) in

Chapter 6.

The regular pattern, whereby the second base -vowel is predictable

in phonetic quality, is shown in the Class 1 items 160 (also 163 -4,

168, 172 -3, 175 -9, 181 -4 and 187 -9) and 161 (also 162, 167, 171, 174,

180 and 185 -6):

203 [alak]- stir with a ladle

636 [ulik]- flay.

In view of these considerations we may (setting aside [bevar]- for the

moment) abolish the distinction between -S- and -C -, and cease to

regard the second base -vowel as a structural component of the base in

underlying representation. Furthermore, we know that, as a base -medial

 239

consonant, the first -C- (including -S- now) after the short root

vowel represents the system of elements at Coda; and it will be

shown later (pp. 264 -271) that base -final -C /-NC represent the system

of elements at Cons Suff. Hence we set up the single structure

(C)VC +C for all the following base types:

(C)VSC (C)VSNC

(C)VCVC (C)VCVNC

now require two rules in respect of this structure: the first,

which will be referred to as the 'epenthetic vowel rule', inserts

the appropriate vowel between the Coda and Cons Suff elements; the

second, which will be referred to as the 'vowel length reduction

rule', ensures that all bases in which Cons Suff occurs have short

root vowels at the phonetic level. That is, we allow for /L/ to

occur at Peak in the normal way for these as for other bases; the

occurrence of Cons Suff then triggers off the vowel length reduction

rule which deletes /L /.

Neither of these rules is given at this stage in a formal way,

since their exact form is dependent upon factors that have yet to be

introduced. For example, they may be formulated either as phono-

logical mutation rules or as realisation rules. They are described

in more detail, and formalised, later on in this chapter (p. 263).

However, the general function of these rules has been given at this

point since appeal will have to be made to them in the following

discussion.

(c) In Classes 2 -5, the second base -vowel of (C)VCVC bases

is predictable in exactly the same way as described above

for Class 1

verbs; the items involved are just the Class 2 items 75 -7 and the

Class 4 items 20 -2:

 240

571 [urid]- roll (intr)

2211 [turd]- be rolled up

4579 [borad]- be in confusion through fear

557/559 [orad]- answer

3751 [porad]- dress (well)

4076 [ morad] - weep

(d) Since the second (base -final) vowel of (C)VCV bases

of Classes 2 -5 cannot be predicted in terms of preceding environment,

we generate Voc Suff in the underlying structure of these items.

Then, we have to formulate the vowel length reduction rule in such

a way that it operates when Voc Suff (as well as Cons Suff) follows

Coda.

(ii) Base -final -C and -CC.

?Te now examine the problem of the contrast of consonant

length after a root vowel, in and This

contrast is of minimal functional load, as it distinguishes only the

Class 1 items 6 and 77:

432 [el]- leave one's position

2767 [tull]- make small jumps

All the other (C)VC bases of Class 1 show base -final [r]; and

(C)VC bases of Classes 3 and 4 show base -final [d] or [r] (which are

thus never in contrast with base -final -CC, which is always [nn],

[nn], [11] or [yy] in these classes). Unlike [d], [r] never occurs

geminate in the language; however, we have to allow for /r/ as a base -

final element in the underlying representation of some (C)VCC bases,

in order to account for certain lexical relationships; see, for

example the Class 1 item 59, and the Class 2 item 77:
 241

4519 [botti- fear

4519 [borad]- be in confusion through fear,

each of which share the same root /bor/ -. This surrTests that the

sequence ,err/ may occur, but is realised as [tt] (we shall have to

refine this statement later, but it will serve our present purpose as

it stands) .

It is also the case that all base -medial consonants are phonetic-

ally short and show no voicing contrast among stops (which are all

voiced). There is a voicing contrast among the phonetically long base -

final stops of (C)VCC bases, but, as was seen in Chapter 2, this

contrast is marginal; -BB (standing for a geminate voiced plosive

cluster) in this position is much less frequent than -PP (the voiceless

counterpart). However, in (C)VC and (C)VCVC bases, voiced and voice-

less stops are in full contrast. This pattern suggests that base -

final consonants are geminate when they stand immediately after a

short root vowel (at Peak); in terms of the phonological system

proposed here, this also entails that stops in this position will be

voiceless. In all other cases (viz. after a long root vowel, or after

a vowel which is not an element at Peak - as in (C)VCVC bases) the

base -final consonant is not automatically geminate. Hence, we require

a rule, which will be referred to as the 'automatic gemination rule',

which geminates those consonants which stand immediately after a short

vowel at Peak and immediately before the base boundary element Il. In

terms of this pattern, (C)VC bases and (C)VBB bases are exceptional;

we shall concentrate on the first type of exception here, and deal with

the second later on.

(a) All the (C)VC bases in Class 1 show base -final [r], or

[1] at the phonetic level. If it can be shown that these are

 242

realisations of base -medial elements in underlying representation,

they will not be exceptions to the pattern noted above. In fact there

is substantial evidence to support this view; consider the following

lexical relationships which involve the Class 1 items 1 -6:

569 [ur]- melt (intr)

[urik]- id. (tr); Class 1 volitive verb

606 [or]- fall asleep

[oraki] sleep; noun

1066 [kar]- be digested

[karak]- digest; Class 1 volitive verb

2655 [tiri- turn (intr)

[tirik]- id. (tr); Class 1 volitive verb

[tirig]- wander; Class 1 verb

[tiri] wick; noun

[tirigani] pulley; noun

4112 [mur]- tighten (intr)

[murik]- id. (tr); Class 1 volitive verb

[muriki] torsion cord (of drum); noun

432 [el]- leave one's position

[elakj- make leave position; Class 1 volitive verb

[elaka] uprooting from position; noun.

We can account for all these cases by setting up the roots /ur/-,

/or /-, /kar/ -, /tir/ -, /mur/-, and /el/ -; we can further account for the

base shape nutation involved in the voice-relation between these verbs

((C)VC to (C)VCVC) by setting up the base -final element /kol for all

the (C)VC verbs. This element is realised at the phonetic level when

it is followed by Volit1, and it is also realised in a number of

related noun forms; it is not realised, however, in base -final position

 243

of verbs, although it is phonologically present at the time of the

operation of the automatic gemination rule. It thus protects /r/ and

/1/ in these items from the effect of that rule. It will be seen later

on that the existence of /k/ in these items facilitates the statement

o

of the operation of Volit1; and it is also a relevant construct in

generating the related noun forms that show final -[ki] (see Chapter 5).

By this solution, we can abolish the distinction between the

canonical base shapes (C)VC and ( C)VCC as far as Class 1 items are

concerned, and set up the single type (C)VC in underlying represent-

ation. stress- assignment rule (p ?'%i 9(i)) marks main stress on all

root vowels, and the following sub-rule (which is the automatic

gemination rule) geminates all consonants which immediately follow a

short root vowel (i.e. the length element /L/ does not intervene), in

base -final position (i.e. before the boundary element 15):

24. íf/ Q (c) (PM 9(i))

25. CT -* cC,/ V 1. (PM ;'(ii)

(b) Classes 3 and 4 also show a contrast at the phonetic

level between the base shapes (C)VC and (C)VCC. Here, however, the

functional load is zero, since all (C)VC bases show final [d] or

[r], and all (C)VCC bases show [nn], [nn], [11], or [yy], Taken by

itself, this night suggest an anal:rsi s for verbs of these classes

that base -final consonants are phonetically long after a short root

vowel, save for those elements which underlie [d] and [r], which are

always short. This is clearly unequatable with what has been discovered

concerning base -final consonants elsewhere in the verbal system,

however, and the correct solution is that base -final [d] and [r] at the

rhonetic level in (C)VC bases are realisations of base-medial /t/ and

 244

lrl, respectively. We shall now consider some evidence supporting

this analysis.

The items concerned are the Class 3 items 2-4, and the Class 4

items 1 -9; among these, the following show certain relationships

w-3-th other lexical items:

4311 [bar]- core

[barat]- cause to cone; Class 1 volitive verb

1614 [hell- be ruined

('edit]- ruin; Class 1 volitive verb

n
['_-e:di] ruin; noun

21E3 [cud]- burn (tr)

[cudigala] cremation place; noun

[tu :di] torch (of reeds); noun

3190 [pad]- lie fallow

{pat]- let lie fallow; Class 1 volitive verb

[padibu:mi] fallow land, noun

[padiña :ri] the west; noun

3191 [pad]- suffer

[padt]- make to suffer; Class 1 volitive verb

4419 [bud]- let go (tr)

[budit]- make to let go; Class 1 volitive verb

[bu:di] toddy tapper's hut; noun

Of these, it may first be noted that the base -shape mutation involved

in the volitive relationship between the verbs cited here ((C)VC to

(C)VCVC) parallels what was found in Class 1 above; that is, base -

final [r] and [d] in the (C)VC bases correspond to base -medial [r] and

[d] respectively, in their volitively related counterparts.

 245

However, there is a second tyre of evidence available with regard

to these items, which is even more compelling, and for which we have

to refer to the vowel length reduction rule. The (C)VCV noun forms
11

1614 [ke:di], 2183 [tu:di 1 (we ignore here the idiosyncratic alter-
nation between /t/ and !c/ at Onset) , and 4419 [bu di ] suggest that

we have to recognise roots of the following forms:

1614 /ke :t /-

2183 /tu: t/-

4419 /bu:t/- ti /bi:t/-.

It will be shown in the next chapter (pp. 320 -321) that there
tt

are good independent grounds for regarding final [i; of the (C)VCV

noun forms as not relevant to the operation of the vowel length

reduction rule: we can therefore account naturally for both the short

root vowel and the single base -final consonant of the related verbal

bases here if we set up a Voc Suff element in their underlying

structure. The vowel length reduction rule deletes /L/ at Peak in

these cases, and the Voc Suff element protects the preceding

consonant (which therefore unambiguously represents Coda) from the

effect of the automatic gemination rule. The Voc Suff element has

no realisation at the phonetic level, unless it is followed by

Volit1. Further discussion of this element is reserved until the

system at Voc Suff is dealt with below (pp. 271 -275). Note the

difference, however, between this form of the solution and that

proposed earlier for the Class 1 verbs; there, a base -final conson-

ant ( /k/, at Cons Suff) was established in the underlying forms to

o

protect the Coda element from the operation of the automatic gemin-

ation rule, while here a base -final vowel (at Voc Suff) serves the

same function.
 246

(c) We now turn back to the exceptional bases of (C)VBB form,

in Class 1. We have seen that, because of the operation of the

automatic gemination rule, base -final stops should yield -PP (voice-

less) geminate clusters after a short root vowel; we account for -BB

realisations in this position by rule. This avoids the necessity of

setting up a row of voiced stops /b, d, g/ in the system at Coda,

thus reflecting the fact that the voicing contrast between stops at

this position is not a regular feature of the phonological system: the

ratio of occurrence of -BB to -PP in verbal bases is only of the order

1 :4 (see Chapter 2, p. 97). Accordingly, we set up the element /+/

at Coda for all these items; its distinctiveness as opposed to /r/

lies just in the fact that it does not satisfy the conditions for the

operation of the epenthetic vowel rule. As a result, /+/ at Coda may

stand immediately before base-final /p/, /t/ or /k/ at Cons Suff;

the resulting cluster is realised as the geminate voiced stop

cluster [bb], [dd] or [gg], respectively (only these -BB types occur

in the verbal system). Further details are given below (pp. 259 -261),

in discussing //.
When 4/ is not followed by an element at Cons Suff, it is

realised phonetically as [r]. Evidence for this solution comes from

the Class 1 item (considered earlier) 72, which shows the following

lexical relationships:

3255 [pabb]- (creeper) twines

[para]- crawl; Class 3 verb

[parat }- spread (grain); Class 1 volitive verb.

(d) Finally, we have noted that there is a regular voicing

contrast among stops in base -final position when they do not immed-

iately follow a root vowel, and it remains to consider how this is

 247

to be accounted for. In order to understand this, it is first

necessary to note that, with one exception, there is no contrast

between voiced and voiceless stops after a homorganic nasal, in

verbal bases. The general pattern, then, is that stops are voiced

when they occur singly in intervocalic position, and after a homor-

ganic nasal; when they occur geminate they are voiceless.

Examination of sets of related items shows that there is a very

common morphological process of stop -gemination in the language,

which serves a number of functions. It will be shown below that one

such function is the derivation of volitive verbs from involitives;

but noun -noun and noun verb relationships are also involved, as in:

1206 [ka:di] forest; noun

[ka:ti] bison; noun

2927 [to:d]- dig; Class 1 verb

[to:ta] estate; noun

In terms of the phonological system proposed here, these items would

be represented as

1206 /ka:ti/ forest

/ka:tti/ bison

2927 /to:t /- dig

/to:tta/ estate.

Note that a low -level phonetic mutation rule (AS 7) will reduce the

length of geminate consonant clusters when they occur after a long

vowel (see Chapter 2, p. 106). This phonological representation shows

the morphological relationship between these items very clearly.

However, there is another sort of morphological relationship, which

is very common in sets of volitively related verbs but which may also

be seen elsewhere in the language; this involves homorganic nasal

 248

plus voiced stop clusters on the one hand, and geminate voiceless stop

clusters on the other:

2680 [ti:nd î- touch; Class 1 verb

[ti:taj excrement; noun.

The phonological representation of these items is:

2680 /ti:nt/- touch

jti:tta excrement.

These two types of morphological relationship are capable of being

expressed as a single process, on the assumption that the sequence

(homorganic nasal _lus geminate voiceless plosive cluster) is

inadmissible in the language: we set up the suffix / */, which is

defined as /p/ when it follows /_/, /t/ when it follows /t /, /t/

when it follows ¡t /, etc., and delete the homorganic nasal segment

when /Y/ occurs. This is handled by the following rule:

v p.
26.
'
Ciao. + *
i i (p = p, t, c, t..
r, r') (PM2('ii))

It is important to note that /r/ and /r'/ also function like

stop consonants in respect of this rule; consider the Class 1 items

22 and 23:

2373 [ne:r;- hang (intr)

[ne:tj- id. (tr); Class 1 volitive verb

2300 [ne:rJ- rise up

[ne:tj- lift; Class 1 volitive verb.

In phonological representation, these are:

2373 /ne :r/- hang (intr)

/ne :rr/- id (tr)

2380 lne :r'/ rise up

/ne :r'r' /- lift.

Note that the occurrence of /*/ does not shorten a long root vowel;
 49

it was seen above (r. 240) that the vowel length reduction rule is

triggered by elements at Voc Suf f and Cons Suf f (although this

statement will have to be refined somewhat in what follows), whereas

/ */ may occur after Cons Suff (which is in paradigmatic relation to

Voc cuff), as in the Class 1 item:

?655 [tirik]- turn (tr); volitive counterpart of

the Class 1 item 4,

and the Class 1 item 167:

2655 [tirigj- grander.

The first shows the occurrence of / */ (here expressing the volitive

relationship), after the Cons Suff element /o /, while the second

shows the Cons Suff element /k/ without following /*/. The base -

final -/kk/ of the first of these items (from -/k + */) is realised

as [ k] by the rules R3 and P4; it is subsequently reduced in length

by the phonetic mutation rule '.S7, which operates on all long

consonants which are immediately preceded by either a long vowel in

a stressed syllable (i.e. at Peak) or a short vowel in an unstressed

syllable (either at Voc Suff, or the epenthetic vowel).

We are now in a position to discuss the individual elements at

Coda.

3.3.2 The system of elements

/p/: this is realised as a voiceless geminate cluster in the

Class 1 item 68:

2498 [tappi- escape

and in 69, which is of similar form. It is realised as a single

voiceless stop in Class 1 item 120:

1341 [ci:p]- comb

In all other cases, it is base- medial and realised as a bilabial

 250

semivowel or as a labiodental voiced fricative, depending on the

environment; these instances are in the Class 1 items 155-6, 157 -8,

and 172, the Class 2 item 38, and the Class 5 item 72:

1973 [cavt]- step on, kick

2454 [tavd]- grope in dark

2386 [ñavnd]- squeeze

4349 [_bavnd]- be bent

4516 [bevar]- sweat

1121 (.kávi ]- lie face down

1927 [cávé]- chew

The last three items show the labial approximant with lip compression

(see Chapter 2, p. 63); this is, of course, not distinguished from

the lip- rounded approximant in the transcription of Appendix I,

since it is predictable with reference to the vocalic environment.

This variation is, however, to be handled by the realisation rules

111 and R4, which have therefore to be supplemented as follows:

lip- rounding] / X( {wr })0 *(L)" IAX()

27. A [bilabial p. of a., XO" IAX(swr1)( )

lip- compression] else. l f (R1)

[occlusion] /
Xj()4(L) A
X()"A"Xj()
2c. I - (R4)
[obstruction] / R06(L) A e XO

The term 'obstruction' is from Halle (1959). It is assumed that

[occlusion] (as opposed to [occlusion, groove friction]) and

[obstruction] automatically specify vibration of the vocal cords.

By 27 above, all implementations of A will have lip- compression,

 251

unless in intervocalic position they are preceded or followed (or

both preceded and followed) by a non-front vowel (i.e. other than

Xj() ); by 28, feature I represents [occlusion] when the bilabial

approximant is preceded or followed (or both preceded and followed)

by a front vowel (i.e. an Xj() ), and [obstruction] when this is not

the case: the complex AI represents the bilabial approximant, of

course, only intervocalically (elsewhere, I is represented as

[contact], in association with which A specifies lip-compression).

The first five items cited above all show the operation of the

epenthetic vowel rule, which may be given at this stage:

29 (i) C1 + C2 C1 + VC2 (CIA +; C2 o) (P?.:3(i)

(ii) V (from (i))

aC+_
of
P113(ii)

i else.

When C1 is /p/, it is realised as an approximant with lip -compression

and fairly close articulation (by R1 and R4) in intervocalic position

if the preceding vowel is /i/ or /e/ (this is the case in [bevar]-,

'sweat'). When this happens, the epenthetic vowel ( &al, by P:3(ii))

is retained. But when /a/ occurs at Peak (as in the remaining cases),

/D/ is realised as a lip-rounded approximant, a d a later rule, at the

phonetic level (AS3),deletes the epenthetic vowel. Note that we have

formulated the epenthetic vowel rule as a type of mutation rule at

the phonological level which inserts a phonological element into

the underlying structure of these items prior to the operation of the

realisation rules; the reason for doing this will be discussed below

(p. 263) .

Finally, only base -cedial cases are unambiguously realisations of

 252

/p/ at Coda; base- final [p] /[pp] may e ^ually well be realisations of

/p/ at Cons Suff.

/c /: is realised as a voiceless stop (geminate or single,

depending on the length of the preceding vowel) in the Class 1 item

30 (also 31), and 93:

4025 [muco]- close

1352 [gi: c ]- make a scratch

[c] after a long root vowel indicates the presence of /*/ in under-

lying structure. Elsewhere, ¡c/ is realised as a palatal semivowel;

it is base-medial in the underlying forms of the Class 1 item 159,

the Class 2 item 4 (also 5 -8), the Class 4-item 10 (also 11 -14), and

the Class 5 items 5 (also 6-11), and 27 (also 72 -3):

S 249 [kuynd]- feel prickly

1142 [kayy]- (tie) passes

1628 [keyy]- work

39 [ay]- send

1809 [koya]- string (flowers)

Note that in the type represented by the second, third and fourth

items here the base -final element in underlying structure is li;

at Voc Suff : this protects the base -medial ¡c/ from the automatic

gemination rule, and the single ¡c/ is later lazed to [y] by R4. 4

later rule at the phonetic level changes the [i] (from /i /) to [y]

if there is a following vowel, and deletes it otherwise (.tiS 5).

/c/ is base-final in the Class 2 item 12 (also 13-17), and the

Class 4 item 16:

306 [a:y]- choose

3103 [ne:y]- spin (thread)

 /J
253

Because of the long preceding vo;rel, it is not affected by the auto-

matic ,_cmination rule, and is lazed by R4.

To handle the lax realisations of /c /, therefore, we supplement

R4 with the following subrule:

w
30. I -4 [obstruction] / 1r( nT )Bj t( )
(R4)

Further, to account for the fact that the place of articulation is

palatal (not postalveolar) for the approwimant, we supplement R2 ,with

the subrule:

B [palatal p. of a.] / C^(

,x()
31 -i )o jl (R2)

Note that all the base -final realisations might result from %c/ at

Cons Suff (followed by / * /).

/t /: this is realised as a geminate voiceless stop in the

Class 1 item 43 (also 44-60), and as a single voiceless stop in the

Class 1 item 113 (also 114 -7):

646 [utt]- ooze

349 [a:t]- become mature.

The last item shows the presence of /*/ in underlying structure; when

it does not occur, /t/ is realised in this position as a voiced stop,

as in the Class 1 item 116 (also 119):

866 [o :d]- read

/t/ occurs intervocalically in base-medial position in the Class 1

item 166, the Class 2 item 44 (also 52), and the Class 5 item 32

(also 35, 45, 61, 71); in all such cases it is realised as a voiced

stop:

4402 [ badile]- live (happily)

 DBIA 212 [ tudi ]- be praised

3686 [poda]- thatch, cover

Thus we supplement R4 with the subrule:

.
32. I -' [contact, vocal cords vibrating] / X()A(L).B .
1
(R4)

This is, of course, very similar in form to the supplement given above

(31) for the palatal approximant realisation of /c/, and to that part

of R4 (28) which handles the bilabial approximant realisations of /p/.

Note that some instances of the voiceless stop articulation may

be the result of /r/ or /r'/ at Coda, or at Cons Suff, followed in

each case by / */; the variant forms of Class 1 item 116, 2857 [ de: t ]- e.+

[de:t] -, 'drive away', may thus be explicable in terms of /r/ (or

/t /) and /r'/, respectively. [d] in base -final position may

similarly be the result of /t/ at Cons Suff.

/t/: this is realised as a voiceless stop (geminate or single

according to the length of the preceding vowel) in the Class 1 item

32 (also 33 -41), and the Class 1 item 97 (also 98-9):

If
1287 [kitt]- be got

1566 [ku:t]- sharpen

It is followed in the latter type of base by /*/, but not in the

following ones, where it is realised as a voiced stop; the Class 1

item 100 (also 101 -112), and the Class 2 item 11:

290 [a:d]- dance

3059 [ni:d]- stretch out straight (intr).

In all the foregoing examples, it might be /t/ at Cons Suff that

underlies the voiced/voiceless articulations. However, the following

voiced stops are all uniquely the result of /t/ at Coda, in the

Class 1 item 162 (also 163, 175 -6, 183), the Class 2 item 32 (also
 255

40, 58 -60, 63), the Class 3 item 13 (also 16, 18), the Class 4 item

1 (also 2 -7), and the Class 5 item 31 (also 34, 39, 42, 53, 56, 63-4,

66 -7):

506 [odik]- sweep

366 [idi]- (wall) falls

929 [Izada]- cross

375 [id]- drop (tr)

3493 [poda]- flutter, tremble.

This requires a further slight modification to R4, so that it

includes the subrule:

`j()
33. I -, [contact, vocal cords vibrating] / R( }A(L }Br (R4)

The feature 'Br' defines the column of phonologically retroflex

elements in the matrix of Figure 9 (p. 236). Hence, we also need the

following supplement to R2:

34. B [prepalatal p. of a., retroflexed tongue-tip]/ r (32)

in order to account for the realisations of /t,/, /n/ and /1/. r is a

further second -order feature (along with j) under B, serving a

contextual function in the realisation of 3. It is discussed further,

in connection with /r' /, below (pp. 256 -258).

/j/ this element is set up to account for the few instances

of [j] in base-final position (contrasting with the regular [y] as

the lax realisation of /c /). It occurs just in Class 1, in the items.

94 -6:

DBI á 274 [1011:j]- do acts of worship

4479 [bi :j]- wave (tr/intr)

3927 rxe: j ]- (clothes) soil

 256

Note that the first item is non -Dravidian, and that the second is an

alternative form (cp. [bi:d] -). ¡j! has to be set up at Cons Suff

also, on the basis of one clear instance, and it may be that all

the items cited here show that element; this would mean that /j/

could be eliminated from the system at Coda altogether.

/n/: this is realised as [nn] in base -final position after a

short root vowel, in the Class 1 item 65 (also 66 -7), and the Class

3 item 6:

1336 [kinn]- tear into strips

2670 [tinn]- eat (not rice)

Otherwise, it is [n], base medially, in the Class 5 item 30 (also

40):

[nena]- think.

/n/: this underlies [nn] base-finally after a short root

vowel, in the Class 1 item 42, and the Class 3 item 5:

678 [enn]- say, tell

516 [unn]- eat (rice).

Elsewhere, it is [n], base-medially, in the Class 1 item 174 (also

178), and the Class 2 item 28 (also 39, 41):

Indo-Aryan [gun.is]- calculate

98 [ani]- wear, enjoy (jewels).

/r/, /r'/: these may be realised as [r] in base -medial or

base -final position: base -medially, in the Class 1 items 1 (also 2 -5),

and 164 (also 165, 167 -8, 170 -71, 177, 181, 186), the Class 2 item

29 (also 30, 34, 37, 42-3, 46 -7, 49, 51, 55 -7, 65 -7, 69, 71), and

75 (also 76 -7), the Class 3 items 2 (also 304), and 9 (also 12, 14,

17, 19, 21), the Class 4 items 8 (also 9), and 20 (also 21 -2), and

the Class items 12, and 20 (also 21 -2, 24-5, 28, 36, 43, 46, 48, 50,
5

54 -5, 57 -8, 62, 68-9):

 257

569 [ur]- melt (intr)

3623 [porik]- pick up small objects

176 [arij- (liquid) is strained

571 Lurid]- roll

407 [ir]- be (in a place)

312 [ ara]- make hoarse noise

2833 [ter]- pay (penalty)

557/559 [orad]- answer

3708 [por]- fight

191 [ ara] - grind with rolling stone.

Base- finally, [r] from these sources may be found in the Class 1

item 121 (also 122 -139), and the Class 2 item 18 (also 19 -27):

346 [ a: r]- become dry

2597 [ta:r]- descend.

therefore require the rule R5:

35. II -+ [ contact, flaÿ ped] (R5)

We assume that Lcontact, fla?ped] implies automatically that the

vocal cords are vibrating.

Further, since [r] is articulated on the alveolar ridge, we have

to extend the operation of R2 so as to include the following subrule:

36. [alveolar p. of a.] / (r)II, (R2)

where the notation '(r)' formalises the fact that both jr/ and /r'/ are

realised, singly, as [r].

ás noted earlier, when either of these elements is followed by /k/

in underlying structure, they are realised as voiceless dental stop

articulations; this would seem to be the case, for example, in the

Class 1 item 59 (which may have either /r/ or /r'/ at Coda); it is

given here together with the Class 2 item 77, to which it is related:
 258

4519 [bott]- fear, be frightened

[boradj- be thrown into confusion by fear.

In order to account for these voiceless dental stop realisations

in geminate clusters, we require the further subrules:

(r)IPB(r)II
37. B > i dental -ID. of a.. J/ (R2)
B(r)II-(r)II

B(r) __"B(r)II
38. II -> Lcontacti /
B(r)II4B(r)

/1 /: this underlies [11] in base -final position after a short

vowel, and Lli elsewhere. Base -medially, it is found in the Class 1

item 173 (also 179 -80, 184 -6), the Class 2 item 31 (also 36, 48,

54 -5, 68, 72, 74), the Class 3 item 10 (also 15), and the Class 5

i te::: 29 (also 33, 38, 52, 60) :

200 [alas]- (curry, rice) spoils

213 [ali]- dissolve (intr)

200 [ala]- crave

1504 [kola]- (plant) shoots against planter

Base -finally, it is found in the following items: Class 1 item 76,

Class 2 item 9 (also 10), Class 3 item 7, Class 4 ite-: 15, and Class

5 item 14 (also 15):

1969 [cell]- sprinkle

1772 [koll]- kill

3043 Lnill]- stand

1641 [gell]- win

2354 [co :l]- become tired

Note that in the last item it is phonetically short after a long root

vowel.
 259

For this element we formulate the following

realisation rule:

39. III - [contact, lateral air escape] (R6)

We assume that tcontact, lateral air escape automatically specifies

vibration of the vocal cords.

/1 /: base -medially, this underlies [1], in the Class 1 items

6, and 160 (also 161, 169, 182, 188 -9), the Class 2 item 33 (also

35, 45, 50, 53, 61 -2, 64, 70, 73), the Class 3 item 11 (also 20),

and the Class 5 item 23 (also 26, 37, 41, 44, 49, 51, 65, 70):

432 [el]- leave position

203 [alsk]- stir with ladle

426 [Ili]- descend

252 [ala]- measure

[ala]- sit.

Base -finally, it is realised as [11] (after a short root vowel) or

[1], as in the Class 1 items 77, and 141, and the Class 5 item 18:

2767 1tull]- make small jumps

341 a: l]- rule

1677 Lke:l]- hear, ask.

This element, like /1 /, also involves the realisation rule R6;

it is distinguished from /1/ by R2, which specifies prepalatal

contact with retroflexed tongue tip when B is marked with the second -

order feature r.

/+/: a preliminary statement of the way this element operates

has been given earlier; thus, it helps to form -BB clusters in base -

final position in the Class 1 items 22 (also 24-6), 61 (also 62 -4),

and 70 (also 71 -2):

DBIA 5 [age]- ( fire) consumes

 260

572 Luddj- rub

1025 lkabb]- seize with open mouth.

It has been noted already that the most important feature of the

behaviour of /r/ is with respect to the epenthetic vowel rule PIi3,

the operation of which is blocked when the element at Coda is /r/.

This is the contextual function of the second -order feature x in the

matrix of Figure 9 (p. 236). Since the feature x has no articulatory

effect in itself, /r/ is realised as the apico-alveolar tap when it

is followed by an element at Voc Suff, since here the feature x has

no relevance, and the specification of /3/ is in all other respects

identical to that of /r/; as for example in the Class 3 item 19

(also cited earlier):

3255 [ para] - crawl.

The form of the realisation rule R5 (35 above) already allows for

this, by not specifying the feature x at all; hence, whether or not

x is present, R5 (and of course R2) will apply, to identical effect

in each case. However, we have to supplement this with the following

subrule which realises the feature II identically to Ib when the

immediately following phonological segment contains I (i.e. not Ib,

Io or In) :

IN

bilabial p. of a.]

dental p. of a.J B

prepalatal p. of Br
40. II `contact, vocal cords a., retroflexed
vibrating, tongue -tip] x

postalveolar Bv
p. of a.]

velar ü. of a.]
J \-=F1i
/

if it is not
Simultaneously, the following element at Coda,
 Lbl

marked for voice already, is supplied with

this feature by R4:

41. I i Lcontact, vocal cords vibrating] / BIIxA()

li (R4)

Note that voicing is only introduced if the element at Coda is not

geminated (by following / * /).

Finally, /r/ also has to be set up in the verbal system for the

Class 3 item 3; this is because this item contracts a lexical relation-

ship with a particular noun form:

2526 [tar]- give (to 1st /2nd person)

L tandi] offer of marriage (to girl) ; noun.

Setting up /r/ for the common root here will yield, in the case of

the noun form, the sequence -/rnt/- which will be realised as elements

at the systematic phonetic level, and then simplified by a phonetic

mutation rule (ÄS7) .

/1 /: this element, like /r/, is defined by feature x,

specifying that it is not affected operation of the epenthetic

vowel rule. Consider the Class 2 item 11, and its related verb and

noun forms:

3059 [ní:d]- stretch out straight (intr)

id. (tr); Class 1 volitive verb

[ni:la] length; noun.

e can set up here the coúron root /ni:1/-, to which is affixed the

element /t/ (at Cons Suff, for the intransitive verb), additionally

/ */ (at Suffix2, for the transitive verb), or, instead of either of

these possibilities, the noun -classifying element /a/ (see Chapter 5).

This will result in the following sequences after all relevant phono-

logical rules have aT:plied: /ni:lt/-, /ni :ltt/-, and /ni:la/

+
(omitting irrelevant details such as boundary elements). These will

then be realised as phonetic elements; in particular, /1/ will be

 262

realised as [1], since the feature x has no articulatory function.

Finally the rule AS6 will simplify the clusters by deleting [1].

Exactly similar considerations hold in the case of the Class 1 item

141, the Class 4 item 18, and the irregular item 4; these are cited

here, together with their relevant related forms:

341 [a:l]- rule

La: ce] day of the week; noun

4402 [ba:l]- (child) lives, rule

[ba:li] way of life; noun

[ba:ce] act of living; noun

2002 [ca:1]- die

[ca:vu] corpse, death; noun.

-Je therefore require a further extension of the voicing subrule

in R4:

[obstruction] / BrIIIxAA
42. I - B() 1] (R4)
[contact, v.c.s vibrating] / BrIIIx

It will be seen that this extension closely corres_onds to what

was proposed for /ter/ above (41); however, /r/ is realised, not as

[b], but as [v], after /1/, and this is allowed for by the first part

of the subrule set out here. It is also clear that 'BrIIIx"A ' and

"B()1 ' function like a preceding vocalic environment in

BrIIIx C

respect of the voicing subrule in R4. The final form of this rule

gives the environment 'BrIIIx A ' first, and this allows the other

environments 'BIIx () and 'BrIIIx' IB() _' to be collapsed as

C

'0x4()

Finally, the form of R6:

43. III [contact, lateral air escape], (R6)

 C)

allows for the realisation of /1/ in all

environments as 1 (i.e.

identically with /1 /).

Since the last tuo items cited above each take the non -retroflex

form of the consonantal past affix, /1/ may be taken to be the con-
+
ditioning element involved in this phenomenon, and hence we must

also set it up in base-final position for the Class 4 items 17 and 19,

and the Class 5 items 16, 17, and 19.

At this stage, we may give the full form of the epenthetic vowel

rule:

(i) C1 + C2 C1 + VC2 (C1 í r 1 ;

C2 / k) (PK3(i))

e
a
a C
(ii) V (from (i)) (PK3(ii) )
0

i else

e may also now formalise the vowel length reduction rule:

45. LCV(C) CV(C) 3 (PM8)

This states that when an element at Coda is followed by either V

(from Voc Suff) in base -final position or VC (where the V is from the

epenthetic vowel rule, and the final C is an element at Cons Suff),

the length element /L/ immediately preceding the Coda element is

deleted (if it is present; otherwise, the rule applies vacuously).

Clearly, if the formulation of PM8 is defensible, then so is the

claim in PI13 that the epenthetic vowel is a phonological rather than

a phonetic entity. In this regard, notice that items such as

/ni:lt/- and /ni:ltt / -, cited earlier, show a Cons Suff element

following a Coda element without an intervening vowel. The crucial

point here is that the Cons Suff element in these items does not

trigger the operation of the vowel length reduction rule; only when
 264

the epenthetic vowel occurs before Cons Suff does PM8 apply. Of

course, it would in principle be possible to formulate PM8 as a

realisation rule, whereupon PM3 could be so formulated also; but

PM8 is probably more simply stated as a mutation rule. The most

important point here, perhaps, is that it is clearly undesirable to

account for the non-occurrence of long root vowels in (C)VCV,

(C)VS(N)C and (C)VCV(i)C bases by preventing the occurrence of

/L/ at Peak in those bases which contain Cons Suff or Voc Suff.

Not only would this complicate the statement of the relationship

between such items as:

II
1614 [ked]- be ruined: Class 4 verb
_ u to,
ruin; noun

by setting up the root variants /ket /- w /ke:t /-; it would actually

fail to account for the relationship between:

3059 [ni:d]- stretch out straight (intr)

Lni:t]- id (tr); Class 1 volitive verb,

and the other items cited above.

3.4 Cons Suff

This is represented by the system of elements in the matriz of

Figure 10, optionally preceded by the one term system N.

It has already been noted, in discussing the system at Coda, that

there are certain items in which it is possible that either Coda or

Cons Suff is represented; only those instances which unambiguously

show elements at Cons Suff will be dealt with here.

,' j this occurs preceded by /IMF/ in the Class 1 items 86 (also

87) and 187 (also 188 -9) :

2735 [dumb- become full

 59Z

g-

AI

aanY;çg tOl eu smaa.slis UO3 s J,T-r.c

ó8 ] qucto -f ai.Saa.â

uI a IT sosto ZuteloAut '/V koz-IA. .:tuo san000 aaoTaq reo-p,°oTouo:.¡d

coq.s +uauosuoo q.akq.) 'sz '`I() au,o, uozTaszlaaa aina a sa<¿ea auq.

IEsau ozuaAsotuot.i ut aoald jo uoTI.alnoT4sa q4.ut aq; Puzmolloj :4oú

Iazqlcy. d Jo i 71-

TaTuaT d Jo ra
laq.aladaad jo á :XL
a ' paxaljoaq.aa
917 li l Toaoruoo ' Mnlan paaanoT ' T4-ariöuoq. Ld

aaloablaq.sod
3 To fE

aalaA d jo ;E
 266

The voicing of the stop (at all places of articulation) after /N/

allows us to state here the following extended form of the voicing

sub rule in R4:

A N.
f()
' r_ \A J)

i
47. I -+ [contact, v.c.s vibratir. _ kThfir

/p/ also occurs with following 1*/, in the Class 1 item 171:

578 [uri.p]- play (flute) .

Final17, note that it is also part cf the source for the -BB

category of base-final consonants in Class 1, as discussed earlier:

when /r/ precedes /p /, [bb] is yielded.

¡t!: this occurs with preceding /T/ in the Class 1 items 83

(also 84 -5), 153 (also 154), and 159:

3696 [pond]- bounce

3944 [ma:nd]- scratch

S 249 [kuynd]- feel prickly.

It also accounts for the base -final element in the Class 1 item 169

(also 170):

725 [eld]- write

When preceded by /r/, it yields [dd] after a short root vowel, as

noted earlier.

/t /: this occurs with preceding /N/ in the Class 1 items 81

(also 82), 152, 157 (also 158), and 186:

2712 [tund]- break (intr)

2680 [ti:nd]- touch

2386 [navnd]- squeeze

2213 [curind]- shrink (intr)

 267

Base -finally (i.e. not followed by /* /), it accounts for [d] in the

Class 2 item 75 (also 76 -7), and the Class 4 item 20 (also 21 -2):

571 [urid]- roll (intr)

557/559 [orad]- answer

/c /: this occurs with preceding IN/ in the Class 1 items 79

and 151 :

3962 [miñj]- be left over

4130 [mu:ñj]- suck (penis) .

/v /: this occurs with preceding /N/ in the Class 1 items 78,

142 (also 143 -50), and 175 (also 176_85):

3537 [purig]- rot, fester

746 Le:/16- lengthen note (singing)

56 [adai g]- submit to.

It is followed by / */ ir? the Class 1 item 160 (also 161 -6):

203 [alak]- stir with ladle,

and not so followed in the Class 1 item 167:

2655 [tirig]- wander

:Te therefore extend the operation of R4 to include the following

subrule:

48. I [contact, v.c.s vibrating] / XO~(b)4C 4 (R4)

Finally, when preceded by /+/, base-final [gg] after a short

root vowel results, as discussed earlier.

/j /: this occurs just in the Class 1 item 168:

3262 [baraj]- grope.

 268

/g/, /o /, /t /, /o it has already been seen that /o/ has to

o
be set up as the underlying base -final element in the Class 1 item 1

(also 2-6):

569 [ur]- melt (intr)

We can utilise /I/ further in the Class 1 items 27 -9:

3658 [poriri]- (grain, stomach) swells

4096 [mu,- dive

3233/3306 [ parui ]- go stooping

In all these items, we set up the sequence -/No/ immediately after

the root vowel. /ko /, as noted earlier, has no articulatory function,

and we therefore extend the realisation rules R3 and R4 as follows:

49. C - [no articulatory effect] / Io (R3)

50. I -. [no articulatory effect] / o (R4)

However, the phonological rule H'17 operates on the sequence -/Nó/ to

yield -/ri/:

t
o

51. N (FM7 )
t
o

The resulting nasal consonants are geminated in base -final position

after a short root vowel by PM9 (the automatic gemination rule). But

when -/TIP is followed by / */ (at Affix2), it is mutated to -/kk/ by

PTi2 (which is required anyway, to account for the volitively related

forms of the Class 1 items 1 -6):

 269

52. (i) -> k/ + * (P 2(i))

(11) (:)Pi + * ->

pipi (= = p, t, t, c, (P1.T2(ii) )
, ,
k, r, r ;

Note that this accounts naturally for the volitive relationship

exhibited by the Class 1 item 28:

4096 [murir1 ]- dive

[mukk]- dip (tr) under water; Class 1

volitive related verb.

Setting up /o/ in this way also avoids the necessity of

including /ri/ at Coda and/or Cons Suff, just for the three Class 1

items 27 -9 cited above.

Consider also the volitive relationship involving the Class 3

items 6 and 8:

2670 [tinn]- eat (not rice)

[ti :t]- feed by hand; Class 1 volitive verb

1209 [ka:n]- see, seem

[ka:t]- show; Class 1 volitive verb.

These pairs of items suggest that (ignoring here the idiosyncratic

length alternance on the root vowel of the first item) -/Nt /, -/Nt/

should be set up in base -final position for the non-derived forms,

and we extend PK2(i) as follows:

53. t t + * (7:2(i))
ñ

Note that /n/, /n/ have to be set up on independent grounds at Coda;

however, apart from the Class 1 items 73-5:

1536 [kumm]- churn

 270

1634 [kemr ]- cough

]
,
2740 [timm]-
sneeze
[ tl -

there are no independent grounds in the verbal system for an element

/m/ at Coda and /or Cons Suff. We therefore derive /m/ via the

operation of P 7, along with /ri/ and the instances of /n/ and /n/

cited above (rule 51). Note that, since /t/, /to / and fig/ only occur
o
after /I,/ in the data under consideration here, we do not require

special realisation rules for them; they are in all cases mutated by

either R 2 or PH7, and never appear in a terminal phonological

sequence.

/T /: this is set up to account for the single Class 1 item 80:

1408 [kunt]- be lame,

which shows, unusually, a voiceless stop after a homorganic nasal.

/T/ is defined by the feature p (indicating that it is marked for

voiceless plosive articulation) in the matrix of Figure 10. When it

occurs immediately after the homorganic nasal element /N /, it is not

voiced by R4 because the feature symbol p intervenes between the

symbol I and the following environment ( MO, or ] ). Hence the

environment ' p' is included in the 'else' in the relevant subrule:

11\73 A () ()
_. )( T ) 3r1 ;,

[contact, v.c.s vibräting!

, ` C ,f
54. b (R4)
[contact] else.
il

¡r/, /r'/: these elements are set up here to account for those

cases of [r] which were noted in the discussion. of Coda, above, to

be possibly realisations of Cons Suff.

 271

/s /: this is set up for the single Class 1 item 140:

4000 [mu:s]- smell (tr)

3.5 Voc Suff

The matrix in Figure 11 shows the elements that occur at this

position. They are set up to account in the first place for all

Figure 11: The system at Voc Suff

base -final vowels: thus /i/ underlies base -final [i], in the Class 2

item 28 (also 29 -57), and the Class 5 item 34 (also 35 -62); /i/

underlies base -final [i] in the Class 5 item 63 (also 64 -71); and

/ a/ underlies base -final ;a]and [e] see the discussion below for

this alternation) in the Class 3 item 9 (also 10 -21), and the Class

5 item 20 (also 21 -33), and the Class 2 item 58 (also 59 -74), and

the Class 5 item 72-(also 73), respectively:

98 [ ani ] - wear, enjoy (jewels)

67 [ adi] - (lightning) strikes

, ..

69 [adi]- approach

312 [ara]- make hoarse noise

191 [ara-j- grind with rolling stone

73 ;,
(gap) is stopped

1927 Lcáyé]-
chew
+ [cávé]-
 272

In addition, we have noted earlier that

/i/ is to be set ulo at
this position for the Class 2 item 4 (also 5 -8), the Class 4 item

10 (also 11-14), and the Class item

5 5 (also 6-11):

1142 L kay-y]- (time) passes

1628 Lkeyy]- work

39 LayJ- send

This has the result that, at the time of the operation of the auto-

matic gemination rule, the element /c/ at Coda in these items is

not in base -final position, is therefore not ceminated, and is hence

lazed subsequently, by the realisation rule R4. A later, phonetic

mutation rule (IS5) deletes the Li]

We have also seen that a vowel is to be supplied at Voc Suff

in the underlying representation of the Class 3 item 2 (also 3 -4),

and the Class 4 items 1 (also 2 -7), and 8 (also 9):

407 Lir]- be (in a place)

375 Lid]- drop (tr)

2833 Lter]- pay (penalty.

For this we set up /i /, for the followinf reasons:

(a) it is one of the unmarked vowels in the system at this point;

(b) it occurs base-finally at the phonetic level in Class 5, but

not in Classes 2 -4; hence, setting it up in underlying represent-

ation base -finally for items of Classes 3 and 4 (the (C)VC base type

does not occur in Class 2) helps to rectify a skewness in distrib-

ution of this element. Notice that the other unmarked vowel /a/

is not a possible candidate for the items under consideration, since

it has a phonetic realisation at this position in the items of

Class 3, and therefore contrasts with whatever vowel is set up for

(C)VC bases;
 273

(c) it explains why the volitive

forms of the Class 4 bases 2

and 5 -7:

1614 Lkedit]- ruin (tr)

3190 [padit]- let lie falloff,

3191 [padit]- make suffer

4419 [budit]- make to let go

all show [i] as the second base vowel, even though only the last item

cited corresponds to the pattern noted earlier with respect to the

epenthetic vowel rule.

This requires that we supplement the general realisation rule

R9 as follows, in respect of the unmarked realisation of the feature

55. X - [back of tongue highest, lips unrounded] / (

ro
1-,
2
It will be recalled that, in discussing the system at Peak,

where /i/ is introduced under the third -order feature r, the environ-

ment given for this particular realisation of X was simply ' 2'

(for /a/). Note that, for the sake of typographical convenience,

the alphabetic transcription used in the text for systematic phon-

emes does not distinguish between the element under consideration

here and /i/ at Peak (i.e. /i/ is used in each case). Because they

occur in different systems, the risk of confusion is slight enough;

however, where it exists, it is avoided by referring to ' /i/ at

Voc Suff', or ' /i/ at Peak', as appropriate.

,fl

Since i/ at Voc Suff has no realisation in items of Class 3 and

4 when it occurs base -finally, we require two further subrules for

R9 and R10:

56. X -4 [no articulatory effect] / 1 1 [ot t, - f+tp] (R9)

 274

57. 1 -a [no articulatory effect] / X - o( p]

]]
1_0(t, (Rio)

The notation '[a( t, - o(p]' specifies Classes 3 and 4; for the

development of these class features (from II, under Class), see

below (pp. 276 -278).

Note that in this system of vowels /i/ is not taken as a type

of /i/ (as in the system at Peak), but as an element in its own

right. The justification for this is fairly clear from an examin-

ation of just the verbs, but the evidence in the language as a

whole is overwhelmingly in sunport of such an analysis. There is

full contrast between /i/ and /i/ at this position in the nominal

system also, as will be seen in the next chapter.

Concerning /a/ at this position, two points have to be made.

The first is that when /a/ occurs as a suffixal vowel, the vowel at

peak cannot be either of the elements /i/ or /u/: this is handled

by the phonological rule PIS 4:

i e
58. /_C+ a (PM 4)
u o

There is no restriction on Peak possibilities when the vowel at

Voc Suff is /i/ or /1/. Note that it would be satisfying if it were

possible to formulate FM 4 as a realisation rule or phonetic

mutation rule; unfortunately, there are just enough exceptions in

the nominal system to make this impossible.

Finally, /a/ is realised as [e] in the simple base forms and non -

past stems of Class 2 verbs, in contrast to the past stems and

volitive verbs of those items, which all show [a]. [e] also occurs

(ignoring the nasality feature) in the Class 5 items 72 and 73:

1927 [ cáyé] -
chew
[cáve] -
 275

2026 [taye]- level a field

PN5 handles all these cases, save for the alternative form [cavé] -:

a e / (' Í -t, -p].# non -past

59. (PN5)
a c ]J [+t, +p] #

Note that the environment ' a c ]] [ +t, +p]# ' is stated in such

a way as to allow for /a/ to be realised as [a] in the Class 5 item

27:

1809 [koya]- string (flowers)

Regarding the alternative form [cávé] -, /e/ would have to be set up

in the system at Voc Suff for those idiolects having this form.

3.6 Suffix2

As noted earlier, the element at this point in the base

structure is /*/; it occurs only in Class 1 verbs, and is set up in

the first place to account for the regular voicing contrast between

stops occurring base -finally after a long root vowel or after the

second base vowel (introduced by PIN 3). It occurs in the following

Class 1 items: 88 (also 89 -92), 93, 97 (also 98-9), 113 (also 114 -7),

120, 155, 160 (also 161 -6), and 171:

593 [u:k]- comb

1352 [gi:c]- make a scratch

1566 [ku:t]- sharpen

349 [a:t]- become mature

1341 [ci:p]- comb

1973 [cavt]- step on, kick

203 [alak]- stir with ladle

578 [urip]- play (flute)

 276

It turns out to be the case that

/ */ from Suffix2 is not important

for stating relationships between

verbs, nor verb -noun relation-

ships; this is the main reason

why it is distinguished from /*/

under Volit1 (see below, p.278 -281) .

3.7 I and II

Discussion of these elements (from Class by rules

LS 3 and 4)

has had to be postponed until this point,

since they embody features

that can only be clearly understood in terms of the

particular
phonological system used here. However, the rules 60 and 61 follow
on directly from 11 above. They expand I and II as feature -pairs

(set off in square brackets), and the set of possible pairings

specifies the verb classes 1 -5.

60. (i) I - [i] (LS 12(i))

(li) [-p] (LS 12(ii))

61. (i) II - L+ti (LS 13(i))

(ii) L±t] -4 [±p] (LS 13(ií))

The table in Figure 12 shows Classes 1 -5 of Appendix I and

Classes I and II of the generative component, and how the feature -

pairs correlate the two systems:

Class in Past Non -past Classifier

Appendix I affix affix element

1 vocalic [i] lax [-p] I

2 lax [ -t] lax [ -p]

lax L -t] tense [ +p]

II

4 tense [ +t] lax L -p]

5 tense L +t] tense [ +p]

Figure 12: The functions of the Classifier elements I and II

 277

The features are set up on the assumption

that the underlying
forms of the syntactic tense affixes are /i/ (the
vocalic past

affix), /nt/ (the consonantal past affix), and /p/ (the non-past

affix). Verbs of Class I (Class 1) all show the vocalic past affix,

and the lax form of the non -past affix; Class II verbs (Classes

2 -5) all show the consonantal form of the past affix, and within

this group there are four sub-groups, depending on whether the form

of the past affix is lax ( [ -ti ), or tense ( [ +t] ), and whether

the form of the non-past affix is lax ( [ -p] ) or tense ( [ +p] ).

The [ +] values for the features set up here are handled by rule in

the following way: the sequences

a /...[ +t, +p] # nt.../

b /...[ +t, +pj p.../

represent what may occur at the lexeme boundary ( # ) between a

verbal lexeme and a following past or non-past affix. In the case of

a following past affix, the feature [p] is irrelevant, and hence is

assigned +j
[ values in (a) above; when a non -past affix follows, the

feature [t] is irrelevant and is likewise assigned [ +] values in (b).

A rule will then rewrite the past affix -/nt/ as -/tt/ in the

environment of a preceding [ +t] feature, and the non-past affix

-/p/ will become -/pp/ in the environment of a preceding L;-p]

feature. For minus values of the relevant feature, no change in the

form of the syntactic tense affixes will be effected, and -/nt/ will

yield a homorganic nasal plus voiced stop articulation, and -/p/

will be laxed to a bilabial semivowel. Further details are outside

the scope of this study.

At this stage, the conditions 62 and 63 apply:

62. NOT: #í ... Volit L+t, +p] (LS 14)

1
 278

63. NOT: #... Volit2 ]] Í,,+tj ...# (LS 15)

62 states that Volitl may occur with verbal bases of all classes

save Class 5; 63 supplements this with the statement that Volit2

may only occur with verbal bases of Class 2 (see above, p. 205).

3.8 Volit
1

All the forms of this suffix show the element /*/, which is

identical in form and phonological function to the element represent-

ing Suffix2, but their respective sources have nonetheless to be

kept distinct. The justification for this lies in the fact that /*/

from Suffix2, as noted above, is found only with Class 1 verbs, and

serves to mediate no lexical relationships, while / */ from Volitl,

is generated within verbal bases of Classes 1 -4 and operates on

these in such a way as to derive Class 1 volitive verbs. The

distinction between the two sources for /*/ in the verbal system,

however, does not obscure their similarity of operation, or the fact

that the verbs in whose base structures they operate are all of

Class 1; the following rule states, in effect, that /*/ from Volitl

can only occur in this class:

64. # ... Volitl 3 [+ t, + p]# #... volitl I Li, -p]# (L816)

The element / */ is the sole representation of Volitl in most

cases; but where the preceding element is a vowel at Voc Suff, a

consonantal element is introduced, by the phonological rule 65, as

a 'vehicle' for the following /*/:

65. V+ *-' V C+ * (pm (i))

 279

66. C (from 65) (P1,11 (ii))

"{PA
t -
elsewhere

67. i i/ o (from 66) (PY1 (iii))

Rule 66 then determines the introduced consonant as /p/ or /t /,

depending on the preceding vowel; 67 also adjusts the -/ip/ sequence

to /ip /. These rules operate in the following instances; the

Class 2 item 64 (also 68, 73, 74), the Class 3 items 4, and 13 (also

18 -19), and the Class 4 item 2 (also 5 -7); and in the Class 2 item

30 (also 33, 45, 51):

2551 [tale]- become weak; Class 2

Ltalat]- make weak; Class 1 volitive verb

4311 [bar]- come; Class 3

Lbarat]- cause to come; Class 1 volitive verb

929 [kada]- cross; Class 3

[kadat]- take across; Class 1 volitive verb

1614 [ked]- be ruined; Class 4

Lkedit]- ruin; Class 1 volitive verb

265 [ari]- find out; Class 2

[arie]- inform, tell; Class 1 volitive verb

(See Appendix II.)

Note that, in the case of the Class 3 item 4311 [bar] -, we need

a fourth subrule within PM 1, 68, in order to state that the second

vowel of the derived base is /a /, not /i/ (as set up for the non-

derived base):

68. i -a a / t * ]] [ -t, +p] (PM (iv) )

Finally, the operation of Volit1 in all other cases is shown in

 280

L.ppendix II, and summarised in the tail of Figure 13:

1
3

-Pr!

-tt
n

-CC

-t -tt

-rr

-r'

-kk

-tt

-17

- att
It
-attl-Itt

Col= 1: elements at Coda/Cons Suff/Voc Suff prior to the operation

of the phonological mutation rules.

Column 2: elements in base-final position in the terminal seauence

(Volit not occurring).
1

Column 3: elements in base-final position in the terminal sequence

(Volit has occurred, andIM2 has apnlied).
1

Figure 13: The operation of Volit1

There is no phonetic correlate of /t/, of course, is the Class 1

 281

items 1 -6. Note that the Class 1 item 127 and the Class 2 item 23

show distinct derivations from comparable underlying structures:

ft
2312 [ke :r]- go up to house; Class 1

[ke:t]- take (person) up to house; Class 1

volitive verb

2380 [ne :r]- rise up; Class 2

[ne :t]- lift; Class 1 volitive verb.

In the first case, we require a phonological rule of the form:

69 r'r' rr (PM2 (iii))

after PTi2 (ii) has operated (to yield /r'r' /) (but before the real-

isation rule R9, which accounts for the retracted articulation of

/e/ before /r' /; in the latter case, however, /r'r'/ remains up to

the terminal sequence, and R9 determines the appropriate retracted

quality of the preceding vowel, while /r'r'/ is realised as [tt].

It should be noted that a number of speakers have the fora [ne:t] -,

'lift'; in general, the present tendency seems to be in this

direction, whereby the phonological rule P12 (iii) is becoming more

and more important for the description of the speech of the younger

generation.

Finally, the Class 4 items 17 and 19 are irregular in respect

of the base -final consonants of their voice -related forms:

723 [e:l]- get up (intr)

[eppl- raise up; Class 1 volitive verb

4457 [bu:l]- fall

.

[bu:k]- fell (tree), pour; Class 1 volitive

verb
 282

3.9 Volit2

' This has its effect solely on the features that determine the

form of the syntactic tense affixes; see LS 17 (and Appendix II):

LS 17 0 .. . Volit2 [-t, -p]* -+ *.... Volit2 [+t, +p]*

This accounts for the fact that all bases formed by the operation

of Volit2 are of Class 5.

The Class 2 item 35 and the Class 4 item 17 are irregular in

that they also show mutation of base structure:

548 [ oli ]- halt for night, dwell

[oy]- exhaust; Class 5 volitive verb

,
723 [e:l]- get up (intr)
fl,
raise; Class 5 volitive verb.

The Class 2 item 63 is also irregular, in respect of the base -final

vowel of the derived form (Li] instead of [a]):

2460 [tade]- be obstructed

[tadi]- obstruct; Class 5 volitive verb.

4.0 The rules cited in this chapter

4.1 The lexeme structure rules

L31 Lexeme - Q Basel Class(ifier)

LS2 Base -+ scot ( +Affix1) ( +Affix2)

,Verb

LS3 Class oun

etc.

L54 Verb

Root -# (Onset) 1;ucleus

LS5
 283

£S6 'Nucleus -+ Peal,: (coda)

ç oc ff
if Tx a
e Æna S ff
S ix
2a= Aff =7 a
voice 2f2

tZ9 Voice Suffix ®

701 t,

îS1O NOT : Peak æ

tq 4 2OT: Suff if7 + TT

ts1 2 (i) I ® [i]

(ii) [£] a [-

L.:1 3 (i) II -- [+ i)

(ii) [+ t] ® [/ 2}

1514 170T: @,,. 7o1iti ]

3
T 5 1:0T: ¥ ,,. 7olitp

131 ª + ... 7cliti \[/ t, + 2}¥ ,,. 70li t / [i, - 2ª

 284
..1 n_ _....
1-onolo4 cal mutation rules
,q
Pm1 * VC #
-i-
C
/ i
(ii) (from (i)) ,
t else.
(111) 1 -a 1 / U (from (ii) )
11
(/ iv) 1 a a / t + x
Ti [- t,
'
2 \ -4 +
o
tn l
P.P. (2 = r, t, t, c, k, r, r')
(iii) rt7`t rr (optional)
+ VC (C r £,
C1 2 1; C, ó)
+
V (fr37. (i)) -
e
 285

IcTT(('-)J) 1] -> cV(()c) !1

IT -4 TT
/ (
c) , (erase T )

- c.c./ r ( erase ] )

Z The ralisation rules

,
s¡
,
{ ir3)¡)A(L)4_ IA X ( )

., ^ TA, jr
--- Sl
)()
B1 -, I.
bilabial P cf
lip
compression] else.

,[palatal 7. of a-] / X04(1))4 jI 1

[ nos talveolar p. of a.] / _.___j

(r)II" B(r)II
[dental p. of a.: ,'

B(r)II4 (r)II

?2 [alveolar p. of a..! / (r)II

[ prepalatal r. of a., retroflexed

tongue tip] / r

[postdental p. of a.] / IV

[dental n. of a.] else.

[no articulatory effect] / Io

{j
[velar p. of a.] else.
 286

[no articulatory effect] / o

YiOA(T )A-1
/
[occlusion]
,q)4.1
_ A:1.4)
, ,

[obstruction] / {
( X( )"(y ,
"

BrIIIx A

[contact, v.c.;, vibrating]

[contact, velum lowered] / n

[contact] else.

B(r) ___,B(r)II
4.1

B(r)II4B(r)
1
bilabial p. of ae]

dental p. of a.] B

jr
prepalatal p. of a.,
R5 II [contct, retroflex tongue tipi
v.c.s
vibrating, Bj
postalveolar p. of a.!

velar p. of a.

`[contact, flapped] else.

R6 III -+ [contact, lateral air escape]

a7 IV -+ [occlusion, groove friction]

_
bilabial p. of a.] A

dental r. of a.]

T
[contact,
7repalwtál p. of a., / I
R8 -4
retroflexed tongue tip]
velum
lowered, postalveolar p. of a.]
C
velar p. of a.]
 287

I
[no articulatory effect] / __--1
] [ cc t, -a p]

(lips rounded] jO^(Lr3r

i W

/[back of
R9 X OP_..._jOA(L)4Br
tongue 43(1
highest, C

lips unrounded]
jr

f'21
[front of tongue highest, lips spread] / j

[no articulatory effect] / X f [c: t, - a p]

R10 1 [constriction, velum lowered] / n

[constriction] else.

`velum lowered]] W n
ir(midopen tract,
- `'(r)1
w
R11 2-
velum lowered] / v n
[open tract,
1[
- J i :-
R12 L [v.c.s continue to vibrate, unaltered tract]

4.4 The articulatory sequence rules

occlusion occlusion

obstruction obstruction

constriction constriction
Aá1. Add: [velum lowered] /
midopen trac 1r,
open tract

t_open tract
r
velum lowered
 288

4P.

contact obstruction contact

placet of constriction placei of

AS2. Add:
articulation, articulation
open tract
velum lowered
velum lowered

bilabial
AS3. (i) p. of a.
-4 lacei of
dental
J. of a. Or
ar t is ul a t ion 1 ~1
contact, contact,
IMP IMP

rrepalatal
p. of a. placei of
(ii) -+ ilabial p.of a]
retroflexed articulation
velum
tongue tip lowered

OP

(iii) dental p. prepalatal p. / contact, contact,

of a. of artic.,
preralatal
retroflexed of a.,
tongue tip velum
velum lowered
lowered
Obi

(-
r I
back of tongue constriction obstruction contact
highest,
ridopen trac
AS4. lips unrounded,
Erase: oren tract
le.
constriction to

open tract

constriction bstructior_ / obstruction 'constriction

AS
5 front of alatal D. midopen
(i) tongue f a. tract
highest,
open tract
lips spread
J

constriction, obstruction

Erase: front of tongue

o
highest,
e

lips spread
ONO
 289

1
contact,
flapped,

alveolar
p. of a. constriction contact

midopen tract

AS6.
contact,
lateral air open tract r01rT
(i) grase: escape,

v.c.s continue [Contact]

dental
to vibrate,
1. of a.
unaltered
tract
prepalatal
p. of a.,
retroflexe [open tract :1

tongue tip
L
0

contact, obstruction / open 'tract obstruction,

lateral ai --

escape bilabial bilabial

n. of a. n. of a.

e
(iii) v.c.s contact, contact, constriction
dd ; vibratir_.^
placei of placei of midopen tract
articulation, articulation.
open tract
velum
lowered
t.

(:B. 6(iii) must precede 7, since otherwise the non -past and past stems

of Class 5 verbs which show nasalised vowels will be voiced.)

v.c.s continue 1
to vibrate,
unaltered tract
contact .1)contact, t

riS7.
Erase: /
onstriction
velum
midopen tract lowered i
r
open tract
 290

obstruction,

[itsotf tongue open tract

Llatal p. of a.

rotruction,
bilabial p. of a back of tongue
highest
 291

CHAPTER FIVE

The Structure of Lexical Items: Nouns

 292

1 .0 Introduction

Having proposed a 'root -and -affix' analysis of Coorg verbal

lexemes, with systems of phonological elements operating at positions

defined within the root (Onset, Peak, Coda) and affixes (Voc Suff,

Cons Suff, Suffix2, Volit1, Volit2), we turn now to consider the

class of nominals. Essentially the same sort of structure is found,

but there are a number of important differences, which are noted in

the following three sections.

1.1 Lexeme -final elements

The nominal lexemes have to be set up in such a way (the reasons

for this are considered in detail, below) that they show a final

element which is one of the set:

, e, i, u, a, a(1), e(n)]

We are here ignoring the nasal -oral contrast between vowels of the

same quality; as was found in the verbal system, this contrast is

marginal, and does not affect the validity of general statements

concerning the structure of lexemes. The elements [1] and [n],

parenthesised in the above forms, appear only in certain phono-

logical environments. We will generalise these final elements as

-V(C). Just a few nominals show final -[y], which may in careful

speech be followed by [i]; this is handled by a late rule (AS 5)

which may delete Li] in this environment.

It will be seen below that there are numerous asymmetries in

the distribution of these final elements; it suffices here to

point

out that the occurrence of V(C) as a regular part of nominal lexeme

lexemes. There,
structure contrasts with what was found in the verbal

all items in Class 1 show an obligatory final consonant, along with

certain items in Classes 2 -5; rather less than one third of verbs show

a final vocalic element.

 293

1.2 Contrast of consonant length after a short root vowel

It is a principle of verbal base structure that any base -final

consonant occurring immediately after a short root vowel is automatic-

ally geminated. The few exceptions prove to be only apparent

exceptions, where the consonant in question is base -medial in under-

lying representation; cf. the Class 1 item 6, the Class 3 item 2, and

the Class 4 item 1:

432 Leu- leave one's position

407 Lin- be (in a place)

375 j,id]- drop (tr)

It was further shown that all base -medial consonants are single, and

lax. In this regard, it is an especially interesting problem that

nominal bases show a regular contrast of consonant length after a

short root vowel, as in the items 60/243, 140/347, and 95/276

(Appendix IV, pp. 459 -99 ):

4524 [boli] light

4524 [bolli] silver

2473 [tani] coolness

n
3299 [panni] frui t

1897 jade] cobra's hood

1910 Ldadde] dull person (fem)

of the nominal base

Clearly, if we treat the final vowel as a part

in all these instances, we have to say that the second

principle

on the
mentioned above cannot apply to the structure of nominais;
have to
other hand, if we treat it as extraneous to the base, then we

However, since each of

say that the first principle does not al ply.
both prove equally valuable
these Principles complements the other and

is not possible to decide a

in the description of verbal bases, it
 294

priori which of these alternatives, if either, ae

is appropriate.

shall return to this problem below (section 3.0).

1.3 Canonical forms

Nominal lexemes show a much greater variety of canonical forms

than do verbal lexemes. It is true to say, however, that the majority

of nominal lexemes show forms that closely correspond to those found

in the verbal system, except in respect of the final -V(C) element;

thus (marking off the final element by + ):

(C)VCC +V(C)

(C)VNC +V(C)

(C)VC+V(C)

(C)VNC +V(C)

(C)VSC +V(C)

(C)VSNC +V(C)

(C)VCVC +V(C)

(C)VCVNC +V(C)

In addition, the type (C)VC +V among the nominals corresponds exactly

in its canonical form to one type of verbal lexeme; with this nominal

type, the final elements [a(1)] and [e(n)] never occur, and hence we

generalise the final element as just V.

But there is a substantial residue of other, more complex forms

to account for in the nominal system (a fuller discussion of these

follows in the next chapter):

(C)VCCVC +V(C)

(C)VNCVC +V(C)

(C)VCCVC +V(C)

(C)VNCVC +V(C)

(C)VCVC +V(C)
 295

(C)VCVNC +V(C)

(C)VCVC +V(C)

(C)VCVC +V(C) etc.

For certain of these, the canonical form covers only a handful of

instances, and this tendency increases proportionately

to the

complexity of the canonical form. Note also that the more complex

forms violate another 2;rinciple of verbal base structure, namely that

only root vowels may occur with the length element /L`, and may only

occur as the first vowel of the base. For nominals it would seem,

prima facie, that length has to be allowed for with respect to other

than root vowels (maintaining that only root vowels may constitute

the first V of the base), or that root vowels may occur as other than

the first V of the base (maintaining that only root vowels may occur

with length); or we could abandon both parts of this principle.

2.0 Preliminaries to the analysis of nominal lexemes

In what follows in this chapter and the next, an attempt is made

to describe Coorg nominal lexemes (particularly their base structure)

in a way which is consistent, as far as possible, with the model

established in Chapter 4 for the verbal lexemes; while at the same time

not Join: violence to the structure of nominal lexemes considered as

elements within their own system. In general, it turns out that,

apart from certain important modifications to the generative base

component of the lexicon, these two aims are complementary rather than

competing, and that occasionally a constraint on lexeme structure set

up on the basis of the verbal system leads to the most natural descrip-

tion of nominal lexeme structure.

In this chapter, the problem of defining the nominal base is taken

up first; then its structure is analysed in some detail; and finally

 296

the lexical relationships that hold,

both between nouns and between
nouns and verbs, are set out as fully
as possible. We reserve consider-
ation of the complex nominal lexemes for
Chapter 6.

2.1 The nominal lexeme

We define the nominal lexeme as that part of the surface noun form

which cannot be described in terms of closed sets of syntactic elements.

In Coorg, there are three such sets: the first comprises the system

which expresses the syntactic category of number; the other two express

the category of case, the first consisting of a set of inflectional

affixes, and the second comprising a number of noun particles. Since

the noun particles occur only with certain inflected forms of nouns,

they are not relevant to our purpose here. The nature of the number

and case systems is briefly outlined in the next two sections, in order

to prepare the ground for a definition of the lexeme in practical terms.

2.2 Number

This system handles the opposition between singular and plural

nouns. It may be considered either as an optional system, consisting

of the single term -/kal/- marking plurality, or as an obligatory two

term system, comprising -/kal /- and a zero element. Plural nouns are

marked by affixing -/kal /- to the lexeme, thus forming the plural

stem; forms which do not show -/kal/- constitute the singular stem

(which, if it also has no case affix, is therefore phonologically

isomorphic with the lexeme; see the preceding section). The various

case affixes are added to the singular or plural stems. The phonetic

realisation of -/kal /- is determined with reference to the final

element of the preceding lexeme; the table in Figure 1 shows the

possibilities at the systematic phonetic level:

 297

Lexeme -final Plural affix

element

-i

-Y -ya(1) -

-e

-u2 -va(1) -

"3
-i xa(1) -

-a(1)4 -yVya(1)-

-ga(1)-

Figure 1: The forms of the plural affix

1
After these elements the initial consonant of the plural affix is

palatalised and lazed to ryj -.

2 After u] the initial consonant of the plural affix is labialised

and lazed to -Lv] -.

> After -J] the initial consonant of the plural affix is lazed to a

velar approximant. Certain idiolects lose the [i1] and following

aprroximant.

4
The final -L1] is lost before the following consonant of the plural

affix; this consonant is palatalised and lazed to -Ly] -. Note that

this set of nouns has not been grouped with the first three showing

plural forms in -Lya(1)j- because it will prove convenient in the

rest of this chapter to discuss the sets in the order illustrated

here.

5 The final nasal is assimilated in place of articulation to the follow-

ing consonant of the plural affix; this consonant is assimilated in

voicing to the preceding nasal.

 298

It is important to note that any

nominal which has an inanimate

object as its referent, no matter what lexeme-final

element it shows,
may not occur with the plural affix. Hence, the occurrence of -/kal/-

is a purely referential matter, and number in Coorg is a 'natural',

not a 'grammatical' phenomenon (this statement will be modified

slightly later on). The other basis on which number is established

in Coorg is pronominalisation; nouns with singular referents pronom-

inalise as (illustrating with the 'remote from speaker' forms):

.. n
Lave(n)] (mast)

[ava(1)] (fem)

[adi] (nonhuman anim; inanim)

while those with plural referents pronominalise as:

[ aynga (1) ] (hum)

L avu] (nonhum anim)

Ladij (inanim)

(Note that in each of these sets the pronominal form for inanimate

reference does not mark distinctions of number:

n
Ladi mara]

may mean 'that's a tree' or 'those are trees'.) Pronominalisation

also is therefore a wholly regular, referential matter.

Having said this, however, it is necessary to point out that

there are just a few nouns which are to be described as 'inherent

plurals'; items 423 -4, 426, 428, etc.:

283 Latta(1)] cattle

401 [ibba(1)] two persons

2435 [takka(1)] chief family

3768 Imakka(1)] children

the -[a(1)] element is

Regarding these, it will be argued below that
not as a realisation of the
to be treated as part of the lexeme, and
 299

syntactic affix -/kal/- (to which it is diachronically related).

This means that our statement above about the purely referential

nature of number in Coorg is not able to cover these few items;

they are plural in referential terms, and pronominalise as plurals,

but do not occur with the syntactic plural affix -/kal / -. However,

the situation is otherwise straightforward, and there is nothing

here to constitute even the beginnings of 'grammatical.' number; we

should talk of 'lexical' number instead.

2.3 Case

This is a variable-term system, depending, like pluralisation,

on purely referential factors. If the noun referent is an inanimate

(including vegetate) object, the case system consists of six inflection-

al affixes; if not, it consists of only four. The situation is shown

in Figure 2:

Stem-final
element :,omin. Accus. Gen. Locat. Ablat. Dat.

n n
Singular ¢ -na -Da -li ñji -ki

tt tt tt rr

-tina -tira -till -tie n« -ki

V -a -da -ki
If
-a -da - gi

Plural Z a(1)] -a -da -ki

Figure 2: The forms of the case affixes

 300

With nouns referring to humans, the locative and ablative affixes

are not possible; with nonhuman animates they are occasionally found,

or at least elicitable, but are considered unusual. It should be

noted here that only case affixes come under this statement; locative

and ablative functions are perfectly Possible with nouns denoting

animate beings, but are expressed through the system of noun particles.

The various forms of the inflectional affixes at the systematic

phonetic level depend on the preceding environment; hence, there is

only one declension in Coorv, with four subclasses within this - each

of which is fully determined with reference to the final element of

the stem. The nominative forms show no affix; the simule singular or

plural stem represents the nominative noun form. Note that the

locative and ablative affixes do not occur with stems showing final

-[e(n)j or -[a(1)]. This is because all such stems have animate

referents; this point will be discussed further below, when the matter

of gender is dealt with. It may be noted here, however, that what is

true of stems of nouns having animate referents is automatically true

also of plural stems; this follows from the fact that only nouns with

animate referents may take the plural affix. The accusative case

affix is La], preceded by an infix -[n]- when the stem -final element

is one of the set Li, e, i, u, a]; in the case of stem-final [a],

moreover, this infix is preceded by another, Lti] -, which also occurs

in the genitive, locative and ablative forms of this class of stem.

The presence or absence of these infixes, together with their form,

is entirely predictable with reference to the stem -final element, and

hence they form part of neither the stem nor the case affixal system.

The genitive case form is -da] or -,raj; where -,Da] is entered

in the table in Figure 2, it indicates free variation between these

 301

alternatives. Only Lra] may occur after the infix -Lt j-, and
only

-[da] after stem-final -La(1)] and ,e(n)]; in the latter case, the

nasal consonant assimilates in place of articulation to the following

retroflex stop.

The locative and ablative case affixes are represented everywhere

by -[li], -Lñji], respectively, preceded by the infix -[ti] in the

case of stems with final [a].

The dative case affix is realised as Lki] everywhere, except

after stem -final -[e(n)]; the nasal consonant assimilates in place of

articulation to the following velar stop, and the stop assimilates in

voicing to the preceding nasal.

2.4 Gender

Gender in Coorg is established on the same sort of basis as

number, namely the facts of pronominalisation and pluralisation. The

third person singular and plural proforms comprise the following

system:

Singular Plural

A.
Eas cui in e ave(n)
ayñga(1) Human

Feminine ava(1) .:mimate

avu

Nonhuman adi Inanimate

Figure 3: The third person (remote) proforms

That is to say, there is a three -term system in the singular (masc-

uline, feminine, nonhuman) and another in the plural (human, nonhuman

animate, inanimate) , where the only isomorphism between the two

 302

systems lies in the fact that the masculine and feminine terms
(in

the singular) jointly exhaust the same ranee as the human (in the

plural), and that the nonhuman animate and inanimate terms (in the

plural) jointly exhaust the same range as the nonhuman (in the

singular). The feature 'animate' (in the plural) excludes vegetates

as well as inanimates: only nonhuman animals may be pronominalised

as Lavu] (with insects as a problematic border area).

As noted above in discussing number, what we are dealing with

here is apparently a purely referential phenomenon. The state of

affairs may be represented by the feature matrix (ignoring

distinctions of number) in Figure 4:

aynga(1) aynga(1) avu

.rn ,i adi
ave(n) ava(1) adi

Animate + + + -

Human + + -

Masculine + -

Figure 4: Feature representation of gender distinctions in

third person proforms

Given this gender system, we have now to enquire how far it is

reflected in the formal properties of the noun forms of the language

(i.e. in the nominal system, as opposed to the pronominal forms).

The answer is clear; the gender features 'animate', 'human' and

'masculine' are marked fairly extensively in the nominal system, and

the relevant part of the noun form is the lexeme -final element -V(C):

the situation is shown in the table in Figure 5:

 503

Lexeme (C)vCC+ (C)VC+ (C)VCVC+ (C)VC

-final V(C) V(C) V(C) +V Nasc Hum _.nim
elemen.a,

1 ajji mu:di gavdi - - ,

-i '
2 palli e: ri karadi nari (:- ) -+ +
¡
3 dadde cu:le ( ) --

-e
4 enne ka:ke kudire age

5 ubba a:ka kadaga mara s' - --

-a .

6 okl:a pa: pa karapa mona

!1 II fl ,I

-,.. 7 etti to :di eraki iTi --

-u E payru ca: vu koravu kuru

II II fl

It
ajje(n) me:de(n) gavde(n) + + -

-e(n) fl ,1 It 11

10 akke(n) ko:de(n) kurike(n) - + +

11 obba(1) mo:va(1) mayma(1) - + +

-a(1) . }

12 atta(1) ( ) ( i +

Figure : The representation of gender in the nominal s stem

The first column shows the lexeme -final elements, on the basis of

systematic phonetic contrast (here again, though, the nasal versus

oral distinction between vowels is ignored). The next four columns

show the canonical forms which account for the majority of the lexemes,
verbal
and which correspond closely to those already recognised in the

system. However, base -final -NC has not been distinguished, and

(C)VsC- is not distinguished from (C)VCVC -. The last three columns

terms of the
indicate the properties of the referents of noun forms, in
Thus, a - value in the
gender features we have recognised for Coorg.
 304

'Kasc' column indicates that all the nouns entered in the correspond-

ing row have female referents; similarly, a + value in the 'Hum'

column indicates that all the nouns in that row have human referents.

:;here + values are assigned together, this means that a number of the

nouns represented in that row have referents which possess the

relevant property, and a number of others do not. Finally, notice

that, in the case of an item such as 218:

1371 [kuññi] (Coorg) child

the + masculine properties of the range of referents is irrelevant

to the lexical meaning of the item; this situation is distinct from

that where + values have to be entered simultaneously, and it is

indicated in the table of Figure 5 by a blank in the relevant column.

It will be seen from the table that there are a number of nouns

showing final -[i], -[e], [a(1)] which have specifically human

female referents; these are represented in the table by items 207,

554, 791, 276, 583, 425, 779, and 814:

46 [ajji] great grandmother

S 806 Lmu:di] girl

[ gavdi ] milkmaid

1910 [dadde] dull person (f em)

2262 [cu:le] dancing girl

834a [obba(1)] one person (fem)

3768 [mo:va(1)] daughter

4189 [ mayma (1) ] cross -niece

It should be noted that, apart from the person /number proforms, which

are outside the nominal system proper, the canonical shape (C)VC +V

is not represented in this set; this is apparently a systematic gap,

since no nouns of this form show gender Parking. This is an important

point, which will be taken up in more detail in the next section. By

 305

contrast, the lack of an entry for type (C)VCVC- with final

-Le]

is probably an accidental gap (marked by parentheses), due to the

relative infrequency of this canonical shape.

However, a large number of nouns, including those of the form

(C)VC +V, show final -Li], -[e], and have referents of which some are

inanimate, others are nonhuman animate, and the rest are human;

among these last, there are no specifically male or female referents,

and for this reason no entry is made in the masculine column for

either of the rows concerned (the second, and the fourth), except for

the parenthesised + entry in the second row for the single item 829:

3623 [poriki] mean fellow

The fifth row represents the large majority of nouns with lexeme-

final -[a]: this group shows only inanimate referents. The sixth row

represents a smaller group showing animate or human referents; among

the latter, the male /female distinction is irrelevant, and most of

these instances are accounted for by the (C)VC +V type in this row;

here also, therefore, the (C)VC +V type may be treated as not marked

with respect to the animate/inanimate distinction. The next two

rows contain nouns with final -[i], u]: these final elements are

actually in complementary distribution for the most part, as [payyu]

n rr
is the only example of its type ([etti] shows the regular pattern),

[ca:vu] represents a small group of nouns all with the sequence

-[a:vu] which renders the final -[u] predictable, and [koravu] is the

only example of its type, where the -[u] is again predictable. Only in

the (C)VC +V form of nouns is there contrast between Li] and -[u]; but

for presentation of the data here the two rows have been kept separate

across all four canonical types. The marginal contrast between -[i]

and -[u] will be taken up in detail in the next section. Row eight
 306

represents relatively few items, and covers a referential range

which includes inanimates, animates, and humans; row seven in

addition covers male and female referents, as in items 357 and 616:

3608a LDonnij wife, female

342 La:ni] man, male

Rows nine and ten contain the class of nouns with final -[e(n'),
' ].

All have animate referents, and a large number of these are

specifically human males (row nine). In addition, a proportion refer

to nonhuman animals, and a very few to human females; these last two

groups are entered together, in row ten. Note that -Le(n)j, which

seems to be a marker of essentially human male reference, and only

seconaril:.- a marker of animacy, cannot occur with nouns of (C)VC +V

form, even where the referent is an animal rather than a human.

Finally, row twelve contains that group of nouns, among those with

final La(1)j, which have plural referents; only a few lexical items

are involved.

By this analysis, there is a marked as well as an unmarked

function for the elements -[i], e], and -[a]; a marked as well as a

relatively unmarked function for - Le(n)]; two distinct marked

functions for -`a(1)]; and the maximally unmarked function within the

system is represented by -[i] /[u]. In no case do any of these

elements show marked function in (C)VC +V nominals.

In its marked function -[i] is the realisation of a feminine

suffix, which we set up as FEIN; for the marked function of -[e] we

also derive it from FEM. In each of these cases, the instances of

marked function are much less frequent than those of the unmarked

function, for which we set up the suffixes -/i /, -/e/, respectively.

However, in the case of -[a], the marked function accounts for most of

the items concerned, for which we set up an inanimate suffix INÀ1;

 307

just a few nouns show the unmarked suffix, which

we shall write as

For all instances of -LW-[u], we set up, at this stage, the

unmarked suffixes /i/ and -/u/, respectively. This proposal will have

to be modified somewhat, in the following section.

For items showing final -[e(n)] and having specifically male

referents, we set up the masculine suffix hASC, and for all other

instances of this final element, ANIM; it is an arguable point that,

for the purposes of a semantic statement, we should derive some

instances of Le(n)] from the suffix FFM (for those cases where a

specifically female referent is involved), but this is not a crucial

point for our purpose, and the distinction is accordingly not reco

nised here. Finally, we derive -[a(1)] also from FEN, and from

FLUR, for those items which show final -[a(1)] and have specifically

female, and plural ('inherent plurals') referents, respectively. For

further details of these suffixes, see the discussion of Noun, below

(pp. 366 -368).

We have thus far assumed that the final element V(C) forms part

of the lexeme (i.e. does not represent a system of syntactic affixes)

and we have just now established that it does represent that part of

the noun where distinctions of gender are marked. To what extent is

it justifiable to treat these gender distinctions as not specified

within the syntactic base component?

In order to answer this question, we shall examine the patterns

of gender-relationships between Coorg nouns. This will require citing

a number of nouns which show complex canonical shapes and which

strictly fall outside the scope of this chapter, therefore; in partic-

ular, We shall recognise abstract nouns ending in -[atana], feminine

nouns in -Lati], and masculine nouns in -[ka:re(n)]. The structure

of these will be dealt with in detail in the following chapter. Other

types of gender elements will be illustrated where necessary.

There are a number of gender relationships, all more or less

related to each other, between Coorg nouns:

A ( i) ïasc /Fem/Ábstract: -[(n)] L i] / -[ atara] .

1156 [kalle(n)] thief (mast)

[kalli] id. (fem)

[kallatana] theft, state of being a thief

(ii) IYiasc/Fem/Abstract: -[e(n)] / -[ ati] / -L atana] .

3223 [bande(n)] shameless person (masc)

[bandati] id. (fem)

[bandatana] shamelessness

(iii) I:asc/Fem/Abstract: Ze(n)] / -lati] / -[i].

3508 [punde(n)] quarreller (masc)

[pundati] id. (fem)

[pundi] quarrel

n
B (i) Iiasc/Fem: -Le(n)] / -[i].
n
[pucce(n)] madman

[pucci] mad woman

(ii) East /Fem: -[e(n)] / Lati].

1374 [kodave(n)] Coorg man

[kodavati] Coorg woman

(Cf. [kodavi], Coorg (the land))

(iii) Masc/Fem : -[ka :re(n)] / -[karati].

37 [angadika:re(n)] shopkeeper (masc)

[añgadikarati] shopkeeper's wife

(Cf. [axigadi], shop, stall)

 309
( iv) Hasc/Fem: -[ ga: re(n) ] / -[ garati]

3980 [mundiga:re(r_)] paramour

[mund.icarati] mistress
rt
(v) Masc /Fem: da:re(n)] / -[darati].

[patteda:re(n)] head of family

Lpattedarati] head of family's wife

(Cf. [patteJ, title deeds)

n
(vi) Iiasc/Fem: -[va:le(n)] / -[va:lati].

3804 [madiva: le(n) ] washerman

[madiva:lati] wife of washerman, washerwoman

(Cf. [madi], (ritual) cleanliness)

(vii) Lasc/Fem: R a:re(n)] / -[arati].

[kumba:re(n)] potter

[kumba:rati] wife of potter

(viii) h;asc/Fem: -[ e(n) -[ ] / ici]

-L ici].

[maccine(n) ] younger cross-cousin (mast)

[maccinici] id. (fem)

(ix) Masc/Fem: -[i] / -[ici].

rr

3623 [poriki] mean fellow

tt

[porikici] mean woman

(x) h:asc/Fem: -[e(n)] / [a(1)].

All
_ N rr

3768 [mo:ve(n)] son

[mo:va(1)] daughter

hiasc/F em: -[(n)] / -[e].

11,
orphan (mast)

[ana:te] id. (fem)

From the foregoing data, the following observations may be made:

(a) If the gender suffixes illustrated above are to be specified

 310

in the syntactic base component, then a problem arises over the rather

large number of them (the examples above are only a representative

selection; no instances of [e(n)] with nouns having animals as refer-

ents have been given, where there is no corresponding feminine form),

and particularly over the fact that most of them are restricted to

rather few lexical items. For example, the forms *rañgadiye(n)],

*[añgadida:re(n)], etc. are not the result of phonologically ill -

formed sequences, but are not possible alternatives to [arigadika:re(n)],

'shopkeeper'. Further, no general statement of complementary

distribution is possible with respect to the gender suffixes; consider

the following:
dite 9
3945 [ma:ve(n)] mother's brother

[má:vi] father's sister

N
but 3768 Lmo :ve(n)] son

;mo:va(1)] daughter

and. 1156 [kalle(n)] thief (mast)

[kalli] id. (fem)

but [calle(n)] wastrel (mast)

[callati] id. (fem) etc.

In the face of such evidence, the existence of just a few cases where

there is equivalence between certain alternative forms, as in:

[gavde(n)] Gowda man

[r
gavdati
Gowda woman
[ gavdi]

simply entails a further complicating factor which the syntactic

component by its nature is ill- equipped to cope with. Consider also,

in this regard, the complicated situation with respect

to

[kumba:re(n)], which, besides !potter' may also have the meaning

 311

'stupid fellow'. '_ccordin :

to my informants (none of whom are

potters, it should perhaps be pointed out) , it is the 'same' lexical

item in each case; but with the first meaning the feminine form is

Lkumba:rati], and with the second it may be Lkumba:ra(1)].

(b) 3ven within a single formal gender relationship - say,

-Le(n)] / -L ati ] - semantic specialisation has typically occurred to

such a degree that the syntax would have to recognise at least two

types of suffix -L ati] ; one with the meaning 'wife of , the other

with the meaning 'feminine of'. Thus, Lpattedarati] is not 'head of

family (fen)', but 'wife of head of family'; and Lbandati] is not

'wife of shameless man', but 'shameless woman'. In a number of other

instances, moreover, the distinction between these senses is unclear,

and informants hesitate between them; thus, La_ñgadikarati] may be the

wife of a shopkeeper, or possibly refer to a widow (perhaps) who now

supports herself by running a shop. It is clear that these possibil-

ities are dependent on highly idiosyncratic ontological or socio-

logical factors, which it is not the business of the syntactic com-

ponent to handle.

It should be noted here that a 'lexicalist' approach to the

statement of gender distinctions in the 0oorg nominal system, such as

is emer,~in-r from these considerations, does not lose sight of the tie -

up between form and function, where this exists. For example, the

fact that -[a(1)] is fairly typical of nouns having human female

referents is not lost by deriving it from F21 (which is one source of

-La(1)])in the lexicon rather than the syntax; its frequency of

occurrence in those lexical items of the language whose referents are

female humans, and its relative constancy of function (as opposed,

say, to the function of the unmarked suffix -/i /) are both reflexes of
 311

'stupid fellow'. According to my informants (none of whom are

potters, it should perhaps be pointed out), it is the 'same'

lexical

item in each case; but with the first meaning the feminine form is

[kumba:rati], and with the second it may be [kumba:ra(1)].

(b) Even within a sin le formal gender relationship - say,

-[en(n)] / -[ati] - semantic specialisation has typically occurred to

such a degree that the syntax would have to recognise at least two

types of suffix L
ati] ; one with the meaning 'wife of', the other

with the meaning 'feminine of'. Thus, [pattedarati] is not 'head of

family (fem)', but 'wife of head of family'; and [bendati] is not

'wife of shameless man', but 'shameless woman'. In a number of other

instances, moreover, the distinction between these senses is unclear,

and informants hesitate between them; thus, [añgadikarati] may be the

wife of a shopkeeper, or possibly refer to a widow (perhaps) who now

supports herself by running a shop. It is clear that these possibil-

ities are dependent on highly idiosyncratic ontological or socio-

logical factors, which it is not the business of the syntactic com-

ponent to handle.

It should be noted here that a 'lexicalist' approach to the

statement of gender distinctions in the Coorg nominal system, such as

is emerging from these considerations, does not lose sight of the tie -

up between form and function, where this exists. For example, the

fact that [a(1)] is fairly typical of nouns having human female

referents is not lost by deriving it from FEE (which is one source of

-[a(1)])in the lexicon rather than the syntax; its frequency of

occurrence in those lexical items of the language whose referents are

female humans, and its relative constancy of function (as opposed,

say, to the function of the unmarked suffix %i /) are both reflexes of

 312

the correspondence between form and function. Thus, while -[a(1)]

is not a very widely distributed suffix, it occurs on nouns that

appear to have a fairly high frequency in the language, and hence it

is not surprising that informants tend to volunteer its meaning.

It was stated earlier, with respect to number, that any noun

which has an inanimate object as its referent, no matter what lexeme-

final element it shows, may not pluralise (p. 298). It has also been

claimed that number and case in Coorg are 'natural' (i.e. determined

by wholly referential factors) insofar as the person/number proforms

are concerned (p. 299); and that pronominalisation does not take

place with reference to features of noun morphology (p. 302). These

are good reasons for treating gender as not (or, no longer) a

syntactic category in Coorg. Thus, item 769:

1820 [ko:de(n)] monkey

pronominalises as [adi] not because of the fact that Le(n)] is in

this item the realisation of the suffix ANIM, but because the referent

of this noun is a nonhuman animal. Because of this, it is possible,

and even advantageous, to dispense with the notion of 'gender suffix'

in retard to the nominal system, with all that it implies. Thus, we

need not recognise in item 104:

Ldore] master, European

a suffix -/e/ which is distinct from that in item 102:

[tore] gourd.

By the same token, we do not have to analyse -[e] in the first of these

items as a co- realisation of a single morphological unit which is also

realised as -[e(n)] in certain other items. The referential meaning

of the nouns these suffixes occur in fulfills those functions (deter-

mining pronominalisation, pluralisation, and co- occurrence with verbs

denoting certain activities or states, etc.) which such a unit might

 313

serve. On the other hand, we have seen that it is necessary to

recognise a distinction between the sources of Li] in the items 529

and 554:

Lcu:di] coir rope

S 806 [mu:di] girl,

partly because of the relative constancy of function of -[i] as a

final element in nouns having human female referents, and partly

because of its peculiar distribution in this function (never in

(C)VC +V nouns). Hence, in some respects the notion of 'gender suffix'

is valid; but there is no evidence in favour of its treatment in the

syntactic component.

3.0 The generative component

It has been a working assumption thus far, and this will be

justified in the rest of this chapter, that the nominal lexemes that

we are concerned with here comprise the (C)VC+V type, which is

strictly comparable to the corresponding type of verbal lexeme, and

eight other types which are all made up of the following base shapes

plus the final element -V(C):

(C)VCC

(C)VNC

(C)VC

(C)VNC

(C)VSC

(C)VSNC

(C)VCVC

(C)VCVNC

Granted, at this stage, that this is so, we may suppose the underlying

structure of these base shapes, which are identical to those recognised

 314

in the verbal system, to be generated from the same node Base, and

in a similar way, as described in the preceding chapter (pp. 222 -224).

Figure 6 shows the generalised schema for the introduction and

development of the node Base, this time with Classifier rewritten

as Noun:

h Lexeme

IL Base j]

Root ( +Affix1) ( +Affix2) Noun

xi

Nucleus
Voc .Suff . Suffix2

(Onset) Peak (Coda)

Cons.Suff. horn

Figure 6: Generalised structure of the nominal lexeme

Note that Voice Suffix is not relevant here, and that Nom is required

for the statement of certain lexical relationships involving nouns

(this is dealt with below). Hence, we have to extend LS8 as follows:

Voice Suffix / Verb

Affixe --3 (LS8)
Nom

:ufflx2

In other respects, the schema is generated by the same LS rules that

were given in the preceding chapter, however. ale also need to include

the following constraint in condition LS10:

2. NOT: Peak + Nom (Lsio)

This yields the following base strings (for nominals):

 37 5

(Onset) .*Peak°%Coda.I' Voc Suffix : for (C)VC +V

(Onset) Peak (Coda1 Cons Suffix) : for (C)VCC / (C)VNC

(Onset) Peak Coda ^ (

Suffix
2
for (C)VC (C)VNC

Cons Suffix Nom

(Onset) Peak ^Coda Cons Suff Suffix : for (C)VSC / (C)VSNC

2

Nom (C)VCVC / (C)VCVNC

(NB: linked parentheses indicate that one of the elements so linked

must be present, but that neither is obligatory.)

We now take up for more detailed discussion the other problem

outlined in the introduction, namely that the nominal system shows a

significant number of items where there is contrast of consonant

length immediately after a short root vowel. We have seen, in the

preceding chapter, that the Classifier element Verb, rewritten as I,

II, was an integral part of the lexeme, accounting for the

appropriate forms of the following past and nonpast affixes. Regard-

ing the nominals, we have seen in this chapter that

(a) the 'gender' suffix V(C) has to be specified for each

lexeme, prior to the insertion of the lexeme in the preterminal

syntactic string;

(b) there is only one declensional class in Coorg, with four sub-

groups within it being determined by the form of the suffix -V(C).

These two considerations suggest that the 'gender' suffix peculiar

to the nominal system be introduced under the node Noun, fulfilling a

function there analogous to that served by I, II under Verb. This is

supported, in the framework of this analysis, by the general consider-

ation that the -V(C) element must be simultaneously intra- lexemic and

extra -basal: in other words, unless the structure of nominal and

verbal lexemes is fundamentally dissimilar, there is no other place

 316

for the element -V(C) to fit. If it can be shown, therefore, on

independent grounds, that this is an appropriate solution, then this

will confirm that nominal and verbal lexemic structure is fundament-

ally similar, and that constraints on the form of verbal lexemes are

relevant to that of nominal lexemes. This solution will also account

automatically for the fact that no (C)VC +V nominal is marked for

gender.

In this regard, note that this analysis of nominal lexeme

structure allows for two distinct sources for lexeme -final -V; one

under Noun (i.e. in that majority of cases where there is no final

consonant), and the other under Voc Suffix (within the base). If,

for presentation purposes here, we set off the intra -basal suffix

by + , and the extra -basal suffix by = , we can now account for

consonant length alternation after a short root vowel as follows:

4524 [boli] light ¡bolli!

4524 [bolli] silver /bol =i/

n
2473 [tani] coolness /tan +i/

3299 [ panni] fruit /pan=i/

1897 [jade] cobra's hood /jat +e/

1910 Idadde] dull person (fe °a) /clad =e/

w

2463 [tatte] flat plate /tat =e/

By this solution, which arises naturally in the context of the

analysis suggested thus far, we adhere to verbal lexeme structural

principal that a consonant standing in base -final position immediately

after a short root vowel is automatically geminated; also, that, if

it is a stop,it is thereby tense and voiceless. Thus, in underlying

representation, the true length contrast between stops is illustrated

 31 7

by [jade] and [tatte] (not [daddc], which shows a derived base -final

/d/, by the operation of R2, R4, and R5 on the seduence /T ti). All

single consonants are lax and voiced intervocalically.

How far is this solution justifiable? Evidence i:i favour of it

would consist in one or more of the following conditions being met:

(1) if the sy- tom of elements operatin_r, at one source could be shown

to be significantly distinct from that operating a1V the other, in

respect of its inventory;

(2) if the elements of the system at one source could be shown to have

a distributional pattern distinct frco that of the 'same' elements at

the other source;

(3) if cases could be shorn where both systems are simultaneously

operative.

it is not difficult to show that all these conditions are net, although

the evidence from (3) is not as full as might be expected. Concerning

(1), Figure 7 shows the generalised schema for items such as [boli],

[tani], and [Jade]:

# Lexeme #

[ Base T Classifier

Root + 1fÎ;; :1 PToun

N
Tucleus

(Onset ) Peak Coda Voc.Suff

l
Figure 7: Generalised structure for (C)VCi -V nominals
 31 0

The syster at lee Suffix consists of 5 elements; /u/ and /i/ are in

contrast, though minimally. IToun is not developed.

By contrast, Figure S shows the schema reóuired for [ bolli],

[panni], [dadde], [tatte], etc.:

e w w

# hexene #

Classifier

`
Root

Nucleus
(-; Affix1 ) Noun

(Onset) Peak
f \
(Coda) (Cons. cuff.)

Figure 8: Generalised schema for (C)VCCV(C) nomin_als

Noun here is developed; the system operating at it consists of 5

elements, of which just 3 are identical in form to those found at

Voc Suffix; however, two of these, /iV and /e/, represent FL';N in sone

items, which they never do in (C)VC-14 nominals. (The presentation of

Noun elements is simplified here, but the details that are omitted

are irrelevant to our purpose; for a fuller treatment, see the

discussion .of Noun below, pp. 366 -368). The justification for setting
,r

up -AR/ as the phonological source for -[e(n)] must be postponed until

the last part of Chapter 6; /a/ is the underlying form of lexeme-

final -[a(1)]: see pp. 424 -427 for both these elements.

Note that the elements /1/ and /u/ are not included in the system
here; the reason for thin ?'Till bocor e apparent when we consider the

evidence of type (2), which is precontc . in the table in Figure 9:

u
Canonical forr i e a i u a(1) e(n)

1 (c)vc+v 68 48 9 11 - - 197

2 (C)VCC = V(C) 40 54 22 76 1 6 25 224

3 (C )VidC = V(C ) 16 17 10 25 - 2 6 76

4 (C)VC = V(C) 51 53 59 Q1 4 2 10 270

5 (C)Vr1C = V(C) 4 1 _ 4 - - _ 9

6 (C )vCT7c = V(C) 16 35 26 32 1 1 17 128

7 (C)JCj1NC = V(C) 4 2 - 8 - 14

E (2 to 7) 131 162 117 236 6 11 5 721

oJ 1l1 to 7`
.
192 230 165 245 17 11 918

Figure 9: Lexeme -final elements in the nominal system

This shows the distributional pattern of elements at the systematic

phonetic level, counted for a total of 918 nouns representing all the

lexeme types under consideration.

The first row contains those entries which, if the analysis under

consideration is correct, show the suffixal system at Voc Suff,

while the remaining six rows contain entries for all those lexeme

types where the analysis posits a suffixal system at Noun. If row 1

is compared with row 8 (¡1 (2 to 7)) across the first 5 columns, it

u_
will be seen that the relative frequency of [±] and [u] changes

markedly between the two sets of entries. If the data is organised

in terms of (C)VC +V versus non- (C)VC +V lexemes on the one hand, and
 320

[i] /[u] suffixal elements versus non- [i] /[u] elements on the other

(as in the table in Figure 10), it is quite apparent that there is

an association between the non- (C)VC +V lexeme types and the

occurrence of the [i] /[u] suffixal elements; the Chi- square value

for the table in Figure 10 is in fact extremely high (X2 = 41.583).

Hence, it is justifiable to recognise a fundamental distinction

between the two systems of suffixal elements here (one occurring with

(C)VC +V lexemes, the other with all the rest); and a natural way to

formalise this would be to establish a distinct source in underlying

representation for each of these systems.

(C)VC+V 20 177 197

(C)VC+V 242 479 721

L 262 656 918

i/u = neither DJnnor [uj

(C)VC+V = canonical forms other than (C)VC+V

Figure 10: The distributional pattwrn of the

lexeme -final elements [ili[u]

Furthermore, since it is possible to state a complementary dis-

tribution of [i] and [u] for all the lexemes of rows 2 to 7 (this

is discussed in detail in the following section), and since they

together account for the largest number of lexeme -final elements in

those rows (see row 8), it is most convenient to exclude them from

the system at Noun altogether; they are not present in underlying

 321

representation at all, therefore, but are introduced as surface

elements after a consonant at Coda or Cons Suff which stands in base -

final position (i.e. where no element at Voc Suff or Noun occurs).

This accounts naturally for the maximally unmarked status of [i] and

[u] noted earlier. The rule that handles this is the realisation rule

R13:

constriction, I

II
back of tonge
3. Insert: B # (R13)
highest, () III
C IV
lips unrounded

Note that the form of the enunciative vowel from R13 is [i]; it is

mutated to [u] subsequently by the phonetic mutation rule AS9 when it

is preceded by the sequences [av]- or [a:v] -.

By contrast, [i] and [u] have to represent underlying elements

(at Voc Suff) in base -final position in (C)VCV nouns, since otherwise

the consonant at Coda would stand base -finally after a short root

vowel, and be geminated.

Regarding condition (3) above, consider the schema shown in

Figure 11. This represents the generalised underlying structure of

items such as 914, 918, 920, 933, 943, etc.:

2957 [nadeya] passage to shrine

3295 Lpaluva] coral

3497 [podiya] sari

1379 Lkudiye(n)] man of toddy tapper caste

4270 Lbadave(n)] poor man

Both systems of suffixal elements are present simultaneously; the labial

or palatal approximant in each of these items is a predictable surface

element, breaking up the vowel cluster. It is not clear why the system
 322

* Lexeme*

-----
LL Base 3 Classifier
__-
Root 1,._ffix1 N o un

i/
¡ Nucleus

/
(Onset) Peak Coda Voc Suff. a

an

u
a

Figure 11: Generalised schema for nominals of the

Lnadeya] / [paluva] type

at Noun should be so restricted; nor why this lexeme type (which is

included in the table in Figure 9 above as a type of (C)VCVC =V(C)

for simplicity of presentation of the data) should be relatively rare.

However, we may briefly anticipate some of the discussion of Voc Suff

(below), where this lexeme type will be dealt with in some detail,

by noting the following lexical relationships which support the

analysis proposed here:

2957 [nadeya] passage to shrine

[nade] limb

[nada]- walk; Class 3 verb

1379 Lkudiye(n).j man of toddy tapper caste

[kudi] family of servants living in one

hut

u
4270 Hbadave(n)_j poor man

'bada] poor
 323

In all such cases, there is evidence that the correct structural

analysis has to recognise a (C)VC +V base (which may constitute a

lexeme, if no element occurs at Noun), which may be followed by an

element at Noun.

Finally, note that the analysis proposed here also preserves

another principle of verbal base structure, namely that a vowel at

Peak cannot be realised as long if a suffixal element from Affix1

(Voc Suff, Cons Suff)occurs in the same base; the fact that this does

not apply to extra-basal suffixal elements (from Noun) allows for

lexeme -types such as (C)V(N)C =V(C), and provides the explanation of

why these pattern like (C)VCC =V(C), (C)VNC V(C), and (C)VCV(N)C =V(C)

lexemes in respect of their final elements (see the table in Figure 9).

We turn now to consider the systems of phonological elements

which operate at the places defined in the generalised structure

yielded by the LS rules.

3.1 Onset

The system at this place is shown in Figure 12:

A B C

I k

ITT (1 )

IV ( ~,

Figure 12: The system at Onset

 324

In most respects, it is similar to the verbal Onset system.

/p /, /t%, /k/ are among the most frequent elements, set up to

account for the voiceless stops [p], [t], [k], respectively. The

voiced stops [d], [j], [g] are much less frequent, and for them

the elements /d/, /j /, /g/, respectively are set up, under the second -

order feature b (defining plosives marked for voicing). However, [b]

is almost as frequent as [k] and [p] in this position, and about as

frequent as [t], and hence belies the marked status of /b/ in the

matrix of Figure 12. An explanation for this state of affairs may

be found by examining in other South Dravidian languages cognate forms

to those showing- initial [b] in Coorg; in a large number of cases

Coorg initial [b] corresponds to (and diachronically has developed

from) initial [v]: cf. 4254 [boli], 'light', [ bolli], 'silver', and

Ma. [veli], [velli], id. Rather than derive [b] by rule from /v/,

we set up /b/ for all these cases; note that it would be a complicated

matter to derive [b] from [v], since the two are in contrast at the

systematic phonetic level (however minimally; see the following dis-

cussion), and /v/ is required to account for [v].

The most frequent nasal consonant is /m/, followed by /n /;

/ñ/ occurs in only four items, 230, 474, 731, and 732:
« MM.
3097 [nenni]
chest (body part)
. [neñnl]

2362 [ñandi] crab

3014 [ ña: na] shame

DIS! 182 [ña:ya] dispute, justice.

This is consistent with the general pattern whereby the elements in

the column defined by the second -order feature j are less frequent

than the elements in the corresponding rows.

 325

/s/ occurs in sixteen items, which is considerably more than was

found with the verbs, but this element is still a relatively

infrequent one in the system.

Thus far, the inventory of the system is identical with that

set up for the verbs; however, in addition 12 items show initial [r],

2 show initial [1], 3 show initial [s] and 7 show initial [v]. For

these we set up /r/, /1 /, /s/, and /v /, respectively. The last

element is particularly interesting, since it underlies voiced labial

articulations that contrast with [b] at the systematic phonetic

level. The items concerned are 203 -4, 569 -70, 761 -2, 818, and 877:

[vara]
rice -paste extruder
.r ora J )

[ visa] poison

896 [va:re]
slantness
. r o:re J

903 [va:le]
ear ornament
.,. Lode]] I
4389 [va:rai rent

[va:ra] week

827 [vanali]
sieve
L onali ]
Ï I
11_

606 [varakij
11
sleep
'
L oraki]

Of these, four show initial [va(:)] alternating with [o( :)] in free

variation; as described in Chapter 2 (pp. 90 -92), it is convenient

to regard the [o(:)] forms as basic at the phonetic level, and derive

the [va(:)] forms from them by a low -level rule. This leaves only 3

items for which we have to set up /v/; hence the contrast between

/v/ and /b/ is minimal. /v/ is placed in the row defined by the
 326

feature IV in the matrix, along with the other relatively infrequent

items /s/ and /s/. /s/ is found just in the items 613, and 763 -4:

[sa:le] school

[sá:pa] curse

[su :la] trident

which are all loans (via Kanarese). Hence we have to supplement the

realisation rule R7 as follows: -

occlusion, groove friction] / B(j)

4. IV a [occlusion] / A ^Kj() (R7)

obstruction] / A

/1/ occurs in just two items, 397 and 760:

[lekka] account, sum

[lo:ka] world

which are also loans (via Kanarese).

/r/ is found in the items 200 -1, 296 -7, 495, 559-60, 612, 759,

777, and 813:

[raja] holiday

[rasa] sweetness,

2133 Eratte] upper arm

2133 [ rekke ] wing

4237 [rampa] hubbub

[ ra: gi] sp. grain

[ra:ni] queen

[ra:te] board for pounding rice

[ro :ma] body hair

[ra:je(n)] king

[rayte(n)] cultivator
 327

Only those supplied with a DED reference number are part of the

native vocabulary. In the case of [ratte] and [rekke] the Ka.

forms erake, rakke, 'wing' suggest that these Coorg items are

unusual in showing loss of an initial (root) vowel; the %r/ here

is, diachronically, a Coda element.

3.2 Peak

The system at this point is shown in Figure 13:

Peak

X (L)

Figure 13: The systems at Peak

It will be seen that this is quite similar to the system set up at

Peak in the verbal system. The most commonly occurring vowels are

Li, e, a, o, u], for most cases of which the elements /i, e, a, o,

u/, respectively are set up here. In just a few cases, [o] and [u]

are representations of /i, e/, respectively, as described in the

preceding chapter for the verbs; and, by the same process, these

systematic phonemic elements also account for certain instances of

Li] and Le]. This is examined in more detail now.

It was seen in Chapter 2 (pp. 67 -70) and Chapter 4 (pp.231 -234)

 328

that there are two types of exception to the general rule that front

unrounded vowels are retracted in a following retroflex environment

(and additionally rounded, if a labial consonant simultaneously

precedes): the back quality may occur without a following retroflex

environment, or a front unrounded Quality may occur even though a

retroflex consonant follows. The latter type does not occur in the

verbal system, but it is found with nouns; cf. items 28, 32 -3, 168,

226, 328, 471 -2, 478, 499, and 514:

1318 Lgili]

N
n,
[gin!]
_ n
gini]
} parrot

1281 [ cedi]

0.1 [kedi]
rr } spark

2269 [cedi]
anger
[codi;

1613 [gida]
n
plant
[ gida]

2135 [ cinni] smallness

2296 [celli]
._ n n .,
flea
celli]

2081 [cindi] scent (in hunting)

rr .

2275 [cendi]
n rr
ball
h [ cendi]

26j7 Ldindi] stem of plantain

2135 [cinde(n)] small man

DIA 41 [i:ti]
_n
spear
.r Li:ti]

It will be seen that a high proportion of these items show an initial

palatal consonant; it seems to be necessary to restrict the vowel

 329

retraction rule (R9), therefore, so that it does not operate when a

palatal consonant precedes the relevant vowel. However, this

restriction, while on the whole well- motivated, is not absolute, as

can be seen from the alternative forms given above, where the back

quality does occur on the vowel in question. In general, such forms

seem to be less common, and quite variable between one speaker and

another. We do not take account of this variation here; we set up

just /1/ and re/ at Peak for both front and back rounded qualities in

this particular environment, and write the appropriate restriction

into the vowel retraction rule; and we note, informally, that this

restriction is optional. The relevant subrule, without the

restriction, has the form:

B(
5. X --. [back of tongue highest, , ()g j()A(1,)"Br (R9)
lips unrounded]

This allows for the consonant preceding the vowel to be defined by

the feature B, Bj (Br never occurs at Onset) or C. If we rewrite the

environment as

( iA j ( )6(14)ABr,

the restriction we are concerned with is effected. We therefore allow

for both these environments, as follows:

¡CB()
1
111
B ()
w
_() A A
(L) Br

We may note here two further instances where the restriction does not

apply; the items 331 -2:

 330

n n
2184 Ljitti]
tuft of hair

dirt (of body)

For the remaining items, it should be noted that all of them have

alternative pronunciations with the expected retracted quality of

vowel, and that these forms are more common. We setup /I/ at Peak

for their non -retracted variants, which is entered in the system

under the second -order feature f and is thereby marked for front

articulation in all environments.

Still within this class of exceptional cases, we find that,

while more than 50 items show the expected [u] or [o] in the environ-

ment of a preceding labial consonant and a following retroflex, there

are 3 items where this is not the case; these are 391, 596, and 604:

4503 [betta] hill, mountain

D?: 276 [perte] market, town

3639 [be :le] halves of dividable seed

For these, we set up /E/ under the feature f in the Peak system.

This requires a further slight modification to the vowel retraction

subrule in R9. Up till now, the feature X has been stated as follows:

6. X a [back of tongue highest] (R9)

If we now modify the first of these stated environments, yielding the

following rule:
 331

A (-J_. r

7. X - [back of tongue highest] / (R9)

{:1

then the feature f will block the operation of the vowel retraction

subrule, since it will stand between j and (}/r. In these cases,

:s will therefore specify [front of tongue highest, lips spread] .

Concerning the other type of exception, where the back unrounded

quality occurs before a non -retroflex environment, we have the

following examples to consider; items 304, 636, 661, 671, 673, 713,

718 and 875:

, ff

698 [etti]
bull, bullock
0+ [etti]
n n
1348 [ke :ki] the K`ast
n f ,
2841 [te:ki]
hiccough
u
w. [te :ki]
It ft
3072 [ne:ri] bundle of paddy seedlings
fr

776 [e :ra]
more than enough
N [e :ra]

1348 [ki:da] place below, down

n n
449 [eraki]
,r eaves
w [ e raki

Concerning all the cases where [r] follows the retracted vowel, we

follow the same solution as was adopted in the verbal system, and

set up ,r'% as the following element (either at Coda or Cons Suff).

Note that the two alternative forms [e:ra] and [eraki] are evidence

for certain idiolects not having (probably, losing) the /r'/ element.
 332

Concerning 1348 [ke :ki] and [ki :da], we also have to take into

consideration the related item, 631:

H n
1348 [ki:li] state of being lower, below

and accordingly recognise a root /ki :1/ with the alternative form

lke:ll-; the relevant terminal sequences will therefore be /ke:lkk/,

+ +
lki:ltal, and lhi :ll, respectively. R9 retracts the root vowel in
+ +
each case, and a later rule, AS6, simplifies the clusters (see the

discussion of /1/ at Coda in the preceding chapter; further details

will be given below).

For the remaining two items, [etti] and [te:ki], we set up /e/

at Peak, under the second -order feature r, as described in the

preceding chapter; in each of these cases, it is in free variation

with lel.

It was stated earlier, in discussing /I/, that one of the items

showing this element at Peak is 168:

1613 [gida] plant

It was also noted that this has the alternative form:

H
1613 [gida] plant

which was regular with respect to the vowel retraction rule. The

statement regarding these items now has to be revised somewhat,

since each of these items is irregular in respect of the vowel height

of their Peak realisations: it was found in the verbal system that

lil and /u/ at Peak are adjusted to lel and /o/, respectively when

lal occurs at Voc Suff in the same base (lil at Peak was not involved

in any of the items in the data), and accordingly the rule F4 was

formulated:

C + a ( Pi?4 )

[1]
 333

Essentially the same situation is found in the nominal system, but

item 168 (both forms) is exceptional, to: ether with items 156, 177,

190, and 203-4:

353 [ija] trouble

[dina] day

[buda] origin, base

[visa]
poison
[viln]

[sima] lion

For just these items (of which the last four are part of the non-

native vocabulary) we have to set up elements at Peak which are

marked for high tongue position; accordingly, we establish a row in

the matrix of Figure 13 (p. 327) which is defined by the feature h.

For the items [ija], [diva], [visa]~ [visa], [sima] and also [gida]

we set up the element /1/ in this row: it is marked for high tongue

position and hence does not mutate to fe/ by rule P:ä4. The feature

h is subsequently irrelevant to the operation of the realisation

rules: thus, fJ behaves in all other respects like ¡i 1, and is

Il TI

realised as [i] in [gida] by the same rule (R9) that realises /i/
.

11

as [i] in the appropriate environment. However, for [sida] we have

to set up. /I/ at Peak,which is marked both for front articulation and

high tongue position. Last, the item [buda] possibly has // at Peak,

which is then retracted and rounded by R9; but we also have to allow

for the possibility that its root vowel is marked for lip- rounding

in underlying representation, and accordingly we set up the element

/w which is marked for both lip-rounded articulation and high tongue

position. Setting up the feature h does not require any further

modification to the realisation rules; it has neither articulatory

nor contextual function, serving only to distinguish a class of elements

 334

at the phonology ca.l level that remain unaffected by the phonolo ical

mutation rale PM4.

Finally, six iters show nasalised vowels at Peak; 4, 23 (2

form) , 163, 551, 774, and 776:

w
2894 [tuy ] hiss of snake
,, N_
1 645 [Devij ear
ft. M,
1644 [keva j wall

3945 [ma:vi,j father's sister

Al
3945 [ma:ve(n)] mother's brother
N
3768 [mo :ve(n)] son

In all these cases the following consonant is an anproximant, and

nasalisation iw spread through the rest of the item by the phonetic

mutation rule A51. This parallels the situation that was found in

the verbal system. Accordingly, we establish the elements /7/,

N N
u/, /e /, /gal, and /oM/ in the system at Peak, under the feature n as

described in the preceding chanter. Note that %i/ as well as /e/

has to be set up, as a possible source for [é] in [kevá] (through the

operation of P144)-

3.3 Coda

'le follow here the heuristic approach outlined in Chapter 4 for

determining the inventory of elements at this position; in particular,

we assume (once more) that all base -medial consonants at the system-

atic phonetic level must represent Coda elements in underlying

representation. ?;re have also seen, in the preceding section, that

/r'/ has to be set up together with /r/; as in the case of the verbs,

we allow for these elements to be representations of both Coda and

Cons Suff. On the basis of this. evidence, we may establish all the

elements in the system at Coda, save for /+/ and /1/; these will be
 335

discussed separately at tre end of this section. The full syster is

shoEr? in Figure 14:

B
r j

I t t c (k)

b (J)

(T) (1)

n (m) n n

II r (r')

4 (T)

III 1 1

X (1)
T

ITT (s) \°/

w .

Figure 14: The system at Coda

/p /: this is realised as [v] in items such as the following:

23, 790, 815, 892, 907; and 693, 706, 720, 776, and 779:

1645 [kevij ear

DIA 80 [kavdij shell, cowrie

1034 [kaviki] vital spot

4516 [bevari sweat

DIA 230 [nevana] pretence

[be :vi] neem tree

3009 [na :vu] tongue

1590 [ku:va] shallow well

 336

3768 L mó ve ( n) .J son

3768 [mo:va(1)] daughter

The realisation rules R1 and R4 handle the realisation of /p/ as the

lax, voiced labial articulation. In the first 5 items cited above,

/p/ is base -medial and intervocalic in the terminal sequence; in the

remaining 5, it is base -final, but intervocalic also when -/a/,

-/a/, or -/a/ follow. [be:vi] and [na:vu] show /_/ in base -final

position which is also lexere- final. [i] is the form of the

'enunciative' vowel supplied by the realisation rule R13; it occurs

after the lip -compressed approximant articulation of /p/ (by R4)

which is conditioned in [be :vi] by the preceding front vowel. '"here

R4 yields a lip- rounded approximant articulation of /p/, as in

[na:vu], the form of the enunciative vowel mutated to is [u] (AS9).

Note also that the epenthetic vowel is deleted in [kavdi] and

kavi'ki] before the consonant at Cons Suf f (by AS4) after the lip-

rounded approximant articulation of /p/, but is retained in Lbevari]

and [nevanaj, after the lip -compressed approximant articulation of

/'/ This is identical to the situation found with the verbs; the

only new factor here is the way the particular realisation of /p/

determines the form of the enunciative vowel.

/p/ is realised as [pp] base- finally after a short root vowel,

and as [p] after a long root vowel when followed by /*/ at Suffix2;

cf. the items 221, 256, and 581:

DIA 108 .[kuppi] bottle

1027 [kappe] toad

2139 [ci:pe] broom

All base -final articulations of /:p/ at the systematic phonetic level,

whether stops or laxed, may result from /p/ at Cons Suff.

 337

/t /: this is realised as [d] when it occurs base -medially at

the systematic phonetic level, as in the items 26 and 837:

1735 [kodi] love, desire

680 [edike] state of being in front

It is also realised as [d] when it occurs base -finally after a long

root vowel in underlying representation, as in the items 518 and 681:

884 [o:di] share

3724 [po:d] auspicious ceremony

It is realised as [tt] when it occurs base -finally after a short root

vowel, as in the items 297 and 337:

4157 [mette] mattress

1 272 [ titti] fire

Finally, it is realised as [ t] when followed by /*/ after a long root

vowel, as in the items 545 and 695:

3746 [po :ti] niche over door

3920 [ma:ti] afterbirth

Note that, in the last three cases, the realisation of /t/ in base -

final position at the phonetic level is not affected by whether an

element at Noun follows (as in [ó:di], [ mette], and [po :ti]) or not
n_ n u.
(as in [po:di], [titti], and [ma:ti]). This is true for all elements

in this position.

Base- finally at the phonetic level, these articulations may

result from /t/ at Cons Suff.

/t /: this is realised as [d] when it occurs base -medially at the

systematic phonetic level, as in the items 7, 843, and 946:

63 [adj.] place below

929 [kadape] steps over fence

938 [kadandi] wasp

 333

It is also realised as Ldj when it occurs base -finally in underlying

structure after a long root vowel, as in the item 528:

DI.;. 157 [ ca:di] slander

It is realised as a voiceless retroflex stop of long duration

when it stands base -finally after a short root vowel in underlying

structure, as in the item 318:

1713 [kott] shed

and as a voiceless retroflex stop of short duration when it stands

after a long root vowel and is followed by /*/ in underlying structure,

as in the item 574:

1564 [ku:te] basket

Base -finally at the systematic phonetic level, [d], [t], [tt] may

represent /t/ at Cons Suff.

/c /: this is realised as a palatal continuant (by R2 and R4)

wherever it is single in the terminal sequence, base -medially or base -

finally, as in the items 1, 176, and 799 (base -medially), and 509

(base -finally):

479 [oy] nail

[daya] kindness

1469 [kuyli] bee's sting

4385 [ba:y] mouth

In the case of [oy] (and others of this form, items 2 -6), /i/ at

Voc Suff is posited in the underlying structure, in order to account

for the fact that /c/ is not geminated after the short root vowel

by the automatic gemination rule; it is deleted by AS5. The

epenthetic vowel is deleted in [kuyli] after [y], before the element

at Cons Suff, by AS4.

/c/ may be realised as Leo] or [c] in base -final position at the

 339

systematic phonetic level, depending on the length of the preceding

vowel; [c] after a long root vowel indicates the presence of /*/

following A/ in underlying representation.

[cc], [c], and [y] in base -final position at the systematic

phonetic level may represent /c/ at Cons Suff.

/k/: this is realised as [g] in the following items: 68, 115,

178, and 181:

16 [age] paddy seedling

3483 [pose] smoke

DIA 221 [naga] jewelry

3061 [nogal yoke

These represent all the cases in the data where /k/ at Coda is

followed by an element at Voc Suff. It is also realised as [g] when

it is followed by an element at Cons Suff, as in the item 857:

DIA 236 [pagade] chess

and when it stands base -finally in underlying structure after a long

root vowel, as in item 753:

3456 [bi:ga] lock

It is realised as Lk] when it is followed by /x/ after a long root

vowel, as in the item 709:

285 [a:ka] paddy (especially nursery) field

and as [kk] when it stands base -finally in underlying representation

after a short root vowel, as in the item 426:

2435 [takka(1)] chief family

Base -finally at the systematic phonetic level, velar stops may

represent /k/ at Cons Suff.

/j /: this is set up to account for just 8 instances of Lj , in

the items 153, 200, 537, 539, 548, 644, 744, and 767:
 340

353 Lija] trouble

[raja] holiday

3740 Lbo:jij goodness, beauty

3927 [ma: ji j soiled clothes

[me:ji] table

DIA`_ 172 [ ju: ji] bet, gambling

3740 tibo:ja] altern. form of [bo :ji]

ra:je(n)] king

The first two items show clear instances of /j/ at Coda; all the

rest are possibly the result of /j/ at Cons Suff. In either case,

/j/ is a marginal element, accounting for [j] which cannot be

derived from /c/ in terms of the phonological system proposed here.

Lija] is in D;D, but is not known to my informants; [raja], Lme:ji],

Lju :ji], and [ra:je(n)j are part of the non -native vocabulary; and

[bo:ji]» [bo :ja] really constitute just one instance of [jj. Note

that the remaining item, Lma:ji], is related to the Class 1 verb

Lma:jj- '(clothes) soil', which was one of the few instances for

which /j/ had to be set up in the verbal system.

/T /, /T /: these elements are set up here to account for the

intervocalic voiceless stops [t] and [t], respectively, in just the

5 items 63, 82, 185; and 182 and 183:

[miti] limit

[kate] story

3226 [pata] softness

DIA 240 [pata] picture

Lpata] (paper) kite

These stops are unusual in respect of their being simultaneously short

and voiceless; if we treat the fact that they are voiceless as their

exceptional feature (rather than that they are short), we can account
 341

for them by setting up at Coda the elements /T /, /T/ which are

defined by the second -order feature p (i.e. marked for voiceless

plosive articulation) just as described for the system at Cons Suff

in the preceding chapter (pp. 264 -271). All the items cited above,

save for [pata], are loans.

/m/, /n/, /n/: these elements are set up in the first instance to

account for base- medial [m], [nj, [n], respectively.

/m/ is present just in the items 10, 23, 204, 909, and 926:

548 E umi]
paddy husks
we [ ummi ]

1645 [kemi]
N ear
LkeviM ]

[sima] lion

[gamana] odour

[samaya] time

Note that the first two of these are alternative forms, while the

last three are all loans (via Kanarese) .

/n/ and /n/, however, are fairly common, in items such as 133,

818, 192, and 948:

4119a [ mone j sharp roint

827 '[onali] sieve

3817b [mane] sand

[manañji] tacky secretion of jackfruit

/n/ and /n/ at Coda may, in addition, underlie [nn], [n]; Lnn]

[n], respectively, depending on the length of the preceding vowel,

when no other element stands after them in the base; cf. the items

358, 563, 365, and 602:

3732 [pon:n ] gold

 342

4235 [ a: ne] ele7Thant

3817a [manni] mud, land property

3354 [ba:ne] open grazing land

As in the verbal system, base-final [m(m)] and tin(n)] are not to

be accounted for by the system at Coda, but by that at Cons Suff (see

below).

Finally, note that the items 193 and 365:

3817a [mana] sand

Lmanni] mud, land property

both share the same root, /man/-. The first is formed from this by

the presence of the Voc Suff element /a/; the second shows no suffixal

element at all, the base -final consonant is geminated, and Li] is

written in by R13.

/r/, /r' /: the reasons for setting up /r'/ beside /r/ have been

described in discussing the system at Peak. These elements at Coda

account for all instances of base-medial Lrj in the data, as in the

items 69, 836, and 875:

192 [are] half

614 [ uripe] small bag for betel quid

449 Leraki] eaves

Base -finally, they also underlie Lr] after a long root vowel, as in

the items 577 and 673:

1869 [ko :re] tusk

u u
3126 (ne:ri] truth

In these cases, of course, the apico - alveolar tap may represent /r/,

/r'/ at Cons Suff.

Finally, as was found in the verbal system, these elements may be

realised as voiceless dental stop articulations: this may happen as

 343

the result of the automatic gemination rule, where /r/, /r'/ stand

in base -final position in underlying structure, after a short root

vowel; or they may be followed by / */ after a long root vowel. In

the first case, they will yield [tt], in the second, [t]. Exemplif-

ication of the first case is hard to find in the data, but the item

144 and its related form are certainly relevant:

n_
3903 Lmari] state of being opposite

[mattiyandij day after tomorrow

(Cf. 1 [andl], that day)

As far as the operation of /*/ is concerned, consider the following

lexical relationships, involving the items 629 and 743:

1353 [ki:ti] torn piece

[ki: r]--

n
tear (intr); Class 1 verb
^+ [ki:rj-

2379 [na:ta] smell

[na:r]- smell (intr); Class 1 verb.

/1 /, /1/: these are set up here to account for all instances of

base -medial [1] and [1], as in the items 13, 863, 161, and 881:

710 [eli] rat

3531 [ polace] dawn

1160 [kala] threshing floor

1523 [kulr] cold (climate)

In base -final position in underlying structure, /1/ accounts for

[11] after a short root vowel and [1] after a long root vowel;

similarly, /1/ accounts for most instances of [11] and [1] (however,

/1/ and /1/ have also (marginally) to be set up at Cons Suff, and

may underlie some instances of these lateral articulations). Cf. the

items 235, 639, 217, and 642:

 344

3294 [palli] lizard

1852 [ko:li] stick

1156 Lkalli] thief (fen)

1886 [go:li] agony, struEEling

Finally, /1/ also has to be set up at Coda, and underlies certain

+
instances of [1]; these are dealt with separately, below.

/s!: this is realised everywhere as [s]; it never occurs long

in the language. It occurs base-medially in the items 58, 62, 867,

etc.:

[bisi] heat

4187 Lmasij

. Ì`
N
masi ,- ,
charcoal

4055 LmosaïeJ
crocodile
v Lmosale]

and base - finally in such items as 566, 609, and 766:

[a:se] wish, desire

3996 [mi:se]
moustache
M
[mi:se]

[ sa: sa] effort

/s/: apart from the items just cited (alternative forms), this

occurs base -medially just in the item 22, and base -finally just in

the item 741; it is realised as [s] in each case:

[kusi]
happiness
N [kusij l'

[de:sa] country, land

Each of these items is a loan, via Kanarese.

In addition to the foregoing, the elements /r/ and /1/ are set

up in the system at Coda, and are dealt with here.

 345

/r/: one reason for settin- this up in the verbal syster was

in order to derive the relativelzT infrequent -BB type of cluster in

(C)VBB bases ( -BB standing for a seminate voiced stop cluster).

This had the advantage of avoiding the necessity of establishing the

voiced stop elements /b/, /d /, and /g/ at Coda and /or Cons Suff,

just for these base types. The ratio of the occurrence of -BB to -FP

in (C)VCC bases was noted as 1 :4 (p. 246). In the nominal system,

the ratio is just less than 1:4, and hence such an approach is viable

here also. Furthermore, in the nominal system this not only avoids

the necessity of specifying at Coda /b /, /d /, and /g/, but also /d/

and /j /; cf. the items 249, 301, 343, 207, and 368:

[ ubbe] washermants boiling -pot

[uddi] comron pulse

u_
2705 [duddi] money, 4 pice

46 Lajji] great -grandmother

[ muggì ] mildew, smell from mildew

However, further evidence comes from a consideration of lexical

relationships. The first of the examples below has already been

given in the preceding chapter; the rest are given here for the first

time:

Item 476 2526 [tandi] offer of marriage to girl

[tar]- give (to 1st /2nd person); Class 3

verb

Item 205 178 [ a -i] husked rice

omari] in [puttari], harvest festival

-[eri] in [tanneri], sheaf of ripe paddy

Item 228 3089 [nucci] broken rice grains

[nuri] small bits of broken rice

 346

Item 406 Lobbe(n)] one person (mast)

[obba(l)] id. (fem)

r
dorme] harmony, unity

Lori] one, a

It was noted in the ;preceding chapter that /r/ has two main

phonological characteristics; first, it is realised as Lr] when it

is operated on by the realisation rules (i.e. the feature x in the

matrix of Fig=ure 14 has no direct articulatory effect), and secondly

it does not satisfy the condition for C1 in the epenthetic vowel

rule PI i3 :

9. C1 + C2 -- C1 + V C2. (C1 / r, 1) (PT.13 (i))

This means that it may come to stand immediately in front of an

element in the system at Cons Suff in the terminal sequence. Hence,

if we set up the roots:

Item 476 /taa/-

Item 205 /aT/- A.+ /er/-

Item 228 /nur/-

Item 406 /or/ -,

the following terminal sequences will be yielded:

Item 476 /tarNt/ offer of marriage to girl

!1

-
/tari/ give (to 1st /2nd person)

Item 205 /arkki/ husked rice

- /ari/

-/eri/

Item 228 /nu+ cci/ broken rice grains

/nuri/ small bits of broken rice

 347

Item 406 /orpan/ one person (mast)

/orpal/ id. (fem)

/orme/ harmony, unity

/o+i/ one, a

The realisation rule R4 then realises /t/ as Ld] after /N/, and /p/

as Lb] after /,r/. R5 realises /r/ as Lb] before the single base -

final element /p/, but as [r] otherwise: thus /r/ is Lr] inter-

vocalically, before /N /, before /m/, and before the geminate conson-

ants /hl/ and /cc /. Finally, R8 provides the appropriate realisation

of /17/. At the phonetic level, AS6 simplifies the 3- consonant

clusters by deleting the first consonant (Lr]) in each case,

yielding L tandi] , [akki] and Lnuccij.

Note that item 205 shows variant root forms; they differ only

in respect of vowel quality, but this variation cannot be accounted

for by rule. External evidence for this particular analysis comes

from entries for related languages at DED 178: "Ta. ari 'rice, paddy,

ear of paddy'; iia. ari 'grain of rice freed from chaff, seed, grain ".

/1 /: this element was set up at Coda in the verbal system

primarily in order to allow for the statement of certain verb -noun

lexical relationships. Since the concern of Chapter 5 was mainly

with verb-verb relationships, the function of /1/ was only briefly

considered. Further details regarding this element are accordingly

given here. Consider the following sets of related items:

Item 153 3537 [pulu] worm

L pig] - rot, fester a little; Class 1 verb

Item 461 4098 L munde] screwpine

[mull ] thorn
 348

Item 561 341 [ a: ce] day of week

[a:l]- rule; Class 1 verb

Item 600 4402 [ba :ce] act of living

Lba:ke ] in [ba:ke ku :d] -, (woman) marries

u H_
_

Lbadiki] property

[badik]- live (happily); Class 1 verb

[ba:li] way of life

[ba d]- (child) lives, rule; Class 4 verb

Item 705 2002 [ca:vuj death, corpse

[ca:l]- die; Irregular verb

Item 718 1348 Lki:da] place below, down

the East

state of being below

Item 744 3059 [n :la] length

[ni:d]- stretch out straight (intr);

Class 2 verb

First, it should be noted that not only noun -verb but noun-noun

relationships are involved in these sets. Secondly, /1/, like /+/,

has two principal phonological characteristics: it is realised as

[l] when it is present in the terminal sequence of systematic phon-

emes; and it fails to satisfy the condition for C1 in PM3(i). Hence,

it is possible for /1/ to stand immediately in front of an element

from the system at Cons Suff. Note that, if PM3 were able to insert

an epenthetic vowel between /1/ and the following Cons Suff element,
+
the condition for PK 8 would now be satisfied:

10. LVCV( (i")C ) ] - CV( (N)C ) ] (PPiB)

 349

and the long root vowel in items 561, 600, 705, 718, and 744 would

be incorrectly shortened. In fact, no phonological mutation rules

apply to these sequences.

Hence, if we note that in certain cases the Cons Suff element is

followed by /*/ (triggering the operation of P1w12), and if we set up

the following roots:

Item 153 /pull-

Item 461 /mul /-

+
Item 561 /a:l /-
+
Item 600 /ba:l /-

Item 705 /ca:l /-

Item 718 /ki :l /- N /ke:l /-

+ +
Item 744 /ni:l / -,
+

then it will be seen that the following terminal sequences will be

yielded (omitting irrelevant details):

Item 153 /pulu/ worm

+
/pulNk /- rot, fester a little

Item 461 /mull`lte/ screwpine

+
/mull/ thorn
++

Item 561 /a:lcce/ day of week

/a:l/- rule

Item 600 /ba:lcce/ act of living

+
/ba:lkke/ in (ba: ke ku: d] -
+
/ba:l/ way of life

/ba:l/- (child) lives, rule

 350

Item 705 /ca:lp/ death, corpse

+
/cad/- die
T

Item 718 /ki:lta/ place below, down

+
/ke:lkk/ the East
+
/ki:l/ state of being belong
+

Item 744 /ni:la/ length

/ni:lt/- stretch out straight (intr)

+'

At this point, the realisation rule R6 realises /1/ everywhere as

[1]; R1 -4 also rewrite /k/ and /t/ as [g] and [d] after the nasal

/ißì /, and rewrite single /p /, /t/, and /t/ after /1/ as [v], [d],
e

T
and [d], respectively. That is to say, /1/ functions like /N/ and

the phonological vowels in respect of the voicing of following single

stops. Finally, [i] is inserted by R13 after lexeme-final consonants,

and AS9 mutates this to [u], after [a(:)v]-. The phonetic mutation

rules AS6 -7 subsequently adjust the length of consonant duration after

a long vowel, and simplify those clusters in which [1] is involved as

the first element by deleting it.

3.4 Cons Suff

The system here is shown in Figure 15. It represents a rather

fuller version of what was set up at this position in verbal base

structure; the additional elements are all of relatively low

frequency, while those that are common to the two systems account for

roughly three -quarters of all those included here. s noted above,

with respect to /+/ at Coda, certain instances of /n /, /t /, /t/, /c /,

and /k/ at Cons Suff underlie -BB base -final elements in (C »iCC bases.

It has also been seen that they may underlie base -final consonants in
 351

Cons Suff

(-e) C
r j

I a t t c k

b (b) (j)

o () (ó) (ó)

_ (P) (T) (C) (K)

n m n n

II r (r')

III I. 1

IV s

Figure 15: The systems at Cons Suff

other (C)VCC and (C)VC bases; in what follows, only unambiguous cases

of elements at Cons Suff are considered, for the most part.

/p/: this accounts for Lb] after a homorganic nasal, as in the

items 440, 788, and 956:

2731 Ltumbi] flying beetle

3361 [pa:mbi] snake

..
3316 Lparambi] large flat pasture land

It is realised as [p] when followed by / */, as in the item 838:

400 [erape] beggar

Otherwise, it is lazed, but there are very few instances of this

in the data. 'le have seen, in discussing Coda, that item 705:

2002 [ca:vu] death, corpse

shows [v] which must be a realisation of ¡p/ at Cons Suff since it

 352

follows /1/ at Coda). Additionally, item 900:

1537 [ koravu] defect

also shows /p/ of this system. In each of these cases, the preceding

environment does not contain a front vowel, and /p/ is therefore

realised as a bilabial continuant; this in turn provides the environ-

ment for the [u] -form of the enunciative vowel, by the operation of

AS9 .

Two further points have to be made here: first, that [v] in such

items as 907:

S 167 [karava] clay pot

probably does not represent any element in underlying structure (see

pp. 321 -323); and secondly, that items such as 840, 878, etc.:

799 [odeve] break

265 [arivi] knowledge

show an unusual second base vowel which cannot be accounted for by

the epenthetic vowel rule in the usual way. It will be seen, in the

following chapter, that these are complex lexemes, and that the

vowels in Question_ are in fact realisations of underlying elements.

It will also be shown that the item 881:

1374 [kodavi] Coorg (the land)

which shows final [i] in contrast (at the systematic phonetic level)

with the [l3] of [koravu], is also of complex structure, and that

there are independent grounds for stating that [i] in this item

represents /i/ (at Voc Suff) in the underlying structure. This has

the advantage of accounting for this one case of systematic phonetic

contrast between [i] and [u] outside the (C)VC +V lexeme type.

/t /: this is set up to account for [d] after the homorganic nasal

 353

in such items as 467, 787, and 949:

n
1548 [kundi] mountain

2021 [ca :ndi] sans- alwood

1700 [kodande] double handful

It is realised as [t] when followed by / */, as in the item 827:

3280 [parati] cotton cloth

There are no unambiguous instances in the data where it is realised

r
as [d following a vowel; in each case, the [d] may result from /t/

at Coda.

/t /: this is realised as [d] after the homorganic nasal in the

items 449, 780, and 950, etc.:

962 [gende]
spleen
Lga_ndej

106 La:ndij mango stone

[parande] gizzard of fowl

When it is followed by /V /, [t] is yielded, as in the item 869:

/,

2102 riti ] darkness

and it underlies [d] otherwise, as in the item 820:

1059a [karadi] bear.

/c/: this is established in this system to account for Lj]

after the homorganic nasal in such items as 431 and 945:

,

363 [inji] ginger

842 Lolarji] fly

"

(No instances of LnjJ after a long root vowel occur in the data.)

It is realised as LcJ when followed by /*/, as in the item 821:

1184 [kenacij dream

Otherwise, it is realised as LT], as in the item 926:

[ samayaj time
 354

Note that [y] here cannot be a low-level glide,

since that would

have to be labial, between [ a] and a] . This is the only unambiguous

L

instance of [yj from /c/ at Cons Suff.

/k_ /: this accounts for '[ g] after the homorganic nasal, in the

items 453, -782, and 954, etc.:

2445 [ tarige] younger sister

2368 [ne:ñgi]
plough
Lne:ngij

[kadaiigi] ditch without water

It is realised as Lk] when follows, as in the item 842:

1277 [kadake] bed, bedding

and otherwise as [g], as in the item 883:

1382 [gudigi] thunder.

/b/, /j /: these are set up here to account for instances of

intervocalic [b], and [j] (where /p /, /c /, would yield [v], [y]),

respectively. Cf. the items 798, 935, and 823:

n.
1083 [kaybi] sugarcane

1530 [kurube(n)] shepherd

[koraji] marshy place

Possibly the -[yb]- sequence in the first of these items represents

a regular derivation from underlying -/cap / -, where the medial vowel

is inserted by the epenthetic vowel rule: /p/ is realised as [ ],

which then mutates to [b] when it stands immediately after [y] (from

/c /), after the epenthetic vowel is deleted. Since this is the only

example of the type, however, deriving the [b] by rule is not an

appropriate procedure, and /b/ is set up, marked for voice, in the

system at this place. The next item shows the base -medial vowel

-[ ] -, which is possibly accountable for by rule as the appropriate

 355

realisation of the epenthetic vowel /i/ in this environment (i.e.

when followed by /b/): but this main is the only example of its

type, and it seems much more probable that [u] here represents under-

lying /v./ in the structure of what is a complex lexeme. Complex

lexemic structure is certainly the case with regard to item 919:

t

3613 [perija] increase, excess

which shows a second vowel Li] which cannot be accounted for by the

epenthetic vowel rule.and which therefore has to be derived from

/i/ in the underlying representation of a complex structure. These

two items therefore fall within the scope of Chapter 6.

/p /, /oc /, /o/: it was discovered (pp. 260 -270), when analysing

verbal base structure, that the system of elements required to

account for all clear cases of Coda representations contained only

/n_/ and /n/ of the row of elements defined by the feature n . It

was also found that base -final nasal consonants were in any case

rare, only -[mm], -[ nr_], [nn_], -[] occurring, in (C)VCC bases,

and -[n] in one (C)VC base. Of these, some -inn] and -Lnn] could

be accounted for by /n/ and /n/, respectively, at Coda. For the

others, it was decided to set up the sequences -/Np/, -ATV, -/W,

o ô
and - /W, and some evidence in support of this was considered.

PK7 operates on these sequences:

MP

ó
11. N (P1.17)

to yield the appropriate nasal consonants, in those cases where the

element following /J/ has not had PM2 applied to it (i.e. where it is

followed by /*/) .
 356

As far as the nominal system is concerned, the situation is

much less constrained and the picture correspondingly less clear.

It has been seen that /m/ is reouired at Coda, on the basis of 5

clear instances; /n/ and /n/ are also set up in that system, but

/7/ and /A/ are not required. The /m/, /n/, /n/ at Coda therefore,

may account for final -[mm], -[nn], -[nn], respectively, in

(C)VCC bases; and for final -[m], - n], -[n], respectively, in

(C)VC bases. But we also have to account for a few instances of

base -final -[m], -[n], -[n], in ( C)VCC and (C)VCVC bases in the

nominal system, and for these the marginal elements /m/, /n/, and

/n/, respectively, may be required at Cons Suff (see below). There

is also the possibility, however, that certain -[mn] /-[m] result

from -/NR/, and (in particular) that [ññ]/ -[ñ] result from -/Nc /;
0

note that there is no unambiguous case for setting up /n/ at

either Coda or Cons Suff. The velar nasal consonant does not occur

in the nouns under consideration here.

Concerning ¡Np /, there is some supporting evidence in the

o

nominal system from the alternative form of item 222:

1450 ['_{ummi]
n
coomb teak
PO ' ktmbili ]
,

This suggests that -[mm] in the first form is, diachronically, a

development from the homorganic nasal plus stop type of base -final

element shorn in the second form.

Concerning /N8 /, the following items are involved: 216, 218,

2.30, 258, 595, and 742:

DIA 85 [kaññi] virgin

1371 [ku ij (Coorg) child

w
3097 [nenanl]
chest (of body)
N Lnersi]
 357

Nw1
kunne] unmarried girl

3572 Lpu:ñe] cat

N
na:na] coin

For certain of these, evidence from cognate forms in related languages

supports the analysis proposed here; DED has the following entries

for 1371:

"Ta. kuñcu 'young of birds and various animals'; Na. kunnu,

kuñci, 'young, small, infant".

For 3097, it has the following entries:

"Ta. neñcu, neñcam, 'mind,..., breast; bosom, chest,...'; Ia.

neñcu, neññu, 'heart, breast".

The alternative Lalayalam forms in each case are particularly relevant

to Coorg.

Internal evidence of a sort is available in respect of item 395,

which has a number of related forms:

4119a L miñña] state of being in front

L muñgay] forearm

(Cf. [kay], arm, hand)

[mumbi] predominance

[muppode] previous

[mane] sharp point

ß_119b [moniyandi] day before yesterday

We have to recognise here the root forms /mi / -.4 /nun/ N /mu/- ...

/mon /- (see the following chapter for further details); Lmiñña]

shows /Nova/ after the first of these. We therefore extend the

operation of FM7 to include the following subrule:

12. Fcc n (w7)

o

Finally, // is set up here to account for the items 877, 897,

and 903:
 358
_

606 [orakil sleep

4112 [ muriki ] torsion cord (of drum)

432 [ elaka] uprooting from position

For these forms, see the discussion of Coda in the preceding chapter.

It is of course possible that they show /k/ at Cons Suff followed by

/*/; but it is more natural to set up /o/ in their underlying

structure, which thereby corresponds closely to that for the related

verb forms, and explains why it is that all these nouns show a

voiceless velar stop in base-final position.

/P/, /T /, /C /, /Y /: these are set up to account for a relatively

small number of items which show a voiceless stop after a homorganic

nasal. Two of these items, 436 and 498, show the identical element

/T/ which had to be set up at this position for their related verb

form:

1408 [kunti] lame person (fem)

[kunte(n)] id. (mast,)

[kunt] - be lame

The other elements may be illustrated by the items 452, 429, and 815:

[jompe] bunch (keys, flowers, etc.)

3597 [anti ] tile

1034 [kavìiki] vital spot

/m/, /n/, /n/: these elements are set up here to account for

instances of base -final -[m], -[n], -[n], respectively, in (C)VsC and

(C)VCVC bases, as found in the items 814 and 824; 796 and 915, and

831 (the only instance):

4189 [mayme(n)] cross -nephew

2529 [talami] hair (of head or body)

4339 [bayne] eggplant

 359

DIA 230 [nevana] pretence

3260 [barani] large clay pot

There is no evidence to suggest that any instances of a nasal

consonant in this position derive from the NP0 type of cluster

considered earlier, though of course this possibility may be realised

after further data collection.

Note that item 860:

3511 [pudume] wonder, marvel

is probably to be taken as a complex lexeme, since it shows a second

vowel Lu], in respect of which exactly similar arguments hold as

were given above re;.:arding [kurube(n) ] .

/1 /, /1 /: these are set up to account for the relatively few

cases of base -final [1], [1], in (C)VSC and (C)VCVC bases, as in the

items 792 and 844:

1125 [gavli] big lizard

941 [kadale] Bengal gram

¡rI: this element is realised as [r] in itsms such as 789 and

849:

DIA 30 [ ayri ] low caste Coorg, carpenter

1423 [kudire] horse

/s /: this is realised as [s], in base -final position of (C)VSC

and (C)VCVC bases, as in the item 896:

[manasi] conscience, mind.

3.5 Voc Suff

The system of elements operating at this place in the base

structure is shown in Figure 16. As with the verbs, this system is

established in order to account for the final vowel in (C)VC=V bases;

 360

(u)

e a

.Figure 16: The system at Voc Suff

however, as we have noted in the introductory section to this chapter,

an element at this place in nominal lexeme structure may be followed

by an element at Noun. This situation of course could not arise in

verbal lexeme structure where I and I (from Verb) are rewritten

simply as abstract class features, with no direct segmental realis-

ation at the systematic phonetic level. As was seen earlier (pp. 303-

306), none of the elements at this sÿstem expresses distinctions of

gender.

/i /: this accounts for all instances of base -final [i] in

(C)VCVC lexemes, as in the item 7:

63 adi] place below

As noted earlier, it is also set up in the underlying structure of

item 1 (and. 2 -6) :

479 [oy] (finger, toe) nail

and is deleted by the phonetic mutation rule AS5. It also accounts

for base-final [i] in those (C)VCVCV(C) lexemes where the second vowel

is a base -final element (i.e. not inserted by the epenthetic vowel

rule), and the following consonant is not an element at Cons Suff, but

the palatal approximant Ey] inserted by the phonetic mutation rule

AS8. The following are the items concerned: 911 -2, 917, 920, 933, and

936:
 361

DI.: 199 [ tali ya] plate

2869 [todiya] small garden

DI! 257 [paliya] wooden stool

3497 [podiya]
sari
[ podeya]

1379 [kudiye(n)] man of toddy tapper caste

994 [keniye(n)] trickster (mast)

Note that the last two items cited here are related to items 20 and

24 which show no element at Noun:

1379 [kudi] family of servants living in

one hut

994 [keni] bird trap, cunning

/e/: this is set up to account for base-final [e] in (C)VC +V

bases, as in the item 68:

16 e] paddy seedling

It also accounts for base -final [e] when it is followed in underlying

representation by an element at Noun; in all such cases, the palatal

continuant [y] is inserted by AS8 after [e]. The items involved are

914, 928 -9, and 940:

2957 [nadeya] passage to shrine of temple

436 [eleye(n)] youth

510 [odeye(n )] husband

3714 [poleye(n)] low caste man

The first and last of these items are related to (C)VC -± -V lexemes

(i.e. not showing an element at Noun), namely the Class 3 verb 18,

and item 118 of the nouns:

2957 [nada] - walk; Class 3 verb

3714 [pole] pollution

Note that the continuant inserted by AS8 is palatal only after the
 362

front vowels, [i] and [e]; after [a], [i] and [u] it is the bilabial

continuant. Hence, item 930:

510 [odeve(n)] master, god

shows base -final [v] from /p/ at Cons Suff.

/a/: this is set up to account for base -final [a] in (C)VCV

bases, as in the item 157:

560 Lora] mortar of stone or wood

It also underlies [a] in this position when it is followed by an

element at Noun, after the bilabial continuant [v] inserted by ß1S8.

The items concerned are 907, and 943:

S 167 [karava] clay pot

4270 [badave(n)] poor man

The last of these is related to the following iterd, which is a complex

nominal lexeme in Ltana] (the first part of the complex shows no

element at Noun):

4270 [badatana] poverty

/i /, /u /: the contrast between [i] and iu] at the systematic

phonetic level is marginal; for most items (20 altogether), we can

set up the element /i /, which is realised as [u] if the preceding

Coda element is /r/ and the root vowel is either /u/ or /o/ and as

[i] in all other cases. This yields 5 instances of [u], in the items

147 -50, and 155; and 15 instances of [i], as in the item 139:

1483 [kuru] seed

1537 [koru] defect

[ ko ru] management

[ guru] teacher

[suru] beginning

1070 [kari ] intestines

 363

The a11Sati Oi 'lac R9 hanrIles this CO^d; t'on,,d variation;

tion; we

therefore extend ?9 with the fo=1o:

13. :i - [back of ton:-.t. hi :hest, 1i-is

On the other hand, /'_/ _:t 'Too 3u 7'f is set u;) to _ecount for the

base -final vowel in the items 145 -6, and 151 -4:

202 L alu] shell, cowrie

953 Lkadu] mustard seed

2460 [tadu] lateness

2959 Lnadu] middle

357 L ouiu] worm

3986 Lmalu] chillies

/u/ is set up in the matrix shown in Figure 16 under the second-order

feature 1tß (i.e., is marked for lip - rounded articulation).

However, when an element at Noun follows, the contrast between

[i] and [u] is lost; this may be treated as a result of the fact that

the bilabial continuant is inserted after [i] or [u] in this environ-

ment by AS8 (since they are not front vowels, defined by the feature

j); a further phonetic mutation rule adjusts [i] to [u] in the

environment of following [v] (S9). Just two items in the data show

[u] in this position (which may therefore be the realisation of /i/

or /u/ at Voc Suff) : 918 and 932:

3295 [ paluva] coral

1280 [kidtuJe(n) ] hawk

Finally, note that in this system /1/ occurs not e a type of

/i / marked for retracted articulation but as a distinct, unmarked

element; the same was four.:_ regarding /i/ at this position in verbal

base structure.
 364

3.6 Suffix2

This is represented by /-/ as described in the previous chapter.

It is set uî, here, as in the verbal system, to account for the

occurrence of base -final voiceless stops in Positions other than

immediately after a short root vowel; see for example, the contrast in

these positions in the items 640 and 648; 801 and 800; and 883 and

888:

1563 Lgu:di] pigpen, fowl- house, nest

2323 (,cu:ti] thumb-forefinger span

2537 [tavdi] inner skin of rice grain

u _

1973 Lcavti] footprint

1382 L;udigi] thunder

3451 ipuriki] mosquito

As with /*/ from Suffix2 in the verbal system, this element plays no

part in the statement of lexical relationships.

3.7 Nom

This is represented by /*/.

We may recall here the general considerations set out in the

previous chapter concerning the stating of lexical relationships

(_ .214 -215 ). It was argued that it is necessary to have a model

which is generatively linear where this is indicated by the data, and

generatively parallel otherwise. It was later shown that certain

voice -related verbs could be stated in a linear way, and to this end

/*/ from Volit1 was set up and distinguished from /*/ from Suffix2.

In the nominal system also, it is possible to generate certain

lexemes linearly; that is, they are relatable to other lexemes as

derived forms to directly- generated forms. This, of course, represents

the strongest hypothesis in respect of any lexical relationship; while

 365

the weakest is that formalised in the generatively parallel approach,

which simply states that two lexemes are each derived from a common

root. The nominal le::emes which it is possible to derive linearly

are set out below. It will be seen that each of these save one is

related to a verbal lexeme, from which it is possible to say that it

is derived, by the process of suffixing Nom, and the consequent

operation of LS18:

i Ì

14. # ...Nom ] -p, # a # ...Nom ] Noun # (LS18)

: +t, +v]

Note that this rule formalises the fact that these are deverbal nominal

lexemes; it applies after the full development of the node Verb

(under Classifier) has taken place, and reclassifies the lexical

structure in which Nom is operative as a nominal lexeme. This is

allowed for by LS8, which does not restrict the occurrence of Nom to

those underlying structures in which Classifier is rewritten as Noun

(see the discussion of the generative component, pp. 313 -315).

However, both LS8 and LS18 allow for Nom also to mediate noun -noun

relationships, and one example of this is found in the data: see the

last pair of items below.

After all LS rules have applied, and before the rules of the

phonological component begin to operate, the underlying structures of

the derived nominal lexemes are as follows (omitting irrelevant

details):

Deverbal nominal lexemes:

432 #L el +
ó + ' - 3 a# uprooting from position
606 # R or + k + * ]] # sleep
1 21 9 #[ ka:c + * 1 a# heat
 366

r'
1353 #(l ki: + *1* torn off piece
r

1562 #[ ku:t + * ] (a)# gathering (quarrel)

2680 #E[ ti: + Nt ± * 1 a# excrement

2927 #E to:t + * 1 a# estate

2379 #1 na:r + * 3 a# (bad) smell

3348 4 pa: t + * 1 # song

4112 #4 mur + '6 1 # torsion cord (of drum)

4179 #(L me:c + *1 1# fodder

Denominai nominal lexeme:

1206 #E: ka: t + * ]] i# bison

3.8 Noun

It was seen earlier that the system of elements which operates

at this Place in lexemic structure is relevant in a not wholly

thoroughgoing way to the determination of certain referential pheno-

mena; to the extent that gender exists in Coorg, it is expressed at

this place in the lexeme, and hence the development of Noun is

relevant to the semantic specification of nominal lexemes. We have

also seen that the systematic phonetic elements occurring at this

place may have a marked or an unmarked function with respect to refer-

ential properties; hence the first rule which develops the node Noun

is as follows:

15. Noun -> (LS19)

UNk

Note that this rule, applying. after LS18, rewrites Noun in

identical fashion regardless of whether it is yielded directly from

Classifier (by LS3) or from the operation of Nom on a verbal class

specification from I, II.

 367

Marked function may be in respect of human male, human female,

plural, inanimate, or nonhuman animate properties of the referent,

so the next rule expands Ira as follows:

IrîA.SC

FIi;I,'i

16. Mk -+ PLUP (LS20)

INAN

AN

;There there is no gender marking, the nominal lexeme may show

final [i], [e], or [a] at this position, or the enunciative vowel [i]/

[u]. For the first three of these elements, we set up ¡i /, /e /, /a/,

respectively, in the following rule:

17. mac (LS21 )

The parentheses around the braces indicate that Mk may have no real-

isation at all, in which case [i] /[u] will be inserted when the

realisation rules operate.

Finally, we require the following rules for the marked cases:

18. 1_kSC +masc ]- (LS22)

-"7 -masc

19. FEY -P /e/ -mas c 4 (LS23)

/alJ -masc

20. PLUN -+ t /al/ +plur - (LS24)

21. INAN - .[ /a/ -anim 4 (L825)

 368

22. AN -+ t Aa/ +anim 4 (L326)

Note that, for the gender marking to be made available to the

semantic component, the specification of the appropriate phonological

forms for these abstract elements has to be carried out within a

labelled domain, defined by the boundary elements t 4 .

4.0 The rules cited in this chapter

4.1 The lexeme structure rules

Voice Suffix / Verb

LSB. Affix2 -4 I; om

Suffix.,

Suffix2

Volit1

LS10. NOT: Peak + Volit2

Voc Suff

` Nom i

{Li, - P
LS18. ..Nom ..Nom Noun
+
t , + uj

LS19. Noun
UT11c

IíASC

FEIN

LS20. ilk PLUR

IìúAN

A.N
 369

LS21. UM: )

LS24.* TLASC +masco

-E /1/ -masc .-

LS25. FEE -+ . /e/ -mast;

-E Al/ -as c.

LS26. PLUR -+ -E Al/ +plur f

LS27. IMAN -, t /a/ -anim4

LS28 AN - -t An/ +anim-

For LS22 -3, see the list of rules at the end of Chapter 6
(P. 427).

4.2 The phonological mutation rules

P'3(i) C1 + C2 - C1 + VC
(C1 ; + , 1 ; C2 1 ok)
+

fe
PM4. 0+ a
0

PT'_7 . 1`+ t
O
to

P1I8. ICV( (I'T)C )

CV( ( )C )
 370

4e3 The realisation rules

[occlusion, groove friction]/ B(j)

R7 IU -+ [occlusion] / A "Y,j()

[obstruction] / A

Lno articulatory effect] 1! [m t, -op]

' ()A j ()" (L )ABr

lips rounded] / +
w
Xw() B(r)II()

R9. X - lback of tongue w

highest,
I B(1
lips / 1

unrounded] B ( ) j ( )" (L)4Br

2
C

[front of tongue highest, lips spread] / j

constriction,

back of tongue
R13. Insert:
highest,

lips vnrounded

4.4 The articulatory sequence rules

 571

^:S9. [lips unrounded] Llips rounded]

OP

f constriction, obstruction, open tract,

back óf tongue bilabial place

highest, of articulation

l VP

o_en tract, v.cs obstruction, constriction,

continue to
vibrate, bilabial p. back of
of a. tongue
unaltered highest,
vocal tract

n
M
constriction, constriction

back of tongue midopen tract

AS10. Erase:
highest,
open tract
lips unrounded
eh
 372

CHAPIÚR SIX

The Structure of Complex Lexical Items

 373

1 .0 Introduction

''re now consider those forms (nearly all of them nouns) which do

not show the canonical shapes on the basis of which the analysis in

the previous chapter was established. However, it turns out that

the notion of a Base (with loot, and optional .! ffix1 and .ffix2) and

Classifier, is relevant to the more complex noun forms also. That

we investigate here is how far the analysis presented earlier may be

extened to include other base- shapes, and what modifications are

required to the generative component of the lexicon to alloy: for

such extension. After this, the four main types of complex nominal

are examined in some detail; and it is shown also that the modified

generative component is able to account for two of the three verbal

lexemes which were noted as falling outside the scope of the

discussion in Chanter 4 (see p. 208 above, and pp. 392 and 416 below).

Finally, a further case of possible linear generation is considered

(pp. 420 -427) .

1.1 Preliminaries to the analysis of complex nominal structure

The noun forms we are considering here fall into four groups.

These are briefly considered here in order to throw. light on the ways

in which the generative power of the model of lexeme structure has to

be increased.

1.1.1 Set 1

Ioun forms in this set are composed of (at least) two lexemes,

conjoined into a unit which behaves syntactically like a single

lexeme and which in many cases has a single reading assigned to it

in the lexicon which is not analysable simply in terms of the combined

readings of its component lexemes. The first element of such a

 374

complex may consist of a verbal or a nominal lexeme; it stands in

relation to the second element, which is always a nominal lexeme,

as modifier to head. In the syntactic string, the complex receives

a single set of case and number affixes, which occur in lexeme -

final position; hence the first lexeme of such a complex, even if it

is a nominal, never shows such inflectional marking. It should be

noted here that we are using the tern 'lexeme' in a _purely formal

sense; a liven lexeme may occur in such a complex as we are consider -

inn but nowhere else in the langua.e, and in this case its status as

a lexeme is determined by the requirements for its formal generation.

One formal property of the simple lexeme is that it shows one main

stress, on the first syllable, which contains the root vowel. In the

examples that are cited_ in this chapter, main stress is indicated at

the systematic phonetic level by ['] before the syllable concerned;

stress assiEnment in complex nominals is frequently a valuable guide

to their internal structure .

E amples of Set 1 nominals are:

['atta'betta] stuff and nonsense

1709/1012 ['kotti'katti] billhook

178/805 ! ' akki'yotti] rice roti

3919/106 I'ma:ñge'ya:ndi] mamo stone

302/3856 ['a:nake'mara] castor-oil plant

u
302/726 L'a:na'kenne.j castor oil

1348/4230 L'ki:'la:ndi] last year

(wherever possible, DED, DEDS, and DBIÂ references are given; where

only one lexeme in the compound is thus identifiable, the reference

number is placed before or after the slash according to whether the

lexeme concerned is the first or second member of the complex.)

 375

In the first of these items, neither

[ 'atta] nor r

'betta]

occur as independent lexemes elsewhere in the languapme; the second

item is made up of L'ko.ti] which occurs only here, but is

apparently related to

1709 Lkodali] sickle

by the root /kot/ -. L'katti] occurs independently, with the sense

'knife'; however, most informants regard L'kotti'katti] as a sort

of L'matti], 'axe'. Thus far, stress has been seen to occur with

the lexeme -initial element of each of the component lexemes; but

the next two items show stress occurring with -L'yo] -, and -L'ya: ] -,

of the second lexemes, where the palatal continuant is not a part

of either lexeme but a low -level glide introduced by rule (Â38) .

The next two items show how stress marking, which is determined with

reference to the syllable at the systematic phonetic level, may fail

to coincide with a lexeme boundary; in the first of these it does so

coincide, starting first with L'a:nake](only occurring in this

complex), and then with ['mara] (an independently occurring item,

with the sense 'tree'), but does not do so in the second item with

respect to the second element; L'a:nak]- in this item is related to

a' a:nake] of the preceding item, and the second element is identif-

iable with ['enne], an independently occurring lexeme, with the

sense 'gingily oil'. The same situation is found in the last item

also: the first element of the complex corresponds to the lexeme

n fl, n_
'state of being below'; while the second lexeme i a:r_di],

'year'. Concerning the first elements of each of these items, we

may set up the lexemes ['a:naki] and L'ki:li]: the first of these

occurs only in the complex form illustrated here, while the second
n +r

may be identified with the independent lexeme L'ki:li], 'state of

 376

being below'. In each case, the enunciative vowel is lost when the

second element of the complex has an initial vowel. Note that in

such cases the lexeme is isomorphic with the base at the systematic

phonetic level; but these items are nevertheless treated as

constituting two lexemes in order to avoid the peculiar distributional

pattern that would unjustifiably exist otherwise, where no lexeme

which did not show an element at Noun could occur as the first

element of a complex nominal where the second element was a lexeme

having an initial vowel.

1.1.2 Set 2

Nominal lexemes in this set show a base -like first element,

followed by a second element which shares certain properties in

common with the lexeme. Thus for example

726/726 ['ellene] gingily oil

looks as though it is composed of a first element ['ell]- plus a

second element -gene], and, apart from the fact that there is only

one stress for the whole item, its composition looks rather similar

to that found in nominals of Set 1: thus, the first element could be

taken as a lexeme rather than a base (with the enunciativo vowel

lost because of the initial vowel of the following element, by the

operation of riS10), while the second could be analysed as a type of

(C)VC +V nominal (such as L'mara]). However, this not only fails to

account for stress assignment in such cases, but also ignores the

relationship that holds between the second element -[ene] and the

independent lexeme ['enne] (with the same sense as L'ellene]).

There is a further fact to account for in this case, however; inform-

ants regarded the form ['ell'enne] as an alternative pronunciation

of ['ellene] (the first alternative is recorded by DED, as

I, Ir

"ell enne ").

 377

Consider also the item:

2513/112 L ' tammane (n) ] younger brother

The last element here is apparently -Lane(n)j, which is related to

the independent lexeme L'anne(n)], 'elder brother'. Note that if

we recognise this relationship by treating -Lane(n)] as in some

sense derived from (as a 'reduced' form of) E'anne(n)], we shall

be able to maintain the restriction formulated in the previous

chapter against -[e(n)] occurring at Voc Suff. [stamm]- does not

occur elsewhere in Coorg, but is recognisable as having the formal

properties of a base, and is cognate with independently occurring

lexemes in a number of related languages; cf. the DED entries for

2513:

"Ta. tampi 'younger brother...'; Tia. tampi, tampan 'younger

brothers; Ka. tamma 'younger brothers".

Consider also:

3112/178 L'nellaki] (inL'nellaki 'bolica], sacred lamp).

Each of the component elements is identifiable here; L'nell]-

corresponds to L'nelli], 'paddy', while -Laki] is a 'reduced' form

of L'akki], 'husked rice'.

In the last two items cited above, no alternative pronunciation

'tam'manne(n)], *L'nel'lakki] is possible.

Finally, note that in an item such as

1/S 630 L'appara] that side

no 'reduced' form of the second element is possible, in segmental

terms; it is related to the independent lexeme L'para], 'side,

area', with a short medial consonant. This is the case with all

(C)VC +V lexemes in this construction. However the fact that the

item above belongs in this set of nominals is signalled by the lack

 378

of main stress on -[para] in this construction; it shows its

'reduced' form suprasegi entally.

Thus, with regard to this set of nominals, we have to account

for the notion of a 'reduced' lexeme, and also show the similar-

ities that hold between this type of construction and that found in

Set 1. In a large number of cases, the forms of lexical items in

Set 1 are identical to those of Set 2, save for the distinct

construction involved; this gives rise to the following altern-

ations (the list is not exhaustive, as not all informants agree

concerning certain items):

Set 1 Set 2
n n u u
726/726 'ell'enne] ['ellene] gingily oil

4057/46 ['mutt'ajje(n)] ['muttaje(r_)] great grandfather

2190/710 L 'cind'eli] ['cindeli] (house, field) mouse

768/178 ['e:laki] cardamom seed

3740/133 L'bo:'jappe(n)] ['bo:jape(n)] father's younger

brother
n

3740/232 L'bo:'javve(n)] ['bo:jave(n)] mother's younger

sister

/178 [ ' avl' akki ] L pounded rice

218/ ['udi.'tale] [' uditale] ironwood tree

3804/DI.'_318 [' madi' batte] ['madibate] clean clothes

929/1371 ['kade'kuilñi] ['kadekuñi.] youngest child

1705/423 ['kodi'yele] ['kodiyele] betel leaf

167/3911 L'ara'mane] ['aramane] palace

['kadi'kone] L'kadikone] sheaf of paddy

?1683/ ['kay'kane ['kaykane] spoon

1522/3911 ['kuli'mane] ['kul imane] menstrual hut

2655/ ['tir'gani] ['tirigani] pulley

 379
Set 1 Set 2

3322/3893 L'pani'male] L'panimale] drizzle

3796/1683 L'mana'kay] ['manakay] elbow

3608/3768 ['pom'makka(i)] ['pommaka(1)] women, wives

II
1348/3558 ['kip'puni] [' kippuni] lower level in field

4173/3558 ['mep'puni] ['meppuni] lo:,:er level in field

4402/3911 ['ba:'mane] ['ba:mane] house into which bride

marries

Finally concerning Set 2, the first element of the complex never

shows an element at houn, and never shows the enunciative vowel [i]

even when the second element begins with a consonant, as in the case

of ['pommak.a(1)]. Hence the highest node that we need set up to

account for the first element is Base.

1.1.3 Set 3
number of items can only be analysed as composed of a base

(first element) plus lexeme (second element). This set has been

kept as small as possible; thus all items which are ambiguous as

between Sets 1 and 3, by virtue of the first element being of

(C)VC-FIT form (which may be analysed either as a base or as a lexeme)

have been included in Set 1. Since Set 3 would in any case be much

smaller than Set 1, this procedure only follows the trend of the

data. Items in this set show the same stress pattern as those in

Set 1, but comprise a first element which is like that of Set 2

items in that it never shows the final enunciative vowel (even when

followed by a second element with an initial consonant) or element

at Noun. For example:

3302/3057 ['pan'ni :ri] perfume

4057/308 ['mut'ta:y] great grandmother

4173/3930 ['me:'ma:di] top storey

 380

The second element in each of these items is an independently

occurring lexeme in the language: l' rr

ni :ri] , 'water'; ['ta:y],

'grandmother'; and ['ma :di], 'upper storey'. The first element of

the first item corresponds to the lexeme L'pani], 'fever'; that of

the second to L'mudi], 'state of being old'; and that of the third

to the first element in another Set 3 item:

4273/1238 ['me:r'ga:li] top Part of foot

n_
(cf. ['ka:li], foot, leg).

In the case of certain items of the form (C)VCVCV, it is not

possible to decide whether they represent the Set 1 pattern or that

of Set 2. For example, the loanwords:

sky

Da:'sa:mi] individual

L'sa:'ma:na] things

may be analysed as showing an initial element (C)VC which is a

lexeme, followed by another lexeme VCV, with predictable loss of

the enunciative vowel by LS10. In all cases where such an analysis

is possible, the items concerned have been included in Set 1. Just

the item ['me :'ma:di], and the following two:

4057/DIA 219 ['mu :'de:vi] quarrelsome woman

jDII 179 [ 'me :'jo:di] pair of socks

can definitely be placed in this set, as comprising the initial

elements ['mu :], ['me:], respectively, followed by the lexemes

['de:vi], 'goddess', ['jo :di], 'pair'.

1 .1 .4 Set 4

This is the final set: it contains those items which are

apparently to be analysed as comprising a lexeme (first element)

followed by Classifier. Thus for example

 381

N 11

['kandave(n)] stranger (masc)

is entered in DED at 985 (with a query as to whether it is related

to 1209), but my informants' intuitions linked it only with:

1209 ['ka:n]- see, seem; Class 3 verb

By this analysis, it is probably made up of the first element

['kind`,] -, itself made up of ['kand]- (past stem) plus -[a] (the

negative suffix), followed by -[e(n)], with a bilabial continuant

inserted
11

by the low level rule AS8 between the vowels [ aj - ]

Against this, however, it may be argued that it does not account for

the nasality on the last two vowels, and that the structure of this

item is therefore as follows: a first element L'kand] -, followed by

y N
11

an element related to the masculine singular proform L'ave(n)].

In this case, of course, it would be a Set 2 nominal, thus accounting

for the stress reduction on the second element.

But consider the following:

DIA 106 L'kupiya] Coorg man's coat

1388 L'kuttuva] big copper pot

1905 L'cattuva] flat wooden spoon

3637 ['po :liya] basketry box (gift to bride)

n
L'su:riye(n)] sun

where the second element seems to be one of the Noun elements -[a],

-Le(n)], preceded by a predictable glide. Further evidence which

may be construed as supporting this analysis is found in forms such

as:

834b L'okkace] state of being together

3115 L'nellike] gooseberry

3467 L'pi:reke] very bitter small cucumber

where the first item has the alternative form ['okka], the second is
 382

recorded by D D in the same entry as:

"Ta. nelli 'er<blic ryrobalan...; 14a. nelli id. ",

and the third is related in my data to L'pi:re], 'sponge gourd'.

It will be noted, however, that this analysis for these items

entails setting u- elements at Classifier with an initial consonant,

and this fact may be taken as a contingent disadvantage of the

proposal. On the other hand, there is evidence to suggest that this

type of Classifier element is reeuired in the system anyhow;

consider nominals such as:

799 L ' odeve] break

3613 ['perija] increase, excess

3686 L'podike] cover

3686 L'podepi] covering, blanket

3714 jpolati ] low caste person (fem)

4524 L'bolica] lamp

In these items, the medial vowel cannot be accounted for by the

epenthetic vowel rule (see the previous chapter, especially the

discussion of Cons Suff), and hence has to be derived from an element

at Voc Suff. In a number of cases, in_ciudin all those illustrated

above, there is internal evi...ence to support such an analysis:

799 t' ode]- break (intr) ; Class 2 verb

3613 ['peri]- increase (intr); Class 5 verb

3686 [':;ode]- thatch, cover; Class 5 verb

3714 ['pole] pollution

4524 ['boll] light

These lexical relationships clearly suggest that the correct analysis

of, for example, E'pociike] is ['podi]- followed by -[ke], where the

final element is apparently a Classifier element with an initial

consonant. However, it should be stated that this set of nouns

 383

re :presents the area of greatest uncertainty in the fra..e;:or'_ of this

stud:.

1.2 The lexeme

The foregoing considerations require certain modifications to

be made to the . enerative component of the lexicon. Taking the

simplest case first, nominals of Set 1 must be accounted for by a

derivation which starts with the rule:

Lexeme -* # Lexeme# #Lexeme# (L1 )

This straightway accounts for those nominals of Set 1 which

consist of just two conjoined lexemes (the vast majority); a schema

involving three conjoined lexemes, or more (the data shows no

evidence that more than three are ever required) is of course

derivable from a reaPplication of 1 above, and requires no further

modification to the rules.

Nominals of Set 3 require a rule of the form:

2. Lexeme -+ Base] #Lexeme# (LS1)

Nomina.ls of Set 4 require the rule:

3. Lexeme #Lexeme# Classifier (LS1)

Thus far, the proposed rules for expanding Lexeme (including

LS1 set a in the previous two chapters) may all be collapsed as:

*Lexeme* Classifier
4. Lexeme -4 (Lsi)

f Base]) #Lexeme #

where all options are taken up.

However, nominals of Set 2 are still to be accounted for. -;Te

have seen that their second element is a 'reduced' type, and we

formalise this here by setting up the node Postbase, marked off by the
 384

distinct boundary elements / / . This requires the rule:

5. Lexeme - Dase] /Postbase/ (Lä1)

The node Postbase has a rather limited set of realisations; at the

systematic level the only possible form for elements derived from

it is (C)VC -+-V. In this connection, two points have to be made.

(a) There is no possibility of a long first vowel in postbase

elements at the systematic phonetic level. This could be analysed

as either the result of a restriction on the occurrence of the

length element /L/ under Postbase, or the result of an obligatory

length- reduction rule. The following items su:jest that the latter

alternative is correct:

993 `' kannadi] glass, mirror

1192 nka:pale(n)] (palace) guard

37 ['a_ñgadi] shop, stall

Concerning the first item, cf. the DED entry at 993:

"Ta. kannati, kannati, 'mirror of metal or glass...(( kan, 'eye'

+ ati, 'mirror, crystal');...^lu. kannadi, kannadi, 'glass, mirror... ".

The second item shows L'ka :p]- (as in ['ka:'pa :d]- 'protect'; Class

1 verb) followed by Lale(n)] which appears to be related to a:le(n)]

in

4045 mutta:le(n)] fool (mast)

and certain other nominals with human male referents. Finally, the

last item here is cognate with the following forms recorded by DED

at 37:

"Ta. aii:ati, 'bazaar, bazaar street';... Ka. aiigadi, 'shop, stall'".

(b) There is a voicing contrast between stops in the medial -C-

position of postbase elements; see for example:

 385

178 -[aki] (in 768/178 ['e:laki], cardamom seed)

-L ise (n) ] (in 959 Ì ' gettige(r_) ] , clever man) .

This is best handled, in view of the sort of relationship that holds

between -L aki ] and Paled] 'husked rice', ir_ terms of the automatic

gemination rule formulated in the previous two chapters. Hence

-Lige(n)] may result from a -BB type of cluster (derived from frk /)

in the following terminal sequence: -/iggan /. Alternatively, we

have to alloua for the possibility that it derives from the terminal

sequence -Akan/. In the first case the reduction is with respect

to the consonant cluster -/gg/-; in the second it is with respect to

the long vowel. But if we allow for long vot :els in this way (see

also (a) above), then we also have to set up Suffix2 in order to

preserve the voicing contrast in stops in cases where long vowels

occur in the terminal sequence.

Everything that has been noted above about postbase elements is

consistent with allowing Postbase to be expanded as Base plus

Classifier: hence we require a further LS rule of the form:

6. Postbase -+ ,Base]/ Classifier (LS1)

With this, the modification of the generative component is

completed, and the final form of the first LS rule, superseding that

given in the earlier chapters, is, as follows:

r
Lexeme Classifier

Base Lexeme

7
Lexeme } (-
Postbase Basel /Postbase/

ÌL Base] Classifier
 386

2.0 The structure of complex nominals

We now examine in rather more detail the formal properties of

the complex nominal lexemes, particularly with respect to the later

rules (realisation rules and phonetic mutation rules) required; we

take largely for granted the phonological systems and rules for

generating the component Base Classifier structures which are

involved in these nominals, since these are essentially the same as

described in the previous chapter. '.e are concerned here mainly

with nominals of Sets 2 -4, since those of Set 1 are quite straight-

forward for the most part.

2.1 Set 1: Lexeme Lexeme

The : eneralised schema which accounts for most of the nominals

of this type (Set 1) is shown in Figure 1:

#Lexeme #

Lexeme* #Lexeme#

base]]
i .\_ ,
Classifier 1Baséll

'
Classifier

/ 1 -.
Root ( - 5 ffixl) (Affi x2) Noun Root (+Affixl ) (+Affix2) Noun

Verb
(onset) Nucleus Onset) Nucleus
/

Perk (Coda) Peá_k (Coda)

Figure 1: The generalised structure of complex

nominal lexemes, Set 1

2.1 .1 Lexeme = Base'4 Classifier

It will be seen below that certain nominals in this set require

that a constituent Lexeme node be expanded as Base Postbase,

 387

Base Lexeme, Lexeme "Classifier, or Lexeme "Lexeme, as allowed for

in the form of LS1. For the moment, however, we concentrate on the

more common exPansion of each constituent Lexeme node as

Base Classifier. Within this type, each Base node is represented

by the full set of possible canonical forms:

(C)VCC

(C)VNC

(C)VC

(C)Vi1C

(C)VSC /(C)VCVC

(C) VSi;C /(C) VCVNC

(C)VCV

The examples discussed below are arranged in this order, first with

reference to the first element, and then to the second.

1. (i) (C)VCC- and (C)VCC-.

3726/3112 'polle'nelli] grainless paddy ear

178/805 L'akki'yotti] rice bread

rr

Cf. 3726 L'polli] emptiness (of seed Pod), light

weight (of money)

3112 L'nelli] paddy

178 L'akki] husked rice

805 Fotti] cake, bread

(ii) (C)VCC- and (C)vNC -.

porcupine
4098/3326 L'mulli'pandi]

L'b.uddi'vante(n) ] wise man

rr

l'matti'yandi] day after tomorrow

3903/1

1756 `'kop'pumbi] ear ornament

 383

The loan ['buddlVante(n)] shows the unusual element '[v]

(from /v/ at Onset) in its second element, and the relatively rare

-[nt] - (from /ITT/ at Cons Suff) .

['kop'pumbi] shows no element at Classifier either in the

first lexeme or the second; the enunciative vowel is lost before

the initial vowel of the second lexeme.

Cf. 4098 ['mulli] thorn

3326 ['pandi] pig

['buddi] wisdom

3903 ['mari] state of being opposite,counter

1 ['andi] that day

1756 DED records "Ta. koppu, ',omen's ear ornament... " .

(iii) (C)VCC- and (C)VC -.

4524/3999? ['bolli'mi:ni] morning star

2002/341 ['cav'va:ce] Tuesday

3608a/342 ['pon'na:li] woman

Informants related ['mi:ni] in the first item to 3999 ['mi:ni],

'fish', but also related each of these to 3994 ['mina]- 'lighten,

flicker'; Class 1 verb. This would require setting up root variants

here, as /mi:n/- ^+ /min /-.

In the case of the second item, informants recognised a

relationship between the first element and ['ca:vu]; this would

require a root variation /ca:l/- N /cal /- (see the previous two

+ +
chapters for the root -final element here). The systematic phonetic

sequence -[lv]- after a short vowel is adjusted to -[vv]- (by AS6).

Cf. 4524 ['bolli] silver

3999 ['mi:ni] fish

2002 ['ca:vu] death, corpse

 389

341 ['a:ce] day of week

u.
3608a [ ' ponni] wife, female

342 ['a:li] servant

(iv) (C)VCC- and (C)57ÌC -.

856/3361 ' olle' pa: mbi ] water snake

3109/2381 'nin'na:ndi] yesterday

3106/2021 'netta'ca:ñc] sp. red sandalwood

['nin'na:ndi] shows a first element ['ninn]- which occurs in

the data only here.

The last item shows the relatively infrequent Cons Suff

realisation -Die]- (from /NC/).

Cf. 856 DED records "Fa. olle, 'kind of amphisbaena'".

3361 ['pa:mbi] snake

3109 DED records ninne, 'time lately passed'".

2381 [ 'a:ndi] also in ''moni'ya:ndi], day

before yesterday

3106 DED records "Ka. nettra, 'blood ".

2021 L'ca:ndi] sandalwood

(v) ( C)VCC- and (C)vCVC -.

If

3608a/1423 ['ponni'kudire] mare

2092/78 ['jib'badake] young, tender areca nut

Cf. 3608a ['ponni] wife, female

1423 L'kudire horse

2092 ['jibbi] young, tender areca nut

78 ['adake] areca nut

(vi) (C)VCC- and (C)VCVNC -.

No examples.
 06C

(TTA) -ÛC1(0) 11D1,_(0)

LC l
LZ Ç TTTei, rnsn::, aoT.z ti: Ta ata =ls-nt1

'Jû [TTTau,] puc .p

L r8t [nana, pees

; (T) -DHAA(0) puy -011(0)

Ï
[-P4o=ltTpu-y1 :121/TuaAa al4s
ll

san000 ;sng aaatl puu ut 'rt;a:au,Tpunm,tpuu,1 aßT;), (jo

-! [Tnrnro:1, san000 ;sup aaatl puu uT Î ;nr(»liuToc_,'[T SuTuaou,

(TT) -311A(0) puy oiLA(3) '-

o1,1
saTcImuxa

(TTT) -0.11(0) PUY -011(0)

LZ6Zf`9C . [uT:onr,TCNt, aa°utZ ose

í1V9Z9Z L 7 cuT,yEuTnr, [ao FupuoT1

06Z/88L 7
[u.:ep,úo:4 Tead as

enrol: T"eqnr ¡,TO;r ryuti`o T^v::oA ut sTtja. suo0 uOT;OPanr sT au0 Jo aua

squamuoatAUa aaatlrl at1; Ty--T3 nruemaTa Jo /Te/- ;u aaT,TTssuTo sT

pastToaa

.TD `v9C 1 [TPwT, aaSur.2

LZ6Z anruTsa
[ui.:onr,j

i [ao:u, Sup To

9E9,7. ]
Ta-, (T)uDy [ uoom

L 88L ; :a.7e1 aATa.o9za-0 qaaA

-[TTox'

06Z l -[;:ul a:Te[a ar_oM :1:TnruaTotA ssyTo l

'q.aaA paATaap fq
ITTo11

2utpau5,,a17. auT ;out o1e; 'sMal.t '30 ata Q3Q SA;ua :922A

ss ;u-;uoq 'n o 1ÇOpua ssuos.zad '49Toos astuad

 391

(iv) (C)VTTC- and (C)VirC -.

No examples.

(v) (C)VîTC- and (C)VCVC -.

No examples.

(vi) (C)VIC- and (C)VCVITC -.

No examples.

(vii) (C)vi c- and (C)VCV.

2606/3556 ['te:ipi'mara] coconut tree

Cf. 2806 ['terige] coconut

3856 ['mara] tree

3. (i) (C)VC- and (C)VCC -.

2375/3299 L'ne:ra'panni] navel fruit

The first element here shots Lr] from /r'/.

Cf. 2378 DED records "Ka. neral(e) , 'Eugenia jambolana Lam.'"

,r

3299 ['panni] fruit

(ii) (C)VC- and (C)VITC -.

ido examples.

(iii) (C)VC- and (C)VC -.

" "
/3999 L'e:ti'mi:ni] sp. flat edible fish

2352/ E'so:'ma:ri] lazy person (masc/fem)

rr
Cf. 3999 ['mi :niJ fish

2352 DED records "Ta. comperi, comari,

Here also may be noted the complex verbal lexeme cited at the

beginning of Chapter 4 (p. 208):

 392

1192/290 ['ka:'pa:d]- protect; Class 1 verb

Cf. 1192 _'ka pale(n) ] (palace) guard

290 a:d]- dance; Class 1 verb

(iv) (C)7c- and (C)7- C -.

1673/3361 ['ke :re'pa :mbi] sp. non -poisonous snake

1348/4230 ['ki :'la :ndi] last year

Cf. 1673 DED records "Ka. kere, 'rat snake".

3361 L'pa:mbi] snake

1348 L " -i : li] state of being below

4230 L'a:ndi] year

(v) (C)VC- and (c)vcvc -.

1826/76 L'go :'tadake] full.;ro;rn, tough areca nut

Cf. 1826 DED records "Ka. gotu, 'state of being full -drown
but hard"

78 L'adake] areca nut

(vi) (C)VC- and (C)VCVNC-.

No examples.

(vii) (C)VC- and (C)VCV.

768/3882 ['e:la'male] cardamom plantation

Cf. 768 ['e :laki] cardamom seed

3882 ['male] hill, jungle land

4. Here there is only pattern (iv), (C)VdC- and (::) 7 -C -.

3919/106 ['ma:ñge'ya:ndi] mango stone

Cf. 3919 ['ma:ñge] mango

106 ['a:ndi] (mango) stone

 c--
.1.)%

5. (i) (C)VCVC- and (C)VCC -.

957/1091 L'kadace'kalli] mortar and pestle

C_°. S57 ;rind with mortar and nestle;

Class 2 verb

1091 stone

(ii) (C)VCVC- and (C)VI C -.

.o e__?n -eles .

(ili) (C)7CVC- and (C)VC-.

It It
1523/ ri 1. cold season

,r;T 1525 cole_ feelir_ is; Class 5 verb

season, time

(iv) (C)VCVC- and (C) V'i c -.

'_';o e:,ramples .

(v) (C)VCVC- and (C)VCVC -.

No e°m-nles .

(vi) (C)VCVC- and (C)VCVi C -.

:jo e1_am les .

(vii) (C)VCVC- and (C)VCV.

168/3856 L'arali'mara] pipai tree

Cf. 160 DD records "Ka. arali, 'P. religiosa t".

3856 L'mara] tree

,`_ C-.
G. (i)-(vii) ()
ITo e._am2les.
 394

7. (i) (C)VCV and (C)VCC -.

800/1012 ['odi'katti] Coorg sword

['oy'patti] evening

The second item here shows the operation of A.S5 which deletes the

Voc Suff realisation [i] immediately after [y] (from /c/ at Coda) .

Cf. 800 DED records "Ka. udi, 'hip, waist'".

1012 ['katti] knife

(ii) (C)VCV and (C)V'TC -.

3411/3831 ['pada'mande] back of the head

1379/ ['ku'tumba] household

The second item is illustrated here as a tentative classific-

ation only; we have to derive -[t]- from /T/ at Coda, followed by

[i] from /i/ at Voc Suff, which is deleted (like the enunciative

vowel, to which it is phonetically identical) before the following

vowel by A.S10. This would require reco ri sing the root alternation

/'_tut /- nv /kuT/- for this and related items.

Cf. 3411 DED records "Ka. peda, 'the back"".

3831 ['mande] head

1379 ['kudi] family of servants living in

one hut

(iii) (C)VCV and (C)VC -.

ft

3911/ ['mane'ka:re(n)] man of a house

DIA 289 ['mari'ya:di] respect

DIA. 196 ['dala'va:y] military commander

3531 / ['po'la:ka] morning

4487 ['bi'ra:ni] pagoda (= 4 rupees)

['ví' ca: ra] trial

L'ka'ce:ri] office
 395

-Lka:re(n)] is by far the most frequent (C)VC- second element

in complex nominals of Set 1; it is a masculine suffix in all cases

(feminine form f[karati]) .

L'dala'va:y] has the alternative form L'dala'ba:y] which shows

the more usual Lb] as the realisation of Onset in the second lexeme.

The remaininc four items may be analysed as the result of

deletion of [i] (from /±/ at Voc Suff); this is the same solution as

was advanced for L'ku'tumba] above, and has the advantage of account-
u_
inE for the non- occurrence of (i] in this position when a vowel

follows. That is, [i] is deleted in the environment of a following

vowel whether it is enunciative [i] or from /i/ at Voc Suff. Note that

the items 1_'vi'ca:ra] and L'ka'ce:ri] thereby require that /C/ (marked

for voiceless articulation) be set up at Coda.

Cf. 3911 ['mane] house

3531 L'pola]- dawn; Class 2 verb.

(iv) (C)VCV and (C)VNC -.

4119b/2381 ['moni'ya:ndi] day before yesterday

fit
madman

The first element of the first item represents a root variant

/-
il

/mon with /mun/- (as in L ' mumbi ], 'predominance') and /mi/ -- (as

in 1 ' miíífia] , 'state of being before?).

L'pi'ra:nde(n)] represents the same type of structure as that

proposed for L'ku' tumba] and ['bi'ra:ni].

Cf. 4119b __
DED records "Ta. mun-nal, 'yesterday "' (i.e. 'before-
todáy') .

2381 ['a:ndi] also in ['nin'na:ndi],

'yesterday'.
 (y) (c)vcv and (c)vcvc -.

1163/78 ['kali'yadake] wholly ripe areca nut

Cf. 1163 DED records "Pa. kali, 'to ripen well ".

78 L adake j areca nut

(vi) (C)VCV and (C)VCVIC -.

A; o examples.

(vii) (C)VCV and (C)VCV

701/3537 L ' ere' pulu] earthworm, tapeworm

Cf. 701 ['ere] earthworm, tapeworm

3537 E'pulu] worm

2.1.2 Lexeme Base Classifier

As noted earlier, a number of Set 1 nominals show constituent

lexemes which are expanded in other ways than as Base Classifier:

these are detailed here.

1. First nominal: Base Postbase

Second nominal: Base Classifier

(i) First nominal is of type 1 in Set 2 (see below).

11

/3564 L'ummati'pu:vi] sp. flouer

1915/3856 cappaye'mara] campaka tree

(ii) First nominal is of type 2 in Set 2 (see below).

834/ L'ondane'ja:ga] headstation of district

(iii) First nominal is of type 3 in Set 2 (see below).

302/726 ['a:na'kenr.e] castor oil

302/3856 Pa:nake'mara] castor oil plant

L'ji:rige'sanna] type of small- grained paddy

 397

(iv) First nominal is of type (vii)

in Set 2 (see below).

37/ L'aiigadi'ka:re(n)] shopkeeper

516/3856 L'umbala'mane] kitchen

Cf. 3564 L'Pu:vi] flower

1 91 5 DAD records "Ka. camp aka, 'Nichelia champaca " .

3856 ['mara] tree

834 L'ondi] one

['ja:ga] place

302 DFD records "Na. amanakku, 'castor oil plant'".

726 L'enne] gingily oil

[' sarna] also in ['sanna'ba: ta], ghee

paddy

37 ['añgadi] shop, stall

516 r'unn]- eat (rice); Class 3 verb

3856 ['mane] house

A
2. First nominal: Lexeme Lexeme

Second nominal: Base Classifier

(i) First nominal is of type 2(ii) in Set 1 (no examples above).

4119/3128 ['andi'mundi'ne:ra] (time of) twilight

Cf. Landi] also in L'andi'kotti], evening

star

4119 L'mund]- precede; Class 1 verb

3128 L'ne:ra] time, sun

The first two lexemes make up a unit which then combines with

r
L 'ne:ra]. The sense is 'evening -precedence time', i.e. the time that

precedes the evening.

(ii) First nominal is of type 7(i) in Set 1 (see above).

 398

L'oy'Datti'mal'i; sn. jasmine

(iii) First nominal is of tyDe 7(ii) in Set 1 (see above) .

/9:4 L'i'la:ti'kande.J _potato

Cf. L'oy'patti] evening

['i'la:ti] England

984 ['Mande] root- stock, bulb

3. First nominal: Base Classifier

Second nominal: Base Postbase

The second nominal is of type 1 in Set 2 (see below).

112/2513 L'anr_e'tammane(n) ] brother (older/younger)

Cf. 112 :_'anne(n)] older brother

2513/112 L'tammane(n)] younger brother

4. First. nominal: Base Classifier

Second nominal: Lexeme Lexeme

The second nominal is of type 3(iii) in Set 1 (see above).

fl,
DIA 179/ turo -horse carriage

[' jo:di]
rr

Cf. DIA 179 pair

5. First nominal: Base Classifier

Second nominal: Lexeme "Classifier

The second nominal is of type 1(i) in Set 4 (see below).

/DIA 106 L'kala'kuppiya] (woman's) blouse

Cf. DIA 106 ['kuppiya] Coorg man's coat

6. First nominal: Lexeme Classifier

Second nominal: Base Classifier

 399

The first nominal is of type 3 in Set

4 (see below) .

2363/1174 L'navili'gari] Peacock's tail feather

It

/341 L'padi'na:ce] ';ednesday

R It
/3025 L'kadipati'na:li] forceful (astroloical) day

/399rß? L'ba:nati'mi:ni] star

Cf. 2363 L'navili] peacock

1174 ['sari] feather

341 ['a:ce] day of the week

3025 ; 'na:le] tomorrow

L'ba :na] sky

3999 See discussion of ''bolli'mi :ni] under 1(111) above.

2.2 Set 2: Base Postbase

have noted earlier three factors concerning the structures to

be generated under Postbase:

(i) the lack of main stress on the first syllable

(ii) the lack of length contrast on the root vowel

(iii) the lack of contrast in consonant length in the position

immediately following the root vowel.

Each of these is to be accounted for with reference to the boundary

symbol /, and the realisations of phonological elements as affected

by this symbol (including a number of sandhi forms) will be discussed

in a detailed way.

First, however, the canonical forms that occur with this

constructional type are illustrated. The arrangement of the items

below is with reference to the base element of the construction, in

the order followed above for Set 1; further ordering is with reference

to the Postbase element, such that items showing an element at Noun

 0017

t) a ug-ual.xaTus ' a.as o jj aaot;

Sq (= eau paaapao aaoJaq esoun.

2ut2sous s Tautj q.uasaTe a.0 ooA zrns at) 'reseg-aq.ut 3jo aaau

Lq + ;

a,zn:-'-Fd
- Z s_o;s au, tTaaaua. pas suiaLas r.lúaa cz ,,
p^. aOT ua- a . O TT nLiTP!O

saTduexa :

# #aMaxaZ

-------._-1.._
7a7Ta soá/ suga. /a
`
.
oo2ì
/ _
,;
-1--'
..,
``aat,tssTO
;.,
..,
\ (xt,,r+) (zxt) /assg
/ . ,

(q_asu0) aonil s ;oog tfiTxt

.j'=) := (z=
moll

rLaad (-epos) /
(a.asu0) snaTonr1
/\
=reed (aPoO)

aanam :z euJi pastTuaaua° aan;onaq.s jo xaTduioo

Troutuzou saulaxeT ;a;; z

sadSS j,-L r.oTag aq.aaq.snTTt STVao sessm.sod tia-tv. ou T'et;tut -uosuoo

4ue asauq. axe sTaAt-aTea gpanl.zojlLpteass aours auT STuo adS1 jo

uot;sq.nu pa1Tor.ut ssoao-a aqa. / Saspanoq st vaq; aaaum 'a reVeTed JO

TuZquTTq aptTZ st paTaasut 04. 31paag En a aouanóas jo 'sTaaon su paTou

aetTaua 11.4TA saa q..oa: q.as sasuqasod uTtm Turnuz ureuosuoo

TTTI`i eq q-TueP I?7-ti4 aa;nT uo dd) (SOt-Ot '

l ' (T) -00A(0) Puu 11=0A

II
j Tto
/9ZL 9ZL Tja [aua STr..911tO

9VL0tr ] (u)aCual.nm, r a.uaa° aauvejrueas

aauxeaJ
/O6LZ [pTzddo,,
 401

Cf. 726 L' elli ]

gir_gi ly seed

726 L' en, e] gin_ily oil

4057 :'mudi] state of being old

46 L'ajJe(n)] ;ran :'.father

2790 DED records "ha. toppa, 'wool, animal's hair; toppal,

'feather'".

The second element in :'toppita] is obscure; it probably shows

`a] at T`oun, with reduction of the base -final consonant or root

vowel. However, it is possible that this postbase is of VC +V type

with Lt] from /T/ at Coda, though this is statistically less probable.

(ii) (C)VCC- and VC+V.

85 ['addige] gold necklace

No internal or comparative data is available to ,ermit the

component elements here to be identified with related items or

cognates; note that the postbase element nay show -e] at _noun, but

that where there is structural obscurity of this sort the VC -V

postbase ty pe is assumed.

2. (i) (C)vrc- and VC =V.

DIA 71 L ' kandi ta] certainty

(ii) (C)Vii C- and VC +V.

21áO/710 'cindeli] (house, field) mouse

2975/ L'nambige] hope, trust

Cf. 2190 DED records "Ta. cuntan, 'grey musk shrew".

710 ['elí] rat

2975 ['namb]- trust; Class 1 verb.

-Lige] corresponds to Ta. -ikkai (e.g. DED records "Ta.

 402

nwm--11ial, 'hoDe, trust, faith, vo;t" ) , a very freruer t nomir:aLsing

suffi::.

3. (i) (C) ;TC- and VC=V.

3740/232 L.'bo :jave(n)] mother's younL-er sister

768/178 L'e:lal:i] cardamom seed

Cf. 374C L'bo:ji] beauty, goodness

232 ,'avve(n)] mother

768 L'e:la'malej cardamom plantation

178 ['e7 ki] husked rice

(ii) (C)VC- and VC.4V.

2674b/441 ['te:neri] honeycomb

4377/ L'ba:dige] rest

Cf. 2674b ['te :ni] honey

441 DED records "Ta. iravu, irai, 'honeycomb ".

4377 ['ba:d]- fade, wither; Class 1 verb

4. (i) (C)NC- and VC=V.

No eyamples .

(ii) (C TTC - and VC-V.

No examples.

5. (i) (C) VSC - and VC =V .

1976/178 ['avlaki] pounded rice

Cf. 1976 DED records "Ta. aval, 'rice obtained from fried

paddy by pestling it' ".

178 ['akkki] husked rice

 403

( i) (C)VCVC- and VC +V.

11

/218 ['uditale] ironwood tree

Cf. 218 DED records "Ta. alli, 'ironwood tree'; Ka. alamar,

alle, 'a sp. of Nemecylon_ ".

Note that the comparative data cannot be used to support a

postbase analysis VC =V here, althouvh this would male -[ale] strictly

comparable to the second Ka. form.

6. (i) (C)VCVI<C- and VC =V.

No examples.

(ii) (C)VCVNC-/(C)vsNC- and VC+V.

200/ L'alambali] hurry

t,
555/1 o;rn.d ad i] place higher up

Cf. /200 [ala]- crave; Class 3 verb

555 ['oynie] high (adj.)

u
1 L'adij 3rd sin« -2roform (inan; rem)

7. (i) (C)VC +V and VC =V.

No esam.11es .

(ii) (C)VC+V and VC +V.

1705/423 ['kodiyele] betel leaf

Cf. 1705 DED records "Ta. koti, 'tree er, umbilical cord' ".

423 in ['ela'kanda] leaf

(DED records "Ta. liai, 'leaf, petal' "; but collocuial

varieties of Tamil show Lela].)

The following items all show postbase elements with an initial

consonant. They are numbered in series with the foregoing.

 404

8. (i) (C)7C- and CVC=V.

11 I1

1607/ 'cernbike(n) ] s'. bird with red feathers

Ir_
Cf. 1607 L'co:pi] redness

',:Te have here to set ul:, the root variants /co: /- ^+ /cen/ -; we

also require /ken/- for items such as Ì "_yen' :leñge], 'brown coconut'.

(ii) (C)VC- and CVC -V.

516/3537 ;_'umbilli] jungle leecr.

1348/3558 ['kippuni] lower level in field

Cf. 516 i'unn]- eat (rice); Class 3 verb

3537 _' u1u] worm

1348 ] state of being below

3558 L'meppuni] higher level in field

Note that the first item requires setting up the root variants

/,cul/- w /y il / -, since, while the voicing of the initial consonant of

the second element can be accounted for (see below), the vo-.:el

alternation cannot.

9. (i) (C)V- and CVC =V.

No examples.

(ii) (C)V- and CVC +V.

1592/ ['ku:kala] rice pot

1592/ L'ku:padi] food wra,Dped in-plantain leaf

Cf. 1592 L'ku:li] (cooked) rice

The loss of the final consonant in the first element of each of

these item is accounted for below.

10. (i) (C)VC+V and CVC =V.

929/1371 L'kadekuñi] youngest child

 405

tI II

4402/ L'barikati] prosperity

Cf. 929 L kade] end (row, event)

1371 ['kuññi] (Coors) child

tt
4402 Eba:li] way of life

Note that in the last item /bar/- is a root variant with /ba:l /-
+
(for the final element here, see the previous two chapters), and
" ".
/bad/- as in r'badiki], 'property'.

(ii) (C)VC-1-17. and CPC +V.

167/3911 L'aramane] Palace

Cf. 167 DED records "Ta. aracan, araican, araiyan, 'kinc. "'.

3911 ['mane] house.

Entries 1 -6 above show a first element with a final consonant,

followed (across the / boundary) by a postbase with an initial vowel;

entry- 7 shows a final vowel of the first element followed (across

the / boundary) by an initial vowel of the postbase; the situation in

entry 8 is where a final consonant of the first element is followed

(across the / boundary) by an initial consonant of the postbase; and

finally in entries 9 and 10 a final vowel of the first element is

followed (across the / boundary) by an initial consonant of the post-

base. However, we have seen, with respect to entry 9, that the

first element probably has to be set up with a root-final consonant,

which is later lost. We now turn to the formulation of the mutation

phenomena across .the / boundary.

First, the statement of consonant mutations across the /

boundary is greatly complicated if the automatic gemination rule is

allowed to operate in respect of the base -final consonant of the first

element in this constructional type; hence a condition is placed on

 406

the operation of this rule, that no postbase must follow in the same

leY.eze:

i
8 (i) V > V / [ (C) .._, (Erase L ) (iL°(i ) )

9
(ii) Ci > CCi / V , NOT / V /.. ./ (P i ) )

( ; ras e ])

n ii

Thus the representation of 'umbili at this stage will be

e n

/ ún / pili / / (omitting irrelevant details), rather than the

9 9
unnecessarily complex / u°.nn / pili / /.

However, the first part of Pig has operated, assigning stress to

the root vowels; it would involve too many complicated restrictions

on PP_9 if stress reduction and vowel length reduction of postbase

elements were to be accounted for by it alone. The question of stress

and vowel length reduction is returned to later on.

This means that the representation of L'muttaje(n)] at this

stage will be (again, omitting irrelevant details) /milt / jjan /.

If base -final /t/ of the first element were to be realised here, it

would be as Ldj, which is not what is required. Hence, we need a

phonological rule which will, in effect, operate as follows:

0 I
á. VCi /V -a VCi / CiV (PM 0(i))

thus eminatin the base -final consonant of the first element just

in the case where (i) it is preceded by a short root vowel, and (ii)

is followed by a vowel across the / boundary.

'`e also have to account for the mutation involved in [' kippuni];
n n

although this has the same canonical form as L'umbili], it

represents rather a different type, where the first element has no

final consonant in underlying structure. Thus, the postbase is

 407

-[puni], also found in L'meppuni], and the first

element is L'kij -,

from /ki / -, a root variant with /ki :l /- (see the previous chapter).
1!

¡ki/- also occurs in L'k-ikkire(n)], 'junior member of household'.

For such cases as these, we require the phonological rule:

10. V ci --)
(H 10(ií))

No phonological rule is reouired in the case where a base -final vowel

of the first element is followed by an initial vowel of the postbase.

Finally, we require a phonological rule to reduce the stress on

the initial vowel of nostbase elements:

11. v / / (c) , (PN11(i))

and also to reduce the length of unstressed vowels:

12. VL -+ V (11411 (ii))

With this, the phonological mutation rules are complete. The

realisation rules apply in the ordinary way to all the sequences of

phonological elements, regardless of whether they have been generated

as simple or complex lexemes. ":e now consider the systematic

phonetic development of those consonants that represent systematic

phonemes at the / boundary. The table in Figure 3 contains all the

sequences across / that occur in the data.

The first two rows represent the situation in the terminal

sequence as a result of the operation of PM10(i) and (ii), respect-

ively. Since PI10(i) is similar in its mode of operation to

PN9(ii), and is illustrated in entries 1 (i)/(ii) above, we shall not

cite further instances here. PX10(ii) is illustrated only in one

item (in 8 (ii) above), however; but it also operates in the following

items:
 408

Systematic _,honemi c Systematic phonetic

level level

Base-final Initial
consonant consonant
of initial of postbase
element element

C.

ífter
short
root
vowel

5 yp

6 vt

:after
long
root
vowel

Figure 3: Consonant mutations across the postbase

boundary

1/S 630 [' appara] that side

1r n n
1348/1326 L'kikkire(n)] junior member of household

4173/3558 L'meppuni] higher level in field

4119a/ ['muppode] state of being previous

Cf. 1 Wadi] unmarked 3rd singular proform

S 630 L ' para] side, area

 409
R fl
1348 L'ki:li] state of being below
" "
1326 L'kirike] small (adj.)

4173 !'me:'ma:di] top storey

3558 'kippuni] lower level in field

4119a L'mumbi] predominance

The roots represented in the first elements of the complexes cited

here are /a/- (variant of /a: / -, in Pa :le], 'by that way') , /'_Ti /-

(variant of /ki :l / -), /me/- (variant of /me:n/-), and /mu/-

(variant of /mun /- w /mon/ -), respectively. Note that the postbase

in the second item above must be set up with a long root vowel, in

order to account for the occurrence of -Le(n)]; the vowel is

reduced by Phi 1 .

Row 3 in the table in Figure 3 is illustrated by L'cembike(n)]

in entry 8 (i) above, and also by the items:

1607/2077 ! 'keñjeri] red squirrel

4518/S 630 Pbembara] place behind

8/1683 P añgay] palm of hand

Cf. 1607/2806 L'ken'der_ge] brown coconut

2077 MD records "Kol. sidde, Nk. sidde, Kur. cirra,

cidra, 'squirrel'".

r "
4518 'benni] back (body part)

S 630 'para] side, area

8/1238 j'a_*iga:li ] sole of foot

1683 ['kay] arm, hand

For the first elements here we set up the roots /ken/- (variant of

Icen/- N /co: / -), /ben/ -, and /an/ -, respectively. For the terminal

sequence of the postbase forms, we postulate -/bikken/ or -/bi:kken/

njeri]: note that the first

for ['cembike(n)], and -/ceri/ for :'k
 410

of these may in fact show initial /p /, but that there is no

evidence

for this. The last two items show -/para/ and -/kaci/ in this

position, respectively.
" "
To;: 4 j.s illustrated by L' u*_nbi li] in entry 8 (ii) above, and by

3608/3768 L'pommaka(1)] women, wives

Cf. 3608 L'ponni] wife, female

3768 L'makka(1)] children

The next two rows show particularly clearly the importance of

restrictinz the operation of the automatic gemination rule to those

cases where no postbase follows in the same lexeme: if it were allowed

to operate in the case of the two items cited below, it would yield

base -final -[cc] and -[pp], respectively:

4566/ !L ' baygala] abuse

4255/2529 bavtale] hair parting

Cf. 4566 L'bayy]- abuse; Class 4 verb

4255 DED records "Ka. bage, 'division' n

2523 ['tale] end

Note that here, unlike after a base -final nasal consonant of the

first element, the voicing contrast at Onset in the postbase is

Preserved.

Rows 7-8 show the consonant combinations that occur after a long

root vowel of the first element: the first is illustrated by:

41 73/1 683 ['me:ñ ay] back of hand

Cf. 4173/1238 ['me:ri'ga:li] top part of foot

1683 E'kay] hand

For the first element here we set up the root /me:n/- (variant of

/me/ -; we also require /me:l/- for the noun particle [me :le], in such

constructions as L'me:jira 'me:le], 'the top of the table').

 411

Finally, row 8 is illustrated by the items

L'ku:kala] and

['ku: padi] in entry 9 (ii) above, and also

by:

4402/3911 ] house in to which bride a_ ri as

Cf. 4402 ba :li] way of life
e,
3911 1
[' mane]
a house

The first element here shows the root /ba:l / -, as described in the
1_

previous two chapters. Note that after [1], as after [y] and. [v],

the voicing contrast at Onset in postbase is preserved.

For the rest of the consonant combinations that occur in this

construction, the phonetic nutation rule :ß.S3 adjusts the place of

articulation of the base -final nasal consonant to that of the post -

base- initial consonant, and :ïs6 adds a voicing feature to the latter

if it is a single voiceless stop (the effect of these rules is given

here in alphabetic notation):

r w
_,b mb

13.
fl nj
(-.s3, 6)

1136 simplifies the consonant clusters involvin E1] by deleting it

after a long vo;:el. Clusters involving [y] or [v] immediately

before another consonant, preceded by a short vowel, are retained.

Thus far, we have dealt with stress and vowel length reduction

in postbases, and also with consonant mutations across the /

boundary. 'ie have seen that the first two are phonological phenomena,

and that certain consonant mutations involve phonological gemination,

but that the remainder are handled at the systematic phonetic level.

Finally, we need the phonetic mutation rule AS7, which reduces a

 412

geminate consonant cluster to the duration of a single consonant

when it immediately follows an unstressed vowel.

2.3 ^et 3: Base Lexeme

The generalised schema for the items that occur in this set is

given in Fiere 4:

# Lexene#

# Lexeme#

f fix Base] C1 __ °_ier

///
(Onset) Root
\C°',
( +A ffix1)(-

rea (Cc=_a) (Onset) Nucleus

/\
Peak (Coda/

Note that Affix2 is not required under the first base node; and that

Affix1 under this node is never Voc Suff.

Figure 4: The generalised structure of complex

nominal lexemes, Set 3

This accounts for all the structural types that are illustrated

below.

1. (i) (C)VC- and (C)VCC -.

n
1607/3556 L'kem'butti] red anthill

4524/3528 ['bov'pilli] dry grass, straw

Cf. 1607/2806 `'ken'denge] brown coconut

n.,

3556 '_utti] anthill

u fl

4524 L'bolipi] white

 413

rr
3528 L'pilli grass

(ii) (c)v(c)- and (c)vNc -.

1607/2806 'ken'deñe] brown coconut

2283/3326 L' eD' par_di] -raorcupine

Cf. 1607/3022 ['ken'na:y] wild dog

2806 L'teñve] coconut

2283 DED records "Ta. ey, ey-ppanri 'porcupine'".

3326 ,.'pandij pig

(iii) (C)VC- anc.

8/1238 L'a$'Ea:li] sole of foot

1607/3022 ['ken'na:y] wild dog

3608/3022 L'pon'na:y] bitch

4057/308 L'nut'ta:y] great grandmother

Cf. 8/1683 i.'anx,ay] palm of hand

1238 L'ka:li] leg, foot

r M
1607/2077 L'kenjeri- red squirrel
rr

3608 P onTli] rife, female

3022 ['na:y] dog

4057 L'mudi] state of being old

308 a:y] grandmother

2. (i) (C)7(C)- and CVCC -.

3087/DI2,_ 269 L'nu:'putti] vermicelli rice

ft,
civet cat
3538/4520

4400/DI1, 233 L'pa:'pakki] bat

rr _.

3087 nu: li] thread

Cf.
rr

putti ] rice preparation

DIA 269 L'
 414

3530 D3D records "Ta. puruku,

civet'".

4520 'bekk i] jungle cats of various species

4400 ba:vali] bat

N
DI'. 233 ['paksi] bird

(ii) (C)7(C)- and CV:-C-.

3161i3792 ['pa:'mañji] scum, lichen

4394/2734 L'ba:'turbi.J bird's tail feathers

Cf. 3161 L'paccej green, blue, cold (of water)

3792 L mañji] dew

4394 L ba:li] tail

2734/1683 L'tumbi'kay] elephant's trunk

(iii) (C )V(C)- and CVC -.

tr,
4173/1238 ['me:ñ'ga:lij top part of foot

4173/3930 L'me:'ma:di] top storey

4057/DIA 219 'mu:'de:vi] Quarrelsome woman

Cf. 4173/1683 'me:ñ`:ay]

i

back of hand
"n
1238 [ 'ka:li,J leg, foot

3930 ['ma:di] upper storey

4057 ['mu :]- grow up; Class 5 verb

DIA 219 L'de:va] god

(iv) (c)c- and CVCV -.

1592/1376 ['ku:'kudike] rice pot

ft,
Cf. 1592 (cooked) rice

1376 ['kudike] cooking pot

Note that this is the only example of its type in the nominal system,

and has a Set 4 nominal as its second element. Here also, however,

has to be noted the complex verbal lexeme cited at the beginning of

 415

Charter 4 (2. 208) :

333/4327 j'a:'valic]- yawn; Class 1 verb

Cf. 333 DED records "Ta. avi n., 'yawn".

4347 break wind; Class 4 verb

Also, DED records "Ta. vali, 'wind, air, windy humour in the
body".

For the same reasons described above with resrect to Postbase,

we require that the automatic gemination rule be restricted from

applying to base -final consonants where the base concerned is

followed by a lexeme in the same lexeme; that is, we formulate Tß'9

as follows:

14. (i) V VI (C) (Erase ) (a9(i))

(ii) Ci -3 CiCi / , _TOT / V ] (PK9(ii ))

(Erase ] )

Yost importantly, this allows for the base -final [y] in the first

element of ['bov'pilli], as described with reference to Pbavtale]

in Set 2. e set up /bo /- as a root variant of /bol / -; base -final

v_I in this item derives from /p/ at Cons Suff immediately following

this root. In the items :_'ep'pandi], ['pu :'bekki], ['pa:'pakki],

_' e:'nF ~ji], and L'mu:'de:vi] the first element shows no base-final

consonant, consisting only of the roots /e/ -, /pu / -, /pa :/- (which :

therefore is a variant of /ba:p/- in 'ba:vali; or possibly this item

shows the root /ba: / -, with /p/ at Cons Suff), /pa: /- (distinct from

the preceding root, and a variant of /pac /- in ,'pacce]; the latter

item may however have the root /pa/ -, with /c/ at Cons Suff), and

/mu :/- (variant of /mu/- in L'muttaje(n)]).

PY10 applies to the first of these items, therefore:

 n16

15. V 16# 1 c.
1
--,
( )

to yield the .geminate voiceless bilabial stop cluster. It does not,

hoever, apply to the other items, since /L/ stands between V and

the / boundary. Note that P 10(i) never applies in the data for

this set of nominale, since in every item the component lexeme

that forms the second element of the complex shows an initial

consonant.

Otherwise, where two consonants occur across the / boundary,

no mutation takes place, as i r the case of Ì 'bov'pilli] ,

_ . n
L'ken'na.:y], 'nu:l'putti], Pba:l'tumbi], or L'ku:l'kudike], as

yielded by the realisation rules. However, the items 'kem'butti],

rr

an j, 'pon'na:y], L'me:ii'ga:li], and

me'ma:dij all show the operation of the honetic mutation rules

AS3, 6 (given here in alphabetic form):

.. ..

n
16. (ßs3, 6)

m?

n nn
..

In all cases except that of i'pon'na:yj the operation of the rule is

functionally equivalent to its operation with respect to Set 2

nominale. Fowever, the assimilation in place of articulation of the

base -final nasal consonant has to be stated carefully; Lnj assimil-

ates to a following [m], but the retroflexion feature is not lost

when the following consonant is Lnj. Hence, the assimilation has

first of all to be stated in terms of the categories Bilabial,

 417

Dental, Palatal, and Velar; secondarily, the feature of rtroflexion

is spread through the geminate cluster.

Finally, the consonant clusters involvinrr ![1] or [1] after a

long vowel are sitnlified by loss of the lateral consonant, by AS7.

2.4 Set 4: Lexeme "Classifier

The generalised schema set out in Figure 5 accounts for the

nominals of this set:

# Lexeme*

# L exeme Classifier

Base11 Classifier Noun

Root ( +Affix1)( +Affix2) Noun

//
(Initial) Noun

í Verb
(Onset) Nucleus

Peak (Coda)

Figure 5: The generalised structure of complex

nominals lexemes, Set 4

It will be seen that Noun here is expanded as an optional Initial,

followed by Noun. At Initial, a system of consonants operates which

will be illustrated in Figure 6 below. At the node Noun under Noun,

the same system of elements operates as has been recognised thus

far. The nominals belonging to Set 4 are illustrated below.

1. (i) (C)VCC-.

DIA 106 'kuppiya] Coorg man's coat

1388 ,'kuttuva] copper pot

3115 L'nellike] gooseberry

 418

Cf. 3115 DED records "Ta. nelli, 'emblic myrobalan".

The first two items show [y], [v], inserted by AS8. Note that

-Lke] in ['nellike] (also L'pi:reke]; see 2 below) probably

corresponds to L'ka :y], '(unripe) fruit'. Clearly, this is suggest-

ive insofar as the diachronic status of these elements at Noun is

concerned. However, an analysis which attempted to derive them from

the node Lexeme (introduced under Noun) would involve a number of

serious complications, especially in stating the vowel mutation in

this example.

(ii) (C)vi;C -.

1209 L'kandave(n)] stranger (mast)

By this analysis, this item shows a root variant /ka/- of /aka:/ -

in 'ka :n]- 'see, seem'; Class 3 verb. The [v] is inserted by A58.

However, the item is unusual in respect of its nasal vowels by this

analysis, and it may be correct to include it in Set 2, with the

N tt

Dostbase element -[ave(n)], related to the 3rd person masculine

N
n
singular proform !' ve(n)
a ] .

2. (C)VC-.

3637 ['po:liya] basketry box (gift to bride)

3467 [' pi: reke] small, very bitter cucumber

Cf. 3467 L'pi:re] sponge gourd

3. (C)VCV -.
tt
694 L'erivi] burnin_ sensation in mouth
"..

3686 L'podepi] covering, blanket

3511 ['pudume] wonder, marvel

n
3613 ['perija] increase, excess

799 ['odeve] break

 419

Cf. 694 ['eri]- (mouth) has burning sensation;

Class 2 verb
799 rodo]_
break (intr); Class 2 verb

3686 ['poda]- thatch, cover

3511/178 L'puttari] rice harvest festival

3613 ['peri]- increase (intr); Class 5 verb

4. CVCC- + CVNC -.

This structure, found in only one item, shows expansion of the

first component lexeme as two lexemes:

['buddi'vantige] state of being wise

Cf. ' buddi] wisdom

Finally, Figure 6 shows the system of elements operating at

Initial; all these elements are of very low frequency in the data.

Figure 6: The system of elements at Initial

/t, t, k/: these are set up to account for [d], [d],

respectively, as in:
ir tt

3724 [ 'polidi] auspicious ceremony

 420

L' garud.a] eagle

66 ['adige] cooking

this is realised as [ vj, by R1 and R4 (Ly , from /c /,

does not occur in the data at this position), as

in:

799 ,'odeve] break

/b /, /j /: these account for [b], [j], respectively, as in:

rr

1530 L'kurube(n)] shepherd

rr
3613 L'perija] increase, excess

/?, T, C, IC/: are realised as the voiceless stops [p], [t],

L , L'_-], respectively, as in:

952 LIkadipaj quickness

2132 ['ci rate] leopard

4524 ['bolicá] lamp

1376 ['kudike] cooking pot

/m /: this underlies the one instance of [m] in:

3511 [ ' pudume] wonder, marvel

/r/, /1/: these are realised as [r], [i], respectively, as in:

2363['navili] peacock
9,
554 breath

2.5 A further case of linear derivation

However, there is a further modification to be made to the

system at Noun, in view of the rather large number of nominals that

show a lexeme -final element -[e(n)] or -[e(n)] and have human male

referents, and the relationship between these and nominal lexemes

showing the lexeme -final element -[ati] and having human female

referents. For example:

4270 !'badave(n)] poor man

 421

_'badavati] poor woman

Arguments have been presented in the previous chapter to show that

such gender distinctions as exist in the Coors nominal system are

not appropriately to be handled in the syntactic base; however, we

have also seen that this does not rule out the possibility of syntag-

matic derivation of lexical items, and it has been argued that where

this possibility of formulating lexical relationships exists (linear

generative capability) it should be implemented, as representing a

stronger hypothesis than does pa.radi7matic derivation (parallel

generative capability). That is to say, it states not only that

certain items are related, but also that one is basic and the other(s)

derived in some way. Finally, we have seen that such a formulation

allows for economy of statement regarding such items; since only the

basic form has to be entered in full in the lexicon. A certain type

of voice relationship between verbs has been treated in an operation-

ally linear way (ChaPtsr 4) , and so has a certain type of derived

nominal (Chapter 5). Moreover, it has been noticed that these two

derivational relationships involve processes which are, in part,

identical to each other (in respect of the operation of /; /) . This

study will conclude with a further example where a linear derivation,

also involving /- /, seems to be indicated.

With a large number of gender- related nominals, a paradigr_atic

approach is the most reasonable; thus

87 L'ajje(n)J grandfather

L'ajj;] grandmother
n
e(n)]
are related to each other through the occurrence of either

or -Li] with the common element [ajj] -. Similarly, with

L'obbe(n)] one person (mast)

['obba(1)] id (fern)
 4 ^7

and with

L'dadU.e(n) j
stupid person (masc)

L'daddej id (few)

there is no justification for settim7 up such rules

as:

17. an/ /i/ in env.

All the evidence suC_,ests that -Li], -[e] and - ..(1)] form art of

the same :aradigmatic system as -[4n)]; for example, the pattern

of t',_eir distribution, with respect to the base types they occur

with, and the fact that not all feminine nouns in -` , -Le.] or
r if
-La(1)] show corresponding masculines in -Le(n)].

On the other hand, if a paradimatic analysis is adopted with

rez%rd to nomi nais in -[ e( n)

r+

] / -[ ati] , then the system at Noun

has to be modified in order to allow for an element of the unusual

form VCV (Lati]). Furthermore, this canonical Noun form covers just

the one type. So it might be argued that for this reason the para-

digmatic approach has to be abandoned here. However, even if a

syntagmatic approach is followed, the system at Noun will have to

undergo some modification, so the argument above is not sufficient

as it stands. The real reason why the paradigmatic generation of

-Lati] nominals is inadequate is because this is unable to formalise

the fact that all such lexemes have corresponding forms in -[e(n)],

to which they are related as ' feminines' to 'masculines'. There is

only one exce2tion_ to tills statement, in that ['polati], 'low caste

woman' has the corresponding masculine form [' poleye(n) ] . In all

other cases, it is possible to say both that no lexeme with -[(n)]

contracts a relationship with another in -[ati] unless its referent

is a human male, and that all lexemes with -Lati]show this relation-

ship.
 Ezª

o=I zap.,: o$ o00= iü Æe % sTq I2 '=oT%aal @oa}u con &=

lai «¿r4
- v ( Æegu 03 goç a e% y=RJxa pa s72. çe Qeqq.

22 'eQoro zo G2 02;71 rc:s un aeos eJi .z=æsl=Tri eera !}ue GJ;} oj sas%}e:la

aqq. lop; & `:e =I q; esG ó Tm`SIeu $oe3I2 2=0 ao32:Ga Jo GT:

a$Jo Yjao7 an4, .uT, s! i

}¡&o:a !¡au Y:r=o 3ç ç* 2e sse p
a% aJ. '.__$«TI m &aG eq1 zulG$uT=eJ Geie %û ) T%99 13u

J=a yuTe oN. üp IIó :s:W

'2L 3:03 4-

(Oasi)

'Oz Q

lz 11.114 4- 2 D« (=d) (ansi)

0%o4 a: o e}% l p% orna ouL `ea $II eI cI 22Ro 2pa22

e119, G =e%s 2o 2< sarna aaaT2 ú RD ee$j'a G!ª q.n =ou Ozsi T Iou sepn

3 NI1 s 4 = Re e$Ta ;o '= olg $=a a oIn a J:= ue sp QJ

'EE ZIZ\3 ® $/ + /# 2 si) (Ç

e2 `a %I aR æGCod eqg oæ a$I=a »$/ =ó pue /T/

=o !3 Æo§ '=aAa %zzTJ 'no- ouDI aa á7$ G=sEnssa ooano =o

 424

geminate sto_s (other than in base -final position after a short

root voel), and the second would ignore the fact that -[i] in

feminine lo minals is already provided for in the system (,Je work

within the system established thus far, as much as pos sibJ.. e) .

LS24 now a-:iplies, to e;:pan_d

It is at this point that we are able to justify the particular

phonological form An/ for this element, since we now have not only

to account for the vowel quality of the masculine lexeme -final

element, but also for the first vowel of the related feminine

element -[at!]. The -[;(n)] / -[ ati] alternation will be handled

at the systematic 7honemic level as between -/a/ and -/ant + *

where /n/ is the element that accounts for the divergence in

quality of the preceding vowel at the systematic phonetic level,

and is defined by the abstract feature z; its feature specification

is BInz. It functions phonologically as a trigger element in

respect of the mutation rule PI12(iv) but in respect of nothing else;

when it is followed by -/t + * i/, PI_2(ii) applies, and the nasal

consonant is lost, yielding the sequence -/atti/. PII2(ii) thus has

to be modified slightly, as follows:

23. (
l
) Pi
+* - P.P. (Pi / o) (PI2 (ii) )

In all other cases, the /2/ is retained, and PE2(iv) applico:

24. a e/n (Pi`:2(iv) )

/J is realised as [n]; but it does not constitute the necessary

environment for the insertion of the enunciative vowel [i] lexeme-

finally, and we therefore have to restrict R13 as follows:

 425

constrict_o_i,
back of tongue
25. Insert:
hi ches t,

lips unroune'_

Hence, /l/ comes to stand in lexeme -final position at the systematic

_) honetic level, and is deleted unless another element follows under

the same noun phrase node but external to the highest lexeme node.

That is, it has no realisation when a lexeme is followed by another

element (which may be a lexeme or a postbase or an element at

n n
Classifier) within the same lexeme (e.g. ['anne'tam ane(n)], not

* 'annan'tamra_ne(n)] or *['annan'dammane(n)]; and it has no real-

isation when it occurs finally under the highest lexeme node but is

followed by no syntactic element within the same noun phrase (as,

for example the plural affix, in 'annañga(1)]; the case affixes, as

in ['annana], ['annanda], ['annaiigi]; or the ?articles [e:], [o :],

[a:] or. [u:]). Such a situation clearly can be handled by realis-

ation rules but not by Phonetic mutation rules; we therefore need to

supplement R2 and R4 with the following subrules:

26. B -+ [no articulatory effect]/ Inz$- (R2)

?T?7

27. B -+ [no articulatory effect]/B nz# (R4)

]#

-,-
 426

The LS rules 24-8 then follow,

as given below, all save LS25

being identical in form to the corresponding rules in the preceding

chapter:

28. RiASC
/aJ + masc.]. (LS24)

/1-// ant *

29. FEM -+ ± /-/ - masc4 (LS25)

/e/ else
/aJ
30. PLUR -+ ic.
/81/ + plur 4 (LS26)

31. D1aN -+ .6 /a/ - anim 4 (LS27)

32. AN -+ An/ + anim 4 (LS28)

Note that the representation of FER has to be partly context -

sensitive by this analysis, since only -[i] occurs after -fat]- at

this position. Notice also that AN is represented by /a /; this is

because there is every reason, on formal and intuitive grounds, to

derive -Le(n)] in animate nouns from the same source as -[e(n)] in

masculines.

Finally, note that /1/ of the FEM and PLUR elements, which is

also defined by the feature z, having the specification BrIIIz, is

like /a/ in that it does not satisfy the conditions for the insertion

of the enunciative vowel [i] lexeme-finally (see 25 above). However,

its behaviour is in other respects much simpler; it is realised

everywhere as [1], and this is allowed for by the fact that the

feature z only provides the environment for no articulatory effect

with respect to the features B and I when it is preceded by the

 427

feature n (see 26 -7 above). When [1] is followed by a consonant

(of the plural affix, or one of the case affixes), or when it stands

in utterance -final position, it is deleted; otherwise (i.e. when

followed by a vowel, of the accusative case affix, or one of the

syntactic particles), it is retained. This is handled by AS6 at

the systematic phonetic level.

3.0 The rules cited in this chapter

Only those rules are given here which have been introduced or

modified in this chapter; for the articulatory sequence rules, see

the relevant sections of the preceding chapters (pp. 287 -90 and 371).

3.1 The lexeme structure rules

S#Lexemel Classifier

[Lexeme Base #Lexeme#

LS1
Postbase jl Base Postbase

Base ]] Classifier
416
116

HUE

i JA SC

LS20. i'k
PLUR

IIJAIS

AN

LS22. HUIv'I -4 MLi5C (FEIyIIId)

LS23. /t * / FErI
 42fl

/i/ / /ant 4-

L25. 17,1 -4 - 1712.3C4

/e/ else. 1
/a/

3.2 The phonological mutation rules

N
PK2(ii) ) * -4 P
-1

(iv) a -4 e/ _n
PY9(ii) C. C.C. / V 1; NOT:
i

(Erase

PE10(i) VC./V VO., C.V

(ii)
/ 1
V
#
c.i -4 VC.
1 #
C.
1

PL11(i) V -- V / / (a)

(ii) VL -* V

3.3 The realisation rules

 A?ç

[no articulatory effect] / Inzfr ,{h..

x1-...74
r
L
. ]* j

[ palatal p. of a.]

[ postalveolar p. of
/
a.]
()"(L)A

/ j
jI"
ri
R2. B
[dental p. of a.] /
B(r)IP(r)II
[alveolar p. of a.] / (r)II
[prepalatal p. of a., retroflex tongue tip] / r

[postdental p. of a.] / IV

[.dental p. of a.] else.

#...##

t[no articulatory effect]/

o

[occlusion] / =:j ( )^
(L,)s

"-
AxjO
(
X()"(L)A f Bj =i()
-4 [obstruction] / ll

B rI II xr 11

[contact, v.c.s
vibrating] /
Ox
B(r)
- T
X()h(L)4
C

[contact, velum lowered] / n

[contact] else.
 LEO

constriction, I b

back of tongue
[II
R13. Insert:
highest, III n

IV
lips unrounded
 431

APPENDIX I

Inventory of verbal lexemes, Classes 1 -5. Items

marked with an asterisk have a related form in

Appendix II; those with two asterisks have two

such forms. For further details regarding the

arrangement of the inventory, see Chapter 4,

pp. 197 -205. Transcription is Emeneau type B

(see Chapter 2, p. 100).

 Class 1

(C)VC

1 DED 569 ur- (uruv- uri -)* melt (intr)

2 606 or- (oruv- on -) fall asleep

3 1086 kar- (karuv- kari -)* be digested

4 2655 tir- (tiruv- tiri -)* turn round (intr)

5 4112 mur- (muruv- muri -)* tighten (intr)

6 432 el- (ëluv- eli -)* leave one's position

(C)VCC

7 573 ukk- (ukkuv- ukki -) boil over

8 784 okk- (okkuv- okki -) thresh (paddy)

9 849 okk- (okkuv- okki -) float away

10 909 kakk- (kakkuv- kakki-) vomit

11 1689 kokk- ( kokkuv- kokki-) be bent

12 2060 cikk- (cikkuv- become

13 2065 cikk- ( cikkuv- cikki-) become tired

14 2333 cukk- (cukkuv- cukki -) grow stout

15 2333 cokk- ( cokkuv- cokki-) be fat, (liquor)

intoxicates

16 (2625) tikk- ( tikkuv- tikki -) to wipe (slate clear

of writing)
n n
17 (2634) tikk- ( tikkuv- tikki-) strain (at delivery of
child)

18 2945 nakk- ( nakkuv- nakki -) lick

19 3540 pukk- (pukkuv- pukki -) boil (to loosen skin

or husk)

20 (4010) mukk- ( mukkuv- mukki -) strain (at stool)

21 4169 mukk- (mukkuv- mukki -) smear

22 DBI.A 5 agg- (agguv- aggi -) (fire) consumes

23 DED 2334 ogg- (ogguv- oggi -) (child) wriggles on

belly
 433

24 DED 3076 nogg- (nugguv- nuggi -) enter by force

25 4371 bagg- (bagguv- baggi-) bend down (intr)

26 3985 mugg- (mugguv- muggi-) swallow

27 3658 poiui- (poiuiuv- poiuii-) swell (grain, stomach)

28 4096 muiui- (muiuuv- muññ; -)* dive

29 (3233/3306) paiui- (pa.auv- paññi -) go stooping, bending

30 4025 mucc- ( muccuv- mucci -) close

31 3865 mecc- (meccuv- mecci-) be pleased

tt 11

32 1287 kitt- kittuv-

( kitti-) be got

33 1391 kutt- (kuttuv- kutti-) pound

34 961 kett- (kettuv- ketti -) tie, build

35 1717 kott- (kottuv- kotti -) tap, beat (drum)

36 2466 tatt- (tattuv- tatti -) ward, strike off

37 3119 nëtt- (ngttuv- ngtti-) shake sharply (intr)

38 3501 putt- (puttuv- putti -) be born, sprout

39 3674 pott- (pottuv- potti -) explode

40 4041 mutt- (muttuv- mutti -) bump against

41 (4043) mutt- (muttuv- mutti-) reach, touch

42 678 ënn- (ënnuv- ënni -) say, tell

43 648 utt- (uttuv- utti -) ooze

44 S 74 utt- (uttuv- utti -) untie (knot)

45 693 ett- (ettuv- etti-) arrive

46 859 ott- (ottuv- otti -) press, squeeze

47 1014 katt- (kattuv- katti -) blaze

48 1154 katt- (kattuv- katti -) wash (plate, hands)

1429 kutt- (kuttuv- kutti-) thrust, gore

49

1536 kutt- ( kuttuv- kutti-) thrust with fist

50

kett- (kettuv- ketti -) cut

51 1624

kett- (kettuv- ketti -) gore with horns

52 1625
53 DED 1740 kott- (kottuv- kotti -) pick up, peck

54 2238 cutt- (cuttuv- cutti -) wind around (tr/intr)

55 3320 patt- (pattuv- patti -) climb

56 3686 pott- (pottuv- potti -) cover completely

57 4355 batt- (battuv- batti -) go dry

58 4428 bitt- (bittuv- bitti -) sow

59 4519 bott- (bottuv- botti -) fear

60 (4156) mett- (mettuv- metti -) smear, plaster (mud

on wall)

61 572 udd- (udduv- uddi-) rub

62 1536 gudd- (gudduv- guddi-) pound with fist

63 2659 tidd- (tidduv- tiddi-) clean, settle (details

of ceremony)

64 3709 podd- (podduv- poddi-) (good /evil influence)

possesses

65 1336 kinn- (kinnuv- kinni -) tear into strips

66 3322 paon- (pannuv- papa' -) drizzle

67 3994 minn- ( minnuv- minni -) flicker, (lightning)

flashes

68 2498 tapp- (tappuv- tappi-) escape

69 2725 tupp- (tuppuv- tuppi-) spit

70 1025 kabb- (kabbuv- kabbi-) seize with open mouth

71 2543 tabb- (tabbuv- tabbi-) embrace

72 3255 pabb- (pabbuv- pabbi-) (creeper) twines

73 1536 kumm- (kummuv- kimmi-) churn

74 1634 kemm- (kemmuv- kemmi-) cough

n n H
75 2740 timm- (timmuv- timmi-)
sneeze
,.+ tumm- (tummuv- tnmmi-)

76 1969 cell- (celluv- celli-) sprinkle, scatter

77 2767 tull- (tulluv- tulli-) make small jumps

(a) vIic
78 DED 3537 puiig- (puñguv- pu$gi -) rot, fester a little

79 3962 minj- (miñjuv- miñji -) be left over

80 1408 kunt- (kuntuv- kunti -) be lame

81 2712 tuna- ( tunduv- tundi -) break (intr)

82 3440 pund- (punduv- pundi -) squeeze

83 3696 pond- (ponduv- pondi -) be raised, bounce

84 3709 pond- (ponduv- pondi -) be suited to

85 4119 mund- ( munduv- mundi -) precede

86 2735 dumb- (dumbuv- dumbi -) become full

87 2975 namb- (nambuv- nambi -) trust

(C)VC

88 593 u:k- (u :kuv- u :ki-) comb

89 1658 ka:k- (ka:kuv- ka:ki -) call

90 2003 ca:k- (ca:kuv- ca:ki -) rear a young one

91 3083 nu:k- (nu:kuv- nu :ki -) push

92 4370 ba:k- (ba:kuv- ba:ki -) pour off water from

cooked rice

93 1352 gi :c- (gi:cuv- gi:ci -) make a scratch

94 DBIA 274 pu:j- (pu :juv- pu:ji -) do acts of worship

95 DED 4479 bi :j- (bi:juv- bi:ji -) wave (intr/tr)

a+ bi :d- (bi:duv- bi:di -) (wind) blows, grind

with stone

96 3927 ma :j- (ma :juv- ma:ji -) (clothes) soil

97 1566 ku:t- (ku :tuv- ku:ti-) sharpen

It n n

98 2678 ti:t- (ti:tuv- ti:ti-) rub, stroke

no :t- (no:tuv- no:ti -) look at

99 3144

a:d- a:duv- a:di -)* dance

100 290 (

101 877 o :d- (o:duv- o:di -) run

 436

102 DED 1562 ku :d- (ku :duv- ku :di)* join (intr)

103 (2012) card- (ca:duv- ca:di -) throw

104 2247 cu :d- (cu :duv- cu :di) stick (flower) in hair

ft
105 2844 te:d- (te:duv- te:di -) search for

106 2927 to :d- (to :duv- to :di -) dig (a well)

107 2782 du :d- (du :duv- du:di -) push away

ft
108 3121 ne:d- (ne :duv- ne:di -) earn

109 3348 pa:d- (pa:duv- pa:di -) sing

110 4377 bard- (ba:duv- ba:di -)* fade,(face) is downcast

111 4548 bo :d- (bo:duv- bo :di -) beg

112 3931 ma :d- (ma :duv- ma :di -) do

113 349 a:t- (a:tuv- a:ti -) become mature

114 2154 u :t- (u :tuv- u:ti -) suck

115 1847 ko :t- (ko :tuv- ko :ti -) fill by scooping

116 2857 de:t- ( de:tuv- de:ti -)

If tt It
N de :t- (de:tuv- de:ti -) drive away

do :t- (do :tuv- do :ti -)

117 4392 ba:t- (ba:tuv- ba:ti -) take up in hand

118 886 o :d- (o :duv- o :di -) read

119 2310 se:d- (se:duv- se:di -) drink a deep draught

120 1341 ci:p- (ci:puv- ci:pi -) comb

121 346 a:r- (a:ruv- a:ri -)* become dry

122 461 i:r- (i:ruv- i:ri -) saw (wood)

123 650 u:r- (u :ruv- u:ri -) thrust stick to ground

while walking
u II ti

124 776 e:r- (e:ruv- e:ri-) (liquor) rises to head

125 1236 ka:r- (ka:ruv- ka:ri -) vomit

126 1353 ki:r- (ki:ruv- ki:ri -)* tear (intr)

127 2312 ke:r- (ke:ruv- ke:ri-)* go up to house

 457

,1
128 DED (1679) ke:r- (ke:ruv- ke:ri-)
winnow
129 1353 gi:r- (gi:ruv- gi:ri-)
gash
130 2169 ci:r- (ci:ruv- ci:ri-) grumble
131 2048 ja:r- (ja:ruv- ja:ri-) slip, be slippery

132 2796 tu:r- (tu:ruv- tu:ri-) defecate

133 (2793) du: r- (du:ruv- du: ri-) complain

134 2379 na: r- (na:ruv- na:ri-) smell (intr)

135 3311 pa: r- (pa:ruv- pa:ri-)* fly, leap

136 3641 (po:ruv- po:ri-) transport by pack -animal

137 3960 ma:r- (matruv- ma:ri-)* sell

138 3998 mu:r- (mu:ruv- mu:ri-) disregard, be superior

to

139 2312 se:r- (se:ruv- se:ri-) join (intr)

140 4000 mu:s- (mu:suv- mu:si-) smell (tr)

141 341 a:1- (a:luv- a:li-) rule

(C)VhTC

142 746 e:Ag- (e:riguv- e:iigi -) lengthen note

143 2777 ju:rig- ( ju:ñguv- ju:ñgi -)* sway (intr)

144 2573 ta: ñg- ( ta: Aguv- ta:ñgi-)* lean (intr)

145 2777 tu:ñg- (tu:ñguv- tu:ñgi -)* hang (intr)

146 3052 ni:ñg- (ni:ñguv- ni :ñgi -)* go aside

147 4371 ba:ñg- (ba:ñguv- ba:Agi -) become bent

148 4543 be:ñg- (be:riguv- be:ligi -) (rain) clears

149 4571 be:ñg- (be:$guv- be:Agi -) cut (plantain trunks

at wedding)

150 3852 me::ig- (me:riguv- me:ñgi -)* mix thoroughly (intr)

151 4130 mu:ñj- (mu:ñjuv- mu:ñji -) suck (penis)

152 2680 ti:nd- (ti:nduv- ti:ndi-) touch

153 3944 ma:nd- (ma:nduv- ma:ndi-) scratch

154 DED 3054 mi :nd- (mi :nduv- mi :ndi -)
swim

CVSC

155 1973 cavt- (cavtuv- cavti -) step on, kick

156 2454 tavd- tavduv- tavdi -) grope in dark

CVSNC

157 2386 navnd- (ñavnduv- navndi -) squeeze

158 4349 bavnd- (bavnduv- bavndi -)* be bent

159 S 249 kuynd- (kuynduv- kuyndi-) feel prickly

(C )VCVC

160 (203) alale- ( alasuv- alaki-) stir with a ladle

u n u
161 636 ulik- (ulikuv- uliki) flay
u u
162 506 odik- (odikuv- odiki -) sweep

163 1385 kodak- (kodakuv- kodaki -) shake (tr)

164 2365 nerak- (nerakuv- neraki -) groan

165 3623 porik- (porikuv- poriki-) pick up small objects

ft

166 4402 badik- (badikuv- badiki -) live (happily)

167 2655 tirig- (tiriguv- tirigi -) wander

168 3262 baraj- (barajuv- baraji -) feel for something,

grope
u u u
,fl u n u
169 725 elid- elidi -) write

170 3729 porid- (poriduv- porfidi -) undertake an office

n u u

171 578 urip- (uripuv- uripi -) play (flute)

172 4516 bevar- (bevaruv- bevari -) sweat

173 200 alas- (alasuv- alasi-) (curry, rice) spoils

u u u

174 Indo -Aryan gunis- (gunisuv- gunisi -) calculate

 439

(C) VCVNC

175 DED (56) adarig- (adariguv- adarigi-)* be within authority of,

submit to

176 70 adei g- (adariguv- adarigi-)* be piled in order

11

177 439 erarig- (erañguv- erañgi-)* (food) slips down

throat

178 517 onarig- (onañguv- onarigi-)* dry, wither

179 1096 kalarig- (kalaiiguv- kalarigi -)* be stirred up

180 1501 kulirig- (kuliriguv- kulirigi-) * shake (intr)

181 2655 terarig- (terariguv- terarigi -)* move, shift (intr)

182 (2825) telarig- (telariguv- telarigi-) (water) settles and

becomes clear

183 (3796) madarig- (madariguv- madaiigi -)*

.
fold (intr)

184 3875 malarig- (malariguv- malarigi -) lie down, neglecting

work
If

185 molirig- (moliriguv- molirigi -)* be pressed

186 (2213) curind- (curinduv- curindi -)* shrink (intr)

187 839 olamb- (olambuv- olambi -)* gargle

188 dolamb- (dolambuv- dolambi -) gurgle

189 balamb- (balambuv- balambi-) serve (food)

Class 2

CV

no- (novv- nond-) pain (intr)

1 3143

CV

1197 ka:- (ka:v- ka:nd-) caw

2

1551 ku:- (ku:v- ku:nd-) crow

3

CVCC
(time) passes
4 1142 kayy- (kayyuv- kayñj-)*
 44v

5 DED 1143 kayy- ( kayyuv_ kayñj -) be possible

6 1469 kuyy- (kuyyuv- kuynj_)* (hole) is bored

7 S 249 kúyy- (kúyyúv- kúyñj -) feel prickly

8 3982 meyy- (meyyuv- meyñj -)* (paddy) is well pounded,

cleaned

9 1772 koll- (kolluv- kond-) kill

10 3720 poll- (polluv- pond -) stitch

(C )VC

11 3059 ni :d- (ni :duv- ni :nd -)* stretch out straight

(intr)

12 306 a:y- (a:yuv- a:ñj -) choose

13 1219 ka:y- (ka:yuv- ka:ñj -)* be hot, boil, bask in sun

14 1337 ki:y- (ki:yuv- ki :ñj -) become rotten

15 2846 te:y- (te:yuv- to :ñj -)* wear off (intr)

16 3362 pa:y- (pa:yuv- pa:ñj -) dash into

17 4179 me:y ( meyyuv- me:ñj -)* graze (intr)

18 2597 ta:r- (ta:ruv- ta:nd -)* descend, (sun) sets

n t ri
19 2683 ti:r- (ti:ruv- ta:nd-)* be used up, finished

20 2856 te:r- (te:ruv- te:nd-) (man) becomes full-grown

21 2353 to:r- (to:ruv- te:nd-) leak

22 2373 ne:r- (ne :ruv- ne:nd-)* hang (intr)

23 2380 ne:r- (ne:ruv- ne:nd-)* rise up, come to view

24 3124 ne:r- (ne:r- ne:nd-) come to view

25 pa:r- (pa:ruv- pa:nd -) ladle, put on (clothing)

26 4513 bi:r- ( bu:ruv- bi.:nd -) *

(flowers) open
^a bu:r- (bu:ruv- bu:nd -)

27 2998 mu:r- (mu:ruv- mu:nd -) disregard, be superior

to
 441

(c)vcv

28 DED 98 ani- (aniv- ániñj_) wear, enjoy (jewels)

29 176 ari- (arie- ariñj -)* (liquid)is strained

30 265 ari - (arie- ariñj -)* find out

31 213 ali- (aliv- aliñj-)* dissolve (intr)

n n rr
32 366 idi- (idiv- idiñj -)* (wall) falls, (land)
slips
rr n n
33 426 (iliv- iliñj -)* descend, get ready

34 565 uri- (uriv- ariñj -) burning sensation is

felt

35 848 oli- (oliv- oliñj -)* halt for night, dwell

36 843 oli- (olio- oliñj -)* fruit drops from tree

abundantly

37 1073 kari- (kariv- kariñj-)* be singed

38 1121 kávi- (káviv- káviñj-)* lie face down

n n

39 994 keni- (keniv- keniñj-)* get stuck

40 2072 cedi- (cediv- cediñj-) break with a crack

(intr)

41 2473 tani- (taniv- taniñj-) (thing) becomes cold

42 1950 tari- (tariv- taniñj -)* bend to one side (intr)

43 2815 tiri- ( tiriv- tirifj -) come to be known

44 DBIA 212 tudi- (tudiv- tudiñj -)* be praised

n rr n

45 DED 2825 teli- (teliv- teliñj -)* (sleep) leaves one

46 2353 tori- (torio- toriñj -) pour (tr)

47 (2343) tori- (torio- toriñj-) (body-part) itches

48 2988 nali- (naliv- naliñj -) play

2387 neri- (neriv- neriñj -)* get entangled, (cloth)

49
is wrung

50 3297 pali- (paliv- paliñj -) speak ill of

pini- (piriv- piriñj -) ** disperse (intr)

51 3435

podi- (podiv- podiñj -) cover completely in a

52 3686
heap
 442

53 DED 3725 poli- (polie- poliñj -)* break (intr)

54 3530 poli- ( poliv- poliñj -) bark repeatedly

55 3716 poli- (poliv- poliñj -)* increase (intr), (crops,

cattle)

56 4438 biri- (biriv- biriñj-)* open (jackfruit),

dismantle (house)

57 3898 mari- (marie- mariñj -)* roll (intr), overflow

58 73 ade- (adev- adand -)* (gap) is stopped

59 799 ode- (odev- odand -)* break (pot) (intr)

60 957 kade- (kadev- kadand -) grind with mortar and

pestle

61 (1142/1157) kale (kalev- kaland -) lose something (thing,

relative)

62 2294 jale- (jalev- jaland-) sway (intr)

63 2460 tade- (tadev- tadand -)* be obstructed

64 2551 tale- (talev- taland-)* become weak

65 2655 tere- (terev- terand -)* thresh about on ground

66 2211 tore- (torev- torand-)* (string) is wound round

67 3049 nere- (nerev- nerend-)* attain puberty

68 3006 nele- (nelev- neland-) ** (person) gets wet

69 3318 pare- (parev- parand -) utter

70 3297 pale- (palev- paland-) speak ill of

71 4360 bare- (barev- barand -)* be fried

72 bale- (balev- baland -)* (eyes) become blurred

73 4464 bole- (bolev- boland-)* become full-grown

74 3880 male- (malev- maland-)** turn face upwards (intr)

(C )VCVC
n n n

75 571 urid- (uriduv- urind-)* roll (intr)

n

2211 turid- (turiduv- turind-)* be rolled up

76

borad- (boraduv- borand-)* be in confusion

77 4519
through fear
 443

Class 3

CV

1 D3D 1607 co:- (co:p- co:nd -)

become red

(C)VC

2 407 ir- (ipp- inj -)

be (in place)

3 2526 tar- (tapp- tard -) give to 1st, 2nd person

4 4311 bar- (bapp- band -)* come

(c)vcc

5 516 unn- (umb- und -) eat (rice)

6 2670 tinn- (timb- tind -)* eat (not rice)

7 3043 nill- (nipp- nind -) stand

CVO

8 1209 ka:n- (ka:mb- kand -)* see, seem

(C)vCv

9 312 ara- (arap- arand -) make hoarse noise, moo

10 200 ala- (alap- gland-) crave

11 252 ala- (alap- aland -) measure

12 400 era- (erap- erand -) beg

13 929 kada- (kadap- karand-)* cross

14 1166 kara- (karap- karand -) milk

15 1092 kala- (kalap- kaland-) knead.

16 1385 koda- ( kodap- kodand -) shake (body) (tr)

17 2667 tora- (torap- torand -) open

18 2957 nada- (nadap- nadand -)* walk, perform ceremony

19 3255 para- (parap- parand -)* crawl

20 3725 pola- (polap- poland -) open the mouth

21 3897 mara- (marap- marand -) forget

 444

Class 4

(C)VC

1 DED 375 id- ( iduv- itt -) drop ( tr)

it ii 11

2 1614 ked- (keduv- kett -)* be spoiled, extinguished

3 2183 cud- (cuduv- cutt -) burn (tr)

4 2865 tod- (toduv- tott -) touch

5 3190 pad- (paduv- patt -)* lie fallow

6 3191 pad- ( paduv- patt_)* suffer, (something) hits

7 4419 bud- (buduv- butt -)* release, let go

8 2833 ter- (teruv- tett -) pay (penalty)

9 3622 per- (peruv- pett -) bear (child)

CVCC

10 1628 keyy- (keyyuv- kejj -) work

11 1763 koyy- (koyyuv- kojj -) pluck, reap

12 3610 poyy- (poyyuv- pojj -) rain

13 3704 Poyy- ( poyyuv- pojj -) beat

4566 bayy- (bayyuv- bajj -) abuse, scold

14
ti

15 1641 gell- (gelluv- gedd -) win

(C)VC

16 3103 ne:y- (be:yuv- nejj -) spin (thread)

17 723 e:1- (e:v- edd -)** get up

4402 ba:1- (ba:v- bald -) (child) lives, rule

18

bu :1- (bu:v- Budd -)* fall

19 4457

(o)vcvc

557
orad- (oraduv- orat-) answer
20
559
dress (well)
21 3751 porad- (poraduv- porat -)
 445

22 DED 4076 morad- (moraduv- morat-) weep

Class 5

(C)V

1 781 o- (opp- ott-) be suitable, consent

2 2904 to- (topp- tott-) salute

CV

3 1192 ka:- (ka:p- ka:t-) wait

4 4057 mu:- (mu:p- mu:t-) grow up

(C)VC

5 39 ay- (ayp- aye -) send

6 1047 kay- (kayp- kayc -) be bitter

7 1136 kay- (kayp- kayc -) remove, undo (knot)

8 city- (cúymp- cúyñc-) (fly, ant) bites

9 2894 tuy- (tuymp- tuy/1 -) (snake) hisses,

(bull) snorts

10 (4057) mup- (muyp- muye -) increase, multiply, grow

(as worms in fruit)

4297 bay- (bayp- bayc -) (stomach) hungers

11
(3165) pay- (payp- payc -) (stomach) hungers

12 3708 por- (popp- pott -) fight

13 1156 kal- (kapp- katt -) steal

(C)VC

co:l- (co:p- co:t-) become tired

14 2354

15 2936 co:l (co:p- co:t-) be defeated

427 i:l- (i :p- :t-) drag

16

17 904 o:l- (o:p- o:t-) have intercourse with

hear, ask
18 1677 ke:1- (ke:p- ke:t-)
 446

19 DED 3588 pu:1- pu p- pu: t-)

( :
bury

(C)VCV

20 191 ara- (arap- arat -) grind with rolling stone

21 266 ara- (arap- arat-) cut

22 346 ara- (arap- arat -) (water) dries up

23 ala- (alap- slat -) sit

24 1175 kara- (karap- karat -) become black

25 (1305) kera- (kerap- kerat -) scrape (yard clear of

grass, weeds)

26 1321 kala- (kalap- kalat-) dig

27 1809 koya- (koyap- koyat -) string (flowers)

28 1496 kora- (korap- korat -) bark

29 1504 kola- ( kolap- kolat -) (plant) shoots against

planter

30 3050 nena- (nenap- nenat -) think

31 3493 poda- (podap- podat-) flutter, tremble

32 3686 poda- (podap - podat-) thatch, cover

33 3686 pola- (polap- polat -) cover

34 67 adi- (adip- adic -) (lightning) strikes

35 DBIL 44 udi- (udip- udic -) (sun) rises

36 DED 694 eri- (erip- eric-) (mouth) burns

37 853 oli- (olip- olic -) hide (intr)

38 839 oli- (olip- olic -) wash (clothes)

39 945 kadi- (kadip- kadic -) bite

40 kani- (kanip- kanic -) throw

1158 kali- ( kalip- kalic-) play

41

1378 kudi- kadip- kudic -) drink

42 (

kuri- (kurip- kuric-) paint

43 1533

kuli- (kulip- kulic -) take bath

44 1522
 447

45 DED 1735 kodi- (kodip- kodic-)

boil, kiss, feel love
for
46 1798 kori- (korip- koric -) (snake) strikes

47 1917 cadi- (cadip- cadic -) cheat

48 2562 tari- (tarip- tarit -) chop up small

49 2827 tali- (talip- talic-) sprinkle

50 2835 teri- (terip- teric-) totter about

tr rt tt
51 2828 teli- (telip- telle -)
laugh
toli- (tolip- tolic-) JJ

52 2937 toli- (tolip- tolic -) peel, hatch

53 DBIA 242 padi- (padip- padic -) learn

54 DED 3317 pari- ( parip- parie -) pluck

55 3436 piri- ( pirip- piric -)

make (rope) by twisting
r biri- (birip- biric-)

56 3412 pudi- (pudip- pudic -) catch, hold

57 (3705) pori- ( porip- poric-) fry (rice, etc.)

58 3707 pori- (porip- poric-) pull up (plant, peg)

59 4272 badi- (badip- badie -) hammer, pound

60 4322 bali- (balip- balie -) snatch, pull

61 3835 madi- (madip- madie -) (animal) lusts for female

62 4078 muri- (murip- muric -) make a cut

rt tt tr
63 69 adi- (adip- adit-) approach
n u n
64 502 udi- (udip- udit-) put on (sari)

65 1523 kuli- (kulip- kulit -) cold feeling is

66 1708 kodi- ( kodip- kodit-) give to 3rd person

67 3494 pudi- ( pudip- pudit-) (body -part) swells

H n rr

68 3613 peri- (perip- perit -) increase (intr)

69 3729 pori- ( porip- porit-) carry

70 4524 boli- (bolip- bolit -)

tt t' tt become white
 448

71 DED 285 modi- (modip- modit -) be sweet

72 1927 cáye- (cáyémp- cáyént-;

N chew
44 cáve- (cavémp- cávént-)

73 2026 taye- (tayep- tayet -) level a field

 449

APPENDIX II

Inventory of derived verbal lexemes. In each entry

the basic form (from Appendix I or III) is given, followed

by its volitively related form. The order of classes in

Appendix I is preserved; in addition, derived forms of

Class 1 are presented first, followed by those of Class

5. For further details, see Chapter 4, pp. 205-8.

Transcription is Emeneau type B (see Chapter 2, p. 100).

 450

Class 1
Class 1

(C)VC (C)VCVC

1 DED 569 ur- melt (intr) urik- id. (tr)

3 1086 kat- be digested karak- digest

4 2655 tir- turn (intr) tirik- id. (tr)

5 4112 mur- tighten (intr) murk id. (tr)

n
6 432 el- leave one's elak- make leave
position position

CVCC CVCC

28 4096 mutui- dive mirk- dip (tr) under

water

(C)VC (c)Tic

100 290 a:d- dance a:t- make to dance,

drive

102 1562 ku:d- join (intr) ku:t- id. (tr)

110 4377 ba:d- fade (intr), face ba:t- fade (tr)

is downcast

121 346 a:r- dry (intr) a:t- id. (tr)

126 1353 ki:r- tear (intr) ki:t- id. (tr)

127 2312 ke:r- go up to house ke:t- take (person)

up to house

135 3311 pa:r- fly, leap pa:t- winnow

137 3960 ma:r- sell ma:t- change

CVNC CVC

sway (intr) ju:k- id. (tr)

143 2777 ju:iig-

lean (intr) ta:k- id. (tr)

144 2573 ta:iig-

hang (intr) tu:k- id. (tr) , weigh

145 2777 tu: rig-

go aside ni:k- put aside

146 3052 ni:Aig-

mix (intr) me:k- id. (tr)

150 3852 me:ñg-
 451

CVSNC CVSC

158 DED 4349 bavnd- be bent bavt- bend (tr)

(C)VCVNC

175 (56) adaiang- submit to adak- hold in closed

hands

176 70 adaAg- be piled in order adak- heap in order

177 439 eraiig- (food) slips down erak- let slip down
throat throat

178 517 onaAg- become dry onak- dry (tr)

179 1096 kalaAg- be stirred up kaiak- stir up, churn

180 1501 kuliiig- shake (intr) kulik- id. (tr)

181 2655 teraAg- shift (intr) terak- id. (tr)

183 (3796) madaisg- fold (intr) madak- id. (tr)

If n
185 moliñg- be pressed molik- press

186 (2213) curind- shrink (intr) curit- id. (tr)

187 839 olamb- gargle olap- wash plates,

pots

Class 2 Class 1

CVC CVC
11

11 3059 ni :d- stretch (intr) ni:t- id. (tr)

13 1219 ka:y- boil (intr) ka:c- id. (tr)

18 2597 ta:r- descend ta:t- lower, close

(window,door)

ne:r- hang (intr) ne:t- id. (tr)

22 2373
t,
0
ne:r- rise up, come to ne:t- lift
23 2380
view

bi :r- bist- open (bag,

26 4513
(flowers) open knife)
ti bu:r- t-
 452

(c)vcv (c)vcvc

30 DED 265 ari- find out arip- inform, tell

tt

33 426 ili- descend ilip- lower

11

45 2825 teli- (sleep) leaves telip- cause (sleep)

one to leave one
t,
51 3435 piri- disperse (intr) pirip- make disperse

64 2551 tale- become weak talat- make weak,

exhaust

68 3006 nele- get wet nelat- wet

73 4464 bole- become full -grown bolat- grow (tr)

74 3880 male- turn face upwards mal at- id. (tr)

(intr)

(C)VCVC (C)VCVC
tt

75 571 urid- roll (intr) urfit- id. (tr)

76 2211 turid- be rolled up turit- roll up (tr)

77 4519 borad- be confused by borat- frighten

fear

Class 3 Class 1

CVC CVCVC

4311 bar- come barat- cause to come

4

CVCC CVC

6 2670 tinn- eat (not rice) ti:t- feed by hand

CVC
CVC

see, seem ka:t- show

8 1209 ka:n-

CVCVC
CVCV

kadat- take across

13 929 kada- cross

nadat- make to walk

18 2957 nada- walk

parat- spread (grain)

19 3255 para- crawl
 453

Class 4 Class 1

CVC CVCVC
tt
it
2 DED 1614 ked- be ruined kedit- ruin

5 3190 pad- lie fallow padit- let lie fallow

6 3191 pad- suffer padit- make suffer

7 4419 bud- let go budit- make to let go

VC vcc

17 723 e:l- get up (intr) app- raise up

CVC CVC

19 4457 bu :1- fall bu:k- fell (tree),

pour

Class 2 Class 5

CVCC CVC

4 1142 kayy- (time) passes kay- pass (time)

6 1469 kuyy- (hole) is bored kuy- bore (hole)

8 3982 keyy- (paddy) is pounded key- pound (paddy)

CVC CVC

15 2845 te:y- wear off (intr) te:y id. (tr)

17 4179 me:y- graze (intr) me:y id. (tr)

Il II

19 2683 ti:r- be used up ti:r- use up

(c)vcv (c)vcv

ari- (liquid) is ari- strain (liquid)

29 176
strained

dissolve (intr) ali- id. (tr)

31 213 ali-

idi - knock (wall)

32 366 idi- (wall) falls
over

oli- beat (fruit)

36 843 oli- fruit falls from
from tree
tree
 454

37 DED 1073 kari- be singed kari- singe

4.40
38 1121 kavi- lie face down kavi- turn face down
(tr)
R
It
39 994 keni- get caught keni- entangle
42 1950 tari- bend to side tari- id. (tr)
(intr)

44 DBIA 212 tudi- be praised tudi- praise

49 DED 2387 neri- (cloth) is wrung neri- wring (cloth,

etc.)

51 3435 piri- disperse (intr) piri- spread (leaves)

53 3725 poli- break (intr) poli- id. (tr)

55 3716 poli- increase (intr) poli- (god) increases

(crops, cattle)

56 4438 biri- open (jackfruit) biri- spread (leaves,

blanket)
dismantle (house)

57 3898 mari- roll (intr) mari- id. (tr)

58 73 ade- (gap) is stopped ada- stop (gap)

59 799 ode- be broken (pot) oda- break (a pot)

62 2294 jale sway (intr) jala- make sway

n
63 2460 tade- be obstructed tadi- obstruct

65 2655 tere- thresh about tera- make to thresh

about

66 2211 tore- (string) is. wound tora- wind (string)

round around

67 3049 nere- attain puberty nera- fill, load

68 3006 nele- get wet nela- wash (clothes)

bare- be fried bara- fry

71 4360

(eyes) blur bala- id.

72 bale-

turn face upwards mala- id. (tr)

74 3880 male-
(intr)

VCV VC

oy exhaust, finish
35 848 oli- halt for night,
dwell
 455

Class 4 Class 5

VC VCV
t,

17 723 e:l- get up (intr) edi- raise

 456

APPENDIX III

Inventory of irregular and defective verbal

lexemes. Certain of these (marked with an

asterisk) have derived forms, which are given

at the end. For further details, see Chapter 4,

p. 208. Transcription is Emeneau type B (see

Chapter 2, p. 100).
 457

Irregular verbs.

DED 282 a:g- (a:p-, a:y -, a:n -, a -)* : become

2 592 u :l- (upp-, utt -) plough

3 1788 konda- (kondap -, ko :nd -)

bring
ko :nda- (ko :ndap-, ko :nd -)

4 2002 ca:l- (ca:v, catt -) die

5 2958 nadi (nadip -, natt-) plant

6 3531 pola (polav -, poland -)* dawn

7 3686 poda (podap -, pott -) cover

8 3716 pola (polav -, poland -)* live happily

9 3734 po:g- (po:p-, po:y -, po :n -, po -) : go

10 4540 be :y- (be :yuv-, beñj -)* (rice) cooks

11 4565 bey- (bepp -, becc -) keep, appoint

Defective verbs.

1 DED 599 ull- (mid -) be (in a place)

2 737 *enn- (emb -, end -) say

3 1788 koll- (kovv -, kond -)* take

4 2007 *ca:y- (ca :ñj -)* be straight

5 3755 po:r- (only in negative, po:ra) be sufficient

6 4548 bo :nd- (bo:ndu, bo:nda) be wanted

Bases derived from irregular and defective verbs.

Class 1

DED 282 a:k- make stay in a place

3531 polat- spend night till dawn

3716 polat- make live happily

1788 kolit fasten

 458

DED 2007 ca:c- straighten (hand)

Irregular

DED 4540 bey- (bepp-, becc -) cook (rice)

 459

APPENDIX IV

Inventory of nominal lexemes. This contains the data

discussed in Chapter 5, and consists for the most part

of simple lexeme types (in terms of the model described

in Chapters 4 and 5). The arrangement of items largely

follows the pattern of Appendix I, except that here the

ordering with reference to final elements is as follows:

-Li], -Le], -[i], -Lub La], -1.;(n)], -[a(1)] (C)VS

and (C)VS types, where S = -Ly], are a special case of

lexemes with final /i /, and are ordered immediately before

the entries showing -Li]. Transcription is Emeneau type

B, with the modification proposed in Chapter 2 (pp. 92-3).

 460

(C)vs

1 479 oy nail

2 1683 kay hand, arm

3 1469 kuy sting, hole

N
4 2894 tuy the hiss of a snake

5 3104 ney ghee

6 4162 may body

(c)vcv
7 63 adi place below, foot measure

8 373 idi the whole

9 udi rope -hung dish for keeping vessels in

10 548 umi
paddy husks
ummj

11 565 uri burning sensation

12 695 eri Eleusine coracana

13 710 eli rat

14 837 oli noise of a metal bell

15 945 kadi bite

16 1073 kari black

17 1171 kari curry

18 1158 kali play

19 1324 kiri lip

20 1379 kudi family of servants living in one hut

21 1533 kuri drawing, spot on forehead

22 kusi
happiness
N kusi

23 1645 kemi
9v
ear
kevi
 461

24 994 keni bird -trap; cunning

25 1704 kodi top, flag

26 1735 kodi love, desire

27 1799 kori sheep

28 1318 gili

parrot

.gini

29 1533 guri aim

30 gedi
boundary
gadi

31 2223a culi
leaf, shoot
tisuli 7

32 1281 cedi
it spark
kedi

33 2269 cedi

» cedi anger

N codi

34 2484 tadi trunk (of tree, body)

35 tari small bits (of rice after grinding)

36 2655 tiri wick

37 2709 tuni cloth

38 tedi end, top

39 2764 tuli drop of liquid

It

40 2828 teli
laughter
toli

41 2699 dudi drum

42 2981 nari tiger

43 3089 nuri
very small bits of broken rice
nucci
 462

44 3072 nuli cluster of paddy plants growing in

one hole

45 neri pleat, tuck, fold (in sari)

46 3183 padi door

47 padi measure (2 seers ?)

48 3209 pani work

49 padi
slowness
N padika J

50 3322 pani fever

51 3412 pudi handful, morsel

52 3546 pulì sour, orange, vinegar

53 3667 podi powder

54 pori puffed paddy

55 badi side

56 4308 bari side (of body, river, etc.)

57 bali
fart
...boll

58 bisi heat

59 4502 bodi act of shooting

60 4524 boli light

61 3804 madi ritual purity

.

62 4187 masi
charcoal
N.
masi

63 miti limit

64 3800 mudi heel of foot

65 3975 mudi small unripe fruit

66 4057 mudi old(ness)

67 4092 moli subject of talk or song

 463

68 16 age paddy seedling

69 192 are half

70 193 are waist

71 ede chest

72 701 ere worm

73 1705 ele betel leaf

ti

74 368 ede the company or proximity of a person

Il
75 387 ene double

76 799 ode piece of something hollow

77 618 ore
state of coagulation of milk
vare

78 621 ore sheath of knife

79 2337 ole
hearth
.v vale

80 kade

81 929 kade end

82 kate story

83 1175 kare stain

84 1087 kare bank

85 1104 kale scar, white spot on nail

86 kale skill

87 1157 kale weeds

88 1296 kede white ant

89 1185 kene act of bellowing

90 1648 kere tank

91 1386 kode umbrella

92 1504 kole bunch of plantains

93 1155 gale long stick

 464

94 gere line, margin

95 1897 jade woman's hair -plait; hood of cobra

96 2294 jale long thin pliable stick

97 (2460) tade obstacle

98 2529 tale end

99 2557 tale
loop of rope, noose
r dale

100 2704 tode thigh

101 2710 tone a person's support

102 2216 tore gourd

103 tore circle, going round

104 213 dore king, master, European

105 2957 nade limb, part of body

106 2984 nere greyness)

107 3074 nere

froth
» nore

108 3049 nere that which fills up something

109 3043 nele halting place

110 pade battle

111 3319 pare drum

112 pare measure (10 seers)

113 3296 pale old(ness)

114 3621 pere cream

115 3483 poge smoke

116 3523 pore thatched roof

117 3729 pore load

118 3714 pole pollution

119 3543 pole river

 465

120 4304 bare stripe, brand mark

121 4315 bare steep slope

122 bare mud wall

123 4326 bale net

124 4348 bale bangle

125 4424 bede intercourse of animals

126 4448 bele cost

127 4464 bole crop

128 3911 mane house

129 3897 mare place out of sight

130 3882 male thick jungle land; cardamom plantation

131 3893 made rain

132 3834 mone calf of leg

133 4119a mone sharp point

134 4076 more lamentation

135 4087 mole breast

136 2102 iri night

137 600 oli the inside

138 1002 kadi ear (paddy, wheat)

139 1070 kari intestines

u
140 2473 tani coolness

141 2271 tedi thunder

142 3192 padi swamp

143 4540 bedi

heat of sun
bodi

144 3903 mari counter

145 202 alu cowrie

 466

146 953 kadu mustard seed

147 1483 kuru seed

148 1537 koru defect

149 koru management

150 guru teacher

151 2460 tadu lateness

152 2959 nadu middle

153 3537 pulu worm

154 3986 malu chillies

155 suru beginning

156 353 ija trouble

157 560 ora mortar of stone or wood

158 kada debt

159 939 kada ocean

160 1098 kala big pot

161 1160 kala threshing floor

162 1638 kela belly

163 1644 kévá wall

164 1477 kora gullet, windpipe

165 koda pot, pitcher

166 kola pond or tank

167 1511 kola flute

.

168 1613 gida

plant
ivgl'da

169 cala determination, stubbornness

170 1337 cela Pus

171 jana people

 467

172 tada delay, tardiness

173 tara sort

174 2452 dada noise of a thud

.

175 dada bank (sea)shore

176 daya kindness

177 dina day

178 221 vaga jewelry

II

179 3046 nela

shade, shadow
.r pala

180 3044 nela ground

181 3061 noga yoke

182 DIA 240 pata picture

183 pata kite

184 DIA 250 pana money

185 3226 pata softness

186 5630 para side

187 3265 para pebble

188 3612 peda name

189 4317 bala force

190 buda origin; bottom of anything (e.g. tree)

191 4436 bera finger, toe

192 3817b mana sand

193 mana maund, quarter (= 18 lbs.)

194 3856 mara tree

195 3966 mida state of being on top

196 4053 moda an article belonging to one

197 4071 mona

hare
_ mola
 468

198 4106 mora winnowing basket

199 4093 mola cubit

200 raja holiday

201 rasa sweetness

202 vara small copper vessel for extruding

rice paste

203 visa
. N poison
.visa

204 sima lion

(C)VCCV(C)

205 178 ekki husked rice

206 44 acci jaggery cake

207 46 ajji great grandmother

208 addì obstacle

209 123 atti sp. fig

210 390 itti Ficus gibbosa

211 666 ecci scraps of food

212 805 otti rice bread

213 925 kajji itch

214 DIA 68 kaddi small stick, slate pencil

215 1012 katti knife

216 DIA 85 kaññi virgin

217 1156 kalli thief (fem. )

218 1371 kir i child

219 1390 kutti non-Coorg child

220 1427 kutti tall narrow container

221 DIA 108 kuppi bottle

222 1450 kummi

n a
Gmelina arborea
N kumbili
 469

223 1526 kulli dwarf (fem.)

224 1777 kolli small rocky stream

225 caddi shorts

226 2135 cinni

small
ti cenna I

227 2834 tetti narrow walk to house

228 2643 dimmi stopper

229 3089 nucci broken grains of rice

230 3097 nenni

of body
h NM.
N nenni

231 3118 netti forehead

232 DIA pakki penis

233 ?3207 patti space before house, for spreading

234 patti list

235 3294 palli lizard

It It

236 3309 palli Poleye or Ihïe:de hut; village (in

'place names)

237 pucci madwoman

238 3551 polli fuel, firewood

239 3600 potti box

240 4277 batti basket, measure

241 4278 batti stomach

242 4351 balli creeper

243 4524 bolli silver

244 3782 macci ceiling

245 4025 mucci lid, cover

246 89 atte water leech

247 88 atte bark of tree

 470

248 576 ucce urine

249 ubbe washerman's boiling -pot

250 699 emme female buffalo

251 726 enne

gingily oil
enne

252 849 obbe open drain

253 922 kacce perineal cloth

254 961 katte bund, platform under tree

255 1149 katte ass

256 1027 kappe toad

257 kukke small basket

?58 kuññe unmarried girl

259 1396 kutte log

260 1415 kunne penis (vulgar)

261 kumnme shaped basket; stomach

262 1629 kette wet mud, mire

263 1689 kokke crook, hook

264 1761 komme corn -bin

265 1402 gudde heap

266 1610 gejje small bell

267 1795 golle anus

268 1876 cakke

j ackf rui t
cekke

269 1914 canne buttocks

270 catte market

271 1929 cappe tastelessness, insipidity

272 2341 colle

nostril
N solle
 471

273 1968 jalle cane of sugar

274 2463 tatte flat plate

275 2469 tatte wooden cattle bell

276 1910 dadde dull person (fem)

277 2883 donne club

278 ?3147 namme festival

279 pacce rottan basket

280 3161 pacce green, blue, cold; emerald

281 patte right of tenure of land

282 3206 patte striped or spotted

283 patte bark of tree, esp. cinnamon

284 ?3449 palle female of various animals

285 potte belly

286 batte path

287 batte clothing

288 bille badge, disc

289 4540 bekke heat

290 4511 bonne butter

291 macce mole (on body)

292 4048 mutte egg, testis

293 4157 mette mattress

294 4167 melle lightly, slowly

295 motte egg, hill

296 2133 ratte upper arm

297 2133 rekke wing

298 saune gesture

299 sonne zero

 472

300 569 ukki steel

301 uddi common pulse

302 2201a uppi salt

303 600 ulli inner thought

tl It
304 698 etti
bull, bullock
ti etti

305 726 elli gingily seed

306 784 okki paddy to be threshed

307 katti cot

308 961 katti knot, bundle

309 973a kann; eye, small hole

310 1091 kalli stone

311 1158 kalli liquor

312 kicci fire

313 1528 kutti small piece of wood

314 1440 kuppi dropping of dung

315 kummi mushroom

316 1624 ketti cut, blow

317 1690 kokki beak

318 1713 kotti shed

319 1807 kotti tip, nipple

320 1619 ginni joint in wrist, fingers

321 ginni milk pudding; cheese

322 1393 Butti secret

323 gotti something known

324 2065 cik',Ji numbness of mouth from areca nut

325 cippi shell

326 2188 culli spot, mark

 473

327 2280 ceppi small round metal box with lid

328 2296 celii

t, tt
flea
ti celli

329 2333 cokk intoxication fatness

n
330 2335 colli name (songs)

331 2184 jitti

tuft of hair
jutti
t, t,
332 jiddi dirt (of body)
t,
333 2092 jibbi young tender areca nut
t,
334 DIA 192 takki talk words language

335 (2463) tatti flat land

n

336 2498 tappi a mistake

tt

337 1272 titti fire

tt ,t

338 1839 tekki

,t
south
N tekki
,t
339 2870 totti cradle
tt

340 totti stalk

tt
341 2200 toppi small leaves
,t

342 DIA 209 dikki direction

H

343 2705 duddi 4 pice; money

t, t,

344 nitti portion (land, house)

tt

345 3040 nippi shoulder

346 3112 nelli rice, paddy

347 3299 panni fruit

tt

348 3281 paddi eagle

349 3247 pappi share

n
350 3288 palli tooth

351 3528 pilli grass

n
352 DIA 269 putti wheat, rice preparation
 474

353 3506 purmi sore

354 3556 putti white anthill

355 3652 pokk navel

356 3603 potti a lie

357 3608a ponni wife, female

358 3732 ponni gold

359 3726 polli empty, light

u
360 4428 bitti seed for sowing
n
361 4449 billi bow (archery)

362 4520 bekki jungle cats

363 4518 benni the back

If

364 3676b botti kumkum

r
365 3817a mann mud, land property

366 3839 matti

n axe
macci

367 3863 maddi medicine

368 muggi mould, mildew; smell from this

369 4062 mutti pearl

370 mutti kiss

371 4098 mulli thorn

372 4150 motti footprint, measure, steps

373 saddi noise

374 sotti property

375 Payyu cow

376 83 atta loft

377 532 udda length, height

 475

378 544 ubba poles in slots forming a gate

379 782 okka patrilineal joint family

380 970 gedda

beard
. gadda

381 2135 canna a little

382 2435 takka sufficiency, suitability

383 DIA 233 pakka nearness

384 DIA 246 patta inauguration as king; throne, kingship,

crown

385 3295 pavva

pavala coral

paluva

386 3407 pitta giddiness; bile

387 4340 barena stoutness

388 batta paddy (in Lsanna batta])

389 4335 bayya behind, place behind

390 4421 butta ceiling joist

391 4503 betta big hill, mountain

392 4523 bella jaggery

393 3814 matta base of fronds of coconut tree

394 matta flatness, destruction (in [matta

ma:d])

395 4119a minna in front, further

396 4063 mutta kiss

397 lekka anything substantial enough to be

called into one's reckoning; account,
sum

398 24 akke(n) elder sister, elder female parallel

cousin

399 46 ajje(n) grandfather

 476

400 112 anne(n) elder brother, elder male parallel

cousin

401 133 appe(n) father

402 154 amme(n) Amma Coorg

403 (232) amme(n) mother

404 163 ayye(n) father's brother, male parallel cousin,

mother's sister's husband

405 232 avve(n) mother, mother's sister, female parallel

cousin
fl

406 834 obbe(n) one man

407 DIA 59 odde(n) man of snake charmer caste, man of

stone cutter caste

408 1156 kalle(n) thief

409 1335 knne(n) boy

410 (1526) kulle(n) dwarf

411 1773 kolle(n) blacksmith

412 DIA 120 golle(n) man of cultivator caste, milkman (low

caste)

413 calle(n) wastrel, idler

414 1 91 0 jadde(n)
slow -witted and slow-moving fellow
N dadde ( n )

415 2435 takke( n) chief man, village headman, man who

superintends property and ceremonies
of a god, leader

416 2466 tatte(n) goldsmith

417 DIA 248 patte(n) Brahmin (masc)

418 pukke(n) in Lpacce pukke(n)j, sp. grasshopper

419 pucce(n) madman

420 pedde(n) idiot

if

421 3871 malle(n) cock

422 4020 mucce(n) langur

 477

423 283 atta(1) cattle (pl)

424 401 ibba(1) two persons

425 834a obba(1) one person (fem)

426 2435 takka(1) chief family (pl)

427 3248 pajja(1) Holeya girl

428 3768 mekka(1) children (pl)

(C)V11CV(C)

429 3597 añci tile

430 andi evening

431 363 iñji ginger

432 927 kañji rice-gruel

433 989 kandi passing place

434 1186 kandi

it
mark of blow, bruise
A, kandi
tt
435 1289 kindi
..
small metal vessel with spout
kindi

436 1408 kunti lame person (fem)

437 kundi buttock

438 kondi button, sting (of insect)

439 2670a tindi

food
N tindi

440 2731 turnbi flying beetle

441 nañji
if
poison
N nanji

442 2513 nambi Malabar Brahman

443 3326 pandi pig

444 3792 mañji dew

 478

445 onte camel

446 984 kande root -stock

447 1733 konde tassels of sash; knob

448 gante bell, time, hour

449 962 gende

spleen
aA 7I[3

450 2194 cunde sp. Solanum

451 2322 joiige bunch, cluster (growing flowers, fruits)

452 jompe bunch (flowers, keys, etc.)

453 2445 tafige younger sister

454 DI. 187 tante trouble, teasing

455 2806 teAge coconut

456 2879 tonde throat

457 DID 238 pasce dhoti

458 3440 punde wetness

459 3831 mande head

460 DIA 299 munde widow

461 4098 munde screw -pine

n
462 150 ambi arrow
rr
463 iñgi asafoetida
rr

464 DIA 77 kambi

pillar, pole
kamba
rr
465 1408 kunti lameness

466 1389 kundi pit

467 1548 kundi mountain

if

468 1461 kumbi dry rot, rust

ri

469 1759 kombi branch, born of animal

if

470 gumpi crowd

 479
tt

471 2081 cindi scent (found by dog in hunting)

472 2275 cendi

11 tt
ball
N cendi
tt
473 2282 cembi
n copper, small metal pot
combi

474 2362 nandi scorpion

475 DIA 186 tandi big stick

476 ?2526 tandi offer of marriage (to girl)

477 2712 tundí piece

tt
478 2637 dindi stem of plantain

479 2713 dundi snout, face

480 3296 pandi antiquity

481 3508 pundi quarrel

482 3913 mandi village green

483 minci lightning

484 4052 mundi waistcloth

485 mundi state of being in front

486 4119a mumbi predominance

487 102 anda bank or edge of river

.0

488 69b and a side

489 988 kanda piece or lump of meat

490 cinda odd bit of paddy (grain with husk)

in cooked rice

491 jamba pride

492 3240 panda temporary structure built for wedding

or festival, pandal

493 3507 pueda bamboo

494 3786 manja turmeric

495 4237 rampa hubbub

 480

496 sonta waist

497 986 kande(n) male (of dogs and other animals, mostly
wild; not of cats)

498 1408 kunte(n) lame man

It

499 2135 cinde(n) small man (comical)

500 3508 punde(n) quarrelsome fellow

501 3223 bande(n) shameless man

502 3980 munde(n) paramour

503 2626 tiñga(1) month

504 3608 poiiga(1) women (pl)

(C)VS

505 1220 ka:y unripe fruit, kidneys

506 2027 ca:y beauty

507 308 ta:y grandmother

508 3022 na:y dog

509 4385 ba:y mouth

(C)VCV(C)

510 a :ni hard callous

511 295 a:ni nail (metal, wood)

512 a:di beginning, origin

513 327 a:li hail

514 DIA 41 i:ti

spear
V i:ti

515 763 e:ri parapet of well, wall in paddy field

 481

516 DIA 55 e:ni

ii ladder
IN e:ni

517 881 o:ni lane to house

518 884 o : di share

519 1206 ka:ti bison

520 1344 ki:ri mongoose

521 1586 ku :li wages

522 ku :li devil

523 1669 ke:ri hamlet

524 (1677) ke:li esteem

525 1862 ko :li fowl

526 1258 ga:li wind

527 1597 gu :li bull

528 DIA 157 ca:di slander

cu :di rope

530 1051 ce:ri coconut fibre, coir

531 ja:ti caste

532 DIA 179 jo:di pair

533 DIA 203 ta:di beard

534 ta:ti amulet

535 ta:li
locket
ta:li

536 2958 na:ti act of transplanting

537 ne:di common cold

538 3347 pa:di hut of a Kurumba

539 pa :ni measure (2 seers)

540 3469 pa:li peacock's tail feather

541 3631 po:di fear

 482

542 3746 po :ti niche over a door

543 3708 po :ri male buffalo

544 ba:ki remainder

545 3541 bu:di ashes

546 4556 be:li fence

547 3740 bo:ji

beauty, goodness
A. bo:ja

548 4546 bo:ti stake for threshing floor

549 3927 ma:ji soiled clothes

550 3930 ma:di upstairs, upper storey

N N
551 3945 ma:vi father's sister, etc.

552 3959 ma:li long hole (of rat, jackal)

553 3966 mi:di

offering to a god
mi:da

554 S 806 mu:di girl

555 3077 mu:li gnat

556 4179 me:ci fodder

557 me:ji table

11,
558 4178 me:di woman of [me:de(n)j caste

559 ra:gi ragi

560 ra:gi queen

561 341 a:ce week, day of week

562 a:ne promise

563 4235 a:ne elephant

564 4232 a:me tortoise

565 a:le sugar -mill

566 a:se wish, desire

567 879 o :te reed

 483

568 o :te mango stone (= /korandi /)

569 896 o :re

slanting
va:re

570 903 o:le

screw -pine leaf, writing on same,
va:le ear ornament

571 1197 ka:ke crow

572 1234 ka:re Randia dumetorum & uliginosa

573 1261 ka:le bull (leader of herd)

574 1564 ku: to basket

575 ke:re point in game

576 1831 ko :te palace

577 1869 ko:re tusk

578 go:de wall

579 2030 ca:ce relationship, kinship

580 ca:ne baldness

581 2139 ci:pe broom

582 2164 ci:le bag

583 2262 cu:le

dancing girl
su :le

584 DIA 149 ce:le belt, sash

585 2353 co:re blood

586 2347 jo:ke care

587 2920 do:se dosa

588 3025 na:le tomorrow

589 3126 ne:re directness, straightness

590 2373 ne:le rope for drying clothes

591 3393 pa:te cockroach

592 3392 pa:re stone slab in bath pit

593 pi:re sp. gourd

 4d4

594 pu:je worship

595 3572 pu:ñe cat

596 DIA 276 pe:te market

597 1027 po:ke frog

598 3758 po:le like

599 4402 ba:ke in [ ba:ke ku :d -j, (woman) marries

600 4402 ba:ce act of living

601 DIA 326 ba:de inner hall

602 3354 ba:ne open grazing land

603 4403 ba:le plantain

604 3639 be:le seed (coffee, groundnut, etc.)

605 4547 bo:te hunt

606 4564 bo:re different

607 S 886 bo:le neck

608 3954 ma:le necklace

609 3996 mi:se moustache

610 4140 mu:le corner

611 4145 mu:le bone

612 ra:te plank for pounding rice

613 sa:le school

n
614 290 a:di dance

615 4229 a:di goat

616 342 a:ni man, male

617 349 a:ti state of being full -grown but not yet
ripe

618 324 a:li banyan

ii

619 342 a:li servant, person

620 i:di singing ceremony before harvest festival

 485

621 u:ni gum (mouth)

622 643 u:ri village

n
623 878 o :di shell, skull

624 1206 ka:di jungle

n
625 1240 ka:ti wind

626 1238 ka:li leg

627 ka:si pie

628 1252 ka:li grain

629 1353 ki:ti torn piece

630 1346 ki:li fastening

631 1348 ki:li state of being lower, below

632 1562 ku:ti gathering

633 1551 ku:ti shout, noise

n
634 1592 ku:li cooked rice

635 1051 ke:ri rope

636 1348 ke:ki the east

if n
637 1614 ke:di rottenness
tl

638 1839 ko:pi festivity

639 1852 ko:li stick

640 1563 gu:di pig -pen, fowl -house, nest

641 1552 gu:mi owl

642 1866 go:li struggling, agony

643 ca:ki knife

644 2017 ca:ni span (measurement)

645 2038 ca:li (inherited) characteristic

646 1341 ci : pi comb

647 2158 ci:ri nit

648 2323 cu:ti thumb-forefinger span

649 (2183) cu:di heat

 486

650 DIA 171 cu :di needle

651 cu :ri piece

652 (2855) ce:li scorpion

653 DIA 180 co :di taste

654 1607 co :pi red

655 ?DIA 172 ju :ji bet, gambling

if

656 jo:ri speed, loudness

657 2598 ta:li bolt of door

658 2603 ta:1i stem

659 2183 tu:di torch

660 2674 te:ni honey

661 (2841) te:k

hiccough
te:ki
iT

662 2855 te:1i black insect with poisonous bite

663 to:ki gun

664 2921 to:di small stream, drain, lake

665 2937 to:li skin, hide

666 3 525 du:li dust

667 2793 du:ri information laid against a person

668 3012 na:di district

669 3023 na: ri fibre of plant, string

1,

670 3057 ni:ri water

671 ?3072 ne:ri bundle 9f neveral paddy seedlings (DED

has ni:ri)

672 3087 nu:li thread

673 3126 ne:ri

truth
ne:ri

674 3348 pa:ti song

675 3370 pa:li milk

 487

It
676 3371 pa:li part, share

677 3381 pa:li bad, ruined

678 3591 pu :ri female privates

679 3643 pe:ni louse

680 3747 po:ti male goat

It
681 3724 po:di
auspicious ceremony
N, polidi
tt

682 (3708) po:ri in [texige po:ri], coconut fight

683 po:ri load

684 3151 po :li daytime

1!

685 ba:ti duck

686 ba:ti fingerful of food

687 4394 ba:li tail

it
688 4402 ba:li way of life

689 4405 ba:li in ] long

6s g sword

690 4419 bu :di toddy- tapper's hut

n
691 4554 be:ri root
fl
692 4298 be:li paddy field
t'
693 be:vi neem tree
t'
694 3931 ma:di burned clearing in jungle
It

695 3920 ma:ti afterbirth

696 3947 ma:r1 fathom

697 3999 mi:ni fish

ti
698 4122 mu:ki nose

699 4129 mu:di face

700 4185 me:li body

701 4015 mo ri
:
buttermilk
tt
702 sa:li row
 488

703 se:di tired, pulling feeling in legs

704 so:ki evil influence

705 2002 ca:vu corpse

706 3009 na:vu tongue

707 pa:vu ç seer measure

708 ba:vu full moon day

709 285 a:ka biggest flat land of man's holdings,

nursery field for paddy

710 a:na arena (sixteenth of a rupee)

711 338 a:la depth

712 e:ka field (any type: e.g. sports field)

713 776 e:ra

more than enough, much
e: ra

714 DIA 62 o :na Malabar version of harvest festival

715 1219 ka:ca heat

716 1207 ka:ta trouble, rubbish

717 DIA 97 ka:la year, season, time

II

718 1348 ki:da place below, down

719 1562 ku:ta quarrel, dispute, gathering

720 1590 ku:va shallow well where water can be dipped

by hand

721 ko:ta coldness

722 DIA 90 ga:ya wound

723 DIA 104 gu:ta

peg, post
.r gu:ta

724 ge : na thought

725 DIA 159 ca:na dung

 489

726 2164 ci:la

bag
ci:le

727 ja:ga place

728 ji:ra
cummin seed
ji: rige .1

729 ji:va life

730 2359 jo:la great millet

731 3014 na:na shame

732 DIA 182? na:ya dispute, justice, that which is right

733 DIA 208 ta:la cymbal

734 ta:la bolt on door

735 2680 ti:ta excrement

736 tu:ka weight

737 tu:ta
hole
A. tu: ti

738 2251 tu:ra cane

739 2927 to:ta garden, estate

740 du:ra distance

741 de: sa country

742 na:ÿa coin

2379 na:ta smell

743

744 3059 ni:la length

3128 ne:ra sun, time

745

746 DIA 242? pa:ta lesson

747 pa:pa sin

748 pa:pa small child

3372 pa:la bridge made of tree, etc.

749

750 4410 ba:na

sky
r ba:na
 490

751 ba:na arrow

752 ba:ra heaviness

753 3456 bi:ga lock

754 3740 bo :ja beauty, goodness

755 3946 ma:yaa disappearance

756 mu:la origin

757 4131 mo:da cloud

758 DIA 304 mo :sa deceit

759 ro:ma hair (of body)

760 lo:ka world

761 4389 va:ra rent, system of tenant farming

762 va:ra week

763 sa:pa curse

764 su :la trident

765 2039 sa:la loan, debt

766 sa:sa
effort, hard work
sa:hasa

767 si:ta
coldness
^. si:ta J

768 1660 ke:me(n) barking deer

769 1820 ko :dd(n) monkey

770 cu:de(n) temperamental, peppery fellow

771 2580 to :te(n) grandfather

772 2682 ti:ye(n) Tiyan man

773 ba:ve(n) elder male cross- cousin, brother -in -law

774 3945 ma:ve(n) mother's brother, etc.

775 4178 me:de(n) man of drummer and umbrella- makers'

caste
 491

776 3768 mo :ve(n) son

777 ra :je(n) king

778 DIA 219 de:va god

779 3768 mo :va daughter

(C)VIdCV(C)

780 106 a:ndi mango stone

781 887 o:ndi

chameleon
o :tike:te(n)

782 2368 ne:Agi

plough
.. ne:ñgi

783 3944 ma:ndi act of scratching

784 3919 ma:Age mango

785 4230 a:ndi year

786 1217 keimbi Oxytenanthera monostigma (prob.)

(bamboo)

787 2021 ca:ndi sandalwood

788 3361 pa:mbi snake

(C)VSCV(C)

789 ayri low caste Coorg, carpenter

790 DIA 80 kavdi shell cowrie

791 DIA 82 gavdi

} serving girl, handmaid, milkmaid
nr gavli

792 1125 gavli big lizard

 492

793 1047 kaype gall- bladder

794 1984 cavte cucumber

795 DIA 203 davde cheek

796 (4339) bayne egg -plant

797 831 uyti force

798 1083 kaybi sugar -cane

799 1469 kuyli bee's sting (iiercara)

800 1973 cavti footprint

801 2537 tavdi inner skin of rice grain

802 pavni gold coin of Z1 value

803 3704 PoYtil blow

804 4570 bay ti evening

11

805 bayti age

806 3793 mayli peacock

807 DIA 155 cavka long white headdress, square piece of

cloth

808 paysa sweet rice (dish)

809 DIA 305 moyra ring

810 gavde(n) Gowda man

II

811 tayre(n) cock

812 DIAS 18 mayme(n) cross -nephew

813 rayte(n) cultivator

814 DIAS 18 mayma(1) cross -niece

 493

(C)VSNCV(C)

815 1034 kaviiki vital spot

816 S 857 baymbe(n) hero

(C)VCVCV(C)

817 450 eraci flesh, meat

818 827 onali sieve

Il
819 943 kadici calf

820 1059a karadi bear

821 1184 kenaci dream

822 1702 kodali axe

823 koraji marshy place

824 ?2529 talami hair of head or body

825 2761 tolasi sacred basil

826 3113 nelaci moon

827 3280 parati cotton cloth

828 3290 palaci jack -fruit

tr

829 (3623) poriki mean fellow

830 3714 polati low caste woman

831 3260 barani large clay pot

832 4057 mudiki old woman

833 78 adake areca nut

834 (66) adige cooking

835 252 alate act of measuring

836 614 uripe small bag for betel

837 680 edike in front

838 400 erape beggar

 494

839 381 edate left

840 799 odeve break

841 580 olake wooden pestle

842 1277 kadape bed, bedding

843 (929) kadape steps for getting over fence

844 941 kadale Bengal gram, peanuts

845 976 kánúcé the plains to the east of Coorg

846 (976) kanive valley

847 DIA 99 kudike window

848 1376 kudike cooking pot

it

849 1423 kudire horse

850 1537 korate diminishment

851 1176 garike Agrostis linearis Retz. (sp. grass)

852 2132 cirate leopard

853 2101 cerate coconut shell

854 2819 terake wing

fl

855 tirike temple

856 2957 nadate walking, behaviour

857 DIA 232 pagade chess

858 3257 parake vow

859 ?3297 palame talk, gossip

860 3511 pudume wonder

861 (3686) podike cover

862 3554 porame outside

863 3531 polace dawn

864 4317 balate right (hand)

865 4056 mudire horse -gram

11

866 mudire seal

 495

867 4055 mosale

crocodile
mosale j

868 265 arivi knowledge

n n
869 2102 iriti darkness

870 2102 irili night time

871 502 udipi clothes

872 734 urupi

11 It
ant
uripi

873 554 usiri

breath
ti usiri

874 694 erivi burning sensation in mouth

875 449 eraki

eaves
eraki

876 725 eliti writing education

877 606 oraki sleep

878 954 kadiki man's earring

879 1175 karaki carbon

880 1175 karapi blackness

881 1523 kuliri cold (climate)

882 1374 kodavi Coorg

883 (1382) gudigi thunder

I,
884 2236 curiki speed
n n
885 2473 tanipi
coolness
tampi

886 (2363) navili peacock

887 (3686) podapi

. podepi
covering, esp. blanket
 496

tt

888 3451 puriki mosquito

889 3724 polidi cutting of paddy at festival, etc.

890 4267 badaki north

n If

891 4402 badiki property

tt

892 4516 bevari sweat

893 4524 bolaki lamp

894 4524 bolipi whiteness

895 3796 madaki bend, fold

tt

896 manasi conscience

tt tt

897 4112 muriki torsion cord

898 (4153) medili

brain (food)
medli

899 4074 moradi foot of tree

900 1537 koravu defect

901 agala breadth

902 56 adaka act of bringing within a compass,

restrictions (at festivals),
thriftiness

903 432 elaka uprooting from position, violent shaking

(as when possessed by a god)

904 931 kadaga thick metal bangle

905 952 kadipa quickness

906 1166 karapa milch cow

907 S 167 karapa clay pot with narrow neck

908 1639 kelasa work

909 gnmana odour, fragrance

910 DIA 124 jagala quarrel

911 DIA 199 taliya plate

 497

912 2869 todiya small garden adjoining house

913 divasa day

914 2957 nadeya passage to shrine of temple

915 DIA 230 nevana pretence, false reason

916 padika slowness (of movement, wit)

917 DIA 257 paleya

wooden stool
.N. pal iya

918 3295 paluva

coral
N. pavva
pavala

919 3613 perija increase, excess

920 3497 podeya

sari
podiya

921 3531 polaca dawn

n
922 4524 bolica lamp

923 DIA 285 modira sweetness

924 sarira trunk (of body)

925 DIA 133 sadla

looseness
sadala

926 samaya time

927 (200) alape(n) greedy man

928 436 eleye(n) youth

929 510 odeye(n) husband

930 510 odeve(n) master, god

931 935 kadame(n) sambur

932 1280 kiduve(n) kite, hawk

933 1379 kudiye(n) man of toddy tapper caste

934 1537 kurike(n) jackal, fox

 498
11

935 1530 kurube(n) shepherd

936 994 keniye(n) trickster

11

937 1374 kodave(n) Coorg man

938 garude(n) eagle

It tt tt

939 1280 gidige(n) kite, hawk

940 3714 poleye(n) low caste man

941 DIk 290 manise(n)

tf man
mañs e( n )
ft It
942 4057 mudike(n) old man

943 4270 badave(n) poor man

944 3838 madema female cross -cousin, etc.

(C)VCv CV

945 842 olañji fly

946 938 kadandi wasp

947 1722 korandi mango stone

948 mananji
tacky secretion of jack-fruit
maninji

949 1700 kodande double handful

950 parande gizzard of fowl

tt t,

951 411 irimbi iron

tt
952 509 udumbi iguana
R tt

953 714 elimbi bone

tt

954 1279 kadañgi ditch without water

11

955 955 kadtimbi small ball of rice

 4yß

It

956 3316 parambi large flat pasture land

It

?3255 parambi mat

957
It

3897 marandi forgetfulness

958
 500

BIBI,IO GPHY
 501

Abbreviations

AUPL Annamalai University Publications in Linguistics

BSOS Bulletin of the School of Oriental Studies (University

of London)

FL Foundations of Language

IA Indian Antiquary

IIJ Indo- Iranian Journal

IL Indian Linguistics

JAOS Journal of the American Oriental Society

JAS Journal of the Asiatic Society (Calcutta)

JL Journal of Linguistics

JRASB Journal of the Royal Asiatic Society of Bengal

Lg Language (Journal of the Linguistic Association of America)

TES Transactions of the Philological Society

UCPL University of California Publications in Linguistics

DBIA Dravidian Borrowings from Indo-Aryan. By h.B.Emeneau and

T.Burrow. UCPL vol. 26. Berkeley: University of

California Press. 1962.

DED A Dravidian Etymological Dictionary. By T.Burrow and

Ii.B.Emeneau. Oxford: Clarendon Press. 1961.

DEDS A Dravidian Etymological Dictionary: Supplement. By

T.Burrow and K.B.Emeneau. Oxford: Clarendon Press. 1968.

DEN Dravidian etymological notes: parts I and II. By T.Burrow

and N.B.Emeneau. JAOS 92. 1972.

 502

Dravlingpex Dravidian Linguistics Papers Exchange. Edited by

H.Schiffman, Department of Asian Languages and

Literature, University of Washington, Seattle.

LSI Linguistic Survey of India, I, Parts I and II. Edited

by Sir G.A.Grierson. Calcutta: Government of India

Central Publication Branch. 1927.

Agesthialingom, S. (1967). A Generative Grammar of Tamil (A Fragment

of Tamil Syntax). AUPL vol. 11. Annamalainagar: Annamalai

University.

Andrewskutty, A.P. (1971). Malayalam: An Intensive Course.

Trivandrum: The Dravidian Linguistic Association.

Anon (1857). Coorg and its Rajahs. London: John Bumpus.

Anon (1882). A Selection of Bible Sentences in Kûrg (Kodagu).

Mangalore: Basel Mission Press. (Possibly by J.F.Veil).

Appacha, A. (1906). Yayati RajandaI'ataka. Mangalore: Dharma

Prakasha Press.

(1908). Subrahmanya-Svami-Nataka. Mangalore: Dharma

Prakasha Press.

(1929). Yayati- íRajanda-Nataka. Mangalore: Sharada

Press.

Appiah, K.M. (1887). Sri-Kodagu-Modalaneya-Pustakavu. Bangalore:

Mysore Government Press and Mysore Book Depot Press.

 503

Asher, R.L. (1966). The verb in spoken Tamil. In Bazell, C.L.,

Catford, J.C., Halliday, M.A.K. & Robins, R.H. (eds.)

In Memory of J.R.Firth. London: Longmans.

(1968). Existential, possessive, locative and copulative

sentences in Malayalam. In Verhaar, J.W.M. (ed.) The

Verb 'Be' and its Synonyms 2 (FL Supplementary Series

vol. 6). Dordrecht: Reidel.

Bach, E. (1968). Nouns and noun phrases. In Bach, E. & Harms, R.T.

(eds.) Universals in Linguistic Theory. New York:

Holt, Rinehart & Winston.

Bally, C. (1932). Linguistique Générale et Linguistique Française.

Paris: Leroux.

Bazell, C.A. (1953). Linguistic Form. Istanbul: Istanbul University.

Bhat, D.N.S. (1967). Descriptive Analysis of Tulu. Deccan College

Centenary and Silver Jubilee Series vol. 15. Poona:

Deccan College Postgraduate and Research Institute.

Bierwisch, N. (1970). Semantics. In Lyons, J. (ed.) New Horizons

in Linguistics. Harmondsworth: Penguin.

Bloch, B. (1950). Studies in colloquial Japanese IV: Phonemics.

LE 26, 86 -125.

Bloch, J. (1919). The intervocalic consonants in Tamil. Imo, 191 -95.

Bright, W. (1958) . An Outline of Colloquial Tamil. Deccan College

Monograph Series vol. 22. Poona: Deccan College Post-

graduate and Research Institute.

 504

Brown, G. (1970). Syllables and redundancy rules in generative

phonology. JL 6, 1 -17.

Burnell, A.C. (1873). Specimens of S. Indian Dialects. No. 3. In

the Kodagu (Coorg) Language. Nangalore: Basel Fission

Press.

Burnett, L.D. (1910) . A Catalogue of the Kannada, Badaga and Kurg

Books in the Library of the British Museum. London:

The British Museum.

Burrow, T. (1938). Dravidian studies I. BSOS 9, 711 -22.

and Emeneau, M.B. (1972). Dravidian etymological notes:

parts I and II. JAOS 92, 397 -418 and 475 -91.

Caldwell, R. (Rev.) (1856). A Comparative Grammar of the Dravidian

or South -Indian Family of Languages. London: Routledge

and Kegan Paul.

(1875). A Comparative Grammar of the Dravidian

or South -Indian Family of Languages. London: Routledge

and Kegan Paul (2nd. edition).

Cheng, R.L. (1966). Mandarin phonological structure. JL 2, 135 -59.

Chinnappa, N. (1924). Pattole Palume. Bangalore.

(1929). Bhagavantanda -Pati (Bhagavadgita). Bangalore:

K.B.D. Electric Printing Works.

Chomsky, N. (1964) . Current issues in linguistic theory. In Fodor,

J.A. & Katz, J.J. (eds.) The Structure of Language:

Readings in the Philosophy of Language. Englewood Cliffs,

N.J.: Prentice -$all.

 505

Chomsky, Y. (1 S'65) . Aspects of the Theory of Syntax. Cambridge,

Lass.: ì..I.T. Press.

(1970). Remarks on nominalisation. In Jacobs, R.A. &

Rosenbaum, P.S. (eds.) Readings in English Transformational

Grammar. Waltham, Lass.: Blaisdell.

(1971). Deep structure, surface structure and semantic

interpretation. In Jacobovits, L. & Steinberg, D. (eds.)

Semantics: An Interdisciplinary Reader. Cambridge, and

Urbana, Illinois: Cambridge University Press and

University of Illinois Press.

Cole, R.A. (Capt.) (1867). An Elementary Grammar of the Coorg

Language. Bangalore: Wesleyan bission Press.

(1871). A hanual of Coorg Civil Law. Bangalore:

Iysore Government Press.

Connor, P.E. (Lt.) (1870). Lenoir of the Codugu Survey: Parts I and

II. Bangalore: Central Jail Press.

Doraiswamy, K. (forthcoming). A Descriptive and Historical Study of

Kodagu.

Emeneau, L.B. (1938). Kinship and marriage among the Coorgs. JRASB

4, 123 -47.

(1945). Theravidian verbs 'come' and 'give'. La. 21,

184 -213.

(1955). India and linguistics. JAOS 75, 145 -53.

(1967). The South Dravidian languages. JAOS 87,

365 -413.

(1970a). Kodagu vowels. JAOS 90, 145 -58.

 506

Emeneau, M.B. (1970b). Dravidian Comparative Phonology: A Sketch.

AUPL vol. 22. Annarnalainagar: Annamalai University.

(1971). Kodagu and Brahui developments of Proto-

!L

Dravidian r . IIJ 13, 176 -98.

Ellis, F.W. (1816) . Note to the Introduction. In Campbell, A.D.

A Grammar of the Teloogoo Language. Madras: College

Press.

Fairbanks, G.H. (1957) . Frequency and phonemics. IL 17, 105-113.

Firth, J.R. (1934). A short outline of Tamil pronunciation. In

Arden, A.H. A Progressive Grammar of Common Tamil (4th.

edition, revised by A.C.Clayton). Madras: Christian

Literature Society for India.

Fowler, M. (1954). The segmental phonemes of Sanskritised Tamil.

Lg 30, 360-7.

Fries, C.C. and Pike, K.L. (1949). Co- existent phonemic systems.

LE 25, 29-50.

Fudge, E.C. (1967). The nature of phonological primes. JL 3, 1 -36.

(1969a). Iiutation rules and ordering in phonology. JL

5, 23 -38.

(1969b). Syllables. JL 5, 253-86.

Garman, I4. (1969) . Coorg verbal base structure. Dravlingpex 1,

No. 7.

(forthcoming). Coorg verbal bases. In Eastman, C.N. &

Schiffman, H. (eds.) Dravidian Phonological Systems.

 507

Garman, M. (to appear). An approach to Dravidian derivational

morphology. Paper read at The Second All -India

Conference of Dravidian Linguists, 26 -7 June 1972, S.V.

University, Tirupathi.

Goda Varma, K. (1946). Studies in Coorg vowels. In A Volume of

Indological Studies (Dr. C. Kunhan Raja Presentation

Volume). Madras: The Adyar Library.

Dover, C.H. (1871). The Folk -Songs of Southern India. London:

11

Trubner & Co.

Graeter, A. (Rev.) (1869). A Short Description of Coorg. Mangalore:

Basel Mission Press.

(1870). Coorg Songs: With Outlines of Coorg

Grammar Mangalore: Basel Mission Press.

Halle, X. (1958) . Questions of linguistics. In supplement to

Il Nuovo Cimento 13, Series 10, 494-517.

Hockett, C.F. (1958). A Course in Modern Linguistics. New York:

Macmillan.

Hodgson, B.H. (1849). Aborigines of Southern India. (From JAS 28).

Calcutta: Baptist Mission Press.

(1880). Miscellaneous Essays relating to Indian

It
Subjects. London: Trubner & Co.

Jacobs, R.A. and Rosenbaum, P.S. (1968). English Transformational

Grammar. Waltham, Mass.: Blaisdell.

Kandiah, T. (1967). The Syntax of the Verb in Ceylon Tamil. Ph.D.

 508

Thesis, School of Oriental and African Studies, University

of London.

Kandiah, T. (1968). Transformational grammar and the layering of

structure in Tamil. JL 4, 217 -45.

Katz, J.J. and Postal, P. (1964). An Integrated Theory of Linguistic

Descriptions. (Research konograph Series vol. 26).

Cambridge, Kass.: E.I.T. Press.

Kothandaraman, P. (1969). Kodagu vowels. In Agesthialingom, S. &

Kumaraswami Raja, N. (eds.) Dravidian Linguistics

(Seminar Papers). AUPL vol. 17. Annamalainagar:

Annamalai University.

K rishnamurthi, B. (1961). Telugu Verbal Bases: A Comparative and

Descriptive Study. UCPL vol. 24. Berkeley and Los

Angeles: University of California Press.

Kumaraswami Raja, N. (1969a). Post -nasal voiceless plosives in

Telegu. In Agesthialingom, S. & Kumaraswami Raja N.

(eds.) Dravidian Linguistics (Seminar Papers). AUPL

vol. 17. Annamalainagar: Annamalai University.

(1969b). Post-nasal Voiceless Plosives in

Dravidian. AUPL vol. 18. Annamalainagar: Annamalai

University.

Kushalappa Gowda, K. (1970). Gowda Kannada. AUPL vol. 20. Anna-

malainagar: Annamalai University.

Ladefoged, P. (1971). Preliminaries to Linguistic Phonetics.

Chicago: University of Chicago Press.

 509

Lakoff, G. (1971). On generative semantics. In Jacobovits, L. &

Steinberg D. (eds.) Semantics: An Interdisciplinary

Reader. Cambridge, and Urbana, Illinois: Cambridge

University Press and University of Illinois Press.

Lindsay, A.W.C. (Major) (1874). Report on the Coorg General Census

of 1871. Bangalore: Mysore Government Press.

Lisker, L. (1951). Tamil verb classification. JAOS 71, 111 -14.

Lyons, J. (1966). Towards a 'notional' theory of the 'parts of

speech'. JL 2, 209 -36.

(1967). A Note on possessive, existential and locative

sentences. FL 3, 390 -6.

(1968). Introduction to Theoretical Linguistics. London

and New York: Cambridge University Press.

Magrath, H.M.S. (Major) (1881). Report on the Coorg General Census

of 1881: With Appendices. Bangalore: Mysore Government

Press.

Matthews, P.H. (1965a) . The inflectional component of a Word -and-

Paradigm description. JL 1, 139 -71.

(1965b). Some concepts in Word- and -Paradigm morphology.

FL 1, 268 -89.

(1967). The main features of Modern Greek verb

inflection. FL 3, 262 -84.

McCawley, J.D. (1968). The role of semantics in a grammar. In Bach,

E. & Harms, R.T. (eds.) Universals in Linguistic Theory.

New York: Holt, Rinehart & Winston.

 510

hoegling, H. (Rev.) (1855) . Coorg hemoirs. Bangalore: ',Jesleyan

Mission Press.

(1858) . Rajendraname. Virarajendra Vodeya.

(with English translation by Lieut. R.Abercromby -

Mangalore 1808). Mangalore: German Mission Press.

Neisser, U. (1967) . Cognitive Psychology. New York: Appleton-

Century-Crofts.

Official (1906). Coorg District Gazetteer. B Volume. Mercara:

Coorg District Press.

(1918). Coorg District Gazetteer. B Volume. Mercara:

Coorg District Press.

(1926). Coorg District Gazetteer. B Volume. Madras:

Government Press.

(1934). Coorg District Gazetteer. B Volume. Bangalore:

Mysore Government Press.

Palmer, F.R. (1971). Grammar. Harmondsworth: Penguin.

Partee, B. Hall (1971). On the requirement that transformations

preserve meaning. In Fillmore, C.J. & Langendoen, D.T.

(eds.) Studies in Linguistic Semantics. New York:

Holt, Rinehart & Winston.

Poovayya, M.S. (1966). Sri- Guru-Bodhe. Virajpet: Shri Cauvery Ashram.

(1968) . Sri -Kaveri- CharanaKusuma. Virajpet: Shri

Cauvery Ashram.

Quine, W.V. (1 96 4) . Leaning and translation. In Fodor, J.A. &

Katz, J.J. (eds.) The Structure of Language: Readings in

 511

the Philosophy of Language. Englewood Cliffs, N.J.:

PrenticeHall.

Ramanujan, A.K. (1963). A Generative Grammar of Kannada. Ph.D.

Thesis, University of Indiana.

Rice, B.L. (1878). Mysore and Coorg: A Gazetteer Compiled for the

Government of India. III Coorg. Bangalore: Mysore

Government Press.

Richter, G. (Rev.) (1870). Manual of Coorg: A Gazetteer. Mangalore:

Basel Mission Book Depository.

Robins, R.H. (1959). In defence of WP. TPS, 116 -44.

Shanmugam, S.V. (1971). Dravidian üouns. AUPL vol. 25. Annamalai-

nagar: Annamalai University.

(1973). Piodernisation in Tamil. Dravlingpex 5,

No. 3.

Srinivas, h.1. (1952). Religion and Society among the Coorgs of

South India. London: Asia Publishing House.

Stockwell, R.P., Schachter, P. and Partee, B. Hall (1968). Integration

of Transformational Theories on English Syntax. Los

Angeles: University of California.

Subrahmanyam, P.S. (1971). Dravidian Verb Morphology. AUPL vol. 24.

Annamalainagar: Annamalai University.

Tay, N.W.J. (1970). Hokkien phonological structure. JZ 6, 81-8.

Upadhyaya, U.P. (1968). Nanjangud Kannada (Vakkaliga Dialect).

 512

(Deccan College Linguistic Survey of India Series

vol. 2). Poona: Deccan College Postgraduate and Research

Institute.

Veil, J.F. (Rev.) (1886) . b. Selection of nrg- Proverbs. hangalore:

Easel Iiission Press.