242

;:

..J~ ci. 5~ ~.9- 11~-t:s._

(.\);_r'ia...Olfc,-t:,.P-vJ
V
o 6 ")

[}2,;, IR.. (CJ 1r:i)

Externa/ factors

7 Input and interaction and second

language acquisition

The following are sorne of the articles mentioned in this chapter dealing
with the relationship berween different types of social contexts and L2
acquisition:
T. Skuttnab-Kangas, 'Multilingualism and the education of minority children' in T. Skuttnab-Kangas and ]. Cummins (eds.), Minority Education
(Multilingual Matters, 1988).
] . Cummins, 'Second language acquisition wirhin bilingual educarion programs' in L. Beebe (ed.), Issues in Second Language Acquisition: Multiple
Perspectives (Newbury House, 1978).
A. d'Anglejan, 'Language learning in and out of classrooms' in Richards, J.
(ed.), Understanding Second and Foreign Language Learning: Issues and
Approaches (Newbtiry House, 1978).
A. Davies, 'Is inrernational English an interlanguage?' TESOL Quarterly
(1989) 23: 447-67.
These are the key books and arrides dealing with the three social models of
L2 acquisirion discussed in rhis chapter:
J. Schumann, 'The acculrurarion model far second Ianguage acquisition' in R.
Gingras (ed.),Second Language Acquisition and Foreign Language Teaching
(Center for Applied Linguistics, 1978).
H. Giles and J. Byrne, 'An inrergroup approach to second language acquisition.' ]ournal of Multilingual and Multicultura/ Developrnent (1982) 3:
17-40.
R. Gardner, Social Psychology and Second Language Learning: The Role of
Attitudes and Motivation (Edward Arnold, 1985).
Finally, for those inreresred in a historicaI-strucrural accounr of rhe relationship berween social factors and L2 acquisition, the following 1s
recommended:
J. Tollefson, Planning Language, Planning Inequality (Longman, 1991).

Introduction

)
j

147

Although ali rheories of L2 acquisition acknowledge the need for input, rhey
differ greatly in the imporrancc rhat s attached ro it; the role of input in Janguage acquisition is a controversia! quesrion. In my previous review of SLA
research (Ellis 1985a: 127ff), l distinguished rhree different views about its
role: the behaviourist, the mentalist, and rhe interactionist.
Behaviourist accounts of L2 acquisition propase a direct relationship between input and output. Beca use they rejecr the idea of 'mind' asan object for
inquiry, they ignore the interna! processing that takes place inside the learner.
Input is comprised of stimuli and feedback. With stirnuli, the person speaking
ro the learner models specific tinguistc forms and patterns which rhe learner
internalizes by imitating them. Feedback rakes the form of positive reinforcement orcorrection, depending on whether the learner's output is perceived to
be targer-like. Behaviourist models of learning emphasize rhe possibility of
shaping L2 acquisition by manipu!ating rhe input to provide appropriate
stimuli and by ensuring rhat adequare feedback is always available. Acquisition is thus conrrolled by externa! factors, and the !earner is viewed as a
passive medium.
Mentalist theories emphasize the mportance of the learner's 'black box'.
Although input is stiil seen as essenrial for L2 acqusition, ir is seen as only a
'rrigger' that sets off interna! language processing see Cook 1989). Learners
are equipped \Vith innate knowledge of rhe possible forms that any single language can take, and use rhe information supplied by the input to arrive ar the
forms that apply in rhe case of the L2 they are rrying to learn. As we wiU see
later, a common assertion of rnentalisr rheories is thar the input is 'indeterminate' i.e. the information that it supplies is, by itself, insufficient to enable
learners ro arrive ar the rules of the target Ianguage.
The third type of theory is the interactionisr one. This Iabel has been applied to t\VO rather different types of theory. According to cognitive inreractionisr theories, acquisition is seen as a product of the comp!ex interacton of
the Iinguistic environment and the learner's interna! mechanisms~ wirh
neither viewed as primary. Cognitive interacrionist mode!s of L2 acquisition have been drawn from contemporary cognitive psycho!ogy and, not

......

~~"'

, ...... 1 .............,,.,

surprisingly perhaps, vary considerably. Despite the dfferences, however, a

common assumption can be seen to underlie them, namely that input does
have a determining function in language acquisition, but on1y within the
constraints imposed by the learner's interna! mechanisms.
The second type of interactionst theory is more social in orientation. The
principie that informs these theories is that verbal interaction is of crucial im-

portance for language learning as it helps to inake the 'facts' of the L2 salient
to the learner. We will be concemed pr.imarily with social interactionist theories in this chapter, delaying consideration of rhe cognitive kind until Chapter
9. We should note, however, that man y cognitve inreractionist theories also
see social interaction as the primary mechanism of mental reorganization.
This chapter begins with a brief discussion of the main methods used to
investigate input and interaction. Subsequently ir focuses on two major
questions:

1 What are the characteristics of the input that L2 learners typically receive?
2 How does the input influence L2 acquisition?

Question (1) leads us to consider research that has sought to describe L2

input, while (2) leads us to examine research that has sought to explain how
input affects acquisition. Because much of the work on input in second
language acquisition research has followed in the footsteps of work in first
language acquisition research, attention will be given to the latter whenever
this seerr1s appropriate. As there is a separate chapter dealing with input and
interaction in classroom settings (Chapter 13), this chapter will concentrare
on input and interacrion in non-instructionaI settings, although sorne reference to classroom research will be made when this is of particular relevance to
the issue under discussion.

Methods for investigating input and interaction

Many of the earlier input studies (for example, Hatch and Wagner-Gough
1976; Hatch 1978b and 1978c; Peck 1978) used data that had been collected
to study iearner language. Because these data consisted of transcriptions of
the interactions in which the learners took part, it was a relatively easy task to
transfer attention from what the learners said to what was said to them. The
data were submitted to both detailed linguistic analyses (for example, to discover the frequency with which specific morphemes occurred in the input or

Later, experimental and pseudo-experimental studies were designed to investigate the effect of specific variables on input and interaction. Srudies car-

ried out by Long (1980a), Gass and Varonis (1985a), and Pica and Doughty
(1985a), among others, have made use of input/interaction data elicited for
the purpose of the enquiry. Long, for instance, in one of the most frequently
cited input srudies, asked sixteen pairs of subjecrs (consisting of American
native speakers of English, and Japanese non-native speakers) to perform six
different tasks, three of them involving one-way information exchange (giving instructons, vicarious narrative, and discussing the supposed purpose of
the research) and three involving rwo-way information exchange (conversation and playing two communication games). The advantage of such studies
is that they enable the researcher to manipulare individual variables deemed
likely to influence the quantity or quality ofthe input provided orto investigare the effects of the learners' stage of developrnent on the input provided.

The disadvantage is that it is difficult to determine to what extent the data collected are representative of the kind of communication the learners typically
take part in. In many cases, as in Long's study, the interlocutors had no prior
knowledge of each other - a factor which may influence the nature of the in-

teractions which take place (see the discussion of addressee effects in Chapter
4).
One major advantage of experimental studies is that they make possible the
collection of baseline data, which provide the researcher with sorne kind of
normative point of reference. Usually these consist of conversations between
native speakers performing the same tasks. Such data enable the researcher to
identifywhat is special about the input addressed to the learners. However, as

Long (1980a) points out, many studies have failed to collect baseline data.
Not surprisingly, perhaps, introspective techniques have not been widely
used in input research, as neither learners nor native speakers are likely to be

a ble to comment accurately or reliably on detailed features of the input. However, Ferguson (1975) collected data on foreigner-talk by asking srudents at
Stanford University to rewrite standard English sentences in the way they
thought they would say them to illiterate non-Europeans with no English.
This method suffers from the same drawbacks we noted with discourse completion questionnaires in Chapter 5. We cannot be sure that what people
think theywould say is what they would actually say.
Anothersource of introspective informacion is diary studies. Brown (1985)

used diaries to investigare the kinds of requests for input expressed by L2

development) and discourse and conve.rsational analysis (for example, to

leamers. She analysed the comment~ the diarists made in ~any reference to
input desired, to amount of input given, to type, complexiry or meaning-

describe the ways in which topics were nominated and developed or the
strategies used to deal with communication breakdown). Discourse and

fulness of input' (page 275). Schmidt and Frota ( 1986) also made use of a
diary srudy as a meaos of investigating what one learner notced in the input.

conve.rsational

This study shows that introspection is an important tool for the input
researcher, as it provides one of the best ways of discovering what it is in the

the complexty of input addressed to learners at different srages of

ana!ysis rapidly became the favoured method of analysis, as it provided

researchers with the tools they needed to investigate the nature of the !earning
opportunities made available to learners through interaction.

148

246

Externa/ factors

Input and interaction and second language acquisition

input rhat learners attend to. It is probably not 'raw' input bue 'heeded' input
that works for language acquisirion (see Chapter 9).

detailed nformation on \Vhat constitutes target-like use. Her advanced Ievel

subjects displayed an abiliry to produce 'fully formed quesrions' but often did
not do so-certainly to a lesser extent than native speakers of American
English. 2
Lightbown and d'Anglejan (1985) found discrepancies between the declared norms for three French strucrures (interrogatives, negatives~ and word
order) and the use of these structures in everyday spoken French. They found
that uninverted interrogative forms and quesrions with 'est-ce-que' (which
also adhere to declararive word order) are predominant in native-speaker
input, contrary to the claims of the formal descriptions of linguists and
grammarians and also to native-speaker intuitions. In the case of negation,
'ne' is often deleted in rapid informal speech {for example, 'Elle vient pas
avec vous.'), making French a post-negation rather than double-negation
language. They also note that a lthough French is considered to be a subjectverb-object (SVO) language typologically, spoken French provides copious
examples of alternative orders (VOS, VS and even OSV). This study futther
demor;,;trates the importance of ensuring that leamer language is examined
rogether with accurate information about the input language.
An additional problem facing the researcher is determining which kind of
input to consider when evaluating the leamer's interlanguage. We have already seen in Chapter 6 that leamers vary in their choice of reference group.
Valdman (1992) talks of the 'illusive ideal native speaker', pointing out that
even in cases where there is a single norm (for example, Metropolitan Standard French), 'the norm allows considerable leeway' (page 84 ). Valdman goes
on to argue that 'an invariant TL norm, based on the planned discourse of
educated and cultivared speakers, is an illusory target for learners' (page 94 ).
Valdman is more concemed with the problem this poses in choosing appropriate pedagogcal norms, but it is also equally problematic for the researcher
concerned '\vith examining the role of input in interlanguage development.
Even assuming that accurare in.formation regarding targetlanguage norms
is available, the problem is not overcorne, as in many cases native speakers do
not adhere to these norms when cornmunicating with non"native speakers.
Frequently they modify their input in a number of ways. We will now examine rhe narure of these modifications in input discourse.

The characterisrics of input to language learners

Input studies have focused on two issues. A number of srudies have examined
what might be called 'input texr' by trying ro esrablish what native speakers
actually say or write (as opposed to what reference books daim narive
speaker usage consists of). Second, rhere are studies of 'input discourse', the
special kind of 'register' that is used when speakers address language learners.1 Studies of input discourse have been recently extended to include an
examination of the ways in which interaction shapes the input thar learners
receive.

Input text: native-speaker usage

A number of researchers have warned of the dangers of making assumptions
about the nature of rhe input thar is addressed to language learners on the basis of descriptions of the abstraer sysrem of the target language. Lightbown
and d'Anglejan (1985) argue the case for detailed emprica! studies of actual
usage, pointing out that such studies are particularly mportant before any
claims can be made about universals of language acquisition.
A good example of the kind of ptoblem rhat can arise in SLA research when
the input is assumed rather rhan attested is found in interrogative structures.
Research reviewed in Chapter 3 provided evidence that learners' early interrogatives take the form of declarative word order with rising intonation (for
exampfe, 'You speak English?'). The question arises as to whether this reflects
a transitional constructon typical of interlanguage developmenr, indicative
of the process of the learner's creative construction of L2 rules (Dulay and
Burt 1977). Studies of native-speaker usage suggest thar this may not be the
case. Vander Brook, Schlue, and Campbell (1980) and Williarns (1990) point
out that many native-speaker quesrions in English are also non-inverted, particularly when there s a high presupposition of a "yes' answer:
A: l'm studying poetry this year.
B: You're srudying poetry chis rerm?
Vander Brook et al. re-examined sorne of the data from Cancino et al. 's
( 1978) smdy of Spanish learners' acquisition of English interrogatives and
concluded rhat in man y cases it was not possible to judge the appropriateness
of the learners' question forms because itwas impossible to gauge the amount
of presupposition they were making. In otherwords, no conclusion I'egarding
the use of transitional constructions by the learners was possible. Williams'
srudy of yes/no questions in Singapore English shows how difficult it is to
judge \Vhen learners have achieved a level of target-like production without

247

Input discourse: the description of modified input

149

When caretakers speak to young children who are in the process of acquiring
their Ll, they typically adjust their speech in a number of ways. The register
that results has been referred to variously as 'baby-talk', 'motherese', 'caretaker talk' (the term used in this chapter), and 'child-directed language'. Similarly, when native speakers talk to L2 learners they also modify their speech;
the resulting register is kno'Wil as foreigner talk. Ir is also possible to talk
about interlanguage talk (see Krashen 1981: 121), the language that learners

lrrespective of cultural differences, adults' speech to children has been

shown to be remarkably well-formed, thus refuting the mentalist claim that
the input children receive is degenerare. For example, Cross ( 1977) found few
examples of speech which was not fluent or intelligible, or of run-on sentences
(in each case~ less than 10 per cent of the mothers' total utterances). Newport,
Gleitman, and Gleitman (1977) reponed that only one out of 1,500 utterances was disfluent.
Another rather different rype of adjustment concems the knds of topics
that get talked about. Parents tend to follow the here-and-now principie by
talking about topics which can be understood in terms of objects physically
present and actions that are taking place ar rhe time. They avoid talking about
acrivities that are displaced in time and space until the child has developed the
necessary concepts to understand rhem. Parents have also been found to prefer ropics that are familiar to their children. Ferrier (1978) has pointed out
that much of the communication with young children centres on routine activities: eating, having a bath, gettng dressed, looking at picture books, playing games, etc. These activities involve caretaker and child in joint attention
on a common set of objects and actions.
Caretaker speech is also characterized by interactional modifications. Both
child and adult caretakers make plentiful use of attention-getters (for example, 'Look!' or 'Hey!'). Adult caretakers make specal efforts to ensure that
what they say is understood by their children by frequently checking compre
hension and repeating all or parts of their utterances, as in this example from
Snow (1972):

address to each other. We wiU now examine rhe formal and interactional
characteristics of these registers.

Caretaker talk
Partly as a response to mentalist claims thar rhe input thar children receive
from their caretakers is 'degenerate' (see Miller and Chomsky 1963 ), in the
1970s researchers into Ll acquisition set out to examine the nature of caretaker talk empircally. lmportant collections of papers were published by
Snow and Ferguson (1977) and Waterson and Snow (1978). Other studies
have contnued to be published since, particularly in the journal of Child
Language. We \vill concentrare on the main findings of this research.
A number of studies have shown that caretakers adjust their speech formall y so that the input that children receve is both clearer and linguistically
simpler than the speech they address to other adults. Broen (1972) found that
speech addressed to two year-olds has only ha!f rhe speed used wirh orher adults. Gamica (1975) showed rhat adults use a 1'igher pitch when talking to
children. Sachs (1977) found that mothers tune the pitch, intonation, and
rhythm of their speech to the perceptive sensitiviry of their children. Such
modifications are often linked with additional clues provided by gesrure and
gaze.
Comparative studes of the speech adults address to other adults and the
speech they address to children ha ve shown that caretakers also make adjustments in lexis and syntax (Snow 1976; 1977). They use a higher ratio of content words to funcrors and also restrict the range of vocabulary items
employed (i.e. they manifest a low cype-token ratio). Modificarions in syntax
are evident in a lower mean length of utterance (Iv.LU), a measure that reflects
both the length and the overa!! linguistic complexiry of utterances. Care
takers use fewer subordinate ancl coordina.te constructions, and correspondingly more simple sentences. They avoid sentence embeddings and they
produce sentences which express a limited range of syntactical and semantic
relations.
It ha,s been suggested, however, that these characteristics are not necessarily found in the talk of all caretakers and, in particular, that they may reflect
the particular child-rearing practices of middle-class, English-speakng parents. Harkness (1977) and Ochs (1982) respectively provide evidence to suggest that formal simplifications are not the norm in caretaker talk in Kenya
and Western Samoa. Ochs and Schieffelin (1984) argue that many of the
characreristics of talk observed in white, middleclass American carecakers
are a reflection of a cultural predisposition for experts ro assist novices and,
that ths may not be evident in other cultures. Crago ( 1992) documents how
lnuit chHdren in Arctic Quebee are traditionally expected to leam language
through observing and listening to others rather than through participating in
conversations shaped by questions from adult caretakers.

Pick up the red one. Find rhe red one. Not the green one.
l want the red one. Can you find the red one?
They are also ready to allow the child to initiate and control the development
of topics. Sorne caretakers appear ro be particularly skilful in the strategies
they use to sustain and extend a conversation which their children have
started. Frequently, however, their attempts to communicate are not successfu.l, either beca use the caretaker fails to understand what the child has said or
because the child cannot understand the caretaker. In the case of the former
the caretaker is likely to pro be further by means of requests for clarification
(such as 'Mm?') or requests for confirmation which often take the form of an
expansion of what the caretaker thinks the child has tried to say. When the
child does not understand the Cfiretaker, the caretaker uses repetitions and
paraphrases to sort out the problem. These features, of course, are not unique
to caretaker talk but they have been shown to be especially frequent in com
parison to discourse involving adult addressees. As we will see la ter, the availabiliry of semamically contingent input (i.e. input that is closely linked in
meaning to something that the child has already said) has been found to be an
excellent predictor of the child's rate of progress.

150

250

Input and interaction and second language acquisition

Externa/ factors

Of considerable interesr ro researchers is rhe exrent to which the adjustmenrs are 'fine-' or 'rough-tuned'. Krashen (1980) defines fine-tuning as the
provision of rhe specific linguistic features which the child is ready to acquire
next and claims rhat caretaker-talk s characrerized only by rough-tuning. As
evidence he cites the posirive but 'not strikingly high' correlarions berween
linguistic input complexity and linguisric comperence in children found by
Newport, Gieirman, and Gleirman ( 1977). Krashen is surely right in claiming
rhat carerakers do not aim rheir input ar specific Iinguisric structures, bur he
probably underestimates the exrent ro \Vhich adufr caretakers are sensirive ro
their children's progress. As we will see when we discuss the relationship berween input frequency and acquisirion, Wells {1985) provides evidence to
suggesr rhar adults srep up the frequency of specific linguistic fearures in their
input shortly befare their children first use them in rheir own speech. Certainly, there is plenty of evidence ro show rhar adjusrments in caretaker talk
occur on a continuous scale; that is, rhey are responsive ro the development
evident in individual children at different srages of acquisition. Cross (1977)
3
considers that this justifies claiming rhar rhe input is fine-tuned.
Researchers have also considered rhe purposes served by caretaker talk.
Ferguson (1977) suggests three possible functions: (1) to aid communication,
(2) to teach language, and (3) to socialze the child. lt is the first of these that
seems ro be the most important. Carerakers seek ro communicate with their
child and this leads them to modify rheir speech in arder to facilitare the exchange of meanings. As Brown ( 1977: 26) pur ir, the primary motvation is to
communicare, ro understand and ro be undersrood, and to keep rwo minds
focused on the same ropic. Brown and Hanlon (1970) ha ve shown that mothers do correct children ro ensure that \vhat they say is true." Thus, if a child
mislabels an object {for example, refers to a horse as doggie'), rhe mother is
likely--to respond \Vith either an explicit correcrion {such as 'No, it's a horsie'}
oran implicir correcrion (for example, 'Yes, rhe horsie is jumping'). N1orhers
also pay arrention to their children's pronunciarion of words and draw rheir
artention ro polireness formulae (for example, \\hen to say 'thank you').
However, rhey pay little atrention ro rhe gramrnatical correcrness of rheir
children's speech, allo\.ving even blaranr errors to go uncorrected. In general,
then, if careraker talk serves ro rea ch synrax and ro socialize rhe child in ro the
adult's culture ir does so only indirectly, as offshoots of the atrempt to
communicare. 5
Ho\v do carerakers determine the narure and the exrent of rhe modifications that need to be made? Gleason and Weinrraub ( 1978) suggest tbar rhey
must forma general idea of clldren's linguistic abiliry, parricularlytheir abiliry to undersrand. They argue rhat rhey do not ha vean accurate knowledge of
rhe specific linguistic fearures the children have masrered. Of crucial importance is rhe exrent ro which children comprehend whar is said ro rhem and the
exrent to which they signal rheir comprehension or lack of ir to their caretakers. Chiidren tend to become inatrenrive when rhey do not understand,

251

causing rheir caretakers ro modify their speech until attenton is restored.

Cross (1977; 1978) found litt!e evidence that the mothers in her study were
able to monitor their own or rheir children's synractic levels, but she <lid find
that they evaluated the extent to \\'hich the children 'Nere a ble to demonstrate
an understanding of rheir previous utterances. The fact that careraker modifications have been observed in rhe speech that even four-year-old children
address ro rheir younger siblngs (see Shatz and Gelman 1973), also suggesrs
that they must be based on a very general assessment of rhe learner's communicarive abiliries.
In summary, there is now plentiful evidence ro sho\.v rhat rhe speech that
carerakers address to children is \VeH-adapted to rheir linguistic abilities, particulariy comprehension. Three main features of caretaker talk stand out: (1)
it is more gramrnarical rhan speech addressed ro adu1ts, (2) it is simpler, and
(3) it is more redundant. However, as Snow (1986) has pointed out, ir cannot
be concluded on the basis of these findings that there is no innate component
in Ll acquisition, as the studies do nor enab!e us to determine the relarive importance of innate and input factors. The studies sho\v only. rhar the claim
that children receive degenerare input s unwarranted.

-~

151

Foreigner talk
The foreigner talk (FT) 6 used by native speakers when commnnicating wirh
non-native speakers dsplays many of the characteristcs of caretaker ta!k.
There are also sorne differences, ho\vever, particularly when the non-native
speakers are adults. Freed (1980; 1981) compared the speech of fifteen morhers srudied by Newport (1976) with that of native speakers of American English ro eleven adu1t non~native spe:ikers. She found no differences in the
degree of well-formedness and syntacric complexity (although insrances of
ungrammatical FT are nor uncornmon-see page 253), but she did find sorne
in the distribution of senrence types and rhe interpersonal funcrions they
encoded. In particular, declaratives were much more common in rhe
foreigner talk, and yes/no questions and imperitives less common. Freed
suggests that this reflects a general difference in purpose: whereas the main
functional intent of caretaker talk is that of direcring the chiid's behaviour,
thar of foreigner talk is theexchange of information. Ir should be noted, however, that when FT s addressed to young children 1 ir appears to resernble
careraker talk fairly closely. Harch, Peck, and Wagner-Gough (1979)
analysed the input in Huang's (1970) study of a five-year-old learner and
found that imperatives and quesrions far outweighed declaratives. The
crucial factor, therefore, may be age.
A detailed study of foreigner ralk necessirates a consideration of a number
of issues: (1) rhe exrent to whch FT is grammaticaJ and ungrarnmatical, (2)
the nature of the formal modifications found in FT, (3) the nature of the intFr~

actional modifications found in FT, (4) discourse structure in FT, and (5) the
functions served by FT.
1 Ungra1nmatical input modifications
In one of the earliest discussions of FT, Ferguson (1971) noted that in languages where native speakers employ a copula in equational clauses in normal communication (for exarnple, 'Mary is a doctor') they often omit it in
talk directed at foreigners. Ferguson suggests that this is because rhe absence
of thc copula is considere_d sin1pler than its presence. The omisson of copula
is a clear exarnple of ungramrnatical FT. In subsequenr publications (Ferguson 1975; Ferguson and Debose 1977), Ferguson suggests that ungrammaticality is evident in three ways: ( 1) omission of grammatical functors such as
copula, articles, conjunctions, subject pronouns, and inflectional morphology, (2) expansion, as when 'you' s inserted before an imperarive verb (for
examplc, 'You give me money.'), {3) replacemenr/rearrangement, as when
post-verbal negation is replaced by pre-verbal negation in English FT (for
example, 'No want play'). Frequently utterances will manifest ali three types
of ungrammaticality. As Ferguson (1971) notes, many of the features found
in Ff are also evident in pidgins.
A nurnber of srudies provide evidence of ungrammatical FT (see Long
1980a and Larsen-Freeman and Long 1991 for reviews). lt is particularly
likely in what Ferguson and Debose (1977).refer to as 'talking clown' situations. Thus, Clyne (1978) reports finding examples in the speech that Australian facrory foremen use to address foreign workers. Germans ha ve been
found to address guest workers in the same way (Heidelberger Forschungsprojek;~ 1978). However, ir is also been found in situations of a more neutral
kind, as \Vhen passers-by give directions to tourists (Walter-Goldberg 1982,
cited in Larsen-Freeman and Long 1991) and even in conversations between
friends (Hatch, Shapira, and Wagner-Gough 1978).
There are striking similarities benveen ungrammatical FT and learner language. Table 7.1 compares the speech of an adu!t learner (Zoila) with that of
a native-speaker friend {Rina) in the san1e conversation. Ir shows that Rina's
input matched Zoila's ourput in a number of ways. Hatch, Shapira, and
Wagner-Gough (1978) observe that Rina felt unable to stop herself producing ungrammatical utterances but they also point out rhat her speech was not
an exact copy of Zoila's. The snilarity between FT and learner language
shou!d not be taken as evidem:e in favour of the matching hypothesis (i.e. that
the source of learners' 'errors' is ungrammatical Ff), for, as both Long
(i983a) and Meise[ (1983) have noted, it may reflecta common set of cognitive processes (see the discussion on page 265 later in this section). However,
native speakers may well introduce ungrammatical forros of the kind they
observe in learner language into theii" speech as part of the process of

152

Grammatical structure

leamer language

Foreigner talk

Copula

Regularty deleted - but does

occur in sorne contexts.

Only deleted in 5 out of 43

lnstances.

Pronoun 'it'

Peivasive deletion

Also pervasively deleted.

Verbtense

Uninflected verb form used

for ali time reference; also
V -ing used in apparent free
distribution. No aux-do.

V -ing used grammatcally.

Aux-do regularly deleted.

Negatves

'no+ V' and 'I don't know'

used.

Mixed 'no +V' and

grammatical negativas used;
'no+ V -ing' negativas most
common.

Possessives

No possessive 's.

Nouns consistently marked

with possessive 's.

'For'

'For' used in expressions like

'Is upset for you?'

Similar use of 'for' found in 3

out of 15 instances.

Ta ble 7.1: A comparison of leamer language and foreigner talk (based on data
{rom Hatch, Shapira, and Wagner-Gough 1978)

accommodating to their addressee (see Chapters 4 and 6 for a discussion of

accommodation theory).
Non-standard forros are also evident in other levels of language: pronunciation and lexis, for instance. Epenthesis {the insertion of an additional
vowel), the replacement of reduced vowels by full vowels, and exaggerated
intonation similar to the kind observed in careraker talk have all been noted
in Ff pronunciation. Interestingly, though, native speakers practising FT do
not seem to follow caretakers in using a higher ptch. In lexis, 'ungrammatical' adaptations include the use of names in place of pronouns, a special lexicon of quantifiers, intensifiers and modal particles and, in highly marked
forros of IT, the use of foreign or foreign-sounding words such as 'amigo',
'capito', and 'compris', which Meisel (1980) reports observing in German
FT.
A number of factors appear to induce ungrammatical FT. Long (1983a)
suggests that four factors may be involved:
1 The learner's level of proficiency in L2-ungrammatical FT is more likely
when the 1earner's proficiency is low.
2 The status of the native speaker-ungrammatical FT is more likely when
the native speaker is or thinks he or she is of higher status.
.
3 The native speaker has prior experience of using Ff but only of the 'limited
kind' used to address non-native speakers of low proficiency.
4 The extent to which the conversation is spontaneous-ungrammatical FT
is less likely in planned, formal discourse or in experimental situations.

254

Externa/ factors

Input and interaction and second lans,ruage acquisition

However, ungrammati~al FT can occur both with interlocutors who are familiars (as in the conversations bet\veen Rina and Zoila) and wirh strangers,
suggesring that fa<.."tors other than those listed by Long are ar work. It is, in
fact, not yer possible to idenrify the exacr condirions rhar will result in ungrammatical FT, perhaps because native speakers vary both culrurally and individually in the kind of FT they prefer to use.
As Meisel (1980) points out, ungrammatical Fr is generally felt to imply a
lack of respect. Meisel reports that Italian and Spanish workers in Germany
reacred negatively when played recordings of speech samples conraining ungrammatical FT, claiming rhat ir showed contempt on the part of rhe narive
speakers. Ho\vever, ir may not be rhe presence of ungrammaricai modfications per se thar arouses negative responses in learners, but their a\vareness of
being addressed in a special manner. Lynch (1988) found that sorne learners
objecred ro the speech of a reacher because rhey perceived ir as <talking clown',
even rhough ir did norcontain obvious ungrammaticalmodificarions. Also, in
cases of very close relationships (as that berween Rina and Zoila) ungrammatical adjustments do not appear to be objecred t.

Type of speech

Exampie

Baseline (i.e. speech

addressed to natlve
speakers)

The 747 is a large-sized jet manufactured by Boeing, an

American company, with a seating capadt]' of over 500,
arranged on two decks.

Simplified

FT

Wel!, er, .. the 747's a bigjet. And er ... er ... it's a Boeing, an
American p!ane. Er ... there's over 500 seats with er ... sorne on
top and er ... sorne down be!ow.

Regularized FT

The 747, it isa big jet. Jt is madeby Bae!ngwhich is an

American company. The seats, they are on two levels. There is
a top !evel and a bottom !eveL

Elaborated Fr

The Boeing 747 or jumbo, as it is called, is a very large jet,

manufactured or made by an American company, a firm in
Seatt!e USA. lt has the capacity or space to seat a large number
of passengers, over 500 peop!e. The seats are on two decks or
levels, sorne up on top and sorne dcwn be!ow.

Table 7.2:

2 Grammatical input modificarions

25 5

Contrived examples of FT ilfustrating the proresses of simplification,

regularization, and elaboration

Table 7.3 summarizes the main linguistic modifications that have been
claimed to contribute to simp!ification. One way of simp!ifylng is by adjusting temporal variables such as speech rare (measured usuaHy in syllables
per second), arriculation rate {measured by calculating rhe ratio of the toral
number of syllables to the total articulation time-) and silent pause phenomena (pause duraton, pause disrribution, and pause frequency). Griffiths
(1991a) summarizes the research in this area of IT. His survey shows that al~
though temporal variables have been frequenrly commented on in rhe literature (for example, Freed !981; Dudley-Evans andJohns 1981; Hatch 1983b;
Klein 1986), these comments are nor accompanied by reports of derailed empirical results. Henzl's ( 1973; 1979) studyof rhe teacher talk in three different
languages {Czech) English, and German) addressed to advanced and beginner
L2 learners and native speakers, does give results, ho\vever. The study shows
rhat the teachers adjusred their speech rare in accordance wirh the listeners'
proficiency, as does Hakansson's (1986) study of Swedish teachers' classroom talk. Griffirhs points out a number of methodologcal flaws in borh
studies (for example, failure to control for rhe sequence of audiences and to
account for classroom acriviry silences) and a[so dra\vs attention to the fact
thar in Henzl's study there was enormous variabi!iry from one language to
another and, in both studies, from one teacher to another wirhin the same
language. In the case of syntactic and lexical modifications, simplification
is achieved both by avoiding difficult items in the target language and also by
reduced use of them.
Regularization entails rhe se!ection of forros that are in sorne way 'basic' or
'explicir'. Examples include: fe\ver false starts, the preference for full forms

Ungrarnmarical FT is highly marked. In many siruations ir <loes not occur,

suggesting thar ir constitutes a particular discourse rype. Arthur et al. ( 1980)
recorded sixty telephone conversations between adult non-native speakers of
English and airline ticket agenrs and reported no instance of ungrammatical
input modifications. Studies of reacher talk (for example, Henzl 1973 and
1979; Gaies 1977; Hakansson 1986) also, notsurprisingly, reportan absence
of ungramrnatical modi.fications, alrhough orher studies (forexample, Hatch,
Shapira, and Wagner-Gough 1978) did find insrances in the ianguage that
reachers use ro organize and manage classroom acrivities. Grammatical FT is
rhe norm in mosr classrooms, however.
Grammarical FT is characrerized by modi.fications indicative of three general processes: (1) simplification, (2) regularization, and (3) elaborarion. To
illusrrare rhe differences in rhe FT thar results from thesethree processes, contri ved examples are provided in Ta ble 7.2. lt should be nored thatwhereas (1)
involves an arrempr on the parr of native speakers ro simplify rhe language
forms they use, (2) and (3) are directed ar simplifying the learners'taskof processing the input and can, in facr, result in rhe use of language thar is not always simple in its_elf. This is importanr because ir means that FT provides n_ot
only simple input, corresponding perhaps to what learners already know, IJut
also input containing linguisric fearures that they have not yet Iearnr.
The account of FT rhar follows draws on a number of studies: Gaies 1977;
Arrhur et al. 1980; Long 1981a and 1981b; Scarcella and Higa 1981; Hatch
1980 and 1983b; Chaudron 1982; Kleifgen 1985; Hakansson 1986; Wesche
and Ready 1985; Griffiths 1990 and 1991a.

153

Type of simplification
Tefnporal variables

Speech to NNSs is often slower than that addressed to NSs

- mainly as a result of longer pauses.

Length

FT makes use of shorter sentences (fewer words per T-unit).

Synta.ctic complextty

FT is genera!ly less syntactically and propositionally

complex, Le. fewer subordinte c!auses of all kinds
{adjectival, noun, and adverbial), greater use of parataxis
(e.g. simple coordina.te constructions), and less preverb
moditlcation.

Vocabutary

FT rnanifests a low type-token ratio and a preference for

high frequency lexical items.

1
''!"!!

Table 7.3: Siniplification in gramn1atica/ forei"gner talk

1
1

1
1

over contracted forms; the preference for canonical word order noted by
Long, Gambhiar, Ghambiar, and Nishimura (1982) in English, Hindi, and
Japanese; the use of explicit markers of grammatical relations (for example,
'He asked if he could go' rather than 'He asked to go'); the movement of topies to the fronr of sentences {for example, 'John, l like him'); the avoidance of
forms associated with a formal style (forexample, "tu' is preferred ro 'vous' in
French IT); and the a voidance of idiomatic expressions and the use of lexical
items with a wide coverage (for example, 'flower' rather than 'rose'). Hatch
( 1983 b: 66-7) suggests a number of wa ys leamers might benefir from regulariza ti o ns of these kinds. For exarnple, they help to make the mearungs of utterances more rransparent. This ma'f be achieved by increasing the processing
time available to learners or by 111aking key structural elements more salient,
thereby hclping rhem to identify constituent boundaries in utterances.
Elaboration is the opposite of simplification, but to claim that IT evidences
both is not contradictory, as both processes can occur at different times. Elaboration often involves lengthening sentences in an attempt to make the meanM
ing clear. Native speakers often use analytic paraphrases of lexical ite~ they
consider difficult. Chaudron (1983b) provides a number of examples m the
spcech used by a university lecturer to a dass ofESL learners l'holdon tightly'
is uscd in place of 'cling'; 'there's still this feeling .. .' instead of 'we have this
myth'). Native speakers also sornetimes offer synonyms ('funds or money'}
and they define irems. They may add nformation that helps to contextualize
an itern (for exarnple,'lf you go for a job in a facto~.. .' where '!na factory' is
redundant). Chaudron suggt:sts that such elaboranons are des1gned to make
the message more ~cognitively simple', but he also makes the point that they
can result in too much redundant and confusing information. The lecturer he
.studied sometimes over-elaborated, making the interconnections between
ideas difficulr to comprehend.

As in caretaker talk, these different kinds of adjustments are conrinuous in

nature; speakers make more or less of them depending on their perception of
the learner's ability to understand. Hakansson (1986), far instance, found
that the input provided by a teacher to learners of L2 Swedish increased in
length over time as a result of an increased use of subordinations andan expansion of nominal phrases through increased modification. Kleifgen (1985)
found that a kindergarten teacher's input became more complex over time to
children who sho\ved improvement, but remained static for those who did
not.
The learners' proficiency is not the only factor that determines the extent of
modlication. Age has an effect as well. Scarcella and Higa (1981) suggest that
Iinguistic simplification is in fact triggered more by age than linguistic cOmpetence, on the grounds that they found much more to occur in Ff addressed
to children aged between eight and ten years than in FT to adults.
The linguistic modifications that have been observed to occur in Ff correspond closely to the language found in simplilied texts of the kind found in
graded readers. As such, the input is often considered 'non-authentic' (see
Vincent 1986). lt is worth noting, however, that in FT the modifications occur quite spontaneously as part of the process of trying to communicate with
learners of limted competence.
3 Interactional modifications

"-),

154

Increasingly, studies of foreigner talk ha ve switched their attention frorn linguistic to nteractional modifications. This has been motivated in part by the
finding that interactional modifications occtir even when input modifications
do not (Long 1980a) and also by theoretical claims regarding the importance
of this type of modification for comprehension and acquisition (see following
sections in this chapter).
A useful distinction can be made berween those interactional modifications
that involve discourse management and those that nvolve discourse repair. 7
The former are motivated by the attempt to simplify the discourse so as to
avoid communication problems, while the latter occur when sorne form of
communication breakdown has taken place or in response to a learner utterance that contains an error of sorne kind (factual, linguistic, or discourse). 7
These distinctions are shown diagrammaricaUy in Figure 7.1 on page 258.

Discourse management
One of the most effective ways of managing discourse with native speakers is
to ensure that the topic of the conversation is understood. Long (1983a)
identifies anumber of strategies which native speakers use to achieve this end:
selecting saHent topics, treating topics simply and brefly, making new topics
salient, and, when necessary, relinquishing to pie control, although this last
strategy seems to relate more to discourse repair than discourse management.

258

Input and interaction and second language acqrsition

Externa/ factors

discourse management

interactional
modifications

amount and type of informatlon conveyed

use of questions
here-and-now orentation
comprehension checks

- self-repetition

repair ot communication
breakdown
- negotiation of meaning
{requests far c!arification; requests far

discourse repair

confirrnation; selt- and

other-repetitions)
- relinquishing tapie
repair of learner error
- avoidance of othercorrection
- en-record and offrecord corrections

Figure 7.1:

Types of interactional modifications in fareigner talk

One method used by native speakers ro control topic concerns rhe amount
and rype of inforrnation that is communicated. Arthur eral. {1980) cornpared
the number of 'informarion birs' thar narive speaker airline agenrs included in
their answers ro a specific telephone enquiry from narive speakers and nonnarive speakers ('Whar kind of plane is a_?'). A distincrion was made berween 'simple' informarion (such as 'size' and 'jer') and 'complex' informaran (fo.r example, 'searing capacity', 'name of manufacrurer' J and 'seating
arrangemenr'). There was no difference in rhe amounr of simple information
given ro narive speaker and non-native speaker callers, but significant differences were found in the amount of complex informarion, rhe non-native
speakers receiving far less. Derwing (1989) found that native speakers adjusred the inforrnarion rhey provided about a film rhey had seen when speaking to low-proficiency L2 learners. The information contained in the
speakers' proposirions was classified as belonging ro one of rhree caregories:
( 1) crucial information, (2) non-essential major informarion and (3) minor
informarion, consisting of background or irrelevant informa;ion. There was
no difference in the amounr of crucial information which the native speaker
and non-native speaker addressees received, but differences were evident in
the relarive proportions of major and minor informarion. OveraII rhe narrators included less major informarion and more minor informarion in speech
to rhe learners. There were considerable individual differences however
which, as we will see la ter, hada significanr effect on the non-nativ~ speakers:
comprehension. Ehrlich, Avery, and Yorio (1989) also found evidence of
variation in rhe arnount of information supplied by individual narive speakers

155

259

in a problem-solving task. They disringuish 'skeletonizing', where rhe barest

details are provided, from 'embroidering', where the information is expanded
and embellished. Interestingly, rhey found both str:J.tegies present in interactions wirh borh native speakers and non-narive speakers, suggesting thar
individual native speakers may ha ve preferred inreractional styles which rhey
use irrespecrive of rheir inrerlocutors, Ir is clear, though, that often native
speakers do seek to manage discourse wirh non-native spe3kers by regulating
the amount and rype of information rhey provide.
Narive speakers also make use of questions to establish and control topics.
We ha ve already nored rhat questions appea.r to be 111.ore frequenr in IT rhan
in careraker ralk. Long (1981b) found that in convers::trions benveen narive
English speakers and elemenrary Ievel Japanese learners> rhe native speakers
initiared most of rhe tapies, rypically making use of quesrions ro do so. Ninery
six per cent of all ropic initiations were quesrions, whereas in narive speakernatlve speaker conversations only 62 per cent were. Long suggesrs a number
of reasons why questions were favoured: they compel J.nswers 1 rhey signal ro
the non-native speaker thar a turn is approaching) and rhey lighten the learn~
er's conversarion burden beca use rhey encocle part (and sometimes all) of the
propositional content required ro respond. Long a!so found differences in the
rypes of quesrions used. In conversarions with rhe learners, rhe narive speakers made greater use of yes/no and 'or' type quesrions. One possibiliry is rhat
the use of quesrions in Long's study reflected, in parr at leasr, rhe lack of familiariry between the parricipanrs who made up the pairs. However, Gaies
(1982) obrained similar results, suggesting thar rhis was nor the case. In this
srudy, unlike Long's srudy, the native speakers and non-native speakers were
already acquainred wirh each orher, but rhe narive speakers still used more
questions when ralking wirh the non-native speakers. Ir should be nored,
rhough, thar in ali rhese srudies rhe narive speakers were adulrs; child native
spea.kers seem less inclined to estabish and develop ropics rhrough quesnonmg (see Harch 1978b; Peck 1978).
A rhird straregy of discourse rnanagement is to selecr copies that have a
here-and-now orienrarion (see the discussion of this fearure of caretaker ralk
on page 249). Long (1980a) and Gaies (1982) both report significanrly more
prese.nt-tense verbs in native speaker speech addressed to nonnative speakers
rhan in speech addressed to orher narive speakers 1 suggesting suc.-h an orientation. The here-and-now orientation allows [e::Irners to r11ake use of rhe immediare context ro interpret the me;ining of ntterances.
Final.Jy, narive speakers have bcen noted ro try to mana.ge Jisc%>urse by frequenrly _checking whether rhe learner has understood. Comprehension
checks (tor example, 'You undersrand?', 'Okay?') have been found ro occur
more frequenrly in native speaker-non-narive speaker discourse rhan in
native speaker-narve speaker discourse (Long 1981a; Scarcella and Higa
1981). Teacher talk, in particular, seems to be rich in comprehension checks.

. . . ... , '-
.,,.~

1-~-

Pica and Long (1986) found that ESL teachers were much more likely than
native speakers to check comprehension in informal conversations.

Discourse repair

The need for discourse repar arises in cerrain types of problematic communication. Gass and Varonis (1991) presenta taxonomy of problematic communication rypes (see Figure 7.2). An initial distinction is made benveen
'non-engage1nent' and 'miscornmunicarion'. The forrner occurs either when
there is 'non-communicaton' (for example, \vhen a non-native speaker
avoids talkng to a native speaker) or when there is 'communication breakoff
(for exan1ple, when a na ti ve speaker stops communicating as soon as they disco ver they are ralking to a nonnative speaker). 'Miscommunicaton' occurs
when sorne message other than that intended by the speaker is undersrood. lt
can take the form a 'misunderstanding' or an 'incomplete understanding'
(either 'non~understanding' or 'partial understanding'), depending on
"vhether or not the participants overtly recognize a problem and underrake
repair. In the case of an 'incomplete understanding' remediation occurs, but
in the case of a 'misunderstanding' no repair occurs and the speakers are
likely to lapse into silence. Gass and Varonis also note thar miscommunication can occur both as a result of cross-cultural differences in the way language is imerpreted and because of purely !inguisric difficulties.

L
i'

Prob!ematic com111unication
Miscommunication

Non-engagement

Non-communication

Communlcation

Misunderstandlng

break.off

lncomp!ete
understanding

non-under
standing

Figure 7.2:

partial
understanding

Problenw.tic comrnunication types (from Gass and Varonis 1991)

Rep.air~ then, occurs when there is an 'incomplete undersranding'. It takes

the form of negotiation of rneaning - the coilaboratiye work which speakers
undertake to achieve mutual undersranding. Native Speakers typically use requests for clarification ('Sorry?' ~ 'Huh?', 'l beg yourpardon,} and requests for
cnfi.rroation, which often make use of intonation or tag questions:

NNS: Mexican food ha ve a !ot of ulcers.

NS: Mexicans have a lot of ulcers?
(Young and Doughty 1987).

11'

156

Other conversational modifications that help to repair discourse are selfand other-repetitions, which can be exactor semantic (i.e. paraphrases) and
complete or partiaL It should be noted that not ali repetitions ha ve a reparing
funcrion; as Pica and Doughty ( 19 8 8) point out, na ti ve speakers also use them
to manage discourse (i.e. to try to prevent a communication problem arising).
Native speakers may also abandon the attempt to negotiate meaning by giving upa topic and allowing the non-native speaker to nomina te an alterna ti ve
one.
A considerable amount of research has been undertaken to establish the
conditions that promote the negotiation of meaning, motivated by the claim
that the comprehensible input that can result from it is of particular benefit to
L2 acquisition (see the section on input and acquisition in this chapter). Pica
(1987) argues that the most important factors concern rhe social relationships between the interactants. Interaction involving participants of equal
status ensures that 'learners and their interlocutors share a need and desire to
understand each other.' Conversely, unequal status makes ir difficult and
even unnecessary for participants to restructure interaction. This claim mirrors that made by Wells and Montgomery (1981) for caretaker inreraction.
When mothers act as conversational partners plentiful negotiarion takes
place, but when they adopta tutorial role ir is inhibited.
The dstribution of power among the participants helps to explain a number of other findings, such as the influence that the non-native speaker's age
has on FT. Scarcella and Higa (1981) found that adult native speakers assumed much greater responsibility for conversations when speaking with
nonnative speaker children than with non-native speaker adults. With children they tended to dominare conversations so that less negoriation took
place. Pica and Long (1986) found interactional modifications more frequent
ourside than inside the classroom, a finding that can also be explained by rhe
fact that roles are unequally distributed in the classroom, with the teacher assigned many more discourse rights than the learners. Overall, these studies
suggest thatwhen role relarionships are asymmetrical, rneaning negotiation is
inhibited. Other kinds of non-negotiated input may still be available in such
siruations, however. Cathcart {1986), for _instance, found that adult native
speakers were more likely to respond to requests for informaton from child
non-native speakers than were child native speakers.
A number of other facrors ha ve been found to inflence negotiation: for
example, the narure of the task, the characteristics of the participants, and
partcipant stnucture. Much of the research that has investigared these factors
has involved classroom learners or has been conducted for a clear pedagogic
purpose (for example, to compare interaction in group and locksrep settings).
For this reason it will be considered in Chapter 13.
The presence of higher frequencies of discourse repair functions such as requests for clarification and confirmation has been taken as evidence that
higher levels of negotiation of meaning are occurring. This is, however,

262

Input and interaction and second language acquisition

Externa/ factors

quesrionable. Asron (1986) has pointed out that these discourse acts do not
unambiguously indicare negotiation of meaning, as rhe sarne procedures can
be used in non-problematic conversation. He argues that they are ofren used
'to achieve a formal display of convergence of the participants' worlds' by
allowing rhem to perform 'a ritual of understanding or agreement'. In other
words, negoriation can be morivared by the interacranrs' need to display satisfactory ourcomes rather than to overcome trouble sources.
In addirion to repair \vork directed ar solving problems of undersranding,
there is also the repair of learners' errors. Schwarrz (1980) reports a general
preference for self-correction over other-correction in non-native speakernon-native speaker discourse. The same seerns to be the case in native
speaker-non-native speaker discourse. Gaskill (1980) examined the types of
repair which an Iranian learner of English experienced in both elicired and
naturally-occurring conversations with native speakers. He found only
seventeen examples of other-correction in 50 pages of transcript. ltwould appear, then, thar narive speakers typically ignore learners' errors, a conclusion
supporred by Chun, Day, Chenoweth, and Luppescu's (1982) study of 28
ESL learners of mixed proficiency in Hawaii. In this case, less than 9 per cent
of the rotal errors were corrected. Also, it was 'factual' errors and 'discourse
errors' {for example, inappropriate openings, closings, and refusals) rarher
than lexical or syntacric errors that were more likely to attract repair from
narive speakers. Chun et al. suggest thar the low leve! of repair reflects the
narive speakers' desire not to impair the cohesion of the discourse.
In the case of other-correcrion, a disrinction can be made between onrecord and off-record feedback. Day, Chenowerh, Chun, and Luppescu
(1984) define the former as feedback which occurs when the native speaker
responds to rhe source of a learner's language problems directly and unamR
biguously, by means of a statement with declararive intonation. Off-record
repair is ambiguous and can ha ve more than one inrerpretation; it can consist
of a question in the form of a confirmation check (although not all confirmation checks are corrective) ora statemenr. They found that adult ESL learners
received significantly more en-record than off-record feedback, perhaps
because as friends of the learners, the native speakers felt they could overtly
correct without threat to face.
In the case of linguistic modifications, there is substancial evidence that
their extent in narive speaker speech varies according to the learners' level of
developmenr. Some\vhat surprisingly, there have been few studies that have
invesrigated \vhether this is also true for interacrional modifications. tviy
study (Ellis 1985d) suggests thar sorne, but perhaps not all, of rhese modifications are sensirive to the learner's level. It was found that a teacher used significantly fewer self-repetitions, but more expansions, in interacrions wirh nvo
learners when they had progressed beyond the stage of absolute beginners.
Also, the reacher swirched from topics that required objecr identification to
topics rhat required the learners to make sorne kind of comment about an

263

object. However, no change was observed in other interactional features such

as corn~rehension checks, requests for clarification and confirmarion, types
of ques.nons, and overaH response to communication breakdo\vn. Day et al.,
in the study referred to above, found rhat during game-like activities beginner
and intermediare !earners received significanrly more on-record repair than
did advanced learners. However~ no differences were noted in conversarions.
The effect of the learners' proficiency on interactional modifications is clearly
an area in need of further study.
The study of interactional modifications in FT has blossomed and there is
now a rich literature to draw on. It is probably true to say, ho\vever, that the
research has focused on a rather narrow set of inreractional phenomena
associated with the negotiation of meaning. The categories themselves are
less watertight than researchers sometimes admit-the function of a self- or
other-repetition, for example, is not always clear. Also, the research has been
somewhat restricted in the discourse types that it has studied-many of the
studies ha ve examined data obtained in interview-type situations or by means
of information-gap activities.

4 Discourse structure

157

The study of interacrional modifications involves examining the speech of

both participants in a conversation. l\s Long (1983a) has pointed out, acts
such as requests for clarification and confirmatlon take place within a
discourse context-they have life across utterances and speakers. However,
the srudy of individual discourse acrs <loes not show how sequences of
discourse are constructed. For rhis it is necessary to examine the strucrure of
Fr discourse.
Gass and Varonis (1985a; Varonis and Gass 1985) have developed a model
to describe the strucrure of non-understanding routines' where meaning
negotiation takes place. It consists of a 'trigger' (i.e. the utterance or part of
an utterance that creares a problem of understanding), an 'indicator' which
indicares that something in a previous utterance \vas not understood, a
'response' ro the indicator, and finally a 'reaction to the response', which is
oprional. The 'indicator-response-reaction to rhe response' portian of a nonunderstanding sequence is called a 'pushdovm', because it has the effect of
pushing the conversarion clown rather than allowing it to proceed in a forward manner. The model is recursive in thar it aflows for the 'response' ele
ment irself to actas a ~trigger' for a further non-understanding routine. Ta ble
7.4 provides a simple illustration of the model.
Other models exisr. Ehrlich, Avery, and Yorio (1989) daim that rhe highly
specific nature of the problem-solving task they used yielded discourse wit.h a
'hierarchical constiruent strucrure'. The task required the speaker {the native
speaker in native speaker-non-native speaker pars) to give instructions

...

~-~-----

Utterance

Function

NNSi: My faU1er now is retire.

Trigger

NNS2: Retire?

lndcator

NNS1: Yes.

Response

NNS2: Oh yeah.

Reaction to Response

1977). In fact, it can reflect eirher downward divergence (such as when a

native speaker deliberately employs ungrammatical forros wirh a competent
native speaker ro signa! lack of respect), or downward convergence {such
as when a native speaker approximares the interlanguage forms used by rhe
native speaker as a way of signalling solidariry) . This double funcrion of (2)
may help to explain why ungrammatical Fr can occur berween non-familiar
interlocutors in service or workplace encounters and berween familiar inrerlocutors in ordinary conversation. (3) is only 'impli'cit' because native
speakers do not usually ha ve any pedagogic intent, although Naro (1983) in a
response to Hatch argues that FT can occur with an explicit teaching function
(i.e. with the intention of helping a leamer !earn).
Many of the formal characterisrics of FT are very similar to rhose found in
other simpJified registers such as 1earner language, caretaker talk, and
pidgins. This suggests that i~ reflects universal processes of simplification,
knowledge of which constitures part of a speaker's linguistic competence.
However, as Meisel (1980) has pointed out, there are differences berween the
simplification found in FT and that observed in leamer Ianguage. Whereas
both manifest restrictive simplification (for example, rhe use of an infinitive in
place of inflecred verb forms), only learner language manifests elaborative
simplification (for example, the use of novel verb forms through processes
such as overgeneralization). Meisel suggests that resrrictive simplification in
both registers serves 'the purpose of achieving an oprimal result in communication' ( 1980: 36), but elaborative simplification occurs when learners are
rrying to complexify their interlanguage system.
An interesring question is how native speakers come to be a ble to adjust rhe
leve! of their IT to suit the leve! of individual !earners. Hatch (1983b) considers three ways; (1) regression (native speakers move back through the
stages of development that characterized their own acquisition of language
until they find an appropriate level), (2) matching (native speakers assess a
learner's current interlanguage state and then imitare rhe forms they observe
in ir), and (3) negqtiation (native speakers simplify and clarify in accordance
with the feedback they obtain from learners in communicarion with rhem).
The second is the explanation offered by Bloomfield (1933), but it seems un
likely, as ir is probably asking too much of learners' interlocutors to measure
simultaneously the learners' phonology, lexicon, syntax, and discourse wirh
suffi.cient accuracy to adjust their own language output. The most likely exp!anation is (3), alrhough (1) is also possible.

Table 7.4: A sitnple discourse model of the negotiation of meaning (example from
Gassand Varonis 1985a)

about howro fill in a matrix of 16 b!anksquares with differem objects. Al! the
speakers proceeded through the task in rhe sarne manner beginning wirh an
'orientation', rhen moving on to the 'identification' of an object, before providing a 'description' of it. The information provided in the 'descriprion' was
also hierarchically organized, as sorne items of information depended on
others ha ving been previously conilllunicated. The model was used to distinguish speakers who adopted a'skeleronizing' strategy (i.e. provided only rhe
basic inforn1ation needed to perforrn a task) from those who adopted an 'ernbroidcring' strategy (i.e. providing information that expands and embellishes
beyond what is required to perform the task).
The models that ha ve been developed are essentially data-led models; that
is, they ha \'.e been devised to describe the particular tokens of FT discourse elicited by the tasks used in the studies. 1\lthough there are now a number of
models of IT discourse structure currently a vailab!e (such as those proposed
by Gass and Varonis and Ehrlich, Avery, and Yorio), it is not clear wherher
these are sufficiently powerfu! to describe different discourse types and thus
to pern1if generalizations across studies. Nevertheless, the use of models that
account for discourse structure rather than taxonones of discrete discourse
functions constitutes a definite advance, as it enables researchers to examine
the 'pouring back and forth' whch Brown (1968: 127) consders essential for
investigating how learners acquire Language.
,

5 The functions of foreigner talk

Overall, three functions of foreigner talk can be identified: (1) to promete
communicarion, (2) to signa!, implicitly or explicidy, speakers' attitudes towards their interlocutors, and (3) to teach the target language imp!icitly.
Hatch (1983b) argues correctly that (1) is primary in that most adjustments
are geared to si1nplifying utterances ro make them easier to process or to
clarify what has been said by either the narive speaker or the non-native
speaker. Hatch characterizes (2) in terms of the special kind ofaffecrive bond
that FT can crea te benveen the nacive speaker and non-native speaker, but it
is also rnanifest in FTwhose purpose is 'talking down' (Ferguson and Debose

Interlanguage ta/k
Interlanguage talk (ILT) consists of the language that learners receive as input
when addressed by other learners. ln Chapter 6 we noted that ILT constitutes

158

266

Input and interacton and second language acquisition

Externa/ factors

rhe primary source of input for many learners. The treatment will be brief

267

ungrammarical and gramrnaticaI FT occur, a!rhough it is not yet possible ro

identify the precise social condirions rhar favour one over rhe orher. In rhe
case of grammatical FT, three processes are evident: simplificJtion, regularization, and elaboration. The modifications are continuous, influenced by the
lea.rner's stage of deve!opment and age: FT also displ2ys interactional modifications direcred ar both managing and repairing discourse \Vhen eirher communication breakdo\vn or learner error occurs. Figure 7.1 on page 258
summarizes rhe inreracrional rnodifications associated 1vith rhese broad functions. Ivluch of the research on FT discourse has been concerned with the
negotiation of rneaning-the attempr to remedy mis- or non-undersranding.
A number of models aimed ar describing FT discourse have also been
developed, but no general model is available yet. Foreigner ralk, Iike
caretaker talk, seems to be primarily motivated by rhe desire to achieve
communicative success.
Finally, we looked at inrerlanguage ralk. This, nor surprisingly, tends to be
less grammatical than FT, but ir is characterized by more inreracrional rnodifications associared \Vith rhe negotiation of meO:ning.

beca use rhere is a more extended discussion of rhe research in Chapter 13.

Two issues have figured prominently. The firsr concerns the extenr to
\vhch ILT provides learners with adequate access to the grammatical properties of rhe target language. Nor surprisingly, ILT has been found to be less
grammarical overall rhan FT or reacher talk (Pica and Doughry 1985a and
J985b; Wong-Fillmore 1992). Porter (1986) in a derailed study of rhe !LT

produced by intermediare and advanced L2 learners in pairwork and comparable FT found that whereas only 6 per cent of FTwas 'faulry', 20 per cent
of ILT preved ro be so. Porter also found ILT to be sociolinguisrically defi-

cient. She looked ar a number of speech acrs such as expressing opinions,

agreemenr, and disagreement and found rhat the learners failed to use politeness straregies to the same extent as native speakers. In general rhey did not
generare rhe kind of sociocultural input needed for language learning. On the
plus side, however, both Pica and Doughty and Porter found that the learners
in rheir studies repeared only a very small amount of rhe faulry input they
heard.

The second major issue concerns whether ILT provides learners with rhe
same opportunities for negoriating meaning as occur in FT. Gass and Varonis

The effect of input and interaction on acqnisition

( J985a), in rhe srudy referred to above, reported an average of 0.50 'push-

We now turn to consider what effect input and interaction have on acquisirion. We will begin by considering Ll acquisition research before moving on

down routines' in native speaker-native speaker pairs, 2.75 in native

speaker-non-native speaker pairs, and a massive 10.29 in non-native
speaker-non-native speaker pairs. Porter (1986) found thar learners

to input studies in L2 acquisition.

prompted each other five times more than the narive speakers prompred non-

'

native speakers, while repair frequencies were similar. Overall, these studies
pro vide evidence that meaning negotiation is very extensive in ILT, more so
rhan in c'Omparable FT discourse.
The qualiry of interlanguage talk is of considerable importance given the
currenr emphasis placed on small group work in comrnunicarive language
reaching (see Brumfit 1984). Ir is clear thar ILT differs in sorne respects from
FT, but, if the theoretical arguments relating to the importance of meaning
negoriation for acquisition are accepred, a good psycholinguisric case can be

Input and interaction in first language acquisition

A number of studies ha ve investigated to \Vhat extent the.re Is a relatonship

berween rhe language that caretakers address to children ::ind acquisition. The
results have been some\vhat contradictory, leading to cootroversy regarding
the role of input in Ll acquisition. In an early study, Newport1 Gleitman, and
Gleirman (1977) found little evidence of a close relationship between the fre-

quency of specific linguisr{c fearures in rnorhefs speech and the gro\\'th of rhe
same features in their children. They concluded thar most aspects of language
structure in L1 acquisition were insensirive to individual differences in rhe input, alrhough rhe rate of growrh of certain language-specific features (such as
auxiliary verbs and noun inlecrions} Wa:S serisitive. FUrro\V, Nelson, an'd
Befledict (1979) found much gre::iter evidence of an effect for input. In rhis
study, four aspecrs of Ll deve!opment {mean length of utterance, verbs per
utterance, noun phrases per utterance, and auxiiiaries pe-r verb phrase) we.re
related to a number of input measures. Barnes, Guttfreund, Satterly, and

made for ir (see Long and Poner 1985, and Chaprer 13 of rhis book).
Summary

This secrion has examined three simple regisrers of language addressed to

learners. First, tn srudies of caretaker ralk, ir has been shown rhat the input
rhat children receive when learning rheir Ll is well-formed and \VeH-adjusted
linguisrically ro their leveJ of development. Also, sorne caretakers seem adept
ar helping children to esrablish and develop topics that rhey wanr to talk

Wells (1983) also produced evidcnce ro suggest rhat input influenced lan-

about.
Foreigner talk resembles caretaker raik in sorne respects, but also differs
from ir in others (for example, there are fe,ver ves/no ouestionsL Rorh

guage development. They reported significant correlarions bet\veen rhe frequency of polar interrogatives and subjecr-verb inversion in the input, and

159

~<>~=~ni~~----:-

--

_j

'

"

'~_::

intonation and auxiliary meanings and pragmatic functions. However, Gleitman, Newport, and Gleitman (1984), after reanalysing the data used in their
1977 study, reaffirmed rheir original results, finding few correlations berween
input and ianguage deve!op1nent measures.
As is often the case in acquisition research, these differences are difficult to
interpret or ro reconcile. In part, they can be accounted for methodologically,
as Schwartz and Camarata (1975) and Bennett-Kastor (1988) point out. For
exa1nple, most of the studies u.sed co.tr.!!lational staristics, which are difficult
to interpret beca use they do not pro vide any indication of the directionality of
a relationship. The studies also varied in whetl1er and how they took account
of the chil<lien's age and stage of deyeh_JJ)ment, borh of which have been
sho\vn t-l!JluC;;ce .ca~er_a_ker-i~1put. Snow {1986: 76) ad vanees the argument
that 'the task ~Of lea~niJ.l:g a lan~uage is quite different in differenr stages of acquisitiqn' and thar, conseqUenr1Y, it Ts reasonable ro presume thar c!~.fferent aspecrs of careta~er talk may be importnt ar different stages. It is worth noting
th.lt al! the studies report sorne relationships involvIUg i~p!J~ and acquisition.
Fragile features 9 such as auxiliary verbs seem to be sensitive to inPut:-wens
(1985), for instance, found thatcaretakers increased the frequency of specific
auxHiary verbs in their input just before the same verbs appeared for the first
time in their children's speech. The argument, therefore, is notwhether input
has an effect, but rather what the extent of its effect is.
A somewhat n1ore convincing case has been made out for the facilitative effect of interactional features in caretaker talk Wells (1985) has shown that
car(!takers .do not vary much in the extent t_o which they modifr. ~eir speech
linguistic~y, but-dO vary consider;f>ly in the way they interact .w!~~-JJ.ieir
children. 1,"his suggests that is the interactional rather than input fearures that
may b.crucial in accounting for diff~~ences in children's rate of acquisition.
One way in which interaction may help children learn language is by providing them \Vith opportunities to form vertical constructions. A vertical consrruction is built up gradually over severa! turns. ScoUon (1976) provides a
number of examples of mother-child discourse, where the child produces a
meaningfu! statement over two or more turns:
Brenda: Hiding
Adult: What's hiding?
Bren da: Balloon.

'!

}!e suggests that vertical constructions prepare the child for the subsequent
production of horizontal constructions (the production of a meaningful statement within a single turn).
Other researchers point to the contribution that specific discourse acts
make to acquisition. Interaction ric~-~-4_if~Sfh'..e_a,pp,!f..!9 fos.~~u;:apid lan~g!Jl_<:_q~.b.itiori, at !east in the early stages (Ellis and Wells 1980), perhaps
because children are equipped with an 'action strategy' (Shatz 1978) that
n1akes them particularly receptive to directives, perhaps because directives

160

are frequently related to the here-and-now and perhaps because they are
often linguistically very simple.
SfJ2~$jQn$._also appear to P.~-!!1:2.t_~ __ deyelopm~nt. These are full, correct
versions-of.-telegraphi~.-~hild utterances, which often occur in the form~f'
acknowledgements or requests fo~-Confifmation:
Child:

Pancakes away
Duh Duh stomach.
Mother: Pancakes away in the stomach, yes, that's right.

Snow (1986) cites a number of studies that provide evidence of the positve
effect of expansions (for example, Malouf and Dodd 1972; Nelson 1977;
Cross 1978; Wells 1980), perhaps because they l'.rovide .crucial bits o information a_Qout_ syntax and morphology, althollgh, as Snow notes, the same
information may also be.made avaifabl to children in other kinds of input.
~hjldren also seem to ben~_t_Jrom assistance in building conversarions
abouttopics in which they are interested. Wells (1986a) and Snow (1986)
bothemphaS.~Ze the importance of 'coll'!bora_tive meaning maki~g' and discuss sorne of the ways in which this is achieved. Successful conversations are
achieved when caretakers regular!y check rheir understanding of what their
children have said, when they work hard to negotiate misunderstanding,
when they h.elp to sustain and extend topics that their children ha ve initiated,
when rhey give children opportunities to contribute to the cOnversation, no
matter how minimaJly, and when they are responsive ro feedback cues. The
result is what Snow calls 'semantically contingent speech,. However,
although the provision of such speech is considered to be advantageous for
language acquisition, it has not yet been shown that access to it is crucial for
success.
Input and interaction in second language acquisition
lt is useful to distinguish four broa~~l'l'I():!_C~es in studies that have .investigated the rel~.gn~hip __!J~twee11 input/int{!raction and U ac_gyis_itio_n. Th-hfSt
simply seel(s to establishwlether rhe frequency of i!Ilg,ti;;t~ fearure~ in rhe input is related to the frequency of the same features in the learner language.
The second emphasizes the importance of input that is comprehensible to
learners. The third emphasizes the role of learner output in nteraction. The
fourth looks more holistically at discourse by asking how the process of collaborative discourse construction aids acquisition.

Input frequency and second language acqusition

The fre_P:!-"1!0!lYP.othesis.state.s that th': arder _of_L2 a~n is determined
by the,frequen!:i'. with wh1ch d1fferent lmgmstic items occur m the inpuf:Tle
hyp_oQi_~sis deals with the r~Ji.e.~_eeni@!!~~'!E_c_~~

2 70

Externa/ factors

than that berween input and acquisition. The justification for considering it
here rests on the claim thar the accuracy order mirrors the acquisition order
(see Chapter 3). The hyporhesis was first advanced by Hatch and WagnerGough (1976), who suggested thar rhe limited range of topics about which
learners (particularly children) typically tall< results in certain grammatical
features occurring wirh great frequency in the input. The more frequenrly occurring irems, they claimed, were among those that emerged early in the
learners' ourput. Input frequency, ho\vever, was only one of severa! explanations for rhe morpheme order which Hatch and Wagner-Gough considered.
Frequency, strucrural complexity, and cognirive learning difficulty overlap
and may together influence acquisition. Also, ir should be reffiembered rhat
variario~s in the morpheme accuracy arder were reported, reflecting the-different tasks learners were asked to perform (see rhe discussion on pages 90--6
in Chapter 3). This raises the thomy question of which arder should be used
incorrelationalstudiesinvolvinginpurfrequency. --- - -- - - Ta ble 7 .5 summarizes sorne of the srudies that ha ve invesrigated the
input-accuracy-acquisition relationship. The lin_rations of these studies
need to be recognized. First, they are ali co~~~_l_it10n?:t in natll!e, thus making
it impossible to rnake any statement abut cause and effect. A statistically
significant correlation might simply indicare that le_<I-f.I)ers_and_,1;k1_~j~.i_nterloc
utor_s have_ fl sim_ilar communicative r:ieed to use the same graiffi.ltical
featUres~ S~~Olld, Soine o( thf -SnidieS did not obtan measures of input
frequency from the same data sets- used to examine learner accuracy (for
example, Larsen-Freeman 1976a; Long 1981b), thusmakingthe correlations
they report additionally diffi<;ultwjnr_erpret. Third, as Snow and HoefnagelHohle point out, 'it may be that quanriry has~ rhreshold rather than a graded
effect' (1982: 424). If this is the case, the use of correlational statistics is
inapproprate, as these can only be used if the two sets are both evenly spread
throughout the total range. Fourth, ir can be argued that it makes little sense
to investiga te input-output relationshlps in data collecred at the same time, as
any effect for frequency will not be immediately apparenr. It is unforrunate
that there ha ve been no L2 studies of the kind conducted in L1 acquisition
where input frequencies were correlated with measures of language deveiopment (i.e. gains from one time to anorher). Only Lightbown's (1983) study of
grade 6 ESL learners in Canada provides sorne indication of how input
available at one rime can affect acquisition that becomes apparent ata later
rime. In this study, the learners manifested over-use of V+-ing after having
received massive classroom,-exposure to chis fearure at the end of Grade 5.
Table 7.5 shows that the results have been very mixed. Whereas LarsenFreeman (1976a and 1976b), Lightbown (1980), Hamayan and Tucker
(19 80), and Long (19 81 b) ali found signilicant positive correlations berween
input frequency and accuracy, Snow and Hoefnagel-Hohle (1982), Long and
Sato (1983), and Lightbown (1983) did nor lind any direct relationship. It is
perhaps premature on the basis of such m_ixed results to propose that 'there

Input and interaction and second langttage acquisition

2 71

Study

Design

Mair. rosu!ts

Larsen-Freeman 1976a

L2 accuracy orders of grammatical

morphemes corre!ated with input
frequency orders of same
morphemes in the c!assroom
speech of 2 ESL teachers.

Spearman rank order correlations

genarally positfve and significant

larsen~Freeman 1976b

L2 accuracy orders of grarnmatlcat

morphemes correlated With
frequency order cf sama
morphemes in parental speech to
chi!dren {based on Brown 1973).

Spearrnan rank arder correlations

were positivo and s\gniflcant except
for irnitaton tasi<.

Hamayan and Tucker 1980

The accuracy orderot g French

syntactic structures in speech
elicited from L1 and L2 !eamers
correlated wtt:h frequency order of
same structures in teacher
classroom input

Signiflcant rank order correlatlons

reporte for both L1 and L2. The
frequency arder reported fortwo
c!assrooms differed but was
signrficantly related to accuracy
orders in both cases.

Llghtbown 1980

Order of appearance of L2 French

question words in speech of 2
children correlatOO with frequency
order of same question words in
speech addressed to the chi!dren by
a French research assistant.

Freque-ncy with which question

words appeared in NS's speech
matched clasely the sequence of
questlon word development in L2
!eamers' speech.

Long 1981 b

Krashen's average arder of

acqu!sition of grammatical
morphemes correlated Wlth the
frequency order o! same
morphemes in the speech that NSs
addressed to 36 elementaty
Japanese ESL students.

Spearrnan rank arder correlation

positiva and significant

Snow and Hoefnage!~

Hhle 1982

Achievement and gain seores ot

American children acquiring l2
Dutch in Ho!!and correlated with (1}
quantity ofDutch heard in the
classroom, (2) percentage of speech
directed specifica!ly at !eamers, and
(3) amount of spee<::h directed at
individual leamers.

Ne significant re!ationships found.

Long and Sato 1983

Krashen's average order of

acquisition of grnmmatica!
morphemes corre!ated with the
frequency orderof the same
morphemes in ESL teachers'
speech.

No signif!cant re!ationship found.

However, teachers' frequency order
did correlate signficantly with NSs'
frequency arder, which in tum
correlated significant!y w!th
Krashen's average order.

Lightbown 1983

Grade 6 ESLstudents' accuracy

arder of-s morphemes in ora!
communication activity correlatecl
with frequancy order of same
featu res in samp!es of oneteacher's
speech and in textbook used by
students.

No direct re!aticnship bet,veen

ac-curncy and input frequencies
found, btit sorne evidence that high
leve! offre-quency ofV + -ing in input
at earlier period {Grade 5) !ed to
over-use by the students.

;j

~-

Table 7.5:

Input frequency-output accuracy studie:s in SLA research

exist preliminary data supporting a frequency effecr' (La.rsen-Freeman and

Long 1991: 133 ). lt is possible that frequency may be more important at sorne
stages of acquisition {for example, elementary) than others, but no clear

161

conciusion is possible on rhe basis of these studies. All that can be said is that
the frequency hypothesis has not yet been properly tested.
T\vo other sets of srudies, however, do lend supportto the claim that input
frequency influences acquisition: ( 1) studies that ha ve investiga red the effects
of ungrarnmatical input, and (2) studies of the formulaic speech that many
learners produce.

communicative need to perform the language functions encoded by the

formulas with ease and fluency.
Overall, rhere is little evidence to support the claim that input frequency affects L2 acquisirion, but there is also little evidence to refute it. Perhaps the
safest conclusion is that input frequency serves as one of the factors influencing development, often combining with other factors such as Ll transfer and
communicative need.

Ungrammatical input

1
1

,,:

ll

Evidence that ungran1matical input has a direct effect on acquisition comes

from a srudy by Gass and Lakshmanan (1991). They reanalysed data from
rwo of the leamers investigated by Cazden et al. (1975)-Alberto and
Cheo-and found a striking correlation between the presence of subjectless
utterances in the input and in the producrion of the two learners. Again, the
correlations were derived from measures of input and output based on data
collected at the same time, but th.is study is more interesting than those reviewed in Ta ble 7.5 beca use of its longitudinal design. Gass and Lakshmanan
\vere able to show a strong and consistent relarionship berween input and
output o ver time (i.e. the presence of subjectless sentences in the input closely
matched those in the learners' output from one time to the next). They were
also ablc to demonstrate that in the case of Alberto deviant input in a grammaticality judgcrnent test was foliowed by a jump in this learner's production
of subjecdess sentences shortly a.frerwards. Ir should be noted, however, thar
the leamers' L1 (Spansh) permits subjectl;ss sentences. lt is possible, therefore, that input and transfer work jointly to shape interlanguage development. As Gass and Lakshrnanan pt it, 'the leamer initially searches for
corresp_9ndences or marches in form between the native and the second
language' .10
Formulaic speech
The prevalence of formulaic speech in leamer language was commented on in
Chapter 3. How do learners learn formulas? One possibility is that rhey respond to the high frequency of certain patterns and routines in the input. In
Ellis 1984b, for instance, l suggested that the formulas leamt by the rhree
classroo.tn learners investgated reflected frequently occurring social and organizational contexts that arose during the course of communicating in a
classroom environment. The prevalence of formulaic copula utterances with
subjects such as 'I', 'this', 'that', and 'what', which a number of researchers
have noted in leamer language (Hakuta 1974; Rescorla and Okuda 1987;
Ellis 1988a), may also reflect input frequency. However, although it seems
reasonable to assume that frequency is a major determinant of the acquisition of for1nulas, there is no research that clearly demonstrates that this is
so. Also, other factors are likely to be imporrant-in particular, learners'

162

Comprehensible input and second language acquisition

A number of researchers see comprehensible input as a major causative factor
in L2 acqusition. The most influential theoretical positions are rhose advanced by Krashen and Long.
Krashen's Input Hypothesis (Krashen 1981; 1985; 1989) makes the following claims:
1 Learners progress along rhe natural order by understanding input that
contains structures a little bit beyond their current level of competence.
2 :'-lthough c?mprehenslble input is necessary for acquisition to take place, it
is not suffic1ent, as learners also need to be affectively disposed to 'let in' the
input they comprehend.
3 Input becomes comprehensible as a result of simplification and with the
help of contextual and exrralinguistic clues; 'fine-tuning' {i.e. ensuring that
learners receive input rich in the specific linguistic property they are due ro
acquire next) is not necessary.
4 Speaking is the result of acquisition, not its cause; leamer production does
not contribute directly to acquisition. 11
Long differs from Krashen primarily with regard to (3 ). While acknowledging that simplified input and context can play a role in making input com
prehensible, Long stresses the importance of the interactional modi:fications
that occur in negotiating meaning when a communication problem arises. In
other words, Long's argwnent is that interactive input is more important than
non-interactive input.
Tbese rheoretical claims ha ve led to research into two aspects of input: (1)
the relationship between input (both interactive and non-interactive) and
cornprehension, and (2) the relationship between linguistic/conversational
adjustments and acquisition. To date, there has been far more research directed at ( 1), mainly because of the difficulties of designing studies to investigare (2). In recognition of these difficulties, Long (1985a: 378) suggested the
followmg three steps as a way of gaining insight into how input/interaction
affects acquisition:
Step 1: Show that (a) linguistidconversational adjustments promote (b)
comprehension of input.
Step 2: Show that (b) comprehensible input promotes (e) acquisition.

274 Externa/ factors

Input and interaction and second language ticq:tisition

Step 3: Deduce that (a) linguisticlconversational adjusnnents prornote (e)

acquisition.
Young (1988b) poinrs out rhat although a nurnber of studies have dernonstrated step (1) there ha ve been few which have tackled (2), which is, in fact,
the crucial step.
1 The relationship berween inreractive and non-interacrive input and
cornprehension

What evidence s there that linguistic/conversationaI adjusrments prornore

cornprehension of input, and do sorne kinds of adjustments work better rhan
others? To answer rhese questions, it is helpful ro separare out research which
has investigared non-interactive input frorn that inro interactive input.
One rype of non-interactive adjustment concerns speech rate. A number of
studies (for example, Long 1985a; Kelch 1985; Mannon 1986; Fujimoro,
Lubin, Sasaki, and Long 1986) provide evider:ice to suggest that a slower rate
aids comprehension, but in man y cases speech rate was investiga red alongside
other variables, rnaking ir difficult to assess the effect of speech rate per se.
Other methodological problems also make interpretation of the resulrs difficult. For exarnple, Kelch (1985) clairned to present dictation passages at two
different speeds (191 and 124 words per minute) but in fact allowed 45
seconds between each sentence dictated, thus introducing anorher variable
(pause duration) into rhe study.
However, two other, better-designed studies indicare that rate of speech is
a significanr factor. Conrad (1989) asked 29 native speakers, 17 high-level
non-native speakers, and 11 medium-level non-native speakers to recaII five
time-compressed recordings of 16 English sentences. Each senrence was presenred five times at diffetent speeds: 450, 320, 253, 216, and 196 words per
minute. The native speakers managed nearly complete recall of all the sentences by the second trial (320 words per minute), but the non-native speakers of both proficiency leve Is experienced considerable difficulry even with the
slowest rate after the fifth hearing. Griffiths (1990) investigated the effects of
varying rhe speech rate in three 350-400 word passages on the comprehension of 15 lo\ver-intermediate level adult non-native speakers. The speeds
varied from 94-107 wpm (slow) to 143-54 (medium) and 191-206 (fast).
The subjects manifested signficantly reduced comprehenson, measured by
means of rrue/false quesrions, ar fast rates, but there was no difference between the 'normal' and 'slo,v' rates. These studies suggest rhat there may be a
threshold level - around 200 wpm - below which intermediare and advanced
level learners experience little difficulty in comprehending and above which
they might.
Of particular interest to researchers is rhe narure of the modifications that
are most effective in promoring comprehension. Parker and Chaudron
/1 QQ7\ ,.,...,;,,..,., 1 J ,...,.,...,,.,.;""'"""'""I c..,.,.l;.,,.. ;..,,..J,,,-l;n<'T T,...J..neAn 1 QQ1 JU,.,., 1QQJ,

163

275

Cervantes 1983; R. Brown 1987; Chaudron and Richards 1986) of the effects
of input modificarions on comprehension and condude thar airhough linguisric modifications (for examp!e, sirnpler syntax and vocabulary) helped
comprehension they did not do so consistent!y. In contr::ist. \vhar they call
'elaborarive modifications' hada consistent effect on comprehension. They
disringuish rwo types of elaborative modifications: those rhat con tribute to
redundancy (such as reperirion of consriruents 1 paraphrase 1 use of synonyms,
use of lefr dislocation, and s!ovver speech), and rbose thar he!p ro make rhe
rhematic structure explicit (such as exrraposition and cleft constructions).
However, Parker and Chaudron's own srudy of 43 undergraduate and
graduare ESL learners, \vho read one of two reading passages \vtch differed
according to whether rhey had been modied by incorpornting linguisric or
elaborative modifications, failed ro find any significant differences in comprehension. They explain rhis unexpected result by suggesting rhar rhe overall
lexical and syntactic difficulry of both passages may have been so great as to
negare the effects of rhe elaborative modificarions. If rhis is righr, it suggests
rhat elaborative modificarions will only benefit comprehension if rhe level of
linguistic difficulty of rhe input does not exceed a certain threshold.
lssidorides and Hulstijn (1992) also failed to find that linguistic modifications had any effect on comprehension. This study examined the effects of
simplifying Dutch AdvVSO senrences by replacing them with rhe ungrammatical AdvSVO and AdvSOV senrences. These 'simp!ificarions' were morivated by the fact that learners of L2 Dutch and L2 German ha ve been found to
experience production difficulries \Vith verb-subjecr inversion after initial adverbs (see Chaprer 3). In an experimental study thar required three groups of
subjects (Dutch na ti ve speakers, English and T urkish learners of L2 Durch) to
stare \vhar the subject or agent of a series of senten ces \vas, AdvVSO senrences
were found to be no more difficult ro understand rhan modifed, nngrammarical senrences (providing thatthe messages conveyed bythe sentences were semantically plausible). lssidorides and Hulstijn condude that 'the fact that
non-native speakers have difficultes in producing a cerra.in gr::imrnatical
srructure ... does not imp!y that such a structure is also more difficult ro understand in the speech of others' (1992: 167).
Another quesrion of inrerest concerns the role of the learners' Ll in L2
comprehension. Holobrow, Lambert, and Sayegh (1984) found that 10-yearold Canadian children comprehended betrer when they heard an oral texr in
i::heir Ll andar the same rime read an L2 version of rhe text rhan \vhen they
simultaneously heard and read rexts in rhe L2 or when they just read L2 rexts.
Ho'\vever, when Hawkins (1988) replicared rhis study with British undergraduate students of L2 French, he found that the bimodal condition (i.e. listening to and reading texts in the L2) worked best. No conclusions are
possible on the basis of these studies~ bur they raise rhe issue of the role of rhe
Ll in helping to make input comprehensble.

i1

1
!

In the case of interactive discourse, nvo variables ha ve been found to influence comprehension: che amount and type of information, and the extent to
which the participants engage in negotiation of meaning. Derwing (1989)
founJ that sorne native speakers attempted to include much more information than they used with native speaker interlocutors about a film they had
seen, vith the result that their non-native speaker interlocutors failed to understand their narrarives. The native speakers who were successful in communicating did not differ in the knds of information {crucial, major, and
nnor) that they used when talking to na ti ve speakers and non-native speakers. This study suggests that redundancy involving increased use of background detail is not helpful to cornprehension, although, as Derwing notes,
this does not mean that other kinds of redundancy (such as repetition) are not
helpfu!. Ehrlich, Avery, and Yorio (1989) also report that their 'embroiderers' created problems for the non-native speakers by making it difficult for rhem to identify essential information and rhe source of
communication problems.
Comprehension also appears to benefit from opportunities for negotiation
of meaning. Pica, Young, and Doughty (1987) compared the effect of three
rypes of input on the ability of sixteen lo\.v-intermediate ESL learners to com~
prehend oral instructions. The three rypes were (1) unmodified input (i.e. input of the kind that native speakers use \vhen addressing each other), (2)
premodified input (i.e. input that had been simplified and made more redundant), and (3) interactionally modified input (i.e. the subjects listened to
unrnodified instructions but were given the opportunity to seek clarification).
The results showed that (3) resulred in the highest levels of comprehension.
Subsequent studies by Loschky (1989), Tanaka (1991), and Yamazaki
(1991) confirm these results. However, it should be noted that in all these studies rhe opportunity for negot.iation led to considerable reperition and
rephrasings with the result that (3) provided the learners with much more input than was avaihble in ( 1) and (2). lt is notclear, therefore, whetherthe advamage found for the nteractionally modified input arises from greater
quantity of input or better quality (i.e. input made relevant through the negotiation of meaning). A subsequent study by Pica (1992) found that when the
extent of the input adjustments was carefully conrrolled in the premodified
and interactionally modified conditions no difference in comprehension was
found, u although a post-hoc analysis suggesred that those learners rated as
ha ving lo\ver comprehension ability by their teachers benefited from the opportunity to-interact. Pica's study also found no significant differences in the
comprehension of learners who observed negotiatiOn taking place but did not
actu.ally negotiate themselvesr and those who actively participated.
To sum up, rhere is nlixed evidence regarding che value of linguistically
simplified input for promoting comprehension. Whereas speech rate does
have a clear effect, grammatical modifications do not always result in improved cornprehension. Firmer support exists for the beneficia! effect of inter-

164

actionally modified input on comprehension. We should note, however, that

the presence of interactional modification is no guarantee that comprehen~
sion has taken place. As both Aston (1986) and Hawkins (1985) point out,
learners may choose to feign comprehension after negotiation rather than
continue to demonstrate to their interlocurors that they ha ve not understood.
We will now consider to what extent comprehensible input prometes acquisition (srep (2) in Long's suggested procedure).

Source of evidence

Brief explanation

Caretaker speech

Caretaker speech ta young children is roughty tuned

to the children's receptive abilities and is motivated
by the need to aid camprehension.

Foreigner talk

Foreigner talk to NNSs is also roughly tuned and

functions asan aid to comprehension.

Silent period

Sorne young children go through a silent perlad in L2

acquisiton. During this period they do not produce
but nevertheless leam the L2.

Age difference

Krashen (1985: 12) argues that 'older leamers obtain

more comprehensible input' and that this may explain
why they leam more quickly initially than younger
leamers.

Comparative method studes

The studies show that methods that supply plenty of

comprehensible input {e.g. Total Physical Response)
are more successful than methods that supply Uttle
(e.g. audiolingualism).

lmmersion programmes

lmmersian programmes have genera!ly been found

superior to foreign/secand language programmes again because they supply p!enty of comprehensible
input.

Blingual programmes

The success of different kinds of bilingual

programmes is related to the extent to which they
suppfy comprehensible input.

Oe!ayed L1 and L2 acquisitian

Studies of children in both L 1 and l2 acquisition who

are deprived of comprehensible input (e.g. because
their parents are deat') show that acquisttion is
delayed or non-existent.

Reading and vocabulary acquisition

Studies indicate that children are able to increase

their L1 vocabulary and also develap a 'deep'
understanding of new wards through pleasure
reading.

!f>.

Reading and spelling acquisition

Studies indicate that spelling can be most effectvely

acquired through exposure to the written word in
extensiva reading undertaken far pleasure.

Table 7.6: Indirect evidence in support of the input hypothesis

2 78

Input and interaction and second languagc acquisition

Externa/ factors

279

arrenrion ro the linguistic forrns in the message. The quesrion that remains
unans\vered, then, is what type of comprehension processes are needed for
acquisition to take place. If learners can rely extensive!y on top-clown processing rhey may pay litt!e attention to the form of rhe input and may
therefore not acquire anything ne\V. Fxrch and Kasper :irgue that only when
there is a 'gap' between the input and rhe learner's current interlanguage and,
crucially, when the learner perceives rhe gap as a gap in kno\vledge, will acquisition rake place. Sharwood Smith (1986) argues similarly rhat the processes of comprehension and acquisirion are not the same, and suggests rhat
input has a 'dual relevance'-there s input that helps learners to interpret
meaning, and there is input that learners use to advance their interlanguages,
A second major criticism focuses on the claim rhat comprehensible input is
necessary for acquisirion. 13 White (1987a) has argued that a considerable
part of acquisirion is 'input-free'. As we will sce in Chapter 1O, she claims that
certain types of overgeneralizations which learners make cannot be unlearnt
simply by undersranding input. They require negative evidence (for exarnple,
in rhe form ofcorrecrive feedback) which in naturaHstic acquisition may not
be available to the learner. She also claims that learners are able to go beyond
the evidence available in the input, and develop knowledge of target-language
rules by projecting from their exisring knowledge {a point also discussed
more fully in Chapter 10). Finally, she argues that in the case of sorne strucrures (forexample, English passive constructionsL it may be the failure ro understand input that leads to learning. As she puts it 'the driving force for
gramrnar change is rhat input is incomprehensible~ rather than comprehensible .. .' (1987a: 95). White's idea is that failure to understand a sentence may
force the learner to pay closer attention to its syntactcal properties in order to
obtain clues about irs meaning and, as such, reflecrs the views of Frerch and
Kasper (1986) and Sharwood Smith ( 1986) about comprehension and learning discussed above.
These criticisms can be accommodated if rhe hypothesis is modified in rhe
following way:

2 The relarionship berween comprehensible input and acquisirion

The evidence is primarily indirecr. Together Long (1983b), Larsen-Freeman
and Long (1991), and Krashen (1985; 1989) cire literature relating ro ten
sources thar they claim supporr the claim that comprehensible input contributes ro acquisirion. These sources are described in Table 7.6.
Overall, rhis evidence suppon:s the commonsense assumption that learners
need to understand input in arder to learn from it. Ir is less clear, however,
wherher this evidence warrants rhe claim that comprehensible input is a causarive factor in L2 acquisition. A careful examinarion of each source of evidence reveals various doubts. In many cases, the evidence is correlational in
nature: for example, caretaker and foreigner talk co-occur with successful acquisition; extensive reading is associated wirh vocabulary developme~t and
improved spelling; older learners learn faster and engage in more mean1ng.negoriation. Therefore, ir cannot be said to demon~rr_ate that comprehen:1b.le
input causes acquisition. In other cases, the vahd1ty of the ev1dence is 1n
doubt. For example, rhe comparative method studies have problematic designs and have produced very mixed resu1ts; and many learners do t.ot go
through an extended silent period (see Chapter 3). In yet other cases, t~e
available evidence lends less than complete support (for example, learners in
lona-rerm immersion programmes do not develop full target-language proficie:cy). Long (1983b) considers the fact that deprivation of comprehensible
input resulrs n delayed acquisition the strongest evidence, _b~t LarsenFreeman (1983) argues quite convincingly rhat learners can ass1m1lare useful
information about an L2 without undersranding input (for example, information relating to the phonology of an L2). She also makes the point that
learners can work bv rhernselves on unmodified input-such as rhat found in
TV programmes-~nd so gain input that helps rhem learn. Thus, at least
sorne Ll learning may take place wirhour comprehensible input and, more
importantly, comprehending input may not result in acquisition.
There ha ve been a number of critiques of the input hypothesis (Gregg 1984;
Fa:rch and Kasper 1986a; Sharwood Smith 1986; White 1987a; Gass 1988;
Ellis 1990a and 1991a). Gass (1988) makes the point that it is not comprehensible input but rather comprehended input that is important, but the research reported above does not provide an account of how comprehending
input is supposed to result in acquisition. In fact, little attenrion has been paid
to what 'comprehension' entails. Rost defines listening comprehension as 'essentially an inferenrial process based on the perceprion of cues rather than
straightfor\vard matching of sound to meaning' {1990: 33). This suggests
that understanding does not necessirare clase attenrion to linguistic form.
Frerch and Kasper (1986) also recognize the impottance of 'top-down processes' in which learners utilize contextual information and existing know' understand what is said, but they also point out that they may
ledge to
sometimes make use of 'bottomwup processes', where rhey pay closer

Comprehensible input can facilita te acquisition but (1) is not a necessary

condition of acquisition, and (2) does not guarantee that acquisition will
take place.
The question arises as to wherher such a modified hypothesis has much
explanatory value. Arguably, what is needed is a hypothesis rhat explains
how leamers' attempts to comprehend input can lead to acquisition.

3 The relationship between inpur/interactional modifications and

acquisition
Finally, we will consider the few studies \vhich have attempted to estab!ish a
more direct relationship between input/interactional modifications and acquisition. Pica (1992) illustrates how the negotiarion of meaning provides

165

learners with information about the semantic and structural properties of the
target language (English). Native speakers (NS) responded to non-native
speakers' (NNS) triggers by modifying their utterances semantically and/or
formally through the segmentarion and rnovemenr of input constiruents, as in
rhis exarnple:
NNS: Ok, you ha ve a house which has third floor.
NS: Three floors right.
NNS: Tb.ree lloors.
However, as Pica admits, it is not clear from such data whether learners use
the inforn1ation supplied by such exchanges to adjust their interlanguage
systems.
A nun1ber of experimental studies, however, ha ve been carried out to disco ver whether negotiation leads to interlanguage development. Alrhough the
results of t..1-iese studies have been mixed, they give sorne support to the claim
that FT modifications help acquisition, at least where vocabulary is concerned. The studies are summarized in Table 7.7.
These studies give support to i.wo general conclusions. First, Li's (1989)
study suggests that providing Jearners with contextual cues that help them to
understand the meanngs of words results not only in better comprehension
but also in better retentionof the words. Second, Tanaka's (1991) and Yamazaki's (1991) studies indicare that providing Jearners with opportunities to
modify input interactionally improves comprehension and enables them to
learn more new words than simp!y providng them with unmodified or premodified input. These findings augur well for Long's claims regarding the importance of interactionally adjusted input. In contrast, Loschky's (1989)
study f;led to find any advantage for interactionally-modified input over
baseline or presimplifie<l input where vocabulary retention was concerned. Ir
should be noted, however, that this study, unlike the studies by T anaka and
Yamazaki, hada grammatical focus {locative particles in Japanese), so it is
possible that the leamers' attention was directed at this rather than at the new
vocabulary items. To date there has been no empirical test of the claim that
negotiation of meaning aids the acquisition of new grammatical features.

Main results

U (1989)

Advanced Ch in ese ESL leamers

exposed to sentences under two
conditions: (1) cue-adequate
(meaning made ctear by context)
and (2) cue-inadequate.

Cue-adequate sentences were

better understood and led to better
retention of difficult words.

Loschky (1989}

Three groups of leamers of

Japanese as a second language
were given lessons on locative
expressions under three different
conditlons: (1} baseline input, (2)
premodified input, and (3}
interactionally modified input.

Leamers receiving interactionally

modified input had higher
comprehension seores than leamers
receiving premodified input, who
had higher seores than learners
receiving baseline input. No
differences in the three groups far
retention of the new vocabulary
items introduced in the lessons.

Tanaka (1991)

Three groups of Japanese high

school ESL students were asked
to respond to instructions given
under three conditions: (1}
baseline input, (2) pre-modified
input, and (3) interactionally
modifled input. lnstructions
contained a numberof new
words. Pre-, post- and follow-up
vocabulary tests were
administered.

lnteractionally modified input

resultad in better comprehension of
the instructions. lt also resulted in
more words beng leamt and
retained over time.

Yamazak (1991)

Same designas for Tanaka's

study.

lnteractionally modified inpUt

resultad in better comprehension
and also in more words being leamt
{as evident in post-test seores), but
leaming advantage was not retained
over time{as evident in tollow-up
test scores). 14

Ta ble 7. 7:

hypothesis.

lnteraction also provides learners with the opportunity to talk in the U. According to Krashen (1985), this has no direct effect on acquisition. However,
other researchers have argued differently, viewing learner ourput as contributing to interlanguage development.
Following Krashen (1989), two different hypotheses that allocare a role to
output can be identified. The skill-building hypothesis srates that we first
learn rules or items consciously and then gradually automatize them through
practice. We will consider this when we examine cognitive theories of L2

Design

Studies investigating the direct effects of inputlinteraction modifications

on L2 acquisition

acquisition in Chapter 9 and the utility of formal practice in Chapter 14. Here
we wtll focus on the second hypothesis Krashen considers the output

Learner output and acquisition

Study

166

'

Ths ~o1!1es in tw~ forms, ~ccording to Krashen. First, there is 'output plus
correct1on . Accor~1ng to this, leamers rry out rules or items in productioq
~d rhen use the corrections they receive from other speakers to confirm or
d1sconfir~ them. Schachter (1986b) points out that metalinguistic informat1on relat1ng to the correctness of learners' production is available both directly (tJi:ough corrections) and indirectly (through confirmation checks,
clar1ficanon requests, and failure to understand).

282

11
"
1

1\l
I

11

Input and interaction and second language acquisition 283

Externa/ fcr.ctors

The second form of rhe ourpur hyporhesis involves the idea of comprehensible olltput. S\vain (1985) argues rhat learners need th~ opportuniry for me_aningful use of their linguisric resources to a~h1eve full g~am_manc.al
comperence. She argues rhar when learners expenence.commun1canve fa1lure, rhey are pushed inro making rheir ourpur more precise, coherent, and appropriare. She also argues rhar production may encourage tea~ners to move
from semanric (top-do\.vn) ro syntacric (bottom-up) proc~ss1ng. ~hereas
comprehension of a message can rake place wirh little synracnc ?alys1s o~ the
input, producrion forces leamers to pay atten_rion ro rhe ~e~ns ot expre~s1on.
The evidence indicaring rhat comprehensible output is 1mporran_r tor acquisition is largelv indirect. A number of studies (Harley and Swa1n 1978;
Harley 1988; Harey, Allen, Cummins, and Swain 1990) han shown rhat alrhough immersion learners achieve con:iderable confidence in us1ng rhe L2
and considerable discourse skills, rhey fati ro develop more m~rked grammatical disrincrions, such as that between French pass corn~ose and 1mparta1t,
and full sociolinguisric competence. Swain argues that th1s be cannot be ~x
plained by a Iack of comprehensible input, as immersion classr?o?1s are rtch
in rhis. She speculares rhar ir rnighr be because rhe learners had hm1ted oppo_rruniry to ralk in rhe classroom and were n~t "pushed' in rhe ourput they d1d
produce. Allen, Swain, Harley, and Cummms (1990) conducted an observationa1 studv of inreracrion in French 1mmers1on classrooms in nme grad~ 3
and ren grade 6 classrooms and found rhat the students' respo~ses were typ1cally 'minimaI'. Less rhan 15 per cent of srude_nrs' ut~erances in French we~e
more rhan a clause in length. Also, rhere was little ev1dence of any systemauc
correction of rhe errors learners made in rheir outpur. Only 19 per cent of
overall grammatical errors were corrected and of~en i~ ~ co_nfusing and
unsystemaric \vay'. They conclude rhat "rhere was hrde ind1canon that studenrs were being pushed rowards a more coherenr and accurare use of the targerlanguage' (1990: 67).
.
. .
.
. .
Ir is imporrant ro recognize rhar Swa1n's cla1m is rhar.producnon ":1~l a1d
acquisirion only when the /earner is pushed. Thus plennful opportun1t1es ro
speak, such as rhose rhar Wes undoubredly had in Hawaii (see Schmid_r .1983
and Chaprer 6 of rhis book), will not in rhemselves guarantee acqu1s1t1on.
Learners like Wes who develop high levels of strategic competence may be
a ble ro communicate efficiendy without much grammarical competence, and
so may never experience rhe need ro improve their output in order to make
themselves undersrood.
Borh versio11s of the ourput hyporhesis attribure considerable import~nc;
to feedback~ both direct and indirect. In rhe case of 'ourput plus correct1on,
feedback is necessary ro supply learners wirh metalinguistic informarion,
while in the case of 'comprehensible ourput' it s necessary to push learners to
improve the accuracy of their production .in order to ma~e thernselves ~n
derstood. We have already seen that fore1gner talk prov1des learners w1th
little direct feedback 15 bur \.Vith plenry of indirecr feedback in rhe form of

167

clarificaron requesrs, confirmarion requesrs. ere. Ir \vould seem, rherefore,

that ir is indirecr feedback that is crucial. Three questions arise: (1) Does
indirect feedback actually lead to more accurate lcamer output? (2) Do
cerrain rypes of indirect feedback ~ork better than others in this respect? and
(3) Do learners acruaHy learn as a resulr of pushed ourpur?
A series of srudies by Pica ha ve addressed rhe firstr~vo questions. Pica (1988)
examined the inreractons ber,veen a natlve speaker and ten non-native
speakers of English in order to discover whether indirect feedback led to improved output. She found that rhe learners <lid produce more grammatcal
ourpur when the narive speakers requested confirmation or clarification, but
in less than half of rheir toral responses. In a subsequent study, however, Pica,
Holliday, Lewis, and Morgenthalcr (1989) found rhat rhe crucial factor was
the narure of the feedback signals the narive speaker provided. The learners
(10 Japanese adults) were much more likely to produce ourpur modifications
in response to clarification requests rhan to confirmarion requests and repetitions. This was beca use clarification requests consritute an 'open signa!', leaving ir up to the learners ho\v to resol ve the compr:,ehension prob!ems, whereas
confirmation requesrs and reperitions, vvhere the native speaker models what
rhe learner intended to mean, remove the need for improved output. Interestingly, rhis srudy also showed that man y of the modifications that learners
made were grammarical and not just semantic in nature.
Evidence rhat leamers improve rhe grammaticality of their utterances
\vhen pushed does not of course constirute evidence that acquisition takes
place. In an explorarory study involving just six learners (rhree experimental
and three control), Nobuyoshi and Ellis (1993) soughr to estab!ish whether
pushed output resuired in improved performance over time. In two-way information gap rasks, \vhich each learner performed \Vith the researcher, the
experimental subjects experienced 'focused meaning negoriation', where they
received a clarification request every time they produced an utterance conraining a past tense error. 16 The control subjects experienced 'unfocused
meaningnegotiarion' (i.e. negoriaron only took place 'vhen there was a genuine communicarion problem). Both sets of learners believed they were communicating rarher than pracrising Ianguage. Nobuyoshi repeated the
information-gap task with both groups one \.veek Iarer, this time wirhout a
special focus on pasr tense. Two out of three experimenta11earners improved
the accuracy of rheir use of rhe past rense as a result of the requests for darification and mainrained rhis improvernent one week Iater. The control subjecrs showed no improvement on either occasion. This study intriguingly
suggests thar pushing 1earners to produce more comprehensible ourput may
have a long-term effect, bur not necessarily for aH learners. 17
There is also anorher sense in which learners may benefit from being
'pushed'. The studies reported above examined the effect of clarification requests on the subsequent accuracy of learners' output. Sorne learners, however, may have achieved a high JeveI of accuracy but only with a restricted set

of 'sirnpfe' constructions. Such learners need to experience siruarions that call

for more complex output, as, for example, when they are called on to demonsrrate rnastery of the forms required to communicate in a sociolinguistically
appropriate manner with different addressees (see Chapter 5).
So far, rhere is little hard evidcnce to support the output hypothesis, although there is sufficient to suggest thatit is worth pursuing. lt is possible that
\\.rhen learners produce pushed output they are a ble to develop grearer control
over the features they ha ve already acquired, as demonstrated in Nobuyoshi
and Ellis's study. It is not clear wherher pushed output can result in rhe acquisition of new linguisric features.

. Vertical constructions of rhe kind reported by Scollon ( 1976) in L1 acquisitlon ha ve. also been fou_nd i_n L2 acquisition. Hatch (1978b: 407) provides rhis
example tn an 1nteracnon tnvolving Paul, Huang's ( 1970) subject:
Paul: Oh-oh!
]:
What?
Paul: This (poinrs aran ant)
J:
It's an anr.
Paul: Anr.
The construction that Paul builds overthree utterances is 'Oh-oh, this ant'.

. Of cours.e, these studies are notable to show that 'scaffolding' of rhe kind
illusrrat~d in thes~ examples results in acquisition as opposed to just aiding
product10n. In EllJS (1985d), however, 1 found evidence to suggesr that this

Coilaborative discourse and second language acquisition

We saw thar Ll researchers have produced evidence that suggests that children may acquire syntactical structures as a result of learning how to participare in conversarions. Similar claims have been made by L2 researchers.
Hatch, f~r instance, proposes:

m1ghr tndeed be the case. lnteractions between two learners and a reacher
were examined in order to idenrify learner urrerances which featured 'new'
syntactic patterns (for example, the first occurrence of rwo constituent utterances consisting of noun + noun). I then studied what was going on interactionally that might havo helped the iearner to produce rhem. Ir was found
thar 'new' structures emerged when rhe learners were allowed to initiate a
topic (if necessary by the teacher abandoning his own topic) and when rhe
teacher helped the learners by supplying crucial chunks of language at the
r1ght moment.

One learns how to do conversations, one learns how to interact verbally,

and out of this interaction synractic strnctutes are developed ( 1978b: 404 ).
Researchers ha ve documented a number of ways in which this might happen.
One way is through rhe use of an incorporation strategy. Wagner-Gough
(1975) investigared the conversations which an Iranian boy, Homer, had
w irh native speakers and showed how he put rogether utterances by 'borrowing' a chunk from the preceding discourse and then extending ir by affixing an
elemenc to the beginning or end, as .in these examples:

Collaborative discourse may also work against acquisition, however. A

numbe~ of studies suggest that learners can sometimes exploit discourse

1 Mark: Come here.

Homer: No come here.
Where's Mark?
2 Judy:
Ho1ner: Where's Mark is school. (= Mark is at school.)

~-

This strategy may explain why 'no+ verb' constructions are so common in
early L2 acquisition (see Chapter 3), although other explanations also exist.
Peck 1978) gives examples of a related but siighrlydifferentstrategywhich
she calls 'functions' (a term borrowedfrom Keenan's (1974) study ofLl acquisition). These occurred frequently in child-chld discourse involving
Angel, a 7-year-old iv1exican boy. Functions involve repeating with or without so1ne modification part or all of the previous utterance. There are different types. A substitution function', for example, involves replacing a
constitucnt in the previous utterance, as in this example:
Joe:
You know what?
Angel: You know why?

strat~g1es to compensare for morphological features they have not acquired.

Stud1es of L2 German by Dittmar (1981), Klein (1981), and Meisel (1987)
and of L2 French byTrevise and Porquier (1986) show how learners are able
to refer ro_ time and space without recourse to target-language inflections.
!hey rely 1nstead o~ adverbials, implicit reference, order of menrion, and
mterlocutor scaffoldrng. This research relates to the question of how learners
mo:e from the pr~gmati~ to the grammatical mode (how they grammaticalize
their speech) and IS cons1dered in greater derail in Chaprer 9. Whar is important here ts to note rhat collaborative discourse may be so successful rhat it
inhibits grammatical acquisition.
It would seem, then, rhar collaborative discourse can both facilitare the ac~
quisition of L2 syntax and restrict it. The conrradicrion may be more apparrea~, however. Where basic synrax is concerned, ir appears ro help,
ent
part1cularly in the early srages of acquisition. Where morphological features
such as tense markers are concerned, ir may obviare rhe need for
development.

r:han

Peck suggests that this kind of discourse enables learners to practise syntax
and pronunciation.

'

168

286

Externa/ factors

input ana 111teract1on and second language acquisition

Summary

is evidellCe ro show rh;:_t indirect feedback in the forrn of cL1rification requesrs

pushes learners to irnprove rheirourput, bur lirt!e evidence as ver ro show that
rhis rhen leads to acquisition.
,

In this section we have looked at sorne of rhe research thar has investigared rhe
relarionship benveen inputlinreracrion and acquisition.
First language acquisirion research has shown rhat the frequency of linguisric forms in carerakers' speech is associated with developmenral gains in
children, particularly vvhere fragile features (forexample, auxiliary verbs) are
concerned. Ho\vever, theclaim rhat input frequency determines acquisition is
conrroversial. ivlore imporrant for rapid language dcvelopment are rhe inter~
acciona! features of careraker talk {for example, direcrives, expansions, 'collaborative meaning making', and the negotiarion of misunderstandings).
The relarionship bet\veen inputlinteraction and second language acquisirion has been examined in four major \vays:

4 Collaborative discourse ami acquisition

lllustrative evidence has been provided to sho,,v rhar the> joint efforts of na ti ve
s?ea~ers and non-native speakers ro consrruct discourse promores acquisinon In a number of \.vays: for example, rhrough rhe use of an incorporation
srrategy (the construction of utterances by borrowing from and exrending
elern~nts ~rom the preceding discourse)~ functions (repeating \.Vith or without
mod1ficanon sorne previous utterance), and vertical consrrucrions {building
up an urrerance over severa! turns). However, co[L.1bor:itive discourse may
remove the need for learners to acquire certain morpho!ogicaf fearures.

1 1nput frequency-output accuracy studies

These have produced mixed results and suffer from methodological problems, but the positive correlations berween the frequency n input of specific
grammatical rnorphemes reported by a nurnber of studies ha ve received confirmation from studies of the effects of ungrammarical input on learner
outpur.

Couclusion
Young (1988b: 128) concludes his own survey of input and interaction research with the comrnent that 'there is still a great deal of beating about the
bush', by which he me:1ns thar relarively fe\v srudies ha ve rried to investiga te
to what extent and in whar \vays input and interacrion influence acquisition.
One reason for this is the merhodological proh!erns that input researchers
fac:. In Ll acquisition research ir is not clear how porentlally confounding
variables such as the child's age and srage of development csn beconrrolled so
as.ro invesriga:e the effecrs of input on acquisition. In L2 acquisirion research,
rh1s problern 1s exacerbated by the difficu!ty of obraining represenrarive
samples of the inputthar individual !earners receive, for -.,,vhere the Ll leamer
is siruationally restricted to interacring with a fe,v identifiable interlocutors
rhe L2 Iearner, especia!ly vvhen an adulr, is not. Second, whereas a number 0
input studies on L1 acquisirion ha ve been longitudinal, nearlv a1I those into
L2 acqulsition have been cross-sectional. Third, there has been an over~eli~nc~ on correlariona! studies at the expense of experimental srudies, maktng 1~ d1fficult ro d~termine wh<lt is cause and \vhat is effect. Progress requires
srud1e_s rhat (1) rehably sample rhe input data, (2) are longitudinal, and (3) are
experimental.
Des~ite these n:ethodo~ogical problems, a number of insghts regarding the
role of '?put an? 1nteract1?n h~ve been gained. We have sorne undersranding
of how inte~act1onal rnod1ficanons affect rhe cornprehensibility of rexrs. It is
also becanung apparent that different kinds of input and interaction are
needed t~ facilitare _acquisi:ion at different stages of learners' development,
and that input and tnteracnon mayor rnay not affect acquisirion depending
on the narure of rhe linguistic fearure involved {consider the distinction
between re~il~ent and f:agile features, for example). We are still a long way
from expla1n1ng how input interacrs with the learner's interna! cognitive

2 Comprehensible input studies

287

The input hypothesis and Long's claims about the role of interaction have
motivated a substantial amount of research, although rhey have also been
subject to considerable criticism on rhe grounds rhat rhey fail to specify how
comprehending input leads to acquisirion and thar they overstate rhe imporrance ofComprehensible input by claiming thar it is necessary for acquisition.
.Nluch of the research has concentra red on denrifyng the features that help to
make non-inreracrive and inreractive input comprehensible. In the case of
non-interactive input, speech rate, elaborarive modificarions,_and birnodal input (i.e. borh oral and written) have been found to aid comprehension. In the
case of interactive input, the amount of information and the extent ro which
meaning is negotiated through interactional adjusrments have been shown to
be significant variables. There is subs-canrial indirect evidence linking comprehensible input to acquisition, but much of ir is controversial. There is lirrle in
the way of direcr evidence and what rhere is relates ro the acquisition of
vocabulary rather rhan syntax or morphology.

3 Learner output and acquisition

Swain has argued rhat comprehensible ourput conrribures ro acquisirion in
that learners need ro be pushed into producing output thar is concise, coherent, and appropriate in order to develop full gramma.tical competence. There

169

mechanisms to shape the course of language acquisition, and even. furrher

from being able to assign any weighting to externa! as opposed ro interna!
factors. In all Iikelihood, input combines with other facrors such as rhe
learner's Ll, rhe learner's communicarive need to express certain meanings
and rhe Iearner's inrernal processing mechanis1ns. No explanation of L2
acquisition wiH be complete unless ir includes an account of the role of in~ut,
but, as Sharwood Smith (1985: 402) has noted input should be seen as JUSt
one in a 'conspiracy of factors'.

Notes

\
!

\
''

1 The distinction between 'input rext' and 'input discourse' has been borro\.ve<l from Widdowson's (1974) rnore general distincrion berween 'text'
and 'discourse'.
2 In Ellis 1988a, I also made the point rhat much of native-speaker usage is
variable (forexample, English copula has a ful! and conrracred form) and
the learner's target, therefore, is ro achieve the same pattern of variabiliry
as that found in native-speaker usage. However, much of SLA research
has been based on the assumption rhat rhe 1earner's target consists solely
of categorical rules (such as the 3rd person -s rule in English) that specify
the conditions requiring the obligatory use of a single linguisric form.
3 Snow (1986) makes the poinr rhat n order to consider ro what ext~nt
there is 'fine-tuning' ir is cssentiaI to examine correlarions berween ch1ld
variables and complexity of caretaker talk ar the right srage of development. She also discusses a number of orher factors that may attenuate the
reJationshp (for examp!e, fine-tuning may be situation-specific).
.
4 The claim that children receive no correcrive feedback directed at the1r
irammatical errors is based on only one srudy, Brown and 1:~n1on
(1970). As Snow (1986) notes, this study used a very narrow definmon of
corrective feedback (i.e. explicit corrections). Children may be able t? ~se
other clues in the caretaker's speech, such as caretaker responses that mdicate the need for negotiation.
5 It should be nored, however, thar in sorne cultures adulr carerakers do
seem to give something resembling 'language lessons'. Ochs and Schieffelin (1984) give examples ofhow Kaluli mothers in Papua-New Gumea_require rheir children ro engage in in1itation exercises wirhout any attennon
to the meaning of utterances being imitated.
The
term 'foreigner talk' was iniriaily used by Ferguson (1971) to refer
6
solely to the kind' of ungrarrunatical ralk that native speakers sometimes
address to non-narive speakers. Arrhur et al. {1980) use the term 'foreigner regster' to refer ro grammatcal FT while the term 'foreigner talk
discourse' has been used ro refer ro interactionally modified Ff. Increasingly, though, 'foreigner talk' is used as a general cover term for the

!'

170

JO

J1

modified language native speakers use wirh non-native speakers, and this
is how the term is used in this chapter.
The distinction between discourse management and discourse repair of
communication breakdowns mirrors Long's (1983a) distinction berween
"strategies' for avoiding trouble and 'tactics' for dea1ing with trouble
when it occurs.
This problem is acknowledged by Larsen-Freeman and Long (1991 ), who
point out that the discourse moves that perform functions such as clarification requests often ha ve 'mulriple funcrions and also mu1tiple realizations' (1991: 129). The need for 'finer grained analyses', therefore, is
recognized by at least one researcher (Long) who has been actvely involved in researching rhe negotiation of meaning.
Goldin-Meadow (1982) distinguished resi/ient and fragile properties in
rhe sign language of deaf children. The former are those properties such as
recursion and word order thar appear to develop irrespective of environmental conditions, whereas the 1atter (forexample, plural and verb tense)
need special care.
Another purpose of Gass and Lakshmanan's study was to show the dangers of conducting studies based on descriptions of language which assume that the input to the Iearner is target-like. They comment:
'considering principies of UG, or any other principies, devoid of context is
insufficienr and often misleading in accounting for how L2 grammars
develop'.
Although Krashen claims that speaking does not contribute directly to acquisition, he does allow for an indirect contriburion. He writes:
... output aids acquisition indirectly by encouraging CI (comprehensible input), va conversation. When you speak it invites others to talk ro
you. Moreover, as you speak your output provides your conversacional
parmer with information about your competence and whether he or
she is communcaring successfully. This informarion helps your conversational partner adjust the input to make it more comprehensible.
(Krashen 1989: 456)

12 The premodified input was based on negoriated input and was designed
to incorpora te the main features of ir.
13 Although both Long and Krashen claim that comprehensible input is necessary for acquisition, both acknowledge that it is not sufficient. Learners not only need access to input they can understand; they also need to be
affectively disposed to '!et it in' (for example, to be motivated to learn).
14 The main difference in the results obtained by Tanaka and Y amazaki is
that whereas Tanaka's subjects in the interactionally modified input
group maintained their gains in vocabulary in the follow-up test, Yamazaki's subjects did not. The explanation for this difference may lie in the
attitudes of the subjects. Tanaka's subjects were not highly motivated to

290

Externa/ factors

Input and interaction and second langc;agc acquisition

291

'1'
)

learn English, whereas Yamazaki's were-largely because they needed to

do well in English to gain entry to the universities of their choice. Yamazaki was able to show that many of her subjects took efforrs to learn the
new vocabulary in their own time berween the post-test and the follow-up
test.
15 Formal correction does, of course, occur in classroom language leaming.
Chapter 14 reviews the research rhat has investigated the effect it has on
learning.
16 This kind of 'focused meaning negotiation' can be distinguished from a
more traditional grarnmar exercise by the fact that the learners treat the
activiry as one calling for 'use' rather than 'usage'; that is, they concentrare on conveying a message and not on perfonning a particular grammatical feature accurately.
17 Nobuyoshi and 1 noted that the one experimental learner who failed to
improve past tense accuracy was adept at making himself understood in
other ways. He manifested what Meisel, Clahsen, and Pienemann (1981)
refer to as a 'communicative orientation'.

Further Reading

)
'

There is a rich literature dealing with input/interacrion and L2 acquisition. A

good starting point is toread one or two surveys of this literarure:
R. Young, 'Input and interaction.' Annual Review of Applied Linguistics
(1989) 9: 122-34.
D. Larsen-Freeman, and M. Long, An Introduction to Second Language Acquis(tion Research, Chapter 5 (Longman, 1991).
The role of input and interaction in Ll acquisition is considered in:
C. Snow and C. Ferguson (eds.), Talking to Children: Language Input and
Acquisition (Cambridge University Press, 1977).
G. Wells, Language Development in the Pre-schoo/ Years (Cambridge University Press, 1985).
C. Snow, 'Conversations with children' in P. Fletcher and M. Garman (eds.),
Language Acquisition, 2nd edition (Cambridge University Press, 1986).
For descriptive studies of FT see:
M. Clyne (ed.), Special issue on 'Foreigner Talk' of Internationa/ journal of
Sociology of language (1981) 28.
D. Larsen-Freeman (ed.), Discourse Analysis in Second Language Research
(see articles by Vander Brook, Schlue and Campbell, Arthur et al., and Gaskill) (Newbury House, 1980).
S. Gass and C. Madden (eds.), InputinSecond Language Acquisition (see anicles by Kleifgen, Ellis, Wesche and Ready, Hirvonen, Gass and Varonis,
Long, and Lightbown and d' Anglejan) (Newbury House, 1985).
M. Long, 'Native-speaker/non native-speaker conversation and the negotiation of meaning.' Applied Linguistcs (1983) 4: 126-41.

171

For the effe<."tS of inputlinreraction on comprehension and acquisition see:

S. Krashen, The Input Hypothesis (Longman, 1985).
K. Parker and C. Chaudron, 'The effects of linguistic simplilications and
elaborarive modifications on L2 comprehension.' University of Hawaii
Working Papers in ESL (1987) 6: 107-33.
M. Long, 'Native-speaker/non-native-speaker conversation in the second language classroom' in M. Clarke and J. Handscombe (eds.), On TESOL '82:
Pacific Perspectives on Language Learning and Teaching (TESOL, 1983).
T. Pica, 'The textual outcomes of native speaker-non*native speaker negoria~
tion: What do they reveal about second language learning' in C. Kramsch and
S. McConnell-Ginet (eds.), Text and Context: Cross-Disciplnary Perspectives on Language Study (D. C. Heath and Company, 1992).
L. White, 'Against comprehensible input: the Input Hypothesis and the development of second-language competence.' Applied Lingutstics (1987) 8:
95-110.