ONNO CRASBORN

HI L Il EIDEN UNIV .

Articulatory symmetry in two-handed signs

Onno Crasbom

MA thesis, November 1995

Dept. of General Linguistics
University of Nijrnegen

Supervisors (University of Leiden):

Dr. H. van der Hulst
Dr. V. van Heuven
Oh, enchantante beaute organique qui ne se compose ni de teinture
a l'huile ni de pierre, mais de matiere vivante et corruptible, pLeine
du secret febrile de La vie et de la pourriture! Regarde la symmetrie
merveilleuse de I' edifice humain, les epauZes et les hanches et les
marnelons fleurissants de part et d 'autre sur la poitrine, et les cotes
arrangees par paires, et Ie nombril au milieu dans la rnollesse du
ventre, et Ie sexe obscur entre les cuisses!

Hans Castorp talking to Madame Chauchat about her bilateral symmetry, in

Der Zauberberg, Thomas Mann, 1924, Fischer Verlag. Berlin
Preface

One of the nice things about writing this MA thesis is that it allows me to thank a number
of people for their support. To start with, some people did preparatory work in getting me to
this thesis. Without the stimulating enthousiasm of prof. Kraak and Linda Uyechi it would
no doubt have taken me a lot longer to get excited about linguistics and sign phonology,
respectively. Fellow students Judith Haan and Juliette Waals kept the enthousiasm going; I
enjoyed our mutual support a lot. Wim Emmerik deserves a big thanks for teaching me some
initial skills in his language, NGT; without his help over the last two years this thesis would
not have been possible.

My first ideas about symmetry in two-handed signs grew this summer in Albuquerque, during
Diane Brentari's course at the 1995 Linguistic Institute and in discussions with Chris Miller,
Lorna Rozelle, Kelly Stack, and Linda Uyechi; thanks to all of them for the great time and
the first fruits I was able to steal, with both hands, from their minds. I thank Harry van der
Hulst and Vincent van Heuven for trusting me in this personal enterprise, even though I was
not always able to tell them where I was heading. Jean Ann, Ine Dinklo, Chris Miller, and
Vincent van Heuven carefully read early versions of various chapters; thanks to them for the
many good suggestions. I bear full responsability for all the errors and opaque parts that
remain, and will not even try to come up with the same old argument: there's just no time.

Finally, it has scarcely been comforting to Marianne that she knows what it is like to write
an MA thesis; she put up with my obsession, and lowe her a lot for that. Having written that
down, I will not be able to forget it.

Onno Crasborn
Amsterdam, November 12, 1995
Contents
1 Introduction ......... ... . . .... . .. . . . ..... .... ........... . ... .. ....... 5
1.1 Sign languages ................................................ 5
1.2 Sign phonology . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. 6
1.3 Conventions. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. 7
1.4 Outline of the thesis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. 8

2 Two-handed signs. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. 9
2.1 Early research ...... ........... .. .. . .. . . . .. ....... ............. 11
2.2 Other literature on two-handed signs ................................. 16
2.2.1 Other early research on two-handed signs ... . . . . . . . . . . . . . . . . . . . . .. 16
Friedman. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. 16
Frishberg . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. 17
2.2.2 Recent interpretations of Battison . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. 19
2.3 Remaining problems and questions .................................. 23
2.4 Summary: research questions ..... .. . ................... . .. . ..... . . 28

3 Symmetry ................... . ..... . . . ...................... . ....... 31

3.1 The mathematical sense of symmetry . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 31
3.2 The metaphorical sense of symmetry .. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. 36

4 Phonetic description of signs and their symmetry ... . . . . . . . . . . . . . . . . . . . . . . . . . . .. 39

4.1 Different types of phonetic description . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. 39
4.2 The joints of the arm and hand ... . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . .. 44
4.3 Description of signs and their symmetry in terms of joint angles . . . . . . . . . . . . .. 53

5 Symmetry breaking in sign language . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. 59

5.1 Symmetry breaking ............................................. 59
5.2 Symmetry breaking in gestures ..................................... 60
5.3 Symmetry breaking in NGT signs ................................... 69
5.3.1 Description of data set ...................................... 69
5.3.2 Further restriction of diversity . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. 70
5.3.3 Taxonomy of the asymmetric signs in the corpus. . . . . . . . . . . . . . . . . . .. 71
Shoulder joint ............................................ 71
Elbow joint . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. 72
Radioulnar joint . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. 73
Wrist joint. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. 75
5.3.4 Discussion of the data ...................................... 76
Distribution of symmetry types overall . . . . . . . . . . . . . . . . . . . . . . . . . .. 77
Distribution of symmetry types per joint pair ...................... 78
Combinations of symmetry types . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. 78
A note on the neglected joints . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. 81
5.3.5 Realization of symmetry types. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. 81
Movement or no movement . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. 81
Cooperation of joint pairs ................ . . . . . . . . . . . . . . . . . . .. 82
5.3.6 Summary .................................... . .......... 83
6 Theory and explanation .. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. 85
6.1 The two articulators in a phonological model ........ .. .......... ...... . 85
6.1 .1 The phonological model of van der Hulst . .... . . . . . . . . . . . . . . . . . . .. 85
6.1.2 The representation of asymmetries in the model ... .......... ... .... 89
Time asymmetries: translation ... . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. 91
Time asymmetries: reflection . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. 93
Symmetry types 1, 2 and 6 .... . .......... . ...... ... ........ .. 94
Frequent sign types: 1111,6666,3311,2211 ....................... 99
6.1.3 Movement of the weak hand and weak hand as a place of articulation . . .. 100
6.1.4 Heads. dependents, and manner features .... . . . . . . . . . . . . . . . . . . . .. 103
6.1.5 Final remarks ......... .. ................................ 106
6.2 Phonetics: articulatory complexity . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. 107
6.2.1 Symmetry and ease of articulation ... . ................ . . . . . . . .. 108
6.2.2 Phase differences . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. 109
6.2.3 Synchronization of movements in the two articulators ............... 110
6.2.4 Movement of only one articulator ............. . ............... 111
6.3 Ease of articulation versus ease of perception . . . . . . . . . . . . . . . . . . . . . . . . .. 112

7 Conclusions.. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. 115

References .......................................... . ............... 119

Appendix A: ranges of motion and reference axes for movements at each joint . ...... 124
Shoulder joint. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. 124
Elbow joint ...... . .... . ..... .......... .............. .... ......... 126
Radioulnar joint ..... . ..... . .. . ......................... .. ....... . . 127
Wrist joint . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. 127

Appendix B: corpus of all two-handed signs ....... . . . ...... ........ ... ... . 129

Appendix C: corpus of asymmetric signs . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. 132

Appendix D: corpus of completely symmetric signs .......................... 134

 5

1 Introduction

In this thesis, I will look at the internal structure of two-handed signs in sign languages. Much has
been written about this subject, but a number of questions are still unanswered. It has been clear from
the very first studies of American Sign Language on, that the symmetry of the body has an influence
on the signs that occur in the language. It has often been argued that symmetry facilitates both the
perception and the production of signs. However, the precise phonetic realization of signs has never
been the subject of research; rather, the investigations have focussed on the level of the more abstract
phonological representation, and investigated symmetry there. I will try in this thesis to make a
description of signs that is closer to the level of phonetics. Instead of maintaining the focus on the
symmetry of signs, research in the past decade has stayed very close to one particular classification
of signs that was the result of the initial focus on symmetry. In this thesis, I attempt to investigate
possible modifications of this classification. Another topic that has been left uninvestigated so far is
what exactly is easy for articulation. There have been studies demonstrating the way symmetry can
aid perception, but no such studies indicating why symmetry is easy in the production of signs. This,
too, will be one of the questions that this thesis aims to answer.

In this introductory chapter, I will first briefly discuss general aspects of sign languages and sign
phonological research. These sections are intended for the reader unfamiliar with sign languages and
sign language research, and contain references to introductory works in the literature. I will then
provide information on the sign language used in the Netherlands, from which most of the data I use
originate. Finally, I will outline the structure of the rest of the thesis, and the notational conventions
followed.

1.1 Sign languages

Sign languages are the primary languages for members of Deaf communities all around the world. In
this thesis, I will mainly talk about Nederlandse Gebarentaal (Dutch Sign Language or Sign Language
of the Netherlands, henceforth NGT) and American Sign Language (henceforth ASL). These are the
languages of the Dutch Deaf community and the majority of the North American Deaf communities,
respectively (the use of ASL is less widespread in the French speaking part of Canada). For general
information about the Dutch Deaf community and NGT, I refer to Schermer, Fortgens, Harder & de
Nobel (1991) and Koenen, Bloem & Jansen (1993). For the same kind of information on ASL and the
North American Deaf community, I refer to Klima & Bellugi (1979) and Baker-Shenk & Cokely
(1980).
6

In general, I will assume that ASL and NGT behave the same. In the literature on the phonology of
these two and other sign languages, no major typological differences have been found, such as the
difference between tone languages and pitch accent languages in spoken languages, for example. I will
explicitly note and discuss those few interesting cases where there is a difference between the two
languages in chapter 4. The corpus of data that will be examined in this thesis stems from NGT.

1.2 Sign phonology

It may seem strange at first sight to encounter something called 'phonology' in the study of a silent
language, where body motion (but not the possible sounds produced by that motion) is used to convey
meanings. The term 'phonology', after all, refers to the sounds in spoken languages (ancient Greek
'<!>covTj' meaning 'sound'), as does 'phonetics'. Still, excluding the initial studies by Stokoe (1960) and
West (1960), most sign linguists have talked about phonology and sometimes phonetics when talking
about sign language!. Initially, this probably was influenced by the fact that sign linguists argued
strongly that sign languages were 'real languages' , serving the same functions and showing the same
levels of structure as spoken languages. Later, the use of these somewhat awkward terms was
influenced by (apart from the growing convention) the increasing insight that there is no fundamental
difference between the two types of language, signed and spoken, on a structural level. "For sign
languages, a phonology systematically separates the set of gestures which may represent meanings in
a given sign language from the entire range of gestures which may be produced by the human body.
This involves constraints on underlying forms (morpheme structure conditions) and constraints on
surface variation, expressed by phonological rules", Battison (1974:2) writes. Likewise, sign phonetics
studies the way in which sign languages are produced and perceived and how this influences the
structure of these languages. I will follow this common usage here by talking about phonology and
phonetics with regard to sign language, although admittedly these terms are confusing to people
(including linguists) outside of sign language research.

It is generally assumed that signs can be characterized by describing four different aspects: the .
handshape that is involved, the orientation of the palm and fingers, the location of articulation, and
the movement that is performed (cf. Corina & Sandler, 1993). These aspects are called parameters and
the possible values for each parameter are called primes by many researchers. Although these four
parameters still are considered to be the main formational characteristics of a sign, other aspects have
also received attention as possible, separate parameters, aspects that are sometimes considered to be
part of one of the four main parameters in earlier studies. For example, Klima & Bellugi (1979)

I Stokoe (1960) proposed to use the tenns 'cherology' and 'chereme' to talk about sign phonology and sign phonemes,
deriving from classical Greek 'Xe:(P' meaning 'hand'. However, hardly anybody has accepted this proposal since.
 7

recognize two additional minor parameters, focus and hand arrangement. Focus refers to the place of
the hand that contacts a location, hand arrangement refers to the number of hands present in a sign
and whether both hands move or not. Also, many investigators have emphasized the role of the face
in signs as an aspect that should receive attention from phonologists (cf. Baker-Shenk, 1985; Coerts,
1992; Brennan, 1995). In this thesis, I will not look at the possible interaction of facial and manual
characteristics in (two-handed) signs, although it is potentially interesting (cf. Ebbinghaus &
Hessmann, 1995). There has been a lot of debate about the status of the parameters. For instance,
Stokoe (1960) claimed that the sign parameters were parallel in a sense to spoken language phonemes, .
which has been argued against later (Newkirk, 1981; Liddell & Johnson, 1989; Wilbur, 1990, 1993;
van der Hulst, 1995ab). A related area of heavy debate is the question about the extent to which sign
languages may be in large part simultaneously structured compared to the arguably predominant
sequential structure of spoken languages (cf. Stokoe, 1960; Newkirk, 1981; Liddell, 1984; Edmondson,
1986, 1990; Wilbur, 1990, 1993; van der Hulst, 1995ab).

For a more extensive introduction to sign phonology, I refer to Corina & Sandler (1993) and Sandler
(1995a). The phonology of NOT has been studied by Schermer et al. (1991), van der Kooij (1994),
Blees (1994), and Ballering (1995); I refer to these works for more specific information on NOT
phonology. As for sign phonetics, not much work has been done. The only two dissertations that are
fully devoted to sign phonetics so far are Wilcox (1992) and Ann (1993; see also Ann, 1991, 1992ab).
A lot of knowledge about the production and perception of arm and hand movements, although not
in the context of language, is no doubt present in other than linguistic departments, with psychologists
and physiologists. It remains a goal for sign phonologists and phoneticians in the near future to find
out about these studies and make them useful to sign linguistics. I will try to use some such studies
in this work when talking about motoric constraints on two-handed movements in chapter 5. As it is
my impression that previous work has underestimated the relevance of such phonetic types of
investigation, I will try to pay attention to it in this thesis.

1.3 Conventions

Throughout this thesis I will follow the general convention of representing signs by "glosses". A gloss
is a name for the sign, and is not necessarily a correct or full translation of the sign in, for example,
English. As I will not be concerned with the meaning but rather with the form of signs, I will not
bother to give translations of the signs. However, I will give a translation the Dutch glosses into
English after each gloss. As far as possible, I will give examples from both NGT and ASL. In this way
the points made will be more easily accessible to a larger audience than just those people that know
NOT, and thus it slightly relaxes the need for accompanying pictures, drawings, or notations of the
NGT examples. Ideally, there would be a short piece of video or maybe a picture accompanying every
8

gloss, but unfortunately for technical reasons and due to time limitations that is not possible here. An
alternative would be to make a transcription of every gloss. However, there is not yet any transcription
system that is widely used by the whole research community, like the IPA is for spoken languages.
Moreover, a transcription (just as much as glosses) would only be of use to those people that already
know a sign language to some extent, as reading from transcription requires a lot of interpretation that
can only be made by skilled users of the system. I apologize to all those that do not know any of the
signs involved, and for those cases in which it is unclear which variant of a sign is intended.

Signs from NGT will be represented by a Dutch gloss, signs from ASL will be represented by English
glosses. A gloss is a label for the sign, and does not necessarily cover all of its meaning. Signs from
other sign languages will be represented by English glosses, and it will clearly be indicated from which
sign language the sign stems. Following the general convention, glosses will be written in capitals (cf.
Baker-Shenk & Cokely, 1980). When more than one word is needed for a gloss, different words are
joined by hyphens. If signs with different meanings have the same form, thus in the case of
homonyms, different glosses are separated by a slash.

1.4 Outline of the thesis

In the rest of this thesis, I will discuss the symmetry that can be found in the articulation of two-
handed signs. First, in chapter 2 I will review the relevant literature and conclude that it is necessary
to undertake a more detailed study of symmetry than has been done so far. In chapter 3 I will present
the mathematical view on symmetry. In chapter 4, I will propose a phonetic description of signs in
terms of the joints of the arm and hand involved. In chapter 5, I will consider how symmetry can be
disturbed in NGT signs, using the type of description suggested in chapter 4. Then, in chapter 6 I will
evaluate the implications of these data on symmetry in NGT for phonological models, and I will try
to explain part of the patterns found in terms of general characteristics of human motor control.
Finally, in chapter 7 I will present conclusions of the research and propose questions to be answered
in future research.
 9

2 Two-handed signs

This thesis will study phonetic and phonological aspects of two-handed signs on the lexical level:
morphological or discourse processes, for example, will not be considered. Two-handed signs are
defined as lexical items in which both hands are present during the articulation, though it need not be
the case that both move; 'presence' is defined as being out of the rest position, which varies according
to the posture (standing, sitting) the signer takes. Several phonologists have defined two-handed signs
as signs in which both hands are moving with respect to each other, or alternatively, as all two-handed
signs where both hands are present excepting those signs where one of the two hands functions as a
place of articulation. Here, then, we will use a more phonetic definition, not excluding any group on
non-phonetic grounds2 •

There are various terms used to distinguish the two hands 3• The most common are 'left' and 'right',
which are used in everyday life to indicate which is the preferred hand for manual tasks like writing,
using a pair of scissors, and throwing balls. However, as Battison pointed out, these terms "are
superfluous, since they are arbitrary labels for sides unrelated to function, and because the opposition
of left and right is non-distinctive for signs" (1978:26). Indeed, in all sign language studied so far, it
does not make any difference whether a sign is performed with the left or the right hand, in the sense
that the signs produced mean exactly the same thing. Generally, signers whose preference hand is the
right hand will make all one-handed signs with the right hand, will use the right hand for
fingerspelling (in the case of a one-handed fingerspelling system), and use that same hand as the
moving hand in two-handed signs where one hand is held still. For left-handed signers, the reverse
situation occurs, without any risk of confusion. Thus, one of the two hands in sign languages has a
more active role. It is this functional difference that many terms refer to. The more active hand may
be referred to as 'dominant', 'strong', or 'superordinate'; the less active hand may be referred to as
'non-dominant', 'weak', or 'subordinate'. In those cases where the functional difference indeed
surfaces, the hands can be referred to with the terms 'active' and 'passive' (or 'base'), respectively.
Another pair of terms that is relevant in this context, which is used in specific cases to refer to places
of articulation, is 'ipsilateral' and 'contralateral'; the former means 'on the same side of the body as
the (shoulder of the) articulator', as opposed to 'on the opposite side of the body from the articulator'.

2 At the morphological and discourse levels, it frequently occurs that each hand articulates one separate morpheme. I
will not discuss this here, but I refer the reader to Baker-Shenk & Cokely (1980) and SupaUa (1986) for examples and
discussion.

J Actually, it would be preferable not to talk about the hands when we talk about the primary articulators of Sign
languages, but rather about articulators, as will be argued later: the upper and lower arm play an important role in sign
language articulation. The extent to which the arm plays a role in sign language structure is a different matter, and in
principle it may indeed be the case that only the hands (their form, attitude, and location) are relevant For now, I will stick
to the convention of talking about hands.
10

On many occasions, signers can switch the roles of the two hands, and make the non-preference hand,
which is usually the non-dominant one, the dominant one. This phenomenon is referred to by the term
'dominance reversal', and it can happen in a variety of circumstances (cf. Klima & Bellugi, 1979;
Frishberg, 1985). It can happen for discourse reasons, for reasons of convenience (where the preference
hand is occupied carrying something, or where it is almost invisible), and it happens in poetry.
Dominance reversal can also happen in lexical items (cf. Padden & Perlmutter, 1987). In the ASL
signs JESUS, CRUCIFY, GLOVES, and HANDS and in the NGT signs JEZUS 'Jesus', MARKT
'market', and HANDSCHOENEN 'gloves', the hands switch role halfway through the sign, and the
right hand performs the action the left hand has just done, or the other way round for right-handed
signers. This phenomenon will not be discussed further.

Whereas right versus left-hand dominance is not distinctive at all in sign languages, the use of one
versus two hands to articulate a sign may in some cases be distinctive. As the pairs of signs from
different sign languages in (1) show, minimal pairs occur where the only difference is the use of one
versus two hands4 •

(1) Minimal pairs of one-handed versus two-handed signs, respectively

NGT: DEKEN 'blanket' - KIEZEN 'to choose'
ASL: LIKE - INTERESTING, PARTY - PURPLE, NOTHING - FURNITURE, YELLOW -
PLAY, SEEM - COMPARE, SAME - MEASURE
LSQ: THURSDAY - PLAY, TOO-MUCH - PERMIT/ALLOW, COFFEE - COUSIN,
OCTOBER - UNCLE, TEA - UGLY, TWELVE - #APARTMENT, CHEAP - WHICH/OR,
IT'S-THAT - STAY, FIFTEEN - WALK, YESTERDAY - SKILFUL
SVK: READ - SCHOOL

In a few further cases from NGT, it is not completely clear whether they are real minimal pairs or near
minimal pairs (cf. Van der Hulst, 1995e). As Klima & Bellugi remark, "such minimal pairs are rare
in ASL; the use of one as opposed to two hands rarely distinguishes between semantically unrelated
lexical items, but it plays a major role in morphological processes" (1979:50). In the modulation for
predispositional aspect in ASL, for example, a one-handed sign becomes a two-handed sign with
repeated circling movement (cf. Klima & Bellugi, 1979:247-252). For discussion of this type of
process, see Klima & Bellugi (1979, chapter 12).

In the rest of this chapter, I will review the most important previous research on two-handed signs,
consider some of the problems that still remain, and pose research questions that this thesis will try

4 In the lists in (I), LSQ stands for Langue des Signes Quebecoise (Quebec Sign Language), and SVK stands for
Suomalainen Vittomakieli (Finnish Sign Language). Thanks to Chris Miller for the LSQ data, and to Lorna Rozelle for the
SVK data.
 11

to answer. The overview of previous research is mainly intended to highlight aspects of works that I
assume the reader already to be familiar with to a certain extent. Therefore, it may be that sections 2.1
and 2.2, reviewing the literature on the subject of two-handed signs, are difficult to follow for those
people unfamiliar with the general sign phonology literature. I suggest that they proceed to section 2.3
and further, and return to the review of the literature after reading the analyses proposed in chapters
3 to 5.

2.1 Early research

The earliest reference to two-handed signs that I found was in the studies of sign systems used in
communities of North American Plains Indians by Kroeber (1958) and West (1960)5.6. In a review
of the literature of the nineteenth and early twentieth century, Kroeber (1958) distinguished between
one-handed and two-handed signs, the former being more frequent. Two-handed signs were subdivided
in those in which both hands are stationary, in which one hand is moving, and in which both hands
are moving (either interacting / crossing or bilaterally symmetric simultaneous motion). Unfortunately,
there is no discussion devoted to the reasons for making just this division. West (1960) distinguishes
"symmetrical" from "parallel" motion for signs in which both hands move. In symmetrical motion "the
left hand bears the same positional relationship to the left side of the body that the right hand bears
to the right side, and so also for movements and directions of facing. Thus the left hand moves, faces
and locates as a mirror image of the right" (1960:39). In parallel motion, "both hands are positioned
and move in relation to the right side of the body" (1960:39). Although these are very clear
definitions, West's work, like that of Kroeber, has gone largely unnoticed in the sign language
literature on two-handed signs, where years later the same distinctions would be made.

The literature on the structure of sign language of Deaf communities started with the well-known study
of Stokoe (1960). In the transcription system that he proposed, an implicit distinction is made between
two categories of two-handed signs: those in which the location (to be specified in every sign) of
articulation is on the weak hand, and those in which that is not the case. Although there is no explicit
discussion, the location appears to be the weak hand if the weak hand does not move, is 'inactive'.
The sign is then articulated in "a neutral, unmarked position in front of the signer's body" (Stokoe,
1960:50; this position has since been called 'neutral space', cf. Stokoe et at., 1965), and the strong
hand performs an action with respect to the non-moving weak hand (like approaching, entering,

S The sign systems used by North American Indians are sometimes not considered sign languages as they are not used
as primary means of communication by anyone. nor are they learned as native language by any child; they generally are used
only in special social situations (see also Kendon, 1988).

6 Thanks to Chris Miller for bringing these two studies to my attention.

12

grasping, brushing, etc.). Signs in which the weak hand can be seen as functioning as the place of
articulation receive a different treatment in Stokoe's notation than signs in which the weak hand does
not act as a location. Despite this mention of two-handed signs, Stokoe does not explicitly discuss
them any further.

The first extensive treatment of the patterns we find in two-handed signs is the one by Battison (1974);
all subsequent work on two-handed signs refers back to this first discussion, and to the elaboration in
his dissertation (Battison, 1978). Battison set out to describe the process of deletion of one hand in
two-handed signs. This process was later called "Weak Drop" by Padden & Perlmutter (1987:352);
I will briefly discuss that study in the next section. Their term will be used from now on7 . In certain
situations, some signs that are normally made with two hands can be executed with one hand. Battison
states that this deletion process

"follows arbitrary linguistic principles, and does not depend on whether or not the resultant form
(with one hand deleted) is unambiguously identifiable. For instance, PARTY made with only one
hand is seen quite often, although it is then identical to the sign for PURPLE. With other signs
deletion is prohibited even though enough information would be left to unambiguously identify
the sign. Other constraints operate which take into account other aspects of the phonology, and
not just the question of potential ambiguity." (Battison, 1974:7)

Although this last phrase ("not just") raises doubt as to whether the risk of producing homonyms does
not playa role at all (which is what he suggests at first), the quote makes it clear that there are
phonological conditions a sign needs to meet in order to be able to undergo the process. In order to
describe these conditions which characterize the different groups of signs regarding this deletion rule,
Battison proposed two "morpheme structure conditions" (1974:5) that restrict the possible classes of
two-handed signs. These conditions are well-known by now, but I will quote the relevant passages in
full because of their importance for the rest of this thesis.

"Since we have two hands, two-handed signs are potentially symmetrical, and this symmetry is
exploited to a great extent in simplifying the production of signs. Symmetry is unmarked and
asymmetry is marked. This is well illustrated by two morpheme structure conditions, the
Symmetry Condition and the Dominance Condition.
The Symmetry Condition holds that if both hands move independently during a given
two-handed sign (as opposed to one or both being static), then the specifications for handshape

7 It is not very clear in which situations Weak Drop actually occurs. Banison mentions that "people do not stop signing
when they have their hands occupied with books, babies, or steering wheels, just as people do not stop talking when they
eat, drink, or smoke. (... ) Fatigue is also a factor in one-handed signing, as is posture. The psychological and social variables
related to deletion may include whether the signer was deaf or not, whether the parents were deaf, the age of onset of
deafness, age of acquisition of sign language, sex, age, etc" (Battison, 1974:8). These latter variables were not studied
however.
 13

and movement must be identical, and the orientations must be either identical or polar opposites
(reciprocals). Locations of the two hands in such symmetrical signs must also be specified either
as symmetrical or as polar opposites. Since each hand must be specified separately for location,
"symmetrical location" in this case means that the hands contact corresponding locations on
corresponding halves of the body. So two of the relevant descriptive features are ipsilateral (same
side of the body as the signing hand) and contralateral (opposite side of the body). Ipsilateral
location is unmarked. A sign whose locations were polar opposites would involve the hands
contacting the physically identical part of the body, on one side or the other. There would be no
symmetry to such a sign, since one hand would have to cross over to its contralateral side to
contact the physically identical place.
(oo.)
It is clear that a high value is placed on harmony in articulation: symmetry and ipsilateral
contact.
(oo .)
The Dominance Condition is an implicational statement which works from the other
direction. For those signs which have non-identical handshapes, one hand must remain static,
while the other, usually the dominant one, executes the movement. Examples are THAT,
PRESSURE, IN, THROUGH, CHEAT, LATER, WEEK, TICKET, SODA POP.
These signs are even more severely constrained, however, since (with a few exceptions)
only the six most unmarked hand configurations can serve in the subordinate, stationary position.
This set of maximally differentiated hand shapes is (1) A - the closed fist, (2) B - the flat palm,
(3) 5 - the palm with fingers spread, (4) G - fist with extended index finger, (5) C - hand arcs
in a semi-circle, (6) 0 - fingertips meet thumb, forming a circle. (Battison, 1974:5-6; original
italics, boldface added)

Battison goes on to list a few independent grounds for the unmarked status of this group of
handshapes, but this does not concern us here. (See Blees, 1994, for the NGT class of unmarked
handshapes, and Schermer et ai. , 1991, for evidence that the two conditions are largely valid for NGT
as well.)

Battison recognizes that

"[t]here are other two-handed signs which are covered neither by the Symmetry Condition nor
by the Dominance Condition. These are signs which have identical handshapes, but in which only
one hand moves, as in TO, SEW, SCHOOL, MOST, EVERY, PERFECT, YEAR, and WORLD.
Some can be traced back to symmetrical signs in which both hands moved, e.g. PERFECT,
YEAR (oo.); some of these can be traced to signs which were at one time non-identical in
handshape (oo.). DEPEND had a stationary B hand and an active G, now it has two G's. SHORT-
TIME had a stationary B hand and an active H (index and mid finger extended from fist), now
14

it has two H's. FINAL was once made with an I (little finger extended) and a G, but now it has
two I's. Usually the stationary hand has assimilated to the active hand. This assimilation toward
symmetry also overrides independent considerations of markedness of handshapes, since in
FINAL, G changes to I even though I is more marked, and in WORLD, a static A or S hand has
assimilated to a relatively more marked W (index, mid and ring fingers extended) to produce
symmetry."s

Thus, Battison arrives at three classes of signs (those covered by the Symmetry Condition, those
covered by the Dominance Condition, and signs that are not covered by either), which have different
degrees of symmetry9. He argues that these three classes behave differently with respect to Weak
Drop, the process which he set out to explain. Thus, signs that are covered by the Symmetry Condition
(both hands moving, same handshape and alike orientation) most frequently allow deletion of the weak
hand, and signs that are covered by the Dominance Condition (different handshapes, one hand moving)
never allow deletion of the weak hand. Signs that are in between the two conditions are also in
between with regard to the possibility of deleting the weak hand. In any case, Battison stresses that
the

"generalizations made here which match the hierarchy of deletion to the hierarchy of sign
symmetry are only very general tendencies, since the data has more variation than has been
suggested." (1974:12)

Thus, the generalizations themselves are not that strong in the sense that there is only mention of
'frequent deletion'lO in the class of most symmetric signs (those in which two hands move
independently with the same handshape), it is not possible for all members of that class. Second, the
whole study reflects 'only very general tendencies'. It can at least be called remarkable that no effort
is made to create a hierarchy of signs with classes that better fit the deletion phenomena. For example,
Battison mentions that in the group of most symmetric signs, deletion generally is prohibited in cases
where there is contact with the body, in all cases if there is 'alternating' movement, and in the cases
where the hands cross the midline of the body towards the contralateral side. That might suggest that
aspects like contact and alternating are also important aspects of a classification of two-handed signs.
In other words, although the phenomenon that Battison initially wanted to explain was Weak Drop,
more emphasis in the analysis seems to be placed on distributional patterns.

8 The letters in the above quotation refer to ASL handshapes; for drawings of these handshapes, see Klima & BeUugi
(1979).

9 At the end of the paper, Battison suggests that there is a fourth class of signs which violates both of the morpheme
structure conditions; an example is the sign TOTAL-COMMUNICATION.

10 'Frequent' here means that it applies to many members of the class, not that it applies very often to the whole class.
 15

In his thesis, Battison (1978) more explicitly focuses on lexical patterns (and not on processes like
weak hand deletion), wondering what a possible sign in ASL looks like and what kind of signs are
not allowed. He expands his classification of signs that was partly made explicit in the paper to include
6 types:

"For the purposes of this discussion, we shall posit six mutually exclusive, exhaustive, types of
signs:
Type 0: One-handed signs articulated in free space without contact (e.g. PREACH ( .. .)).
Type X: One-handed signs which contact the body in any place except the opposite hand
(CHINESE, SOUR ( ... ))
Type 1: Two-handed signs in which both hands are active and perform identical motor
acts; the hands mayor may not contact each other, they mayor may not contact the body,
and they may be in either a synchronous or alternating pattern of movement (WHICH,
CAR, (... ) RESTRAIN-FEELINGS ( ... ))
Type 2: Two-handed signs in which one hand is active and one hand is passive, but both
hands are specified for the same handshape (NAME, SHORTIBRIEF, SIT/CHAIR ( ...))
~: Two-handed signs in which one hand is active and one hand is passive, and the
two hands have different handshapes. Note that signs which were excluded specifically in
Type X fit in Types 2 and 3 -- one hand contacts the other (DISCUSS, CONTACT (A
PERSON) (... ))
We shall also need to posit a sixth type, Type C, to account for those compounds which combine
two or more of the above types." (Battison, 1978:28-30)

Apparently, type C functions as some kind of a catch-all category for all signs that do not fit into the
other ones. In the earlier paper, Battison (1974) clearly intended type C to cover clear compounds:
signs derived from fingerspelling such as SIMULTANEOUS-COMMUNICATION and TOTAL-
COMMUNICATION. However, in the present typology this is no longer the case: type C covers all
signs that do not conform to any of the other types. This tends to cast doubt on Battison's claim that
the types are 'exhaustive' and 'mutually exclusive'.

Again Battison (1978) mentions the two morpheme structure constraints, and places them in a more
phonetic framework: the constraints reflect the tendencies for a sign to be easy to perceive and easy
to produce. The phenomenon of optional deletion of the weak hand is only briefly mentioned, and
although linked to the typology, it is not put forward as the central motivation for the typology.

In general, it is not clear whether Battison intended the classification to account for something in
particular apart from the respectable goal of constituting a measure of relative complexity. As he
mentions,
16

"[t]his classification is not intended to be absolute and definitive, as there are other bases for
classification, e.g., type of movement ( ... ) or type of contact. But this classification allows us to
relate signs directly to the relative complexity of certain motor acts. As discussion warrants, this
general schema will be amended and refined." (Battis on, 1978:30)

Summarizing this large presentation, Battison (1974, 1978) tried to relate the amount of symmetry in
a sign to its ability to undergo the process of Weak Drop. There is a continuum of symmetry in signs,
and the extreme edges of this continuum are described by two generalizations. The Symmetry
Condition says that if both hands move independently, then the sign is symmetrical in all parameters.
The Dominance Condition states that if the hand shapes of the two hands in a sign are different, then
all other parameters are asymmetrical (i.e., having different values) too. Roughly, the signs covered
by these conditions are called Type 1 and Type 3 signs, respectively, and the signs that fall in between
the edges of the continuum are called Type 2.

In the next section, I will look at the extent to which the classification presented has indeed been
amended or refined.

2.2 Other literature on two-handed signs

2.2.1 Other early research on two-handed signs

Friedman
Friedman (1976), partly building on the work of Siple (1978; an early version had been circulating
since 1973), emphasizes the role of perceptual and articulatory constraints on the form of signs,
including two-handed signs, stating that

"[t]he language is presumably structured in such a way as to minimize perceptual difficulties and
maximize the efficiency of articulation in order to maximize the information received"
(1976:110).

For example, the fact that there are more two-handed signs in so-called 'neutral space' than near the
head, is explained by the need for redundancies in the perception of objects in areas of sight with low
acuity. As sign viewers are generally assumed to focus their sight on the signer's face (for a different
view see Ballering, 1995), the face is the area where there is highest acuity; in areas away from the
face, like the space in front of the chest, the acuity is low (cf. Siple, 1978). Thus, near the face there
is less need for redundancy, and one-handed signs are relatively easy to perceive. As for signs in
'neutral space', Friedman suggests that if the Symmetry Condition is not met and thus there is little
 17

redundancy, it is likely that the weak hand in these signs assumes a perceptually 'easy' or unmarked
handshape. Friedman, then, sees Battison's constraints on two-handed signs as reflecting

"the difficulty of simultaneously articulating two different configurations and perceiving two
different signals in a low acuity area. (All signs of this type are made in neutral space.) In
addition, the reader will recall that only six neutral configurations may be used for the stationary
hand. The motivation here may be either articulatory and perceptual or both." (1976:111-112)

Concluding, it is clear that, although Friedman may well be right with all comments on articulatory
and perceptual ease, she does not make any substantive comments on what exactly 'difficult' means
in different cases, how difficulty might be measured, and how the difficulty of different aspects may
be compared. Also, Friedman does not make any additional observations to Battison's generalizations
on the form of two-handed signs.

Frishberg
Examining changes in ASL signs between the beginning of the century and the 1970s, Frishberg
(1975, 1976) remarks that

"[c]hanges occur within individual parameters to contribute toward symmetry, fluidity, locational
displacement, and assimilation. These changes, on a level analogous to the phonological, are
motivated by such familiar principles as ease of articulation and ease of perception." (1975:700)

Frishberg comments on several changes that involve two-handed signs. If both hands move in a sign,
the handshape of the non-dominant hand tends to assimilate to that of the dominant hand (if they were
different originally). Signs which have changed from one-handed to two-handed always have the same
handshape for both hands, and the movement of the newly added second hand is "symmetrical" with
respect to the movement of the hand already present. Symmetry is the relevant concept here, as
Frishberg states that

"[s]ymmetry obviously facilitates articulation by allowing the signer to program both hands at
once. Symmetry eases perception in that the viewer can predict many characteristics of the sign
just from seeing the shape and movement of one hand. The restriction against two-handed signs
with non-symmetrical movement is strengthened by the historical tendency for signs to become
symmetrical in both handshape and movement." (1975 :701)

Frishberg does not discuss what she actually means by symmetry. Whereas symmetry with regard to
the handshape parameter may be evident (the hands are in the same shape), symmetry with regard to
the movement parameter is not quite so evident, and it is not discussed in her paper. From the remarks
18

about the facilitation of production we might deduce that she means sameness on phonetic articulatory
level, and not necessarily phonological or perceptual sameness. On the other hand, she says that the
(phonological) "hand-shape and movement parameters are identical to those in the original one-handed
variant" (1975 :701). Thus, the precise meaning of symmetry remains unclear: one does not know
whether it should be taken at a phonological or at a phonetic level.

In Frishberg's dissertation (1976), she gives a classification of all possible signs in the diagram copied
in (2), which very much resembles the classification that Battison (1978) would propose later.

(2) Classification of signs by Frishberg (Diagram 1.1 from Frishberg, 1976:18)

one·handed two-handed

' I~ asymmetrical
symmetnca
(dominant hand acting
on a base hand)
in contact
moving in space with the body moving in space in contact

/\
with the body

simultaneous alternating simultaneous alternating

movement movement movement movement

Apparently, Frishberg considers all two-handed signs where one hand moves asymmetrical, and all
other two-handed signs symmetrical. It can also be inferred from the above diagram that contact is a
relevant aspect for a typology, as Battison (1978) later remarked. However, Frishberg does not really
make clear what this specific typology is meant for; it seems to be 'just a possible classification of
signs, lI .

In chapter 2 (called 'symmetry' ) ofFrishberg's dissertation (1976), we find more specific phonological
definitions of symmetry:

II An alternative to this classification is to focus on the fact that one-handed signs are always asymmetric. and therefore

it were more appropriate to make a first division between symmetrical and asymmetrical. and then divide the asymmetrical
signs into one-handed and two-handed. One-handed signs are not considered at all in this thesis. however. and I will not
further discuss the merits of such an alternative.
 19

"Symmetry of handshape means simply that the two hands are identical in formation. ( ...)
Defining symmetry for the other three parameters: location, movement, and orientation may prove
to be more difficult. Notions such as 'reciprocal', 'mirror-image,' and 'alternating' need to be
introduced. For handshape the terms 'identical' and 'symmetrical' are synonymous; for the other
parameters they may not be." (1976:40)
"Symmetry of movement may mean either that the movement in the two hands is identical with
respect to timing and type of movement, or that the movement of the two hands shows the same
type, but the timing is out of cycle. ( ... ) When we say that the movement is of the same type, we
might also equally well say the hands make an identical trace. That is, if we imagined the fingers
and/or the hands each drawing a line in the place where they are moving, we would (rotate and)
superimpose these lines completely." (1976:40-41)
"Symmetry of orientation can include true symmetry about the line of bilateral symmetry (e.g.
palms facing each other inward toward the vertical plane that bisects the body), as well as
symmetry with respect to a horizontal plane (... ). We should take care to distinguish between
identical orientation and symmetrical orientation. (...) Of course most signs which would be
characterized as symmetrical in orientation are both symmetrical and identical." (1976:43-45)

I will argue later that these terms are still not unambiguous, and more importantly, that they are not
sufficiently related to the bilateral symmetry of the body to get a clear picture of the role of symmetry
in the lexicon.

Another case of change and two-handed signs Frishberg discusses is the relation between the number
of hands and the location of articulation. She found that two-handed signs made near the face tend to
become one-handed, and for signs made below the neck that one-handed signs tend to become two-
handed.

Concluding, although Frishberg explicitly argues that symmetry facilitates production and as such plays
a role in historical change of signs, she does not study symmetry at an articulatory level but rather at
a phonological level, without arguing that there is a direct or equivocal relation between her
phonological parameters and articulatory actions. As we will see below, studying symmetry at an
articulatory level is one of the goals of this thesis.

2.2.2 Recent interpretations of Battison's work

In the more recent literature, which has explicitly and extensively focused on several aspects of two-
handed signs, very specific aspects of Battison's initial observations and analyses have been
highlighted and others ignored. In this section I will consider to what extent the original observations
20

of Battison have been preserved and which new aspects have entered the discussion.

Padden & Perlmutter (1987), who are often cited for the introduction of the terms 'strong hand', 'weak
hand', and the name 'Weak Drop' for the process of deletion of the non-dominant hand, do not add
to Battison's observations on the sorts of signs that can be affected by this process. Rather, their
contribution is the suggestion that Weak Drop is a post-lexical rule, as it cannot feed lexical
derivational rules. Furthermore, they clarify somewhat under which circumstances Weak Drop can
occur: it "is optional and especially prominent in rapid or relaxed signing" (1987:350).

Although Weak Drop has since played a large role in the argumentation for different sign phonology
models, its status has recently been questioned. As just noted, the conditions under which it can occur
are still largely unknown. More importantly, there is no agreement on which signs can undergo Weak
Drop and which signs cannot. Ever since Battison (1974) it is known that more symmetrical signs are
more likely to be subject to Weak Drop than asymmetrical signs. However, no clear match between
categories of signs, like Battison' s four types of two-handed signs, and the ability to undergo Weak
Drop has been found . In recent work (Brentari, 1995; Stack, 1995) it has become clear that there are
many more properties of signs, like the absence or presence of contact between the two hands, that
may determine a sign' s Weak Drop-ability. Also, different signers appear to have different judgements
about a sign's likelihood to undergo Weak Drop, and furthermore in some cases signers are uncertain
about their judgements. All in all, Weak Drop seems to be a highly variable process, both regarding
the (type of) signs that are subject to it and regarding the conditions under which it occurs, the process
being optional. Therefore, I follow Stack (1995) here in saying that Weak Drop does not seem to be
a very solid basis for arguments for or against a certain model; I will disregard it her, but would like
to suggest that a different classification of signs than Battison's might aid in the determination of the
properties of Weak Drop.

A common theme in most subsequent research (Sandler, 1989, 1993; Brentari, 1990; Brentari &
Goldsmith, 1993; Blevins, 1993; van der Hulst, 1993, 1994a, 1995e) is that it has moved away from
Battison's original focus on the amount of symmetry that is present in a sign. All of the above
researchers have started from Battison's two conditions and the division into two or three major types
of two-handed signs that followed from them. The discussion in all of these recent works has centered
around the representation of the weak hand. The question that has been discussed is whether or not
the weak: hand should be represented uniformly for all signs, or whether it should be represented
differently depending on whether or not the weak hand functions as a location, which is the case in
Battison's Types 2 and 3. These different approaches have sometimes been called the 'H2-always' and
'H2-sometimes' approaches, indicating whether or not the weak hand (H2) is always present in the
representation as an articulator, or sometimes as an articulator and sometimes as a place of articulation
 21

(cf. van der Hulst, 1995e; van der Hulst & Sandler, 1994; Brentari, 1995)12.

Sandler (1989, 1993) defends the latter view, i.e. that there is an important and large division between
signs of Type 1 and Types 2 and 3, respectively, which is reflected in the representation. The
arguments for this position are, first, that in several respects the weak hand acts in the same way as
other locations, and second, that there are processes crucially referring to the two groups of signs
differently. These arguments will be discussed in chapter 5.

The 'H2-always' approach is taken by Brentari (1990; Brentari & Goldsmith, 1993) and van der Hulst
(1995e). To them the most important thing to express in a model is that the weak hand shows a very
limited number of contrastive properties with respect to the strong hand, in all types of sign. In the
'H2-always' approaches, it also possible to represent the fact that sometimes the weak hand acts as
a location and sometimes it does not. However, the fact that the weak hand can play two different
roles is not seen as enough reason to design two completely different representations for these two
cases. Instead, the continuity between the different types of signs as manifested in historical processes
and morphological processes like Weak Drop is proposed as evidence for a uniform treatment of the
weak hand (cf. Brentari, 1995). The model of van der Hulst (1995e) will be further discussed in
chapter 5.

In a sense all these studies seem to incorporate Battison's observations on the symmetry in signs. But
this idea of reduced possibilities for contrastiveness of the weak hand no longer has a relation with
the symmetry that was crucial to Battison's original observation, from which his two morpheme
structure conditions stem: the relation with the 'inherent' bilateral symmetry of the body is lost in
these modem approaches. Symmetry in these models simply means that the specification for a certain
parameter is the same for both hands. This only coincides with the bilateral symmetry of the body if
the values for parameters like orientation and movement are appropriately chosen. For example, the
orientation of the palm of the two hands would be 'symmetrical' in these approaches in the case where
they both have the value 'rightward', but they would also be symmetrical in the case where they both
have the value 'inward' . In the latter case, however, there is not only phonological symmetry but also
bilateral symmetry of the body. (I will get back to this issue in section 3.1, where I consider the
different possible meanings of ' symmetry' .)

For orientation the incorporation of symmetry does not seem to cause any problems in the current
models, but with respect to direction of movement there do seem to be problems. It is not always clear
whether the movement of the two hands is symmetrical or not. A difficult group is formed by signs
that move on the left-right dimension (which is perpendicular to the symmetry plane). It is not

12 There have also been some scattered references to an 'H2-never' approach, being developed by Perlmutter. There have

not been any publications about this approach, there are only some handouts or lecture notes; so lacking infonnation I will
not discuss the H2-never approach here.
22

discussed in the literature whether these should be considered symmetrical or not, nor whether these
movements can be alternating or maybe are inherently alternating (i.e., alternating in their symmetric
fonn). In all research since Battison, the symmetry of the body has been acknowledged and sometimes
taken into account when choosing features, but it has not taken the central role it did in Battison's
(1974) work. Its role has not been further explored. It will be one of the goals of this thesis to further
explore the role of the symmetry of the body on the structure of signs and on patterns in the lexicon.

As can be concluded from this short review, the classification of two-handed signs by Battison (1978),
although always referred to, is hardly ever completely copied. Most emphasis has been on Types 1
and 3, essentially incorporating the distinction made by Stokoe (1960), between signs where the weak
hand is a place of articulation and those where it is not. Type 2 has been noted by many researchers
as falling in between the two others; some Type 2 signs look like Type I and others look more like
Type 3. Battison himself remarked (see above) that there is a group of signs that is not covered by any
of the two conditions. Furthennore, it has been noticed that there is "a very small class of signs that
falls between the cracks of Battison's constraints. These are the signs in which the two hands are in
constant contact with each other as they move together, but they may be in different configurations"
(Sandler, 1993:346). Examples are the signs in (1):

(1) NGT: LATEN-ZIEN 'to show', BEGELEIDEN 'to guide'

ASL: HELP, SHOW, LEAD (verb)

According to van der Hulst (1995e), analyzing these signs as Type C (the class of 'compounds') would
not explain the fact that there is continuous contact in these cases, which is not a general feature of
Type C signs. Moreover, the signs are not considered as loans from fingerspelling or compounds, as
many of the other signs in Type C like SIMULTANEOUS-COMMUNICATION and ZONDAG
'sunday'. From the examples in (1), it might seem that there is a classifier involved, although they do
seem to be lexicalized fonns, and not productively fonned.

Various aspects of these models will be more extensively discussed in section 5.1, when the analysis
made here is related to existing models.

Summarizing, recent studies on the treatment of two-handed signs have tried to use Battison's
classification of signs and the distinction between weak hand place signs and weak hand articulator
signs. They have not suggested revisions to such classifications, even though it is recognized that there
are intennediate groups of signs which do not have a clear place in current classifications.
 23

2.3 Remaining problems and questions

In this section I will consider several problems that still remain, which have partly been hinted at in
the above review. I will end the chapter by posing the research questions that this thesis will try to
answer.

The main issue in the study of two-handed signs, to my mind, has been the search for a valuable
classification of signs. Most researchers suggest that there is a major division between signs in which
the weak hand acts as a place of articulation and signs in which it does not, following Stokoe (1960).
However, there are problems associated with that view: different criteria are possible for determining
whether the weak hand acts as a location. It is not all that clear who uses which criteria in which case.
First, one might consider the extent to which the weak hand acts as a location in morphophonological
processes and cases of allophonic variation, like the movement that is introduced for the weak hand
in aspectual modulations of a sign (cf. Sandler, 1993). Alternatively, one might set up a structural
definition for 'H2-place', like 'the weak hand is a place of articulation if it doesn't move', which is
the definition used by Stokoe (1960). It seems generally assumed that the two possible definitions are
saying exactly the same thing: signs in which the weak hand does not move are signs in which the
weak hand acts like a place of articulation in a morphological sense, and the other way round. That
causes problems for signs like LATEN-ZIEN 'to show' and HELP, where the two major types seem
to be mixed, although resulting in a possible sign. In these signs both hands move, the handshapes are
different, and the hands are in continuous contact. The difference in handshape suggests that the weak
hand acts as a location (as difference in handshape is a typical property of weak-hand-place signs);
on the other hand the fact that both hands move suggests that it is a weak-hand-articulator sign. The
two major types (weak hand place versus weak hand articulator) also seem to be united in part of
Battison's Type 2 signs, like BEWIJS 'proof and NAME, where only one hand moves but the sign
looks highly symmetric otherwise. This topic is also discussed in Uyechi (1995b).

Although there may be such a division between weak hand place signs and weak hand articulator
signs, it is often recognized that it is possible to make finer-grained classifications of signs, which may
account for the behavior of different signs in processes like Weak Drop. Such more detailed
classifications may also provide information for answering the question what a possible sign is in sign
languages, resolving the difficult or 'in between' status of a number of signs mentioned in the previous
paragraph. Recently no one has seriously attempted to make such a classification that would add
something new to Battison's. Rather, researchers have kept referring back to Battison's division of
two-handed signs in three groups of signs that are 'natural', 'central', or native to the language, and
one group of signs, Type C, that consists of all the 'unnatural', 'peripheral', or non-native signs. These
Type C signs are unnatural in the sense that they tend to be either morphologically complex or
borrowings from spoken language (often through its writing system), and structurally are combinations
of aspects of the other three types.
24

However, there are serious problems associated with Battison's classification, despite its widespread
acceptance. I will tum my attention to these problems in the following paragraphs. Let us first consider
again the purpose of the classification. Battison put forward the hypothesis that the amount of
symmetry in a sign is related to the chance of a sign deleting the weak hand on particular occasions.
He captured parts of the continuum of symmetry in signs in two conditions, the Symmetry Condition
and the Dominance Condition. These conditions roughly come down to saying 'if there is symmetry
in the movement parameter, then there is symmetry in the other parameters too' and 'if there is
asymmetry in the handshape parameter, then there is asymmetry in the other parameters too'. These
two conditions do not cover all signs however, but only the two ends of what we might call the
'symmetry continuum'. Battison (1974) recognized that there are signs in between too: there are signs
which have some symmetry and some asymmetry. In Battison (1978), he expanded the three types of
signs just mentioned to a 'classification' of signs that included three more types: two types of one-
handed signs (having body contact or not, respectively) and one more type of two-handed signs, that
included "compounds" of the other types. Battison intended this class to cover the fingerspelling loans
like TOTAL-COMMUNICATION, that are infrequent and 'non-natural' in the sense of being
borrowings from another language (and from another modality, the spoken modality). But as the Type
C lacks a rigorous and clear definition, Type C seems to be functioning as a catch-all category for all
signs that do not really fit in any of the other three classes: signs like SHOW might also be considered
Type C signs.

This particular history of the four Battison types might explain the problematic status of the
classification. It is problematic for methodological reasons in the following senses. First, the
classification is derivative of the two morpheme structure conditions; it is not based on independently
chosen formational aspects (such as Stokoe's parameters) that are thought to be of relevance for
classifying signs. Thus, other classifications are also imaginable. (This is recognized by Battison, 1978;
cf. p. 16 above.) Second, the classification is somewhat redundant or inconsistent in that one aspect
(viz. the presence of contact with the body) that is relevant for part of the typology (all one-handed
signs) is not relevant for the rest of the typology (all two-handed signs). Third, the classification is
unclear in that it is nowhere explicitly stated what the definitions of the relevant dimensions of the
classification are; the reader has to infer that it is 'activity of the weak hand' and 'identity of the
handshapes of the two hands' that are the relevant dimensions.

At this point, let us consider what a classification actually entails. A classification involves one or
more independent classifying dimensions (or parameters), each having two or more mutually exclusive
values. Multiplying the numbers of possible values for each dimension yields the number of classes
that the classification can describe. In Battison's classification, the classifying dimensions have not
been chosen in advance, but apparently are the following:
 25

(4) Classifying dimensions (with possible values) in Battison's typology

number of hands present (' I' or '2')
2 number of hands that move ('1' or '2')
3 number of handshapes involved ('1' or '2')
4 contact with the body involved ('yes' or 'no')

The first thing that is striking is that the first dimension is not mutually exclusive with respect to the
other dimensions: if the value on dimension 1 is '1', then dimensions 2 and 3 simply are not
interpretable (or not relevant); thus, there is a certain redundancy in the system. Second, as noted
above, it is striking that dimension 4 is not used by Battison for all signs (note that it is not impossible
for logical reasons), but only for signs that have value' l' on dimension 1. In principle, if this outline
for a classification were methodologically sound, we would get a classification with 16 classes (four
dimensions with each two values; 24=16). However, due to the redundancy mentioned above, the
maximum that we can hope for is 10; this is illustrated by the graph in (5).

(5) Classification resulting from the outline in (4)

no. of hands present 1 2

no. of hands moving 1

~ 2

no. of handshapes
/\
1 2
/\ 2

contact with body yes

A ;1 A n ~
no yes no yes no yes no yes no

Some of the classes found in this classification are clearly not considered to be separate categories by
Battison: contact is not regarded as relevant for two-handed signs. However, in the Symmetry
Condition from which this classification is partially derived, contact is relevant: the condition only
applies to those signs where the two hands move independently, thus without contacting each other.
26

This in itself is something different and independent from contact of the hands with the body.

Finally, there is much to say about dimension 2 in (4) . First, as I discussed above, determining whether
the weak hand moves is only one way of finding out about the status (place of articulation or
articulator) of the weak hand. More importantly, there is something odd about this dimension: it plays
a crucial role in the typology, but no one has ever suggested that the movement of the weak hand can
be distinctive. It has been remarked frequently that if it moves at all, the movement very much
resembles that of the strong hand. But there are no minimal pairs that are distinguished by movement
versus no movement of the weak hand. On the contrary: the weak hand is allowed to lose its
movement in certain signs in some contexts. This process is discussed in Padden & Perlmutter (1987)
who call it "weak freeze" (1987:358). If this is indeed the case, then whether the weak hand moves
or not should be fully predictable. As far as I know, this has not been noted in the literature, although
it may have important consequences. Investigating this topic is one of the goals of this thesis; it will
be treated in chapter 5.

Apart from the technical problems associated with the classification, we may also ask what we expect
a classification to do for us. Battison's classification may not be of much use if we want to consider
what constitutes a possible sign in sign languages, or in other words, patterns in the lexicon, or
phonotactics. The particular classification of Battison may be of use when we want to consider the
differential treatment of groups of signs by processes, be they morphophonological or variation-related
processes. (Uyechi, p.c., calls these patterns of alternation; the other she calls patterns of distribution;
..5
I will use these termY'from now on). However, in this thesis I will be concerned with the
characterization of a possible sign, i.e. on patterns of distribution. Both aspects should be represented
by a phonological model (cf. Clements, 1991). Such a phonological model, then, should characterize
natural classes of signs (groups of signs that are treated alike by processes), as well as allow an
expression of the difference between 'typical' signs and 'unnatural' signs. In general, one demands
that typical signs are easily represented in the model, whereas atypical signs receive a more complex
representation. The task for this thesis, then, is to see what constitutes a possible two-handed gesture,
and which of these gestures are actually found in sign languages. From now on, I will use the term
gesture in the sense of 'a physiologically possible arm and hand movement' .

In order to see what gestures are possible signs in a sign language, one type of classification may be
of use. The thing one would demand from such a classification is not that all classes are exemplified
in the language (as in Battison's typology) but rather two other things. First, we demand that the
classes are mutually exclusive. In other words, every possible gesture, occurring or not, should have
a place in one and just one class. Second, we demand that some classes include possible gestures that
are not actually found in sign languages. This second demand is not so much a methodological
demand like the first, but rather a pragmatic demand: if the classification only contains occurring signs
then the goal of finding out what is a possible sign is not met; to meet that goal we need to contrast
 27

'possible' or occurring signs with 'non-possible' or non-occurring signs. Thus, a classification looking
only at the number of hands present in the sign has two mutually exclusive classes, one-handed signs
and two-handed signs, but does not lead to any insight on the difference between the occurring versus
the non-occurring patterns in the lexicon. The implicit hypothesis of such an approach (at least here)
is that not all signs that are possible are equally likely to occur in a language (cf. Lindblom, 1984,
1990a). The likelihood of occurring is detennined by non-linguistic or phonetic limitations. It will be
a goal of this thesis to find out what the character of such phonetic limitations is.

The two approaches, studying patterns of distribution versus patterns of alternation, could in principle
lead us to independent (groups ot) properties of signs. That is, things about a sign that define its
membership of a natural class might be different things than those that define it as a possible sign. For
example, whether a sign has initial contact with the body or final contact with the body may be
completely irrelevant to its status as a natural sign in the language, whereas there may exist some
process (say, reduction of speed of the movement marking a certain aspectual modulation) that only
applies to signs with initial contact 13 • Although the difference is possible, we would expect the two
types of evidence to lead to a unified model.

Finally, we saw in the review of the literature that researchers have sometimes assumed that symmetry
is easy in perception and in production (cf. Frishberg, 1976). It has been argued (convincingly, to my
mind) with reference to phonetic studies (Siple, 1978) that symmetry is indeed easy in the perception
of signs in peripheral areas of vision. However, no such phonetic studies have been cited to
demonstrate that symmetry is indeed easy in the articulation of signs. Furthermore, symmetry has not
been studied at a phonetic level of detail at all; rather, it has been studied at the level of phonological
representation, without demonstrating that there is a direct relation between phonological features and
articulatory actions. Studying the articulatory (phonetic) aspects of symmetry, too, will be one of the
goals of this thesis.

To summarize, I have listed in (6) the topics in the study of two-handed signs which are arguably
unclear or problematic in the research so far.

(6) Problematic or unclear areas in the study of two-handed signs

the role of symmetry of the body as an influence on phonological structure
the relevance of movement of the weak hand
the signs that fall in between traditional classes: all Type 2 signs, signs like SHOW
the role of contact between the hands and between a hand and another part of the body
the problematic status of the process of Weak Drop as evidence for phonological

13 Of course, characteristics of gestures that are not allowed to occur in the language (say, violently pinching the cheek)
cannot be used by morphological processes as they do not occur in the language. '
28

arguments
the extent to which symmetry is 'easy' in the production of signs, and the precise influence
of this ease of articulation on phonological structure

Considering these problems, it seems clear that there may be other fruitful approaches to the
classification of two-handed signs than that of Battison (1978), which has been so influential and to
which little has been added or modified since. These other approaches include making more consistent
and elaborate use of some of the dimensions he thought to be relevant. In the rest of this thesis, I will
explore an alternative classification, and I will consider its relevance for phonology. The starting point
for this partly descriptive task is the same as that of Battison (1974): there is a certain amount of
symmetry present in the human body, referred to as the bilateral symmetry of the body. The question
then becomes, which of the ways in which possible hand and ann movements can 'break' this body
symmetry actually occur in sign languages? In general, the task is to separate out the possible signs
from the possible gestures, gesture meaning a physiologically possible movement of the ann and/or
hand (cf. p. 6).

The motive for focusing on this 'body symmetry' is that although symmetry is often referred to in
recent literature on two-handed signs, it is never examined in an exhaustive and precise way. In other
words, although everyone recognizes that symmetry sometimes is a useful concept, no one has put it
at the center of his investigations. This is precisely that what I will do here, in order to find out
whether a more detailed account of symmetry may be useful. A secondary (methodological) motive
is that in general phonetic description and analysis has not been given as much attention in sign
language research as it has in spoken language research. It mayor may not be the case that a more
phonetic description leads to interesting results. Symmetry intuitively seems an accessible place to start
phonetic-level research, as symmetry is straightforwardly definable mathematically, as we shall shortly
see. More importantly, the concept of 'ease of articulation', often cited in the literature, is something
that is relevant at a phonetic level, and not at a phonological level.

2.4 Summary: research questions

In summary, the present research will be guided by the following research questions, taking into
account the difficulties mentioned in (6).

How can the 'bilateral symmetry of the body' be described for the sign language articulators on
a phonetic (articulatory) level?

2 In which ways can possible gestures be symmetry-breaking, and which of these types of
 29

symmetry breaking occur in sign languages?

3 Can the study of symmetry in signs lead us to a classification of signs that helps to reveal
phonological patterns of distribution?

4 What are the implications of the answers to questions 1-3 for current phonological models?

5 To what extent can symmetry be considered as 'easy' in the articulation of signs ,?

6 To what extent can the factor 'ease of articulation' be seen to be responsible for the patterns
found in answer to question 2?
30
 31

3 Symmetry

In order to arrive at a sound definition of symmetry in signs, we need to consider what symmetry
precisely means. In this chapter, I will distinguish between two senses of symmetry. I will show that
the sense in which the body has been called symmetric is a different one from the sense in which the
symmetry of signs has been considered in the literature so far. I will conclude that in order for
symmetry in signs to be properly evaluated we need an articulatory description of signs, and not just
a phonological description.

3.1 The mathematical sense of symmetry

I distinguish between two senses of 'symmetry', as listed in (1).

(l) Different senses of symmetry

1 mathematical: symmetry is a transformation which leaves a form unchanged
2 metaphorical: symmetry refers to the sameness of two abstract concepts

These two senses of symmetry are both relevant to the present work. The first is the most precise and
formal, and the other one is related to the first 14 • Symmetry in its mathematical meaning will be
central to our description of the symmetry in signs; symmetry in its metaphorical meaning is what is
most often discussed in the sign phonology literature. I will discuss the mathematical meaning first.

In the mathematical sense, the appropriate branch of mathematics being geometry, symmetry of an
object is a transformation (or group of transformations) that leaves the object looking the same (Weyl,
1952; Stewart, 1975; Stewart & Golubitsky, 1992). Crucially, the appearance of the sameness is
perceived by the visual sense: an object (its shape) looks the same after a certain transformation. A
transformation is "a rule for moving things around" (Stewart & Golubitsky, 1992:28). The type of
transformations that we are interested in here are transformations that leave constant the distances
between the different points in an object; these are called rigid motions.

14 A third sense of symmetry might be considered, 'the layman sense'. In this meaning of the word, "symmetric means
something like well-proportioned, well-balanced, and symmetry denotes that sort of concordance of several parts by which
they integrate into a whole. Beauty is bound up with symmetry" (Weyl, 1969:3). Thus, symmetry can be found in the
paintings of M.e. Escher and in other works of art. However, although this sense of symmetry is obviously related to the
mathematical sense, it does not involve any formalization; it rather involves a personal impression of beauty. Because of this
informal and subjective use of the term symmetry, it can lead to a lot of ambiguity if we try to apply it to signs of sign
languages.
32

Rigid body transfonnations can be of different types, but here we will consider only three of them:
translation, reflection and rotation l5 • First, I will briefly discuss these transfonnations disregarding
symmetry, by giving examples of objects in a two-dimensional space (a plane). After these examples
I will illustrate transformations with respect to symmetrical objects.

Let us consider the triangle in (2)16.

(2) Triangle

Applying a translation to this triangle simply means changing its position, for example, for about a
distance the size of its height to the right and downwards. In this case, the object does not look the
same after the transfonnation: its location is different, and at the initial location there is no object that
looks like the triangle that was there before the transfonnation. Here, I will not discuss objects that
have translational symmetry. In chapter 5, however, we will see that translation is an important
transfonnation in determining the symmetry of movement, thus, symmetry in time rather than in space.

Reflecting the triangle means flipping it over about a line. For example, in (3) the triangle A is
reflected in the vertical line L to result in triangle B.

(3) Example of reflection

A B

IS Other transfonnations can be seen as composites of these three, but they are not relevant here.

16 The triangle examples that follow are modified copies from Uyechi (1995a); the star example that follows is a modified
copy from Stewart & Golubitsky (1992).
 33

Rotating the same triangle means turning it around through some angle about a fixed point. For
example, if we take the upper corner of the triangle A in (4) as the point of rotation and rotate it
clockwise through 120 degrees, we get triangle B.

(4) Example of rotation.

B
A

The same transformations can be applied to an object in three-dimensional space. In that case,
reflections occur about a plane instead of a line, and rotations occur about a line or axis instead of a
point.

The triangle in the examples above is not symmetrical at all: there is no transformation which does
not change its appearance. The triangle possesses neither reflectional symmetry nor rotational
symmetry: the triangle B looks different from the triangle A in (3) and (4), no matter which line of
reflection or point and amount of rotation we choose.

The star in (5), however, does have reflectional and rotational symmetry.

(5) Star with reflectional and rotational symmetries

In fact, it has five rotational symmetries and five reflectional symmetries, a total of ten symmetries.
The rotations are rotations about the point in the middle of the star, and the reflections are reflections
34

about five different lines. One of the five rotations is a rotation through 0 degrees; this trivial type of
symmetry is present in every object, and it is called its identity. The other rotations occur through integer
m ul tiples of one fifth of a turn (72 degrees): 72, 144, 216, and 288 degrees.

The type of symmetry we find in the human body, and in many other animals, is called bilateral
symmetry. An object is bilaterally symmetric if it has at least one reflectional symmetry. In the case of
the human body, there is only one reflectional symmetry. The plane about which reflection is possible
without changing the body's appearance, thus, the symmetry plane, is called the sagittal plane in the
physiological literature (cf. Luttgens & Wells, 1982; Winter, 1979). This is illustrated in (6), with two
other planes that can be distinguished which will become relevant in chapter 6.

(6) The three planes bisecting the body (adapted from Luttgens & Wells, 1982:22)

sagittal plane frontal plane transverse plane

The three planes are positioned at right angles to each other, as is more clearly apparent from the
illustration in (7).
 35

(7) The three planes bisecting the body in relation to each other (adapted from Winter, 1979:11)

Sagittal plane

Transverse plane

Frontal plane

There are planes (not lines) of reflection since the human body is a three-dimensional object. If we
compare this bilateral symmetry of the human body with the symmetries of the star in (5), we see how
limited the symmetry in humans is: it is only one single symmetry, whereas many objects in nature (think
of a five-legged starfish, for example) have many more symmetries. A mathematical form like a perfect
circle even has many more symmetries: it has an infinite number of reflectional symmetries and an
infinite number of rotational symmetries. Still, we do not think of a circle as being the apogee of
symmetry, whereas we do note the beauty of bilateral symmetry in animal forms. So in a sense this is a
"psychological trick": we see symmetry best or appreciate it most where there is only one reflectional
symmetry (cf. Stewart & Golubitsky, 1992).

In one respect, the human body cannot be idealized as a mathematical rigid body: it does not have a
constant form. Parts of it can move, and only certain arrangements ofthe moveable parts result in bilateral
symmetry. That is precisely what makes it interesting here, of course, as in sign language these moveable
parts do indeed move. We expect some of these movements to result in bilateral symmetry, but not others.
Typically, in many rest positions the human body is bilaterally symmetric, as illustrated by the figure in
36

(8) for standing position.

(8) Fundamental standing position (adapted from Luttgens & Wells, 1982:24)

The important thing to realize here is that the both sides of the body are shaped alike: not merely on a
superficial view the body seems to consist of two identical parts, but also on a more 'internal' level. The
bones and muscles of the arm and hand, for example, are shaped alike. The similarity is the reflectional
symmetry. Thus, we could expand our view of bilateral symmetry by saying that the body is symmetric
if all the muscles and bones of the moveable parts that exist on both sides of the body are in bilaterally
symmetric position. In other words, within every pair of muscles, one on each side of the body, the
muscles nccd to be contracted to the same extent, and within every pair of joints, one on each side of the
body, each joint needs to be in the same position: only in that case the body is fully symmetric.

One of the goals of this thesis is to find out in which different ways this symmetry of the body can be
broken (cf. research question 2, p. 28, 115; the concept of 'symmetry breaking' will be introduced more
elaborately in the next chapter). To do this, we need a way to refer to the states and the movements of the
moveable parts of the body, in our case the arm and hand, as it is the different arrangements of these
moveable parts that can cause asymmetries to occur.
 37

3.2 The metaphorical sense of symmetry

So far, I have been introducing the mathematical meaning of symmetry: an object is symmetric if it
retains its appearance after a transformation. As said in the introduction, it is useful to distinguish another
sense of symmetry, because this is the sense used often in the sign literature. I will refer to it as the
'metaphorical sense of symmetry', and I will show how it is related to the mathematical sense.

Symmetry is also often used in a metaphorical sense. In those cases, people refer to the sameness of two
(always two) rather abstract things as symmetrical, or they talk about the symmetric behavior of two
things or groups of things. As it is the sameness in some sense of two things, this meaning seems clearly
derived from the one bilateral symmetry that is so pervasively present in humans and many other animals.
This sense of symmetry may have a very formal side to it, in that we can establish an unambiguous
language or set of terms to talk about it. This sense of symmetry is found frequently in linguistics, too.
For exam pie Ewen (1994), starting a review of dependency relations in phonology, tal ks about "the claim
that elements in a particular domain may be asymmetrically related", meaning that elements behave
differently. In similar vein, Dresher & Van der Hulst (1994) talk about different types of asymmetry
between head and dependent constituents in phonology and syntax. Also, it is this meaning of symmetry
that is used in the literature on sign phonology, for example by Battison (1978) and Frishberg (1976): to
them a sign is symmetrical if its handshape specifications are the same and if both hands have the same
movement specification (cf. chapter 2). This use of the term symmetry can very well have a very precise
and formal meaning, depending on the extent to which the phonological constructs to be compared are
formalized. However, it is metaphorical or abstract in the sense that the two things that are considered to
be related can be anything: they may refer to numbers, to phonological features, to behavior in a certain
process, etc. There is not necessarily a relation between the symmetry in this sense of a certain pair of
abstract things describing a shape and the sameness in the form as we perceive it. That is precisely what
has been alarming in the treatment of symmetry in signs so far: the phonological specification of signs
may be defined symmetrical in this metaphorical sense, whereas the signs as they are produced may not
be symmetrical in the mathematical sense. However, it may of course be the case that the symmetry is
found both in reality (the production of the sign) and in our description of it (a phonetic or phonological
model or construct). This requires that the description is appropriately 'grounded', which in the sense of
the human body means that a description should incorporate its bilateral symmetry.

In order to describe the symmetry of signs in a mathematical sense, then, we need an articulatory phonetic
description of signs. In the next chapter I will determine what such a description should be like.
38
 39

4 Phonetic description of signs and their symmetry

In this chapter, I will consider what a phonetic-level description for sign languages could be like in
section 4.1, and I will choose to make an articulation-based description at the level of joints. Then, in
section 4.2 the relevant joints in the arm and hand will be described, as well as the types of movement
they permit. Finally, in section 4.3, I will discuss how symmetry is definable at the level of joints, and
how the state of a joint can be measured.

4.1 Different types of phonetic description

At the end of the previous chapter (section 3.2), I stated that studying symmetry here will mean making
a phonetic description at first, instead of studying the more metaphorical type of symmetry in phonology.
With ' phonetic' I mean that the precise way in which signs are produced and perceived is studied, without
paying attention to which arm and hand movements occur in sign languages. However, in this thesis, such
a study can only be done to a limited extent and in limited amount of detail. I will attempt a level of
description that is closer to the way signs are produced and/or perceived than phonological analyses have
done until now, and that does not incorporate in advance which types of gestures are found in language
and which ones are not.

In principle, such a description can focus on the way signs are produced, on the signal that results from
the production, or on the way this signal is perceived. In spoken language phonetics, a distinction is
traditionally made between these three aspects (cf. Ladefoged, 1967; Kenstowicz, 1994), the fields of
study called articulatory, acoustic, and perceptual phonetics, respectively. For sign language, the same
distinction can be made, although it is more appropriate to talk of 'optical' instead of 'acoustic': the signal
can be seen, not heard. For the study of symmetry, it is necessary to focus on the articulatory aspect of
the communication process.

A description of the articulators in a sign can be oftwo different kinds. The difference consists of talking
about the position of body parts relative to each other versus talking about locations of body parts with
respect to some absolute reference point in space. To give a brief example, if we want to describe the
motion involved in the gesture in (1), we can say either of two things.
(1) Elbow extension (adapted from Luttgens & Wells, 1982:24)

Making a description of the relevant segments in absolute terms, we could say that there is a downward
movement of the hand, slightly arced maybe, the elbow not changing position. In addition, we might say
that the fingers point forward-upward at start, and forward-downward at the end. Alternatively, focusing
on the position of the relevant body segments relative to each other we could describe the movement by
saying that there is a 70 degree extension of the elbow joint, maybe adding the initial state of the elbow
(and other) joints. The first type of description I will call space-based from now on, the second type I will
call body-based. The key characteristic of these two types of description is that the former is not related
to the body whereas the latter is. In other words, the former describes the position of a certain point or
object (the hand in this case) in absolute space, whereas the latter describes the movement of one segment
relative to another segment. Thus, the "frame of reference" (Mandel, 1981) or "reference system" (Winter,
1979) is different for the two cases. In terms of mathematics, the formalization of these two approaches
would probably lead to the use of a rectangular coordinate system in the case of space-based descriptions,
and of a polar coordinate system in the case of body-based descriptions (cf. Winter, 1979; Cohen, Namir
& Schlesinger, 1977; Nagahara, 1988; see section 4.3). The same distinction is made in the literature on
motor control, where a distinction has been made between descriptions of a movement in "joint space"
versus one in "hand space" (Hollerbach & Atkeson, 1986:41; see also Morasso, 1981). It is not clear to
me whether the distinction between body-based and space-based is in part also a distinction between
production and perception; this will need to be examined in the future.
 41

Although there have been a few mixes of both, in the terminology in models of sign language phonology
the general emphasis has been on the space-based descriptions'. The best-known description that is an
example of a mix is that of Stokoe (1960; see also Stokoe, Casterline & Croneberg, 1965), who talks
about various positions of the forearm in articulatory terms (supinated, neutral, pronated) to describe what
has later been called "orientation" (cf. Battison, Markowicz & Woodward, 1975). Other examples of
body-based terms in predominantly space-based descriptions or feature systems can be found in Brentari
(1990). She talks about the position of the wrist (1990:70) and has a feature [prone] referring to forearm
position, both describing what others have termed orientation of the hand, which most often is described
by saying which side of the hand points to which direction (e.g. Sandler, 19892 ; Uyechi, 1995a). Also,
Brentari talks about position of the thumb and fingers with the term "bent" (1990:72). I would like to
emphasize that I am not claiming that phonological models should be exclusively either body-based or
space-based 3 • However, at the level of phonetic description it is necessary to be consistent in using either
body-based or space-based terminology. It remains open for discussion whether the phonology of a
language can indeed 'use' both ways of looking upon a sign, body-based (e.g., forearm prone - supine)
and space-based (e.g., palm down - up); the distinction made here may aid such a discussion.

An interesting property of all the space-based descriptions so far is that they focus on the hand(s) as a
isolated or separately moveable body part: they discuss the shape of the hand, its location in space, its
orientation (or "attitude"; cf. Mandel, 1981 :12), and its movement through space. The arms hardly ever
enter the discussion, although they clearly need to be part of the production of certain movements, of the
location, and of the orientation. (This is reflected in the frequent phonological references to 'hands', often
to indicate 'articulators'; cf. the term 'weak hand', and the titles of, for example, van der Hulst, 1995e;
Sandler, 1993) Again, the most striking exception is formed by researchers who have talked about
supination and pronation oj the Jorearm, and not just of 'hand orientation', which would be the space-

I Mandel (1981) has also explicitly made the distinction between the same two ways of describing body movement for sign

languages. He talked about "internal" vs. "external" modes of description, instead of body-based vs. space-based (1981 :9).
Clearly aware of the distinction between the two, Mandel argued that both modes of description could be useful for the
description of different phenomena (the internal mode is mainly used for describing direction and orientation). This consequent
use of different ways of describing things is in fact is the most characteristic aspect of his work, and he refers to it as "multi-
based phonology" (1981 :36) . The only other work I know where the distinction between the two modes of description is made
is Nagahara (1988). who cites Marr (1982) in using the terms "object-centered" and "viewer-centered". Nagahara is the only one
who has used a completely body-based phonological model (see also not~.
oX

zSandler(1989) uses the binary features [prone). [up), [in), and [contra); however, [+prone) means 'palm down' and [-prone)
means 'palm up ' . So although the names of the features seem to be body-based, the phonetic interpretation of these features
is not body-based at all. This quite nicely illustrates the difference between phonetics and phonology. Here we are interested
in phonetic description, and the terminology that is used in phonological models generally refers directly to some phonetic
description that is the basis for the analysis. In some cases, as is shown by Sandler's terms for describing orientation, this relation
is not so direct.

J Nagahara (1988) does claim that phonological models should be exclusively body-based.
42

based counterpart.

This difference may be the result of some inherent property of the two modes of description. If we specify
the exact position of the hand in space, there probably is only one likely position of the arm to get the
hand in that position. (This remains to be empirically investigated.) Thus, the position of the arm is fully
predictable. As Mandel states, "the position of the joints outside the hand is (oO.) rarely required in an
underlying specification" (1981 :43). Nagahara writes that

"[t]he position of the elbow for most of the signs in ASL is relatively fixed . Further, even in the
cases in which the position of the elbow is variable, such variability does not seem to be
distinctive. It is my claim that the position of the elbow, like the position of the jaw in oral
phonology, is not relevant, at least at the level of underlying representation." (1988: 18)4

This may indeed be true, but it still remains the case that for many movements of the hand (in a space-
based sense) there is movement at the shoulder joint and the elbow joint. Describing movement in terms
of the joints, as Nagahara (1988) did, requires looking at all the joints ofthe arm and hand, and not merely
at (the joints of) the hand.

Of course, the above references to the literature all concern phonological analyses and models, not
phonetic descriptions. It is relevant to keep this distinction in mind. Most phonological models have
stayed close to phonetics in sign language, and it may not be fair to judge them on their phonetic merits.
However, if we consider them to be studying the abstract phonological categories, we also expect that it
is pointed out how the relation between the actual occurrence of a sign and the phonological categories
is conceived of. In other words, phonologists should be able to explain which abstraction they have made,
for instance what original data they had that led them to make the abstract construct. Or put in still another
way: what is the phonetic description that the phonological categories are based on?

Uyechi (1994, 1995b) is the only researcher to my knowledge who has discussed the relation between
the abstract phonological categories and the concrete body actions. She claims (p.c.) that her model is not
specifically either body-based or space-based. However, it seems easily 'translatable' into a space-based

4 Nagahara (1988) proposes a model of sign phonology (not a phonetic description) that is fully body-based, in my

terminology. Thus, Nagahara claims that the relation between phonology and articulation can be rather direct. I will not discuss
the model here extensively, as it is not a body-based descriptiol/, and also because it only discusses one-handed signs. I would
like to remark however, that it seems to be a problem for Nagahara's analysis that he does not include a 'shoulder movement
tier' at all: many signs only involve movement at the shoulder, and some of the examples he suggests involve two positions of
the 'forearm movement tier' at least phonetically involve movement at the shoulder. He does not discuss the way in which this
shoulder movement can be predicted from the values on his three movement tiers. This seems to be an interesting avenue for
further exploring the possibilities of Nagahara's model.
 43

description of signs, with a more indirect or more complicated relation to a body-based description of
articulation. Uyechi (1995b) suggested that this relation to articulation for her three phonological
parameters might be as indicated in (2).

(2) Relation between phonology and production in Uyechi (1995b; adapted from p.26)

JOINT MOVEME NT SIGN MOVEMENT SIGN PARAMETERS SIGN GESTURE

Shoulder: [Bend-main] ~ [straight movement]

rB<nd-sic»-t0-5ide] ~ [loca1iO"]
[T,,;st] [forearm bend side-t0-5ide)

Ebow: [Bend-main) --n~--'" [foreann circle)

[T,,;st)
[orienta/ioll]

Wrist: rBend-main) ..----7 [hand rotate)

[Bend-,ide-to->ide] ~---"'- [hand-circle)

Base: [Bend-main]

[Bend-si c»-to->ide] - - - - ~: _-_ [hamlshape]

Non-basc: [Bend-main]

In this graph, the 'joint movements' roughly correspond to the possible joint actions that I will distinguish
in the next section. It seems that the category 'sign movement' might be more related to perception than
to production, and thus suggests that perceptual categories form an intermediate category between
production and the three sign parameters (which are introduced in Uyechi, 1995a). Forother models, such
relationships are not discussed at all, so it is not possible to compare models to each other in this respect.
Merely looking at (2), however, one indeed gets the impression that the relation between production and
sign parameters is rather indirect, and it suggests that the relation between the signal and the parameters
is closer. The extent to which that is a correct impression will depend in part on the answer to the question
4

whether space-based descriptions are indeed related to perception or nots.

Initially, here I will not be concerned so much with the question of how abstract phonological categories
need to be. For descriptive purposes, I think it is evidently important to stay closer to phonetic detail.

To return to the question posed above, should we make a space-based or a body-based description of
signs? Because in a body-based description it is easier to take the whole articulator (arm and hand) into
account and as our goal is to describe symmetry of the articulators, I decide here to focus on body-based
descriptions. Then, the question becomes what the level of description should be: should we focus on the
joints, on the muscles, or on motor units (cf. Turvey, Fitch & Tuller 1982)? Again, without further
argument (but see Turvey, Fitch & Tuller, 1982; Kelso & Tuller, 1984), I will assume that the level of
joints (and not muscles, for example) is the most appropriate way of describing body motion. Notice that
the choice to focus on all parts of the arm and hand, and not merely at the hand itself, is not simply a
question of choosing a body-based perspective: within a space-based perspective one might also pay
attention to all parts of the hands and arms. It is, however, a question of making a phonological or a
phonetic description or analysis : as we do not decide in advance to ignore some bones or joints because
they do not seem to be relevant in the language, we are making a phonetic description. A subsequent
phonological analysis may consist of considering the relevance of the different joints and movements in
the language.

In the rest of this chapter, I will describe which are the relevant joints in the articulation of gestures
(section 4.2), and how their state might be described (section 4.3).

4.2 The joints of the arm and hand

In (3), a drawing of the arm and hand hanging straight down is given accompanied by the terms of the
most relevant bones and joints. ('Relevant' here is intended to mean roughly 'moveable', or ' moveable
to a certain extent by muscles'; I mainly exclude the intercarpal joints of the hand in this way; cf. Luttgens
& Wells, 1982.)

5 What emerges even more clearly from this diagram is the particular complexity of the orientation parameter within many
models. The problematic relation between the orientation and location parameters is to my mind the most interesting discovery
of Nagahara (1988). His solution, making a body-based pllOn%gica/ ana/y.\'i.~, merits more serious attention, to my mind, than
it has had so far in the literature. I hope that the discussion of body-based versus space-based descriptions in this thesis
contributes to our understanding of the subjcct.
(3) . . a f the arm an d hand
Bones and Jomts

" Ie
cl3\"IC (coll3r bone)

glenolllllnem IJ"oint

scapula (shoulder blade)

,
proximal
ioulnar joint
hlll11eroulnar joint

••• •• ••••

distal rad"IOU !nar J"oint

••• radiocarpal joint
carpals -----n~ carpometaca rpaljoints

metacarpals - ' ' ' '-'-' metacarpop halangeal joints

interphalangea I J"oint of thumb

phalanges

proximal mterp
" halangeal joints

"
distal mterp halangeal joints
46
 47

The goal of the list of joints and the discussion that follow is to make clear which different joints there
are and which joints can move independently. A joint in itself may allow a certain degree of movement
by virtue of the structure of the bones that meet there, but the movement that it actually does make is
dependent on other factors too: how are the bones attached to each other by ligaments, and which muscles
are there to perform the action. For example, as we shall see, the bones of the fingers (the digits or
phalanges) allow independent bending movement by virtue of their structure, but their independence is
constrained by the specific characteristics of the muscles that move them (cf. Ann, 1993). The aim of the
following survey is not to study all of the joint's properties, including the ligamental and muscular
structure. Because of the limited time and space, I will focus on the independence of movement of the
different joints. Data on the other properties will have to be taken into consideration some other time.
However, for the sheer purpose of making a description of the two articulators in signs, looking at the
joint suffices. For data on the extreme positions (the ranges of motion) that are allowed at each joint, I
refer to Luttgens & Wells (1982:21) and to appendix A.

In (4), the different joints that are relevant in the movement of the arm and hand are Iisted, starting at the
most proximal (closest to the trunk of the body) joint (shoulder) and ending with the most distal jOint6 •
For every joint, the type of joint is listed and also the amount of movement that is allowed. The table is
followed first by a discussion of 'axial movement' (which is the movement we find at joints), and then
by a discussion of every joint in the list.

(4) List of joints of the arm and hand

JOINT NAME JOINT TYPE MOVEMENT

glenohumeral ball-and-socket 3-axial
humeroulnar hinge I-axial
1
humeroradial ball-and-socket 3-axial
proximal radioulnar pivot I-axial
distal radioulnar pivot I-axial
radiocarpal ovoid 2-axial
carpometacarpal (thumb) saddle 2-axial
metacarpophalangeal (thumb) ovoid 2-axial
interphalangeal (thumb) hinge I-axial
metacarpophalangeal (fingers) ovoid 2-axial
proximal interphalangeal (fingers) hinge I-axial
distal interphalangeal (fingers) hinge I-axial

"The discussion about joints and movements that follows in this paragraph is based on Luttgens & Wells (1982).

1 This joint is not straightforwardly classifiable; I follow Luttgens & Wells (1982) in saying it is a ball-and-socket joint.
48

The 'movement' specification in (4) refers to the number of axes about which movement is allowed. This
will also be referred to as the degrees of freedom of a joint (cf. MacConaill & Basmajian, 1969). In
physiological terms, the 'angular' (or 'rotary' or 'axial ') movement of a bone of the body occurs in a
plane and about or around an axis. Movement can be either fle.xional or rotational. In case there is a
flexional movement, the movement of the bone defines the plane, and the axis of movement is
perpendicular to this plane running through the center of the joint where movement occurs8. For example,
consider the case of extending movement of the elbow, as was illustrated in (1). There, the plane of
movement is parallel to the sagittal plane (see p. 34), and the axis around which the movement is said to
occur is at a right angle to that plane crossing the elbow joint. For rotational movement of bones, for
example rotation of the humerus at the shoulder joint, the rotating bone is parallel to the axis, and the
plane of movement is perpendicular to that axis at the height of the shoulder joint. To give an example,
if we nod our heads, the axis of rotation is a vertical axis in the sagittal plane and parallel to the frontal
plane; the plane of movement is parallel to the transverse plane and is located at the height of the neck.
Finally, it is important to note that when we talk about movement in two or three planes, we define the
' extremes '; combinations of the two movements, occurring in an oblique plane intermediate between two
or three of the allowed planes, are possible.

The glenohumeral joint is the place where the head of the humerus (upper arm bone) meets the glenoid
cavity of the scapula (shoulderblade). It is classified as a ball-and-socket joint. In principle it allows
movement around three axes. I say 'in principle', because as said above, movement at any joint only
really occurs if there is a specific muscular structure to execute it, and only if it is not prohibited by the
arrangement of ligaments and muscles. In case of the glenohumeral joint, movement indeed can occur
around three axes: there is flexion-extension in three planes (around three axes) and rotation in three
planes.
I will call this joint the shoulder joint'J. Its possible movements are illustrated in (5).

H Flex ional movement can be either flexion, extension, abduction, or adduction. Abduction and adduction are sometimes also

called' latera I extension' and ' lateral flexion', respectively.

'. The common sense notion of shoulder includes more than what I have just presented: it includes movements of the clavicle
(collar bone) and the scapula. For sake of ease, I will ignore these here, assuming that if movement occurs there during the
articulation of gestures it can be considered part of the main movement of the glenohumeral joint; thus they form one functional
group.
 49

(5) Movements of the shoulder joint (adapted from Luttgens & Wells, 1982:84)

abduction

horizoolal hcrizaltal
c~:'on~__~~~~~~

~
I circumdictioo
flexion

Moving to the more distal joints, we see that two bones are connected to the distal end of the humerus:
the ulna and the radius (see the figure in (3»). These come together at the humeroulnar and the
hum eroradial joints, respectively. The humeroulnar joint is classified as a hinge joint, allowing movement
in one plane; it this joint that is generally known as the elbow joint. However, if there is movement at the
humeroulnar joint the very same movement occurs at the humeroradial joint. This latter joint is classified
as a ball-and-socket joint, but because of the architecture of the ligaments around the distal end of the
humerus, it is only allowed flexion-extension in one plane. Also, because of the particular ligamentous
connections with the ulna, no independent rotation of the radius occurs (see below). Because of their
equal movement, I will consider the humeroulnar and humeroradial joints one functional group, which
50

I will call the elbow joint. Summarizing, at the elbow joint only flexion-extension in one plane is
possible; this is illustrated in (6).

(6) Movements of the elbow joint (adapted from Luttgens & Wells, 1982:116)

.
;S1~
____-- __ /],--1"j
extenslon ----=- --- g~ }
<,
1

The proximal radioulnar joint is the place near the humerus where the ulna and the radius meet. It is
classified as a pivot joint, which means that only rotation is allowed. The proximal end of the radius can
partly roll around the proximal end of the ulna. In this rotary movement (supination and pronation) there
is strictly speaking also rotation at the humeroradial joint, as indicated in the previous paragraph; this is
not considered movement of the elbow complex, however (cf. Luttgens & Wells, 1982). The distal
radioulnar joint is the place near the wrist where the sides of the ulna and the radius meet. Although it is
a pivot joint like the proximal radioulnar joint, hardly any rotation occurs; there is slight movement if
there is rotation at the proximal radioulnar joint. I will therefore consider both radioulnar joints as one
functional group, calling it the radioulnar joint; it only allows rotation of what I will call the forearm,
which consists of the radius and ulna. The rotation that is permitted is illustrated in (7).

(7) Movements of the radioulnar joint (adapted from Luttgens & Wells, 1982:116)

supination
pronation

What is commonly known as the wrist joint and anatomically called the radiocarpal joint is a fairly
complex joint: it is the place where the distal end of the radius meets with three of the four proximal
carpals. The distal end of the ulna does not meet with any of the carpals. The joint type is called ovoid,
 51

it allows bending movement in two planes, flexion-extension and abduction-adduction (or 'lateral
flexion'), as illustrated in (8). I will call this joint the wrist joint.

(8) Movements of the wrist joint (adapted from Luttgens & Wells, 1982:126) .

. \ extension

) ulnar flexion
hyperextension (adduction)

Combining flexion, radial flexion, (hyper)extension and ulnar flexion in that sequence or the other way
round results in circumdiction, which is sometimes referred to as swivelling in the sign phonology
literature. In this movement, the finger tips describe a circle and the whole hand outlines a cone.

I will ignore here the different other joints where the carpals (wrist bones) are involved, as their
movement is extremely limited. For the same reason, I will ignore the carpometacarpal joints of the
fingers , which only perform very small gliding movements.

The carpometacarpal joint of the thumb is relevant though, as it does allow movement. It is the place
where the metacarpal bone of the thumb joins one of the carpals. It is a saddle joint, allowing bending
movement in two planes. This is illustrated in (9). I will refer to this joint as the proximal thumb joint.
52

(9) Movements of the proximal thumb joint (adapted from Luttgens & Wells, 1982:128)

hyperadduction

abduction flexion
e:\1ension

I
'ir:'h~
1-- ~
1-
ft ~l:~
l '\ :~~' - I
\ .;: )
... ..

(~l7f!J
'
1 -( .
\1 ~ ~.
\ "

_. -
;-

hyperflexion oppsosition

The metacarpophalangeal joint of the thumb is classified as an ovoid joint, allowing bi-axial movement.
However, because of the specific details of the joint surfaces of the bones involved and because of the
ligamental and muscular structure, it acts rather like a hinge joint: its only movements are flexion and
extension in one plane. It has this latter property in common with the interphalangeal joint of the thumb,
which is classified as a hinge joint. Although the specific arrangement of muscles relating to the thumb
does allow these two distal joints of the thumb to flex and extend independently to a certain extent, I will
treat them as one joint, calling it the distal thumb joint.

The metacarpophalangeal joints of the fingers are of the ovoid type, allowing bending movement in two
planes. This is illustrated in (10).
 53

(10) Movements of the proximal finger joints (adapted from Luttgens & Wells, 1982:130)

- - - adduction extension
~

I
abduction abduction

Because the extent to which each finger can move independently varies (for all movements -- flexion,
extension, abduction, adduction), and for other reasons (cf. Ann, 1993), there does not seem to be a reason
to consider all four fingers that possess this joint as one functional group. Therefore, I distinguish between
four metacarpophalangeal joints of the fingers, calling them proximal finger joints. Thus, there is a
proximal finger joint for the index finger, the middle finger, the ring finger, and the pinky. It maybe
relevant to point out at this point that the proximal finger joints are from a physiologically different type
than the proximal thumb joint, despite the similarity suggested by the terminology.

The proximal and distal interphalangeal joints are of the hinge type like the interphalangeal joint of the
thumb; they can only flex and extend in one plane. Because of the particular organization of the muscles
and ligaments, for each finger the proximal and distal interphalangeal joints cannot flex and extend
separately: they flex and extend in synchrony (cf. Luttgens & Wells, 1982). Therefore, I will treat the two
joints as one single joint, calling it the distal finger joint. Furthermore, as the degree to which each finger
can move separate from the other fingers varies, I will distinguish between four distal finger joints: one
each for index, middle, ring and pinky finger. On the other hand, it does seem that the four distal finger
joints (interphalangeal joints of the fingers) act more as a group than the proximal finger joints (the
metacarpophalangeal joints of the fingers); just to be on the safe side, not making any generalization that
is not allowed, I will assume that there are four independent distal finger joints. Further examination of
the physiology literature is needed to indicate whether this assumption is justified or not.

Concluding this summary of the joints of the arm and hand, I will sum up the different joints (or
54

functional groups of joints) that I have distinguished on anatomical and physiological grounds. This
summary is given in (11).

(11) List of the functional groups of joints that can be distinguished for each articulator

shoulder
2 elbow
3 forearm
4 wrist
5 proximal thumb
6 distal thumb
7 proximal finger, index
8 proximal finger, middle
9 proximal finger, ring
IO proximal finger, pinky
11 distal finger, index
12 distal finger, middle
13 distal finger, ring
14 distal finger, pinky

4.3 Description of signs and their symmetry in terms of joint angles

In section 3.1, I argued that a phonetic description of signs could best be based on a relative description
of the state of the arm and hand, or a body-based description, as I call it. Also, in section 3.1 I showed
that the bilateral symmetry of the body is defined primarily in a space-based manner: the body is bilateral
symmetric as it looks the same after reflection in the sagittal plane.

As I argued above, there should be no discrepancy in the way the body is described and the way that signs
or gestures are described, if we want to make a simple description of the symmetry in a sign. Thus, I am
faced with a problem here: there is such a difference in comparing a description of the body in terms of
joint angles with a description of the bilateral symmetry of the body in terms of a reflection which leaves
the body looking the same. This discrepancy can be resolved if we make a description of body symmetry
in terms of joint angles too. As I already hinted at in chapter 3 when discussing symmetry, the body
 55

symmetry can be easily reformulated in terms of joints. We would say that the body is bilaterally
symmetric if all the joints (which indicate that there are moveable parts) that are present on two sides of
the body (as in the arms) are in the same position. With the knowledge of body positions of section 4.2
in mind, we can say that the same position means that two joints have the same amount of flexion,
extension, adduction, abduction or rotation.

Before we will discuss (a)symmetry in gestures more extensively in the next chapter, we should consider
how we are actually going to describe the state and movement of the joints. I will initially focus on the
state of the articulators, their static position, and later come back to the movement that is involved.

The bones of the human body can each be considered as bars that are attached to a fixed point. When such
a bar moves, its end shows an arc-shaped linear displacement. Such a movement is called rotary or
angular motion. Its displacement is called angular displacement and is measured in angular units, be they
degrees, revolutions, or radians; it is not appropriate to describe the movement in terms of linear
displacement with units like meters or feet. With this idea of angle and angular displacement in mind, we
can conceive of every joint as the fixed point that the bone distal to it is attached to. Thus, for every joint,
wc describe the angle that the bone distal to it make with respect to some reference state. For example,
describing the state of the elbow we need to describe the angle that the forearm makes with respect to its
reference position. If we consider the fully extended forearm as the reference state (as in (12», being 0
degrees of flexion, then if the upper arm and the forearm were at right angle this would be 90 degrees of
flexion . According to Luttgens & Wells (1982), for movements of the forearm and hand the most
convenient reference position is the 'anatomical standing position', and for all other movements of the
body the reference position is the fundamental standing position, cf. (12).
56

(12) Anatomical and fundamental standing position (adapted from Luttgens & Wells, 1982:24)

Some joints, like the elbow joint, only allow movement about one axis: the elbow can flex and extend
in one plane. For this joint we completely describe its position if we say that it is 45 degrees flexed, for
example. Other joints however may have more degrees of freedom of movement, they permit movement
about two or three axes. For example, the wrist joint shows bi-axial movement (see (8»). It can flex and
extend in two planes. The shoulder joint allows flexion and extension in two planes, and rotation in still
another (see (5». All states of the shoulder than can be considered as combinations of positions for each
axis, and thus can be described by giving three angular values.

In (13) I repeat the list of joints that may be considered for each gesture, with for each joint the number
of degrees of freedom of movement. In appendix A, a list can be found of drawings illustrating the
different joint movements with their reference position and the range of movement allowed.
 57

(13) List of the joints that will be distinguished and the number of axes about which movement is
allowed

JOINT NAME DEGREES OF FREEDOM OF MOVEMENT

shoulder 3
2 elbow 1
3 radioulnar 1
4 wrist 2
5 proximal thumb 2
6 distal thumb 1
7 proximal finger, index 2
8 proximal finger, middle 2
9 proximal finger, ring 2
10 proximal finger, pinky 2
II distal finger, index
12 distal finger, middle
13 distal finger, ring
14 distal finger, pinky

Total degrees of freedom 22

To fully specify the state of the arm and hand on one side of the body, we thus need a total of 22 angular
values; to fu lly specify both hands of a gesture, we need 44 values. In other words, there are 44 degrees
of freedom for every gesture lU • In this thesis, it obviously is impossible to make such a description for
all signs or even a large body of say 500 signs, in view of the limited time available. Also, at present there
is no possibility available to me to measure joint angles using any kind of equipment. So even if there
were time to measure a certain number of signs by hand, these would be rather impressionistic and
unreliable data.

However, for the purposes of evaluating the amount of symmetry in a sign, we can ask whether it is really

IU At the outset of this section, I listed all the joints of the arm and hand and specified the movements allowed by virtue of

the form of the bone ends (which is what determines the 'joint type'; see (15)). From this table, it would be predicted that values
would need to be given for 32 axes for both left and right limbs (the four axes of the fingers each counting four times), thus a
total of 64. Comparing this to the 44 which are in fact permitted by anatomical properties other than bone form, like ligamental
and muscular structure, we see that we have attained a reduction of 20 degrees of freedom .
58

necessary to indeed measure every joint mentioned in (13) for both articulators. As I shall propose in the
next chapter, we are merely interested in whether a joint is in the same position or not in the two limbs,
not in the actual values (for every axis) for the two joints. From now on, I will call two homologous joints
(one on the left side of the body and one on the right side) a pair of joints or joint pair. Thus, there are
14 joint pairs which are relevant to us. Of course, in the end the only way offinding out for sure whether
two values are the same is finding out what the values are and comparing them. In this thesis, as said
above, that is unfortunately unattainable. I will therefore use an impressionistic judgement of whether two
joints that form a pair are in the same position, assuming that that is a feasible approach. In the future a
more precise description of signs in the manner suggested above will be necessari I.

In this chapter, I have proposed a way to describe gestures focusing on their articulatory characteristics.
However, I have only talked about static properties of a gesture: the state of the joints at a particular point
in time. The role of movement will be discussed in the next chapter, where I consider the ways in which
gestures and signs may break the bilateral symmetry of the body.

I I Descriptions of ASL signs in terms of joints have been made, and they resemble the one I would suggest, but they either

use a lower level of detail and phonetic accuracy in terms of the number of joints involved (Nagahara, 1988), or with a somewhat
d ifferent and also less detailed use of a polar coordinate system (Eshkol, in Cohen et al., 1977:43·121). Both these descriptions
do not seem to have actually measured the values either; it is not discussed how the data were obtained.
 59

5 Symmetry breaking in sign language

In the previous chapter, I considered which joints are relevant in the production of gestures, the amount
and type of movement that is allowed for each joint, and the way to describe the state of a joint. In this
chapter, I will look at the different ways in which asymmetries may arise in gestures, and which of these
types actuall yare attested in sign languages. In section 5.1, I will present the mathematical concept of
symmetry breaking. In section 5.2, I will discuss the ways in which symmetry may be broken in gestures.
The factor time which was ignored in the previous chapter will be taken into account, and I will
distinguish between asymmetries of space and of time. In section 5.3, I will look at which types of
gestures are actually exemplified by signs from NGT.

5.1 Symmetry breaking

The ce ntral topic in the study of symmetry is 'symmetry breaking': the reduction of the symmetries in
a form . Symmetry can be disturbed (or asymmetries can arise) to form new patterns which have fewer
symmetries, but sometimes for that reason are more salient to us. As I mentioned in the previous chapter,
a perfect circle (see (1» has an infinite number of reflectional and an infinite number of rotational
sy mmetries.

( 1) Circle: infinite number of symmetries

If this circle is indented on two sides as in (2), the number of symmetries is suddenly reduced to four: the
object can be reflected about two axes running through its midpoint (one horizontal and one vertical), and
it can be rotated through 0 and through 180 degrees without changing its appearance.
60

(2) Indented circle: reduced symmetries

As Stewart & Golubitsky (1992) say, this reduction of symmetries is a

"fundamental process of pattern-formation, known as symmetry breaking. This is in itself a
paradoxical phenomenon: it occurs when a symmetric system starts to behave less symmetrically.
In some manner ( ... ) symmetry gets lost. Curiously, the typical result of a loss of symmetry is
pattern, in the sense of regular geometric form, because only seldom is all symmetry lost. An
oddity of the human mind is that it perceives too much symmetry as a bland uniformity rather
than as a striking pattern; although some symmetry is lost, pattern seems to be gained because
of this psychological trick." (1992:5)
In the next section I will look at the different ways in which symmetry can get lost in gestures. In some
cases but not in all cases will the resulting asymmetry involve the birth of new patterns.

5.2 Symmetry breaking in gestures

The starting point for our consideration of symmetry in the movements of the arm and hands is the fully
bilaterally symmetric state of the human body in standing rest position, repeated here in (3).
 61

(3) Fundamental standing position

This bilateral symmetry is the maximum symmetry that a gesture can attain. As I have indicated in the
previous chapter, I will consider symmetry to be a property of the complex of joints of the arm and hand:
the symmetric state we see in standing position is the result of the members of every pair of joints being
in the same state. As these situations (the members of a pair of joints in the same resp. a different state)
will be referred to so often below, I will from now on use the shorter terms symmetric joint pair and
asymmetric joint pairfor these states. In this way, the perceptual symmetry of the body as a whole has
been broken down into parts and formulated in terms of production and not perception: a symmetric joint
pair need not be bilaterally symmetric with respect to the sagittal plane of the body, the requirement is
that the two bones that meet at the joint are in the same relative position in each articulator, regardless
of the states of other joints. Thus, the sense of symmetry that is used here is of the metaphorical type. I
will discuss the relation between this symmetry in production and symmetry in perception in section 5.2.

The idea of the symmetry in (3) being describable as all joint pairs being symmetric has two implications.
First, for every pair there are an infinite number of completely symmetric states and of asymmetric states,
as for every joint there are theoretically an endless number of possible states (which, of course, we cannot
all distinguish by measuring, due to the obvious limitations of any equipment). Second, the symmetry of
the body as a whole can be disturbed, or broken, in numerous ways: every joint pair can be either
symmetric or asymmetric, and combinations of these may occur. Thus, the body symmetry which we see
in the standing position is broken if there is only one asymmetric joint pair, but also if two, five, or all
fourteen joint pairs are asymmetric.

So far, I have only talked about positions or states of the body being more or less symmetric. However,
since every gesture per definition contains movement (cf. p. 26), we need to consider symmetry during
62

movement, too. Movement can be considered as a (infinite) sequence of positions. To make an extremely
detailed description we would want to consider for every point in time whether or not there is symmetry
in a given joint pair. As this is impossible, at least in practice, we need to find a way to constrain the
evaluation. I suggest that we differentiate between three cases, as in (4).

(4) Possible categories in the evaluation of symmetry

1 a joint pair is symmetric during the whole sign

2 a joint pair is symmetric during partes) of the sign
3 a joint pair is not symmetric at any point in time

Thus, there is a continuum of symmetry, with 1 and 3 being the extremes and category 2 covering
everything in between.

For example, if we look at movement at the shoulder joints, three situations are distinguished. First,
starting from rest position we can abduct the humerus of both joints to the same extent, say 45 degrees,
and at the same speed. Then at all points in time, the pair of shoulder joints are symmetric. The shoulder
joints would also be symmetric if there is no movement at all during a gesture, and both joints are in the
same position. (In that case, there would have to be movement at another joint, because otherwise we
would not speak of a gesture.) If, however, we only abduct at one joint for 45 degrees, then initially the
s houlder pair is symmetric, but during all of the movement (the rest of the sign) the shoulder pair is
asymmetric. This is the situation 2 in (4); it covers gestures in which only one articulator moves but in
which there is a symmetric state once every cycle (so the symmetric state occurs twice if a movement is
repeated), as in the above example. Type 2 in (4) also covers many other gestures in which both
articulators move. For example, if starting from rest position the right arm abducts and the left arm flexes,
there is a symmetric state of the shoulder at start, but not during the rest of the sign. Finally, if at start the
shoulder joint is asymmetric, for example if there is a 30 degree supination in one shoulder but not in the
other, and subsequently only one arm abducts 45 degrees at the shoulder, then there is no symmetry at
any point in time. The same if true if both arms abduct 45 degrees, but the starting position is different.

Thus, I distinguish between three categories of symmetry for a joint pair in a certain time span, viz. the
duration of the gesture: it can be symmetric all of the time, some of the time, or none of the time. The
middle category, symmetric at some points in time, covers a large number of different possibilities
compared to the relatively simple extremes of the continuum. Within the two non-fully symmetric
categories, so everywhere outside the symmetric end of the continuum, a number of subcategories can
be distinguished, in which the factor time plays a special role. So far, I have only considered symmetry
 63

in space: sameness under reflection at a certain point in time. However, it may also be the case that
two joints are in the same position, but not at the same time. In this situation there is no symmetric
joint pair, as up until now this term was reserved for joints being in the same state at the same time.
Apart from sameness under a spatial transformation, we may also think of sameness under time
transformations. I will now first discuss the idea of time symmetries rather extensively, because of the
complexity of the notion. Then, I will explain how an understanding of time symmetries leads to an
expansion of the list of categories in (4).

If there is time symmetry, things occurring at different points in time look the same, and there is a
transformation which relates these different points in time. These transformations can be of two types:
translation or reflection (cf. Stewart & Golubitsky, 1992). One case of time symmetry is when things
look the same (or are symmetrical) under a translation in time. For example, when we think of the
earth turning around the sun in exactly a year, every year the earth returns to the same position relative
to the sun. We say that the earth's motion is periodic: the motion repeats at regular intervals; the size
of the interval or period is one year. This period of one year can be considered a translation in time;
if we look at the state of the earth's motion and look at it again after one year, it looks the same.
When we visualize time as a straight line, with intervals of one year (the period involved) marked on
it by putting a small mark every few millimeters, we can shift the line one inch to the right and see
the same state28 • This is illustrated in (5).

(5) Translation in time, visualized as a time line

position
of the sun
t t t
after translation
t t

(Every time that two time lines are shown one below the other these are intended to overlap in reality,
there being only one time line. For purposes of clarity, I often draw one line for one object and
another for the other, or for the object after the time transformation.)

Again, the most symmetric state appears not to be very salient to humans: the most time symmetric
situations are those in which there is no change of any kind over time, since there are a infinite

28 The fact that the visualization of time as a straight line is tied to current western culture (cf. Geertz, 1973; Gould,
1987) does not render the analogy worthless: we could make a comparable analogy with space based on circles rather than
lines of time, although the mathematics would be different.
64

number of transformations that can be applied without changing the appearance of the object. The
transformation in time could be something like 'look 30 minutes later', but it could also be an infinite
number of other translations. In other words, we can look at something at any time we want, and it
still looks the same: the object has not changed position (or color, or form) . For example, a large rock
which does not change its form or position over a certain time span (which is long to us, but short for
geologists, who look at rocks over much larger periods of time) looks the same to us at every moment
in time, and thus is very time symmetric. As with spatial symmetries, we do not conceive of this rock
as possessing a magnificent pattem in time. However, an object that only looks the same if we look
at it at certain points in time (which are related by a transformation) does seem to us to form a pattern.
Again, what strikes us as having a regular pattern involves fewer symmetries than what does not strike
us as having a pattern. For the above example of the rock, any translation in time leaves the rock
looking the same, not just translations in multiples of one year as in the example of the sun.

The second case of time symmetry involves a reflection in time. In this case, a particular sequence of
events or states occurs again in reverse order after a particular moment: things that first happened in
the order AB subsequently happen in the order BA. This is often used as a special effect in
commercials and TV cartoons: there we can see things like a glass of water falling on the floor and
breaking, and after that we see the glass un-breaking, filling up, and jumping back into a person's
hand again.

An example of body motion which is symmetric under a reflection in time would be if one elbow first
flexes 60 degrees and then extends 60 degrees29 • However, we are not interested here in the
movement of one object, but in the relation between the movements of two objects, the two joints in
a joint pair. The exact movement in one joint is not relevant in determining the symmetry of a joint
pair, we are interested in what relates the movement of the two joints. Given the movement of one
joint or body part, we want to find a transformation in time that gives us the correct movement of the
second joint or body part.

For example, imagine a person riding a bike. The feet, in that case, are both tracing repeated circling
movements; they show a circular type of linear displacement (see p. 54). However, they are not
performing exactly the same movement: we commonly say that the motion of the feet is 'out of phase'
or 'asynchronous'. When one foot is at its highest point in space, the other foot is at its lowest point.
What gives us the impression of the phase difference can be seen as a time translation: if we move
the motion of one foot half a period in time, we see both feet being at the same height at every point
in time. Reasoning the other way round, if we take the movement of one foot and translate it half a
period forwards or backwards in time, we get the movement of the other foot. This is illustrated in

29 This sameness under reflection in time seems to accurately describe what is called 'bidirectional movement' in the

sign phonology literature; see e.g. Perlmutter (1990).

 65

(6).

(6) Phase difference in bicycling

position of left foot ~ ~ ~ j ~

(Ii - highest point in space,
t;;;;l
L - lowest point)
L H L H L
position of right foot
I I I I I
~

Only in periodic motions does a time translation equal phase difference. We say that there is a 180
degree phase difference in this case (360 degrees being one period), or alternatively, that the relative
phase of the two objects is 180 degrees. If the movement is 'in phase', there is a relative phase of 0
degrees (cf. Rosenbaum, 1991). However, even under a time translation of half a period in the cycling
example, the feet do not look the same: there is one on each side of the body, and they are mirror
images of each other. Thus, there is also a spatial symmetry here: the reflectional symmetry of the
body is combined with the temporal symmetry of the out-of-phase movement (cf. Stewart &
Golubitsky, 1992). In our use of joint pairs, the symmetry is already 'built in', in the sense that
sameness of two joints in a pair implies that the joints are bilaterally symmetric in their position. Thus,
it is sufficient to look at the time transformations here as they apply to joint pairs.

Although here I have talked about periodic motions, time translations relating two objects do not
presuppose periodic motion of each object. For example, it can also be the case that one articulator
performs a certain non-periodic action, which is executed by the other articulator some time later, as
illustrated in (7).

(7) Time translation relating two non-periodic motions

position of object 1

position of object 2 __ _____11-_'.1.---11.0000. _ _0

~
66

In the example of the movement of the feet when bicycling, the movement of one foot is translated
in time for half a period to yield the movement of the other foot. If we apply a reflection in time to
a (non-periodic) movement of one joint in a pair to yield the movement of the other joint, we get the
situation in (8).

(8) Reflection in time, two joints moving

state of left joint t I

~f--
state of right joint 1 j

The fact that the movement of one joint occurs after the movement of the other is a property of
reflection in time, it does not imply that an extra translation has taken place. It is, however, possible
to combine a reflection in time with a translation in time. If we take the situation in (9) as the starting
point to which a translation can be applied, we can get the situation in (9).

(9) Combination of reflection and translation in time

state of left joint t r f

c B A
state of right joint I I I

Note that this combination is more restricted if only one object moves: a translation applies to the
whole time line, so that overlapping positions as in (9) cannot occur; in other words, one object cannot
be in two places at the same time.

Things get more complicated if within the movement of each joint considered separate there is a
 67

reflection in time. This is what happened in the above example of one elbow joint first flexes 60
degrees and then extends 60 degrees, or more generally, in all bidirectional signs. If this movement
is repeated and occurs in both hands, and there also is a 180 degree phase difference, we get an
untransparent situation which is illustrated in (10).

(10) Time translation relating the two articulators in repeated bidirectional movement

\\;: B

State of left elbow joint 1 ~ 1 ~ 1 ~ 1

time )

State of right elbow joint ~ 1 ! 1 ~ 1 ~

time )

Here it seems as if a reflection may be involved in relating the two elbow joints. However, since the
reflection is also a property of the movement of one hand, a time translation suffices to get the out-of-
phase pattern between the two hands. A time reflection is present in the sign, but it describes the
bidirectional movement of each hand seen in isolation. I will assume that in cases like this the time
reflection is not a property of the relation between the two hands.

Summarizing, the relevant result of a time reflection is that the sequence of events is reversed, and the
result of a time translation is that the interval of time in which something happens changes its location
in time, keeping the order of events within the interval constant.

How exactly can these types of time symmetry be said to occur in our description of arm and hand
movements? In the case of spatial symmetries, we initially determined that a specific joint pair is
symmetric if its members are in the same state at a certain point in time. By evaluating symmetry at
a point in time, we actually ignored or eliminated the aspect of time. To take into account this factor
time (and thus, also movement), I have distinguished above between three different possibilities of
symmetry of a joint pair over time: it can be symmetric all the time, part(s) of the time, or at no time
at all. When we look at what happens to the symmetry of a joint pair if one of the above time
symmetries is applicable, we see that there is either none or some point(s) in time in every period at
68

which the joint pair is symmetric, but not during the whole sign30 • Therefore, the time symmetries
are subcategories of the second or the third type of spatial symmetry in (4). These two types were
considered above as instances of symmetry breaking, and now we have seen that there are three
situations in which this symmetry breaking involves 'new' patterns: if a time translation, a time
reflection, or a combination of these two is involved. In all the other instances where a joint pair is
symmetric at none or some point(s) in time there is no specific new pattern involved. Thus, they are
somewhere in the middle of the continuum from fully symmetric to fully asymmetric.

I will consider the two types of time symmetry (translation and reflection) and the combination of two,
as three different types of symmetry, that can be added to the three different classes in (4). Adding
these time symmetries to the typology in (4) leads to the final classification of the amount and sort
of symmetry in a joint pair, as in (11).

(11) Types of symmetry of a joint pair during a gesture

fully symmetric at all points in time

2 symmetric at one or some (but not all) points in time, but no involvement of a time
symmetry
3 symmetric (at all points in time) under a time translation
4 symmetric (at all points in time) under a time reflection
5 symmetric (at all points in time) under a combination of a time translation and a time
reflection
6 not symmetric at any point in time, and no involvement of a time symmetry

The first case or type is the 'default' state in which there is maximum symmetry. Types 2 and 6 are
examples of a joint pair being asymmetric in the sense that at some or all points in time the members
of a joint pair are in a different state. Types 3, 4, and 5 are examples of a the members of a joint pair
being in the same state during the whole gesture, thus essentially the members of a joint pair being
symmetric, but not at the same time. In these three cases, after the time transformation has applied the
joint pair is symmetric at all points in time. Thus, we might say that in such a gesture the joint pair
is fully symmetric under a transformation in time.

30 Generally. in oscillating or periodic movements of a joint pair that is out-of-phase. there is one or two points in time
during every period in which the joint pair is symmetric. In certain out-of-phase movements there is no point in time in
which the joint pair is symmetric, in the holistic way of looking at a joint's movement that is used here, where we have
abstracted from the different axes that may be involved in the movement at one joint. In cases where there is movement
about two axes, as in a circumdicting movement of the wrists (combined flexion-extension and lateral flexion-extension; see
p. 50), there is no point in time where the joint pair is symmetric in case of a phase difference. If we would look at the level
of detail of the axes of movement, and not at the more abstract level of the joint as a whole, we would indeed see that for
every axis there is one point in time during every period at which the position is the same. Therefore. I will maintain that
for every out-of-phase movement there is a point in time at which the joint pairs are symmetric.
 69

We can summarize and clarify the inclusion of the time symmetries in the symmetry continuum (which
already included types 1, 2 and 6) by looking at the different ways in which the time symmetries can
be present; they can both be present or absent, leading to the four-way division in (12)31.

(12) Presence and absence of the two time symmetries

time translation

absent present

absent types 1,2,6 type 3

time reflection
present type 4 type 5

Although above I have used specific sizes of the time translations in the examples, in the listing in
(12) nothing is said about the size of the time transformations. This is on purpose, as we cannot
predict in advance which kinds of time transformations wiII occur. In principle, anything can occur,
and the sizes that do actuaIIy occur in sign language remain to be determined by looking at sign
language data.

Concluding, every gesture consists of the coordination of 14 joint pairs. Every joint pair can be in one
of six different kinds of states during a gesture. Thus, the descriptive system that I have developed
here is capable of describing 6 14 types of gesture, a huge number of over 78 miIIion. The actual
number of possible gestures is far larger, as each type of gesture may contain a large number of
different instantiations. The level of detail that we have attained is indeed enormous compared to the
four possible types of symmetry at a phonological level considered by Battison (1978). Now, the
interesting question becomes which subset of these possible types of gesture actually occur in sign
languages. This question will be addressed in the next section.

5.3 Symmetry breaking in NGT signs

5.3.1 Description of data set

In the rest of this chapter, I will look at a corpus of NGT signs and determine the type of symmetry

31 Although I speak here of a 'continuum' of symmetry, I do not wish to claim that the order found in (11) actually

corresponds to a ranking from the most simple to the most difficult to articulate. The ranking in (11) roughly expresses the
amount of symmetry, with 1 being the most symmetrical.
70

for every joint pair. The corpus consists of 320 two-handed signs that are used in the Amsterdam
dialect of NGT, which I gathered over the past few years. I have no idea how large a subset of all the
two-handed signs in NGT this is; these simply are all the signs I could find. Signs whose form I did
not precisely remember with the help of the picture or notation were checked with an informant. Signs
for country names were put aside, because they are often recent borrowings from other sign languages,
as were signs which did not have a distinct citation form, but whose form is highly dependent on the
sentence context (GANG 'hall', STRAAT 'street'). For directional transitive verbs the citation form
was used, which generally is the form for first person singular subject and second person singular
object. The whole corpus can be found in appendix B, where glosses are listed in alphabetical order,
accompanied by a translation of the gloss into English.

5.3.2 Further restriction of diversity

For practical reasons, I will need to restrict the description of each sign to a small number of joint
pairs. Without any phonetic motivation, I will focus my attention on the four most proximal joints:
shoulder, elbow, radioulnar, and wrist joints. I will come back to the other joints in the discussion that
follows the analysis of the data. For every joint pair, I will search for examples of the five different
types of symmetry breaking in (11), see whether they can occur as the only case of asymmetry in a
sign, and investigate with which asymmetries in other joint pairs they may combine. The completely
symmetric state of a joint pair, type 1 in (11), is considered the default situation: if all four joint pairs
are in this state we get fully symmetric signs. All other signs are considered as instances of symmetry
breaking. The main purpose will be to make an inventory of which sorts of symmetry (cf. (12» occur
for each joint pair. Depending on the number of different types of signs that occur, the character of
combinations of asymmetries for different joint pairs will be evaluated. (Now that we have limited the
number of joint pairs to four, the number of possible types of signs is still considerable: 64 = 1296.
The task of evaluating all combinations may tum out to be too large.)

Before we start to look at the data, there is one more simplification that I would like to make. A
certain amount of asymmetry (of type 6) is always present if both articulators cross the sagittal plane:
as the hands are unable to penetrate each other, the hands can only both cross the sagittal plane with
all joint pairs being symmetric if a small adaptation is made. The hands can be above each other or
in front of each other, as in HELPEN and JARIG, respectively, or the fingers of both hands can
'intermingle', as in the signs KERK and MACHINE. In all of these cases, either the elbow or the
shoulder joint pair or both are asymmetric, but the difference is only marginal, just enough to permit
both hands to cross the plane; this implies that in all of these cases the hands are in contact, because
that is the closest the hands can get to each other. Instead of counting these asymmetries as a type 6
asymmetry or treating them as a separate type of asymmetry, I will ignore these minor differences, and
 71

pretend that the relevant joint pairs are symmetric at the point in time where the hands are in close
contact. For asymmetries to be part of this group, the requirement is that both hands cross the sagittal
plane, and to the same extent; in all other cases, there is a real asymmetry which is not necessarily
present. Finally, note that signs with this asymmetry are not excluded from consideration, it is merely
assumed that in these cases the joint pair is fully symmetric.

5.3.3 Taxonomy of the asymmetric signs in the corpus

I will treat every joint pair, one at a time, starting with the most proximal. For every type of
symmetry, a list of signs will be given, following the order of the typology in (11). If a joint pair is
fully symmetric (type 1), then a distinction will be made between cases in which there is movement
involved and cases in which there is not. If a joint pair is not fully symmetric (types 2 to 6), then a
distinction will be made between signs in which it is the only asymmetry (after 'only' in the list
below) and signs in which other joint pairs are asymmetric, too (marked 'combi'). In appendix C, this
list is given in an alternative format, with the signs listed alphabetically accompanied by the symmetry
types of the four joint pairs for that sign. In the list below, all signs in which all of the joint pairs are
fully symmetric are left out. The reason is that this specific type of sign occurred very often: 181 signs
in the whole corpus were fully symmetric. These signs are listed separately in appendix D. Thus, all
the signs that are listed in the taxonomy below are asymmetric to some extent. This taxonomy is given
in this section; a discussion of these data and of the fully symmetric signs follows in section 5.3.4.

Shoulder joint

1 Fully symmetric at all points in time

moving CONTACT, GIDS, VRIEND
not moving AARDAPPEL, BEWUS, COMB INATIE, COMMUNICATIE, DEUR,
DISCUSSIE, ENE-OOG-IN-ANDERE-OOG-UIT, GENEN, HARD,
KEUKEN, OP-SLOT, OVERLEDEN, TOLK, VERANDEREN,
VERTROUWEN, WINKELIBOODSCHAPPEN

2 Symmetric at some point(s) in time

only
combi BOUWEN, BRAND, BROOD-2, DOORGAAN, GELIJK-HEBBEN,
GROEN/GROENTE, HALF, INHOUD, JAMMER, KAAS-l, KOPIeREN, KROKO-
DIL, LAATSTE, MES, NULPAARD, PAARS, PRINCIPE, RECHT-HEBBEN-OP,
STEMPEL, TUSSEN, WERKEN, ZEKERIVAST, ZIELEPOOT
72

3 Symmetric under time translation

only VERKEER
combi AFBLUVEN, AGRESSIE, AUTO, BAD, BALLET, BEER-2. BUS, FABRIEK.
FIETSEN,GEBAREN,GEBEUREN,GEDULD,GEHAKT,GESPREK,HANDEN-
WASSEN, HUll..,EN. IN-DE-WAR, INDUSTRIE, INV ALIDE, KOKEN, LUDEN,
MAAKT-NIET-UIT, MELK, MIMIEK, NAAR-BED-GAAN, ONZEKER,
OVERSTUUR, RECHT(SPRAAK), RELATIE, RESTAURANT, ROBOT, ROL,
ROMMEL, RUILEN, SCHAATSEN, TENTOONSTELLING, THEATERfTONEEL,
VERHAAL,VERTELLEN,VERZORGEN,WEDSTRUD,WINTER

4 Symmetric under time reflection

only
combi ENE-OOR-IN-ANDERE-OOR-UIT

5 Symmetric under time translation and time reflection

only VERHUIZEN
combi

6 Fully asymmetric
only
combi ADRES, BABY, BANAAN-2, BEGELEIDEN, BETALEN, BINNEN, BOOM, BORD,
BOS, BOTER, BRIEF, CAFE, CHIRURG, COMPUTER, CONTRACT, DANSEN,
EGEL, ELASTIEKJElSLAPPELING, FOUT, GEBRUIKEN, GRADEN, HERFST,
ISOLEREN, KAAS-2, KIP, KLETSEN, KOFFIE, LATEN-ZIEN, LESGEVEN,
LEZEN, LUST, LUCHTBALLON, OMGAAN-MET, ONDERSTEUNEN, PADDE-
STOEL, PARACHUTE, PIZZA, POLITIEK, PSYCHOLOGIE, REGELS, SCHOOL,
SCHOPPEN, THEE, THERAPIE, TUD, TUNNEL, VERDELEN, VOETBAL,
VOORBEELD, WEGGAAN, ZIEK, ZIEKENHUIS

Elbow joint

1 Fully symmetric at all points in time

moving BEGELEIDEN; BRIEF, CONTACT, GIDS, VERTROUWEN
not moving AARDAPPEL, AFBLUVEN, COMBINATIE, DEUR, DISCUSSIE, EGEL,
GENEN, KEUKEN, OP-SLOT, TOLK, VERANDEREN, VERHUIZEN,
VERKEER,VRIEND.~KEUBOODSCHAPPEN,OVERLEDEN

2 Symmetric at some point(s) in time

 73

only BEWIJS, HARD

combi BETALEN, BINNEN, BORD, BOUWEN, BRAND, BROOD-2, CAFE, DOOR-
GAAN, GELIJK-HEBBEN, GROEN/GROENTE, HALF, INHOUD, JAMMER,
KAAS-l, KOPIeREN, KROKODIL, LAATSTE, MES, NIJLPAARD, PAARS,
PIZZA, PRINCIPE, RECHT-HEBBEN-OP, STEMPEL, TUNNEL, TUSSEN, WER-
KEN, ZEKERJVAST, ZIELEPOOT

3 Symmetric under time translation

only COMMUNICATIE
combi AGRESSIE, AUTO, BAD, BALLET, BEER-2, BUS, FABRIEK, FIETSEN,
GEBAREN, GEBEUREN, GEDULD, GEHAKT, GESPREK, HANDEN-WASSEN,
HUILEN, IN-DE-WAR, INDUSTRIE, INVALIDE, KOKEN, LIJDEN, MAAKT-
NIET-UIT, MELK, MIMIEK, NAAR-BED-GAAN, ONZEKER, OVERSTUUR,
RECHT(SPRAAK), RELATIE, RESTAURANT, ROBOT, ROL, ROMMEL, RUILEN,
SCHAATSEN, TENTOONSTELLING, THEATERITONEEL, VERHAAL,
VERTELLEN, VERZORGEN, WEDSTRIJD, WINTER

4 Symmetric under time reflection

only ENE-OOG-IN-ANDERE-OOG-UIT
combi ENE-OOR-IN-ANDERE-OOR-UIT

5 Symmetric under time translation and time reflection

only
combi

6 Fully asymmetric
only
combi ADRES, BABY, BANAAN-2, BOOM, BOS, BOTER, CHIRURG, COMPUTER,
CONTRACT, DANS EN, ELASTIEKJElSLAPPELING, FOUT, GEBRUIKEN,
GRADEN, HERFST, ISOLEREN, KAAS-2, KIP, KLETSEN, KOFFIE, LATEN-
ZIEN, LESGEVEN, LEZEN, LIJST, LUCHTBALLON, OMGAAN-MET,
ONDERSTEUNEN, PADDESTOEL, PARACHUTE, POLITIEK, PSYCHOLOGIE,
REGELS, SCHOOL, SCHOPPEN, THEE, THERAPIE, TIJD, VERDELEN,
VOETBAL, VOORBEELD, WEGGAAN, ZIEK, ZIEKENHUIS

Radioulnar joint

1 Fully symmetric at all points in time

74

moving
not moving AFBLIJVEN, AGRESSIE, AUTO, BAD, ' BALLET, BEER-2, BEWIJS,
BINNEN, BOUWEN, BUS, COMMUNICATIE, DEUR, DISCUSSIE, DOOR-
GAAN, LAATSTE, ENE-OOG-IN-ANDERE-OOG-UIT, ENE-OOR-IN-AN-
DERE-OOR-UIT, FABRIEK, FIETSEN, GEBAREN, GEBEUREN,
GEDULD, GELUK-HEBBEN, GESPREK, GROEN/GROENTE, HALF,
HANDEN-WASSEN, HARD, HUILEN, IN-DE-WAR, INDUSTRIE,
INVALIDE, ISOLEREN, JAMMER, KAAS-I, KOKEN, LIJDEN, LIJST,
MAAKT-NIET-UIT, MELK, MES, MIMIEK, NAAR-BED-GAAN,
ONZEKER, OVERSTUUR, PAARS, PRINCIPE, PSYCHOLOGIE,
RECHT(SPRAAK), RECHT-HEBBEN-OP, RESTAURANT, ROBOT, ROL,
ROMMEL, RUILEN, SCHAATSEN, TENTOONSTELLING, THEATERtrO-
NEEL. TUD. TUNNEL, VERHAAL, VERHUIZEN, VERKEER,
VERTELLEN, VERZORGEN, WEDSTRUD, WEGGAAN, WERKEN, WIN-
KELiBOODSCHAPPEN, ZEKERIV AST, ZIEKENHUIS, ZIELEPOOT

2 Symmetric at some point(s) in time

only AARDAPPEL. OP-SLOT
combi BOOM, BOS, BOTER, CAFE, CONTRACT, EGEL, POLITIEK

3 Symmetric under time translation

only COMB INATIE, KEUKEN
combi

4 Symmetric under time reflection

only
combi

5 Symmetric under time translation and time reflection

only VERANDEREN
combi ' BRAND, GENEN, OVERLEDEN

6 Fully asymmetric
only CONTACT, GIDS, TOLK, VERTROUWEN, VRIEND
combi ADRES, BABY, BANAAN-2, BEGELEIDEN, BETALEN, BORD, BRIEF, BROOD-
2, CHIRURG, COMPUTER, DANSEN, ELASTIEKJElSLAPPELING, FOUT,
GEBRUIKEN, GEHAKT, GRADEN, HERFST, INHOUD, KAAS-2, KIP, KLETSEN,
KOFFIE, KOPIeREN, KROKODIL, LATEN-ZIEN, LESGEVEN, LEZEN,
LUCHTBALLON, NIJLPAARD, OMGAAN-MET, ONDERSTEUNEN,
 75

PADDESTOEL, PARACHUTE, PIZZA, REGELS, RELATIE, SCHOOL,

SCHOPPEN, STEMPEL, THEE, THERAPIE, TUSSEN, VERDELEN, VOETBAL,
VOORBEELD, WINTER, ZIEK

Wrist joint

I Fully symmetric at all points in time

moving
not moving AARDAPPEL, AGRESSIE, AUTO, BAD, BALLET, BEER-2, BEWIJS,
BOOM, BOS, BOUWEN, BRIEF, BROOD-2, BUS, COMBINATIE,
COMMUNICATIE, CONTACT, DOORGAAN, LAATSTE, ENE-OOG-IN-
ANDERE-OOG-UIT, ENE-OOR-IN-ANDERE-OOR-UIT, FABRIEK,
FIETSEN, GEBAREN, GEBEUREN, GEDULD, GEHAKT, GELUK-
HEBBEN, GESPREK, GIDS, GROEN/GROENTE, HANDEN-WASSEN,
HARD, HUILEN, IN-DE-W AR, INDUSTRIE, INHOUD, ISOLEREN,
KAAS-l, KEUKEN, KOFFIE, KOKEN, LUDEN, MELK, MES, MIMIEK,
NAAR-BED-GAAN, ONDERSTEUNEN, ONZEKER, OP-SLOT,
OVERSTUUR, PAARS, PADDESTOEL, POLITIEK, PSYCHOLOGIE,
RECHT(SPRAAK), RECHT-HEBBEN-OP, RELATIE, RESTAURANT,
ROBOT, ROL, ROMMEL, RUILEN, SCHAATSEN, STEMPEL,
TENTOONSTELLING, THEATERfTONEEL, TOLK, TUSSEN, VERHAAL,
VERHUIZEN, VERKEER, VERTELLEN, VERTROUWEN, VERZORGEN,
VRIEND, WEDSTRUD, WERKEN, ZEKERIV AST, ZIELEPOOT

2 Symmetric at some point(s) in time

only DEUR
combi BINNEN, BOTER, BRAND, FOUT, HALF, PRINCIPE, SCHOPPEN, run,
TUNNEL,VERDELEN,VOETBAL,WEGGAAN

3 Symmetric under time translation

only DISCUSSIE, WINKELIBOODSCHAPPEN
combi AFBLUVEN, INV ALIDE, MAAKT-NIET-UIT, OMGAAN-MET

4 Symmetric under time reflection

only
combi

5 Symmetric under time translation and time reflection

76

only
combi GENEN, OVERLEDEN, VERANDEREN

6 Fully asymmetric
only
combi ADRES, BABY, BANAAN-2, BEGELEIDEN, BETALEN, BORD, CAFE,
CHIRURG, COMPUTER, CONTRACT, DANSEN, EGEL, ELAS-
TIEKJElSLAPPELING, GEBRUIKEN, GRADEN, HERFST, JAMMER, KAAS-2,
KIP, KLETSEN, KOPleREN, KROKODIL, LATEN-ZIEN, LESGEVEN, LEZEN,
LUST, LUCHTBALLON, NULPAARD, PARACHUTE, PIZZA, REGELS, SCHOOL,
THEE, THERAPIE, VOORBEELD, WINTER, ZIEK, ZIEKENHUIS

5.3.4 Discussion of the data

In this section, I will first look at the joints and at the symmetry types 'in isolation', and then look
at the different combinations that occur. As a quick look at the above listing can tell us, the signs are
not evenly spread over the different categories. Thus, for each joint pair some types of asymmetry are
more common than others. In the table in (13), the number of signs that have been found for each type
of symmetry are listed; as different joint pairs in a sign can be of the same type, the column totals can
be higher or lower than the total number of signs involved, viz. 139. Remember that these only include
the asymmetric signs; the 171 fully symmetric signs, in which all joint pairs are of type 1, are left out.
The numbers for the types are equal to those in the list in (11) above; this list is repeated in (14), in
modified form but with the same order. These type numbers will be used throughout this chapter.
 77

(13) Number of signs for each type of symmetry per joint pair (n=139)

symmetry type 1 2 3 4 5 6 total

II
shoulder 19 23 43 1 1 52 139
"
elbow 20 31 42 2 - 44 139

radioulnar 72 9 2 - 4 52 139

wrist 78 13 6 - 3 39 139

total
11
189 I 76 93 3 8
II
556

(14) Types of symmetry of a joint pair during a gesture

fully symmetric at all points in time

2 symmetric at some point(s) in time
3 symmetric under a time translation
4 symmetric under a time reflection
5 symmetric under a combination of a time translation and a time reflection
6 not symmetric at any point in time

Distribution of symmetry types overall

Looking at the lower extreme of the distribution of the six different symmetry types (the column
totals), we see that the cases where a time reflection is involved are very limited (resp. 3 and 8
occurrences). The two signs that involve a time reflection only (type 4), ENE-OOG-IN-ANDERE-
OOG-UIT and ENE-OOR-IN-ANDERE-OOR-UIT, are peculiar signs in that they may well be 'joke
signs', and not regular parts of the lexicon; I have not checked this with native signers yet. Type 5
joint pairs occur somewhat more often: the signs GENEN, OVERLEDEN and VERANDEREN have
this type of asymmetry for both the radioulnar and wrist joints. In only one case, there is movement
at the shoulder joint which is of type 5, in the sign VERHUIZEN; it is not clear to me whether this
is a directional verb. In the Groningen dialect of NGT, the sign BOOT also has this type of asymmetry
for the shoulder joint, with both hands moving from left to right through the space in front of the
. body. That suggests that type 5 is indeed an allowed type of asymmetry for the shoulder joint in NGT,
and that other examples may be found. Still, signs in which a reflection in time is involved are
relatively rare.

Focusing on the higher extremes of the distribution of symmetry types, we see that both the fully
78

symmetric type I and the fully asymmetric type 6 are very frequent (resp. 189 and 187 occurrences).
For the fully asymmetric joints, the number is more or less evenly spread over the different joints. For
the fully symmetric case, however, it is mainly the radioulnar and wrist joint pairs that are in this
position, over three times as often as the shoulder and elbow joint pairs. As can be seen in the list in
the previous section, in all these cases of symmetric wrist and radioulnar pairs the joints are not mo-
ving, and thus there needs to be movement in either the shoulder or the elbow pair.

In between the two extremes, we find the two remaining symmetry types 2 (symmetry at some point)
and 3 (time translation), resp. 77 and 93 occurrences. Both types occur more frequently for the
shoulder and elbow joints than for the radioulnar and wrist joints, although this difference is largest
for type 3, the time translation type.

Distribution of symmetry types per joint pair

For the shoulder pair, the fully symmetric type I is least frequent, and there are only a few more signs
in which the shoulder joint is symmetric at some points in time (type 2). The slightly more common
state for the shoulder pair is either fully asymmetric (type 6) or translated in time (type 3); however,
the differences are not very large. The same distribution is found for the elbow joints, except that the
differences here are even smaller. For the radioulnar and wrist pairs, the distribution is roughly the
same for both, but different from the shoulder and elbow pairs. Here, the most frequent situation
(about 50% and 55% of all signs, resp.) is that of complete symmetry (type 1). Around 35% and 30%,
respectively, of all signs have completely asymmetric (type 6) radioulnar and wrist joints. As we saw
above, for both types and in most cases the radioulnar and wrist joints do not move. In general, it can
be said that the radioulnar and wrist joint pairs are less frequently asymmetric than the shoulder and
elbow joint pairs.

Combinations of symmetry types

So far, I have looked at frequencies of types for single joint pairs, without considering the
combinations of pairs that do occur. The table in (13) does not provide any information about which
types of symmetry combine in a sign. In (15), a list is given of all the different types of signs that
result from the combination of the four joint pairs which can all be in one out of six different
symmetry states. The numbers for the types refer to the list in (14).
 79

(15)

should. elbow r.ulnar wrist number of

type type type type signs
f:~',;'ri:; .i'3.~~
:e:. ;. _.~.;? .• ~l ~~~:;]tqlre.f'i;; . ~~~~~4 ,K~~lMi'4.~
1 1 1 3 2
1 1 2 1 2
1 I 3 1 I
1 1 5 5 3
1 1 6 1 5
1 2 1 1 2
1 3 1 1 1
1 4 1 1 1
2 2 I I . 12
2 2 I 2 2
2 2 1 6 I
2 2 5 2 1
2 2 6 1 4
2 2 6 6 3
3 1 1 1 1
3 1 1 3 1
3 3 1 1 36
3 3 1 3 2
3 3 6 1 2
3 3 6 6 1
4 4 1 1 1
5 I 1 I 1
6 1 2 6 I
6 I 6 6 1
6 2 I 2 2
6 2 2 6 I
6 2 6 6 3
6 6 1 1 2
6 6 1 2 2
6 6 1 6 2
6 6 2 I 3
6 6 2 2 1
6 6 2 6 1
6 6 6 1 2
6 6 6 2 4
6 6 6 3 1
6 6 6 6 26

As was mentioned above, the number of logically possible different types of gestures is 64 , which
equals 1296. Of these, only 38 (3%) actually occur in my NGT corpus.

We have already seen that where the frequency of symmetry types is concerned, regardless of the
80

exact combinations, on the whole the shoulder and elbow pairs tend to cluster together, as do
radioulnar and wrist pairs: the distribution of signs over the two groups of pairs is comparable. Is this
also true if we look at the exact combinations? In other words, is it the case that the shoulder and
elbow joint pairs tend to be of the same symmetry type, and likewise for the radioulnar and wrist
joint? That need not necessarily be the case; for example, it may be that signs which have a time
translation at the shoulder (43 signs in total) are other signs than those which have a time translation
at the elbow (42 in total). From the above table it can be calculated how many signs have different
joint pairs in the same state. For cases in which two joints are in the same state, the numbers can be
found in (16).

(16) Two joint pairs being of the same symmetry type

combination number of signs (n=139)

shoulder - elbow 124 (89%)
shoulder - radioulnar 45 (32%)
shoulder - wrist 53 (38%)
elbow - radioulnar 41 (29%)
elbow - wrist 48 (35%)
radioulnar - wrist 95 (68%)

Indeed, we find here that the shoulder and elbow pairs and the radioulnar and wrist pairs are of the
same type in a larger number of signs than the other four combinations. In 86 of all 139 asymmetric
signs (62%), both the shoulder-elbow group and the radioulnar-wrist group are of the same type
(though the type of the two groups may differ).

In the list of occurring types in (15), we also see that there is a maximum to the number of different
symmetry types that may be combined in a sign. The maximum possible number of types is four, as
there are four joint pairs and more than four symmetry types, but at most we find three different types
combined within one sign. Signs with three different symmetry types occur only 15 times in the corpus
of 139 signs; in these 15 signs, at most one joint pair is fully symmetric (type 1), and one or two joint
pairs is fully asymmetric (type 6).

Looking at the frequencies of the different types in (15), we notice that there are only a few types that
occur quite often: type 2211 (12 times; e.g. PAARS), type 3311 (36 times; e.g. FABRIEK), and type
6666 (26 times; e.g. CHIRURG). The most frequent type, of course, is 1111, the fully symmetric
signs, of which there are 181. All the other types occur only a couple of times, and many of them
occur only once. Now, what are the characteristic properties of these four types? It seems to me that
what unites these four types is that there is the grouping of both radio-ulnar and wrist pairs and the
 81

shoulder and elbow pairs, which is a characteristic of a large majority of the asymmetric signs, as we
saw above. This grouping also occurs in a number of other types, namely type 1155 (3 signs), 2266
(3), 3366 (1), 4411 (1), 6611 (2), and 6622 (1). These combinations however, all have one of two
specific characteristics that might be involved in their infrequent occurrence: either a very infrequent
type overall is involved (type 4 or 5), or one of the two groups of joint pairs is of the 6 type (2266,
3366,6611, 6622) . Apparently, the occurrence of type 6 is mostly limited to the cases where all joints
are of type 6, that is, signs which are maximally asymmetric.

A note on the neglected joints

I have neglected in this chapter all of the joints of the fingers and thumb. Although I have not studied
these joints systematically in a group of signs, it seems at first sight that without exception, all these
joints are fully symmetric (type 1) during signs that are otherwise fully symmetric (signs of type
1111). In signs that are asymmetric in at least one of the four most proximal joints, the finger and
thumb joints can be asymmetric too. Further, there do not seem to be many examples of any time
asymmetries in the finger or thumb joints. An old version of the NGT sign MELK (cf. Koenen et ai.,
1993) seems to be the only exception, showing out of phase movement not only at the shoulder and
elbow, but also for all of the finger and thumb joints.

5.3.5 Realization of symmetry types

The main purpose of this chapter was not to look at the actual realization of a sign, but rather to focus
on the type of symmetry that is present in the sign. In this separate section, I will briefly mention
some generalizations on the realization of the above symmetry types. These generalizations will all
be further discussed and evaluated in the next chapter. I will pay attention to two aspects: first,
whether there is movement in part of a joint pair or in the whole joint pair, and second, on
'cooperation' of different joint pairs.

Movement or no movement
For joint pairs of!Yill:...!, we can see in the list in section 5.3.1 that in almost all cases the joints are
not moving, but that they are in a static position; the few exceptions to this generalization occur for
the shoulder and elbow joints.

As for the realization of ~ joint pairs being symmetric at some points in time, we see a striking
regularity as well. With the exception of two signs (BOS and BRAND), in all signs (not merely in a
joint pair) where one or more joint pairs are of type 2, only one articulator moves, and the other
82

articulator is held steady. For both exceptions the joints which have movement in both articulators are
not of type 2. Thus, although in principle type 2 asymmetry may mean that both articulators move (but
in different directions, for example), this never occurs: only one articulator moves in signs. Of the two
exceptional signs, BOS may be a morphologically complex fonn: it seems equal to the plural of
BOOM. The other exception, BRAND, is also exceptional in that one joint pair is of the infrequent
type 5.

For the types 1, 3, 4, and 5, per definition both articulators move: the time transfonnations of 3, 4,
and 5 can only occur if both articulators move, and joint pairs of type 1 would not be fully symmetric
if one and only one articulator moves (nor would it be if both joints move but in different ways; this
is covered by type six joint pairs). That leaves us with types 2 and 6, the joint pairs being symmetric
at some points in time or during all of the time. These two types allow equal movement of both joints
in a pair (which necessarily would mean different movement, because otherwise it would be a type
1 joint) or movement of only one joint in a pair. However, just as for type 2, a large majority of joint
pairs of!YQti show no movement at all or movement in only one joint of the pair (164 out of 187).

Concerning the realization of the remaining types 3, 4, and 5, which all involve a time transfonnation
and which can be seen as subsets of type 2 or type 6, I will disregard here ~ in view of the
limited number of occurrences in the corpus. As for the realization of joint pairs with a time translati-
on, thus of!YillL1 it can be remarked that in all occurrences the motion is periodic. Thus, we can
speak of a phase difference in all of type three joint pairs. The phase difference for type 3 pairs is
always half a period, or 180 degrees, never more or less. In principle, all possible sizes of phase
difference could be found in sign language; this possibility appears not to be used. The translation part
of ~ joint pairs concerns non-periodic motions in all cases. Remember that the reflection that is
involved in type 5 pairs implicates that movement of the two joints occur subsequently, and that the
translation which is applied to it can be of any size; it appears that the translation is always such that
the beginning and end of the movement of the respective joints coincide, so the movements of the two
joints occur simultaneously and not subsequently.

Cooperation of joint pairs

In the previous section I already noted that for most signs, the shoulder and elbow joint pairs and the
radioulnar and wrist pairs have the same type of symmetry (for resp. 87 and 67% of all asymmetric
signs). If we look in detail at the movement of the pairs within these two groups, we see that in the
many cases where both shoulder and elbow pairs are of type 3 (phase difference), the shoulder and
elbow cooperate to create one perceptual movement pattern, either the hand moving in circles or along
straight lines up and down. A parallel situation is found in the few cases in which both radioulnar and
wrist joints are of type 5 (GENEN, OVERLEDEN, VERANDEREN). Here, the perceptually largest
movement occurs in the radioulnar pair, and the wrist movement seems to make smaller compensatory
 83

movements, keeping the attitude (or orientation; both are perceptual concepts) of the hand constant on
one dimension, while focusing the attention on the movement on the other two dimensions.

5.3.6 Summary

In this chapter, we have seen that where symmetry is concerned the number of actual types of sign
in NGT is very limited compared to the possible number of signs. A majority of the two-handed NGT
signs are fully symmetric. If a sign is asymmetric, then there are a few types of asymmetry that are
preferred: signs which are asymmetric for all four joints (type 6666), signs with a phase difference in
the shoulder and elbow pairs but otherwise fully symmetric (type 3311), and signs which are
symmetric only part of the time at the shoulder and elbow joints but all of the time at the radioulnar
and wrist joints (type 2211). More in general, the tendency is towards having the same type of
asymmetry at the shoulder and elbow joints, and at the radioulnar and wrist joints, respectively. In
many cases, the shoulder and elbow joint function together to perceptually form one movement shape,
as do the radioulnar and wrist joints. Asymmetries that involve a reflection in time (types 4 and 5)
hardly occur at all.

In the next chapter, I will discuss the implications of the data and patterns described in this chapter
for phonological models, and I will consider possible phonetic explanations for the observed patterns
in terms of minimization of articulatory and perceptual complexity.
84
 85

6 Theory and explanation

In this chapter, I will discuss the consequences of the findings in the previous chapter. First, in section
6.1 I will look at whether phonological models need to be modified on the basis of the outcome of
the empirical study of symmetry in NGT in chapter 5, and by taking into account criticisms that were
raised in the discussion of the literature in chapter 2. In section 6.2, I will consider to what extent the
factor 'ease of production' provides an explanation for the patterns found in chapter 5. Finally, in
section 6.3, I will contrast ease of production and ease of perception, and see how they conflict in
establishing patterns in two-handed signs.

6.1 The two articulators in a phonological model

In this first section, I will look at the implications of findings discussed in earlier chapters for one
particular phonological model, that of van der Hulst (1993, 1994a, 1995abef). Two specific topics will
be discussed. First, in chapter 2 I observed that whether or not the weak hand moves is non-distinctive,
and in most if not all cases predictable; on the other hand movement or no movement of the weak
hand is one of the criteria for determining whether or not the weak hand acts as a place of articulation.
Second, in chapter 5 I studied symmetry in the production of two-handed signs and found that only
a very limited number of possible asymmetric gestures actually occur in NGT.

6.1.1 The phonological model of van der Hulst

A number of different current views on the representation of two-handed signs were discussed in
section 2.2.1. Here I will limit myself to considering the model proposed by van der Hulst (e.g., 1993,
1995ef), which I will now further introduce. The key aspect of two-handed signs that this model tries
to account for is the limited contrastive possibilities of the weak hand. In the representation ally
simplest signs, in other words in the default case, the properties of the weak hand are the same as
those of the strong hand, are fully predictable, and the relevant features, being underspecified in the
lexicon, can be filled in by a default rule. In that case, the sign is fully symmetric. If a sign is
asymmetric, there is at least one feature of the weak hand that is different from that of the strong hand,
and this feature needs to be part of the lexical representation. As we saw above, there are several
different types of asymmetry that occur and that a model needs to account for. In section 6.1.2 I will
therefore look at how the asymmetries that are found in signs can be accounted for in the van der
86

Hulst model.

In the most recent version (van der Hulst, 1995e), the model for one-handed signs looks like the graph
in (1)32.

(1) Representation of one-handed signs in van der Hulst (1995e:8, terminology is slightly adapted)

one-handed sign

[x x1 Sk

Vertical lines dominate heads. diagonal lines dominate dependents. A = articulator (Ao = finger selection),
M = manner of articulation, P = place = major place), XI, X2 = projections of the X node,
of articulation (pO
Or = orientation, FC = finger configuration, Fo = focus, Se = setting. Sk = skeleton

The actual features (not given above) which are dominated by the terminal nodes in this diagram are
linked to the bipositional skeleton. Movement is represented by branching nodes at the lowest
(featural) level in the A2 or p2 constituents. The sign is seen as a combination of three major
properties: features of the articulator shape (or active articulator; this includes the traditional parameters
handshape and orientation, following the proposal of Sandler, 1989, to see these two as forming a
group), features of the articulator position (or passive articulator), and features of manner of movement
or transition, which specifies the relation of the active articulator to the passive articulator. The internal
structure of the manner node is not yet worked out. For a discussion of other aspects of this model
which are not related to two-handed signs, I refer to the various articles of van der Hulst (1993,
1995abf).

32 This tree gives the so-called 'end view' of the representation, in the sense of autosegmental phonology. Thus, no

ordering in time is implied anywhere in this tree.

 87

The model is based on ideas in dependency phonology (Ewen, 1994) and a relative of it, radical CV
phonology (van der Hulst, 1994b, 1995cd). In these approaches, phonological structure is hierarchical,
which is expressed in tree diagrams, as in many phonological models (cf. Goldsmith, 1991;
Kenstowicz, 1994). In the stance taken in this sign language model, the tree structures must be binary
branching. The crucial difference with other non-dependency models is that the branching is assumed
to be headed, as is more general practice in syntax and metrical phonology (cf. Kenstowicz, 1994).
Head constituents have several properties that distinguish them from their dependent constituents. For
a more detailed overview and discussion of these so-called head-dependent asymmetries, see Dresher
& van der Hulst (1994, 1995). These authors distinguish relationships between two constituents of the
same type (a-a dependencies) and between two constituents of different types (a-~ dependencies). An
example from spoken language of the former is the relation between onset and rhyme in the syllable,
and an example of the latter is the relation between two syllables in a foot. Some of the properties of
the head-dependent difference are listed in (2).

(2) Differences between heads and dependents

heads can only spread with their dependents, dependents can spread alone
in the case of an a-~ relation, the head cannot branch, the dependent can
if there is an a-a relation, then the possible values for the dependent constituent are a
subset of those of the head constituent; in other words, the dependent is never more
complex than the head
heads, not dependents, determine the status of the two constituents as a group, thus, the
category of the mother node; in other words, the properties of the head determine the
role of the constituent in higher structure

A further distinction between two dependents of (projections of) the same head is that higher
dependents, being more remote from the head, spread more easily than lower dependents.

These properties are recurrent characteristics of relations between constituents at various levels of
phonological organization (segment structure, syllable structure, and higher organizational levels) and
outside phonology. Van der Hulst's claim is that this head-dependent organization is present in both
modalities of language, spoken and signed.

In the above tree, the articulator node is seen as a dependent of the place node, although both are
considered dependents of the manner node. The place has head properties as it does not branch (there
is only one major place of articulation in each sign), and also because it cannot spread alone from its
88

dependent33 •

The central property of the reduced possibilities of the weak hand as compared to the strong hand is
seen as an instance of the head-dependent phenomenon. Thus, somewhere in the representation the
weak hand is a sister of the strong hand, the former being a dependent of the latter, which is a head.

There are different ways in which the above representation can be expanded to cover two-handed signs
as well. The most obvious way is to simply double the tree diagram in (1), making the weak hand the
dependent of the strong hand (cf. van der Hulst, 1994a). However, as pointed out above, the model
makes the crucial claim that the possible properties of the weak hand are a subset of the possible
properties of the strong hand, whereas this first proposal still claims that the weak hand can have
distinctive properties for all aspects of the sign, even though on the whole the weak hand has a more
limited set of appearances due to its dependent status. In the most recent version of the model (van
der Hulst, 1995e) a stronger claim is made: the only distinctive properties of the weak hand are
articulator features. Thus, we have the model in (3).

(3) Representation of two-handed signs in van der Hulst (1995e:1O)

two-handed sign

/1 M

[ X X 1 Sk

xw= X node for the weak hand

Here, the weak hand is a dependent of the strong hand, but at the level of the articulator node, and
not at the highest level of the representation. This model predicts, then, that the (manner of) movement

33 The representation in (1), where 'manner' is the head of the tree, seems to contradict the statement that it is the
place features that form the head. This situation indeed merits further consideration, and I will discuss it and propose a
different representation in section 6.1.4 below.
 89

features and the place of articulation features of the weak hand must be the identical to that of the
strong hand. Only orientation and handshape can be distinctively specified for the dependent weak
hand node, the possible values are correctly predicted to be a subset of the set of possible values of
the strong hand.

By representing the weak hand in the same way in all signs (as an articulator), the model belongs to
the group of 'weak hand always' models (see p. 20). I will not further discuss this aspect here, nor
how the model contrasts with 'weak hand sometimes' models, but see Brentari (1995) and van der
Hulst & Sandler (1994) for discussion.

6.1.2 The representation of asymmetries in the model

What interests us here is the question whether the model can represent the different types of signs that
are distinguished above. Thus, we are interested to represent the patterns of distribution: describe the
signs that may occur but not those that may not occur. Note that here the question is not so much
whether the model represents differently the different types of asymmetry, thus, whether the
representation distinguishes between them. We do not yet know whether they have any phonological
reality aside from establishing patterns of distribution, i.e., whether there are phonological processes
that crucially refer to some but not other types of asymmetry. This remains a topic for future research,
it is not the subject of this work. To the extent that distributional patterns have emerged from the study
of symmetry in production in chapter 5 they need to be represented in the model, however. The
patterns of distribution are the subject of this subsection.

A problem immediately arises in judging whether the model can account for the asymmetries that we
found : the model is built on space-based terminology and observations, not on body-based ones. All
the observations on symmetry in the previous chapter concerned the detailed production of signs, and
thus were cast in body-based terminology. Moreover, in the model under discussion like in other
phonological models, only the hand is treated as a body part that needs to be described, not the arm.
Thus, there is no direct one-to-one mapping of the categories used in the model and the categories
used in the description. If we try to work out how the phonological parameters in the model, which
are based on the perceptual properties of the hand, correspond to a phonetic representation of the
whole articulator at the level of joints, we would arrive at the complex of relations in (4). (A
comparable diagram for the model of Uyechi (1995ab) was discussed on p. 43.)
90

(4) Relation between articulation and phonological parameters for the van der Hulst model

manner

finger and thumb joints

orientation

wrist
finger selection

radioulnar
f inger configuration

elbow
focus

major place shoulder

setting

Several things can be remarked about these relationships, which certainly demand further study. First,
the only phonological parameters that seem to be 'implemented' by only one joint, finger selection
and finger configuration, in reality are realized by a large number of different joints; the seeming
simplicity results from our abstraction of detail within the hand. More noticeable, the parameter
'manner' is not implemented at any joint. This is because manner specifies the precise details of the
movement of a sign, and is assumed to be valid for all movements in a sign, thus, for change in any
joint. In this view, the interpretation of manner features does not require any joint in particular for its
expression, like the other features do. Finally, one can notice that the parameters orientation and focus
are related to the same joints. This expresses the close resemblance of and relation between orientation
and focus which is noted by van der Kooij (1994) and Greftefreff (1992), but which seems not to be
expressed in this stage of the model. This is an interesting area for future investigation.

How does this complex set of relations shed light on the question we are trying to answer, i.e., are the
categories distinguished in the previous chapter representable in the model? We saw that symmetry
types are most frequently specified alike for two groups of joint pairs: the shoulder and elbow pairs
 91

form a group, and so do the radioulnar and wrist pairs. Shoulder and elbow are treated as a unit in the
parameters major place and setting; thus, we may say that the shoulder and elbow joints express the
phonological parameter place (see also Uyechi, 1994, 1995b)34. Radioulnar and wrist joints are treated
as a group by the orientation and focus parameters, but the group also includes the shoulder and elbow
joints. The influence of the shoulder and elbow joints (or alternatively, the parameter 'place of
articulation') on the orientation parameter has attracted the attention of some researchers recently
(Nagahara, 1988; Uyechi, 1994b, 1995a; van der Hulst, 1995a), but in general it seems assumed that
the space-based term 'orientation' merely concerns the state or action of the forearm and wrist (Stokoe,
1960; Klima & Bellugi, 1979; Sandler, 1989). For now, I will assume that the crucial aspects of
orientation are indeed articulated at the radioulnar and wrist joints, recognizing that this area needs
further investigation. If this assumption appears to be justified, that is if the role of the elbow and
shoulder joints in the orientation parameter is limited, then that would contribute evidence for the
phonetic (articulatory) reality of the two phonological parameters location and orientation.

Thus, the grouping of the four joints that were considered into two groups is expressed in the model.
How about the different symmetry types?

Time asymmetries: translation

Van der Hulst (1995e), following Prillwitz et at., 1989), sees two subtypes within the class of balanced
signs which are not completely symmetric: alternating signs and shadow signs. "In exact mirror signs,
the strong hand and the weak hand are like the practising ballet dancer and her image in the mirror"
(van der Hulst, 1995e:4). Alternating signs are signs with repeated movement and a phase difference.
Shadow signs are signs in which the weak hand follows the strong hand in space, like a suspect is
followed by her private detective; there is no mirror present in this case. Both of these two SUbtypes
show disruption of the absolute bilateral symmetry, but alternating is considered a disruption in time
and shadow a disruption in space. If one describes signs in terms of the joint angles involved,
however, we see that all shadow signs are characterized by a phase difference. Thus, shadow signs (if
repeated) are alternating signs. The thing that imposes the difference between the two groups is not
found on an articulatory level, the difference concerns the movement of the articulators in space and
thus is of a perceptual nature: the movement of the hands in space involves the left-right (or frontal-
transverse) axis which is perpendicular to the mirror plane. In signs with a time translation and
movement on the left-right dimension, perceptually the hands move in the same direction at the same
time. In signs with articulatory symmetric movement on the left-right axis, perceptually the hands
move in opposite directions at the same time. The difference between articulation and perception will
be the object of further discussion in section 6.3.

34 Notice that 'place of articulation' is not a concept that is natural to a body-based description, it is a purely space-

based term.
92

Shadow signs, then, can be represented by the feature [alternating], and the presence of a feature
[shadow] under the Aw node is not necessary. The perceptual difference between shadow alternating
signs and non-shadow alternating signs is already represented in the model because the direction of
movement needs to be represented, which occurs in the model under the place node. If this movement
involves the left-right axis, then the alternating sign is shadow. So far, however, no special
phonological properties have been demonstrated for shadow movements, and there is no reason to see
them as different from other alternating movements.

Van der Hulst (1995e) sees a unary feature [alternating] (and the feature [shadow], which we have
eliminated) as being dominated directly by the A; node, which is otherwise empty in case of fully
symmetric signs . This is peculiar for two reasons. First, a phase difference is a property of the
movement of two objects together, not merely of one object, the weak hand in this case. Vander Hulst
chooses to represent this feature there because it can only be present phonetically if there is a weak
hand present. (Remember that a weak hand can never articulate a sign on its own, as per definition
it would then be the strong hand.) I can see the reason of this argumentation, but I do not consider
the argument very strong. A feature [alternating] specifying the movement of the whole sign could
very well be placed on a higher level in the tree, in which case it simply gets no phonetic
interpretation if there is only one hand present. Second, the location of [alternating] under the A; node
is peculiar in that it is not merely a property of the movement specified under the articulator node.
Rather, [alternating] is generally claimed to be a property of all movement in the sign, including
movement represented as changes in the place constituent. This appeared to be true for my corpus as
well: in all cases where there is a time translation involved somewhere in the sign, this translation is
a property of all moving joint pairs. Considering these two points, it seems that it may be more fruitful
to see [alternating] as a manner feature, manner features being features that characterize properties of
the sign as a whole, not regarding any specific parameter of articulation.

However, there are two powerful argument in favor of putting the feature [alternating] under the A;
node. In the first place, there is the argument that the conditioning environment for the process of
Weak Drop can then be formulated very easily: Weak Drop may occur only if there is no specification
whatsoever under the A; node (this argument is not advanced in this manner by van der Hulst). The
problem with this argument is that it depends on a particular view of which signs may undergo Weak
Drop; as was noted earlier, as yet the Weak Drop process is far from being described exhaustively.
Therefore, I do not consider this to be a major argument in favor of putting [alternating] under the
articulator node.

In the second place, putting [alternating] under the A; node is favorable because it correctly represents
the fact that a sign that has alternating movement cannot have distinctive articulator specifications for
the weak hand. In other words, the two hands in signs which have alternating movement have the
same handshape and the same orientation specification, and this is predicted if [alternating] is an
 93

articulator feature. This argument also has its counterarguments though: although no contrastive
articulator features is indeed a characteristic of morphologically simple alternating signs,
morphologically complex alternating signs can have two different handshapes. Examples are the ASL
compound signs SIMULTANEOUS-COMMUNICATION and TOTAL-COMMUNICATION. These
signs would not be representable at all if [alternating] and articulator features for the weak hand are
mutually exclusive. So that in tum would be an argument for placing [alternating] somewhere outside
the A; node.

If [alternating] is indeed put outside the A! constituent, we make the following prediction: if there is
Weak Drop at some stage in the derivation (not necessarily in the surface form), then the feature
[alternating] remains present in the representation. If subsequently a morphological process like
compounding adds a new specification of the weak hand, then the resulting two-handed sign should
be alternating. This may indeed be the reason why the above ASL compounds not only can have but
must have alternating movement. Examination of more compounds and of other morphological
processes are necessary to bare out this prediction.

Concluding, there are a number of arguments both in favor of and against seeing [alternating] as a
feature of the weak hand or alternatively as a manner feature under a separate node. More knowledge
about Weak Drop, morphological processes, and about the character of the manner node are necessary
to shed further light on this subject. For now, I will make no specific proposal for either option.
However, the subject of the manner node in relation to the general architecture of the representation
will be discussed in section 6.1.4.

Time asymmetries: reflection

There are other types of asymmetry which also involve time transformations: signs where a time
reflection is involved in the relation between the two hands, either alone or together with a time
translation (resp. types 4 and 5). Both of these types of signs require that reference is made to the fact
that there is movement in opposite directions (which is not repeated, in the signs in my corpus). As
it is, there is not a way to express this type of asymmetry for all movements in the current model.
Direction of movement can be distinctively specified for the weak hand in case of changes in the
articulator node, but not in case of changes in the place node. Thus, the signs GENEN, OVERLEDEN,
VERANDEREN, and BRAND (and many ASL signs like SCREWED-UP and BAPTIST) are
representable, as they involve a change in orientation in both hands but in opposite directions, but the
signs VERHUIZEN and VERKEER (and the Groningen sign BOOT) are not, as they involve a change
in place specification, there being movement at the elbow and/or shoulder joints. However, even the
opposite orientation specification which is possible to represent has a strong disadvantage: it does not
express that the changes need to be in opposite directions and of equal 'size', not just in different
directions and of any 'size'.
94

Therefore, I propose to adopt a new feature [reflection] to represent this type of asymmetry. This
feature has the same status as the feature [alternating] discussed above, and it seems evident that it
should be placed at the same location in the tree as [alternating]. As the feature [alternating] is
generally assumed to only cover phase differences, which require repeated movement, and not just any
time transformation, I also propose to change this feature into a feature [translation]. This feature
covers both alternating and shadow signs in the traditional terminology, and does not put any
requirements on the number of times a movement is executed (several times for traditional alternating
signs, once or several times for traditional shadow signs), Thus, it can also describe "non-repeated
alternating signs". Both of these features can occur alone in a sign, or they can combine. I do not have
any preference yet for making these two features either unary or binary. Both features seem to apply
to all movement in a sign, so not just to either movement within the articulator or movement of the
articulator in space (changes in the place constituent).

In the future, some way should be found of expressing the relative markedness (in terms of frequency)
of these two features and of the four possible combinations. Absence of both features (or negative
specifications) seems most frequent (occurring in all one-handed signs and the majority of two-handed
signs), followed by presence of only [translation], which is more frequent than the presence of both
[translation] and [reflection]. The presence of [reflection] on its own seems very marked indeed, only
occurring in two signs in my corpus.

Finally, it should be remarked that these two features have a very specific and constant phonetic
interpretation. First, if [translation] indicates a phase difference, it is always one of half a period (180
degrees). Second, the phonetic realization of a combination of the features [translation] and [reflection]
is always such that the beginning and end of the movements of the two articulators occur at the same
time.

Symmetry types 1, 2 and 6

Completely symmetric joint pairs are easily represented, by the presence of a (partly) empty A; node.
This emptiness implies that there is no contrastive specification at all for the weak hand, and a default
rule can copy all the (relevant) specifications under the A2 node to the A~ node.

The situation for type 2 and type 6 asymmetries is more complex, ~ these types cover a wider range
of different asymmetries. The common property of these two asymmetries is that there needs to be a
contrastive specification for the weak hand somewhere in the representation. In other words,
asymmetry in this model is expressed as the presence of contrastive specification for the weak hand.

These types 2 and 6 are always representable in the current model if they are properties of the
radioulnar or wrist joints, or in terms of the model, in the orientation parameter. For this parameter,
 95

the weak hand can have a specification independent of that for the strong hand, and thus any
asymmetry can be represented there.

As stated above, in the model there can be no contrastive specification for the place node features of
the weak hand. However, there are a large number of signs which have a type 2 or a type 6
asymmetry in the shoulder or elbow joint (75 out of 139 in both cases; cf. (13), chapter 5, p. 77).
These two joints, as we saw above (p. 91), express the phonological parameter place of articulation.
How does the model represent such signs?

First of all, it may be the case that there are signs in which the difference in state within the elbow
pair and the shoulder pair is present at a phonetic level, but not on a phonological level. There may
be very small differences that do not count for the phonology, and are fully predictable, like the
differences that I ignored throughout the thesis in signs where the sagittal plane is crossed by both
hands. However, as yet I have not found any examples of such a situation: none of the differences are
predictable. Thus, the problem still stands: some way must be found to represent signs which are
asymmetric at the shoulder and/or elbow joint.

Secondly, there is one way to represent signs in which the weak hand is not a copy of the strong hand.
All recent models (e.g. Sandler, 1989; Brentari, 1990; Uyechi, 1995a; van der Hulst, 1995e) have a
way of representing signs in which the weak hand is as a place of articulation instead of an active
articulator. In all these models, signs have been divided into two classes, symmetric signs (modulo
time asymmetries) and 'weak hand place' signs 35 • Sandler (1993) argues that the weak hand should
be given a completely different representation in the two groups of signs, to reflect the different
behavior of the weak hand (cf. (5) below). In the model of van der Hulst (1995e) under discussion
here, for all signs the weak hand has the same position in the representation; the difference between
the two groups of signs is expressed by the presence or absence of a feature [weak hand] in the place
constituent, under the AO node.

The question arises, then, whether all the signs that we are trying to represent are of this type. To
answer this question we need to examine which are the defining properties of these 'weak hand place'
signs. So far, we have only seen the 'negative' definition of weak hand place signs, namely that the
weak hand is not a mirror image of the strong hand. Judging only by this criterion it would indeed
be better to talk about asymmetric (or unbalanced) signs instead of weak hand place (or 2P) signs, as
the mere presence of asymmetry does not entail that the weak hand acts as a place. Sandler (1993)
advances several 'positive' arguments for seeing the weak hand as a place of articulation in these
asymmetric signs. These arguments are listed in (5).

3S Various other tenns are used to indicate these groups. For example Sandler (1993) talks about E2 and 2P signs.
and van der Hulst (1995e) calls them balanced and unbalanced. respectively.
96

(5) Arguments for seeing the weak hand as a place of articulation in asymmetric signs (Sandler,
1993)
like other places of articulation, the weak hand does not move
the movement of strong hand is (most easily) defined with reference to the weak hand, its
place of articulation36
no other place of articulation which would need to be specified is present in the sign; the
location of the two hands as a group is a default one and thus the same for all signs3?
there is no hand-internal movement for the weak hand
neither Weak Drop nor Weak Freeze can occur, since the weak hand is not an articulator
(which is what the rules refer to)38
the weak hand behaves like other places of articulation in morphological processes39

Most of these arguments are valid points (though see notes 4 and 5), but can be considered
problematic that some of the structural criteria are not valid for all signs.

First, in some of the asymmetric two-handed signs the weak hand does move: e.g. BEGELEIDEN,
ELASTIEKJE, LUCHTBALLON, ZIEKENHUIS, THERAPIE, LATEN-ZIEN. In some of these
(BEGELIEDEN, ZIEKENHUIS, THERAPIE, LATEN-ZIEN), as in the ASL sign SHOW, the
movement of the weak hand is predictable if we can specify that there is continuous contact in the
sign. In others, the hands move independently, but this movement may either be in the same direction
in space (LUCHTBALLON) or in opposite direction (ELASTIEKJE).

Second, in LSQ there are a few signs in which both hands are located near some body part, but stilI

36 Of course, in principle the movement of the strong hand can be defined with reference to any point in space. The
important point, however, is that the movement of the strong hand with respect to the weak hand is the feature of the
sign that remains constant over various instantiations, where the exact location of the two hands in space may vary
according to body posture, other practical reasons, or stylistic variation.

37 This default location is generally believed to be the space in front of the trunk, called 'neutral space' in the sign
language literature (Battison, 1978; Klima & Bellugi, 1979; Sandler, 1989). This location contrasts with other so-called
'major places' like head, neck, arm, and hand. Differences within neutral space (e.g., high or low) are claimed to be
predictable and non-distinctive. The important point that remains is that it still needs to be specified how the two hands
form a group: what is the location of one hand with respect to the other?

38 As pointed out above, to my mind the Weak Drop and Weak Freeze facts need further investigation before they
can be contributed as evidence for anything.

39 Sandler (1993) argues that the morphological modifications for dual inflection and for the multiple exhaustive

forms of verbs treat the weak hand different from the strong hand. However, I do not understand why this is so. I agree
with Brentari (1990), in that it seems to me that in all signs the whole sign is repeated and an extra sweeping movement
(for the latter process) is added for the repeated sign as a whole. Thus, although this type of morphological evidence
would be very important indeed to distinguish 2P from E2 signs, such evidence has not been conclusively argued to
exist.
 97

the strong hand moves on the weak hand (Miller, 1995). For example, the LSQ sign PIG is articulated
at the nose, with the flat strong hand making circling movements on the fist-shaped weak hand. These
signs seem to have two places of articulation at the same time, which is pretty peculiar in view of
most other signs and the models that have been made to represent these. It would be interesting to see
whether signs like PIG can also occur in other sign languages. In NGT, for example, I have seen
realizations of the sign VERDIENEN where the weak hand was in permanent contact with the belly
and the strong hand repeatedly approaching and touching the weak hand.

Finally, there is one exception in my corpus to the demand that the weak hand in asymmetric signs
has no internal movement for the sign to be a weak hand place sign: in the sign KLETSEN, there is
flexing-extending movement of the fingers at the proximal finger joints in both hands. In this respect,
the weak hand acts as an articulator, and not as a place of articulation.

Thus, there is a small group of signs that has asymmetry at the shoulder and/or elbow joint, but in
which the weak hand does not seem to have all the properties of a place of articulation.

Apart from this type of problems, seeing all asymmetric signs as weak hand place signs is problematic
because, as several researchers have argued in the past, many of the asymmetric signs, in which the
handshapes are the same and only one articulator moves, are closer to the fully .symmetric signs than
to other asymmetric signs in which the weak hand functions as a place of articulation (e.g., Mandel,
1982; Uyechi, 1995b). We may wonder, then, what is the added value of restricting the class of
asymmetric signs to (or seeing this class as) weak hand place signs, if there are two groups of signs
that do not seem to fit this definition.

In the first place, I would like to suggest for now, in agreement with Sandler (1993:362-364), that the
first small group of exceptions is too small to necessitate modification of the basic generalization that
most asymmetric signs form one unified group. Many of the counterexamples are suspicious in the
sense that one may doubt whether they really are lexicalized forms that conform to the
morphologically simple lexical patterns that are being described here. For example, the signs
LUCHTBALLON, PARACHUTE, BEGELEIDEN, and ELASTIEKJE may well be recent productive
classifier forms (see Klima & Bellugi, 1979; Baker-Shenk & Cokely, 1980; Supalla, 1986), and
morphologically complex. Also, the sign BOS is a plural form of the sign BOOM, and as such might
have been excluded in advance, as it is not the goal here to describe patterns of alternation. The sign
HERFST is more symmetric than asymmetric if we look at the level of axes of movement instead of
the level of the whole joint. Finally, the movement of the weak hand in the signs ZIEKENHUIS,
THERAPIE, and LATEN-ZIEN is representable by specifying that the hands are in contact during the
whole sign.

Thus, in view of the size of the exceptional group, the generalization that in (one group of) asymmetric
98

signs the weak hand acts as a place of articulation still stands. However, we may wonder whether the
fact that the weak hand is a place of articulation in these signs is the basic fact that we want to
represent in the model, or whether it is a derivative of other properties. This question becomes
especially interesting when we look at the other group of signs in which the weak hand is a place of
articulation, but which at the same time very much resemble fully symmetric signs. Only by looking
at the formulations we have been using here, opposing fully symmetric signs (modulo time
asymmetries) to weak hand place signs, we can see that the distinction is not very sound: it would be
more natural to oppose fully symmetric signs to asymmetric signs, and weak hand place signs to signs
with other places of articulation. In other words, the fact that the weak hand is a place of articulation
does not exclude that the sign is very symmetric.

This is indeed what we see in the second problematic class, which consists of the in between category,
traditionally called Type 2 signs (Battison, 1978; see also Sandler, 1993; Uyechi, 1995b; Brentari,
1995). This group includes signs like NGT BEWIJS, PAARS, WERKEN, and ASL NAME. These
signs have the property that the handshapes are the same, radioulnar and wrist joints are fully
symmetrical (type 1) and that the shoulder and elbow joint pairs are both of symmetry type 2: they
are fully symmetrical, but only at one (or some) points in time. The reason for this is that only one
articulator moves, which in turn is the main (or maybe the only) reason for also considering the sign
weak hand place signs. Thus, in these signs the two aspects conflict. However, the only reason there
is a conflict is that we have made the 'unnatural' opposition between symmetric and weak hand place
signs, at least in the representation.

I would like to propose, therefore, that the main distinction between signs is that between symmetric
and asymmetric. It is these two categories that need to be distinguished in the representation of signs,
and not the two categories fully symmetric and weak hand place. It follows, then, that whether the
weak hand acts as a place of articulation or not is a derived phenomenon which can, and indeed does,
cut across the two categories. Thus, the group of signs of type 2211, in which the place of articulation
is the weak hand, can be represented as fully symmetric signs.

All the asymmetric signs, then, need a new feature instead of [weak hand], which did not completely
cover all asymmetric signs. The crucial thing that this feature needs to express is that there is an
asymmetry in the sign mainly at the level of the shoulder and elbow, thus in the location parameter.
However, all the signs involved are articulated in neutral space, in which distinctions have claimed not
to be relevant. However, the hands are in a somewhat different location. The way to represent this
group, then, is by the major place feature [neutral space], in combination with a new feature
[asymmetric] which specifies that there is a specific relation between the two hands, which renders
the sign not fully symmetrical. This feature says nothing about the precise relation between the two
hands, called "hand arrangement" in Klima & BeUugi (1979:45ff). I will not have much to say about
this precise relation here, simply assuming that the relation between the two hands follows from the
 99

representation of movement direction and focus, or is predictable in another way. If it appears in the
future that the precise hand arrangement is not predictable, other features replacing [asymmetric] are
needed with more concrete phonetic content than the abstract feature [asymmetric].

It seems to me that these features, either [asymmetric] or alternatives, are in the same place in the
representation as the time asymmetry features [translation] and [reflection]. Just like these time
asymmetry features, [asymmetric] says something about the relation of the weak hand with respect to
the strong hand.

Next, I would like to suggest that whether the weak hand acts as a place of articulation is determined
by the absence or presence of movement of the weak hand: the weak hand is a place of articulation
if it does not move. This is the dominant criterion, despite other criteria like utterance in neutral space,
for example. This brings us to the other main topic of section 6.1 besides the representation of the
symmetries observed in chapter 5, namely that of the non-distinctiveness of weak hand movement,
observed in the discussion of the literature in chapter 2. This topic will be discussed in section 6.1.3
below.

In this new light, the small group of signs in which the weak hand does not seem to act as a place,
but which are asymmetrical, need no longer be considered as exceptional. Although above (p. 97) I
have indicated that some of these signs might be considered morphologically complex or non-native
lexical items, most of them are easily representable in this model. The signs LUCHTBALLON,
PARACHUTE, and ELASTIEKJE remain exceptional in that both hands move.

Symmetry types 2 and 6, then, are well representable not only for radioulnar and wrist, but also for
shoulder and elbow. The modifications proposed for the model of van der Hulst leave unimpaired the
claim that asymmetries at the level of place of articulation are much more limited in possible form
than asymmetries at the orientation level. It remains a topic for future research to find out whether the
restrictions on asymmetries in orientation and handshape, which now are only expressed by the head-
dependency relation of the two articulators, need further restriction.

Frequent sign types: 1111, 6666, 3311, 2211

Now, concluding this section, how about the few sign types that occurred more frequently than all
others; do they get a special treatment in the model? In reviewing the symmetry types above we
already came across some of them. The fully symmetric signs, type 1111, have no specification of
weak hand features nor of special two-handed features like [translation] or [reflection]. The realization
of the weak hand is a full copy of the strong hand, all of its specifications being copied by a rule from
the strong hand articulator node. Sign type 3311 has also been implicitly discussed: this type is
completely like sign type 1111 except for the presence (or positive specification) of a time asymmetry
100

feature (either [translation] or [reflection] or both). Sign type 2211 is completely like type 1111 except
for the fact that the hands in these signs cross the sagittal plane, which influences the movement such
that only the strong hand moves (see the next section for discussion). The sign type 6666 is
characterized by the presence of the feature [asymmetric].

6.1.3 Movement of the weak hand and weak hand as a place of articulation

In chapter 2 it was observed that whether the weak hand moves or not is not distinctive40 • I know
of only one claim to the contrary. Sandler (1993) writes that "[f]or signs in which the two hands are
in different shapes, whether both move or only one moves may be contrastive [... ]. No such contrast
exists for E2 signs." (1993:363) The only example she gives is the minimal pair DOLLAR vs. LEAD
(1993:346). However, one might also consider this minimal pair as minimally contrasting by contact
features: in the former sign there is only initial contact whereas in the latter sign there is contact
during the whole sign. Indeed, Sandler's claim that such minimal pairs cannot occur in symmetric (E2)
signs suggests that the difference between a moving and a non-moving weak hand is derived from
other properties of the sign.

Thus, my claim remains that movement of the weak hand is not distinctive. I would like to repeat here
(cf. p. 26 above) that I find it striking that the difference between a moving versus non-moving weak
hand has played such a large role in discussions of two-handed signs, whereas to my knowledge no
one (with the exception of Sandler, 1993) has ever claimed that movement of the weak hand is
phonologically distinctive.

My claim further is that whether the weak hand moves or not is fully predictable on the basis of other
characteristics of the sign. In particular, the amount of symmetry in the sign determines whether only
one articulator moves or whether both articulators move. For the most typical one-hand-moving signs,
those that have traditionally been called Type 3 signs, one can say that if a joint pair is asymmetrical,
then only one joint in the pair will move (which per definition is the joint in the strong hand or strong
articulator) .

At first sight, there seems to be come circularity in the above statement, because when only one joint
in a pair moves, that is precisely what makes the joint pair asymmetrical. So it would be circular to
state that only one joint moves if the joint pair is asymmetrical. At this point it is useful to follow

40 In most of the literature. movement of a hand refers to so-called 'path movement' . or a change in location. My

claims about weak hand movement pertain to this type of change too, although I do not wish to exclude the possibility
that they are also valid for changes (branching terminal nodes) within the articulator node.
 101

Uyechi (1995b) in distinguishing static symmetry from dynamic symmetry, a distinction which was
also made in chapters 4 and 5. I consider static symmetry the symmetry of a joint pair at one point
in time, whereas I consider dynamic symmetry to be symmetry in a joint pair during the whole sign,
thus, incorporating the factor time. The symmetry that we are interested in here is static symmetry.
Thus, if a joint pair is asymmetric at one point in time, let us say at the start of the sign, then only
one joint will move.

This generalization is valid for all signs in which the elbow andlor shoulder joint pair is of type 6. As
I have argued above, these signs are characterized by a feature [asymmetric] somewhere in the
articulator node. That means that the movement specification under the place node only applies to the
strong articulator.

However, there are other signs in which only one hand moves, but in which there is no feature
[asymmetric] present. This is the case for signs of type 2211, which have traditionally been called
Type 2 signs, the in-between group. As remarked above, what causes them to seem 'in-between' is
that they are fully symmetric apart from the fact that only one hand moves. I have proposed above
to represent these signs as fully symmetric signs. Now what distinguishes these signs from other fully
symmetric signs, apart from the fact that only one hand moves? This is the fact that there is a
contacting movement of the two hands, but the movement is not on the left-right dimension. This is
quite an opaque situation, and I will discuss it in more detail.

Let us contrast the signs ZATERDAG (or SAME) and WERKEN (or NAME). In ZATERDAG, the
hands both are in a fist and both repeatedly move towards each other (towards the contralateral side),
contacting at the point where they both reach the sagittal plane. If we look at the paths the two hands
trace in space, we see that these paths are bilaterally symmetric by reflection in the sagittal plane. If
we idealize these paths as straight lines, we can say that they are orthogonal to the sagittal plane. In
Uyechi (1995b) this is expressed by saying that the 'dynamic plane of bilateral symmetry' is the
sagittal plane. The sign ZATERDAG is fully symmetric at all points in time for all joints. In the sign
WERKEN, there is also contacting movement, but here the movement is only executed by one
articulator. The shape that this articulator traces in space is also a straight line, but in this sign the line
is orthogonal to the transverse (horizontal) plane, not to the
sagittal plane.

Now let us imagine a non-existing sign (maybe an emphatic version of the original WERKEN) which
is exactly like WERKEN but in which both articulators move. If we would be forced to articulate this
non-contrastive and maybe non-occurring variation of WERKEN, I think we would not copy the
movement of the strong hand to the weak hand, but rather make the movement of the weak hand
opposite to that of the strong hand. In other words, what is preserved in this non-existing sign is the
movement to contact, not the movement downwards. When we look at this sign, we see that there is
102

a 'dynamic plane of bilateral symmetry', but that this plane is not the sagittal plane. The articulators
are only fully symmetric at the point in time where there is contact, and at all other points in time the
articulators are asymmetric (at the shoulder and elbow).

In other words, articulating the sign WERKEN with two hands would mean that we are doing opposite
things with the two articulators, whereas in the sign ZATERDAG, which is articulated with two hands
moving, the hands are doing articulatorily the same thing with the two hands.

Thus, we arrive at the generalization in (6), which covers all signs, and which I have termed 'weak
hand movement rule'.

(6) WEAK HAND MOVEMENT RULE

The weak hand moves if and only if that means that the articulators are performing the same
motor actions.

This generalization clearly emerges as a phonetic-level statement, and indeed it is in agreement with
my earlier claim that whether the weak hand moves or not is not phonologically distinctive, but rather
predictable from other phonological information41 • This generalization will be given a phonetic
explanation in section 6.2.4.

Although the statement in (6) is formulated at a phonetic level (which is where it belongs, it seems
to me), we might also formulate it employing phonological terms. In the terminology of Uyechi
(l995b) we would say that the weak hand moves if and only if the dynamic plane of bilateral
symmetry is identical to the sagittal plane. In the terms of the model considered here, that of van der
Hulst, the formulation is more complicated. We would need to say something like 'the weak hand
moves if and only if the feature [asymmetric] is absent and if when there is contacting movement, it
is on the left-right dimension42 •

Concluding, since movement of the weak hand is predictable from the other specifications (the feature
[asymmetric], movement direction features, and contact features) we need not represent it separately.

41 In the model as presented in van der Hulst (I995e), a rule is required that says that phonetically there is no weak
hand movement if the feature [weak hand] is present in the p2 constituent. As the weak hand in this model cannot have
specifications of its own for location and movement, the movement needs to be identical to that of the strong hand. If
the weak hand does not change location but the strong hand does, then the required identity of movement is missing, and
an extra rule is needed.

42 It also seems that in all signs of type 2211 the hands cross the sagittal plane, and that this in tum determines the
movement direction. However, it may also indicate that my assumption that these sagittal plane crossing configurations
are fully symmetrical (cf. p. 70) is unjustified, and that in all cases where there are differences in arm configuration due
to the crossing of the hands, the feature [asymmetric] is present. This is in part a different issue, concerning redundancy
of movement specifications, and it will not be discussed here.
 103

On the other hand, movement or no movement of the weak hand is very important in that it seems
the decisive clue for determining whether the weak hand acts as a place of articulation or not. Future
examination of the claim that weak hand place signs behave as a group (e.g., Sandler, 1993) may shed
light on the question whether a separate feature [weak hand] is still necessary. My claim is that at the
lexical level, describing patterns of distribution, a place feature [weak hand] is not necessary but
instead obscures the fundamental division in signs between symmetrical and asymmetrical.

One pattern of alternation, the phenomenon of Weak Freeze, may be more simple or more natural in
the treatment of two-handed signs proposed here. Weak Freeze is the phenomenon that in some cases
a sign which normally has movement of both hands can be uttered with both hands present but only
one hand moving. As movement of the weak hand is a matter of phonetic implementation in the
revisions proposed here, Weak Freeze is no longer considered a phonological process, but rather a
phonetic process. The same might be true for Weak Drop too, which would be in line with remarks
by Stack (1995).

6.1.4 Heads, dependents, and manner features

In section 6.1.2 three new features were introduced: [translation], [reflection], and [asymmetric],
roughly corresponding to symmetry types 3, 4, and 6. It was suggested that all three features specify
the relation between the two articulators, either in space or in time. Therefore they have the properties
of manner features, as manner features express (the details of) relations between two constituents. They
seem to specifically apply to the articulator node, and not to the relation between place and articulator.
For that latter relation, the model of van der Hulst already had a node called 'manner' (cf. (1), p. 86).

I will propose below that there is a second manner node, which relates the two articulator nodes. The
new features introduced here are dominated by this manner node, in a organization yet to be explored.
In order to determine the location of this second manner node, let us look again at the basic
architecture of van der Hulst's model.

Van der Hulst (1993, 1995abef) basically follows the traditional models in assuming that there are
three major parameters in a sign or constituents in its representation: articulator, place, and movement.
Movement itself is represented by branching place or articulator nodes, but the details of this
movement, the manner of movement, still fulfills the role of a third parameter. (Following Sandler,
1989, the fourth parameter orientation is seen to form one constituent with handshape; this constituent
is called 'articulator' in van der Hulst, 1995e.) Manner is the way in which the articulator (or active
 104

articulator) relates to the place of articulation (or passive articulator)43. The basics of the relation
between the two constituents articulator and place is clear: they are in a head-dependent relationship,
place being the head, for reasons outlined earlier (see p. 87). Manner features specify the details of
the relation, like whether there the articulator contacts the place of articulation during the movement
or at the end of the movement, and whether the movement is repeated or not.

Thus, we have a complex of a head and a dependent, and a specification of the relation between them.
There are different ways of expressing this in a feature tree; for example, consider the representations
in (7).

(7) Different ways of representing relation R between head constituent A and dependent
constituent B

R~ B~
R

A B A
~B
R A
1. R as a speC ifier [spec,A) 2. R dominating the 3. R as the head
canst it ue nts

I will assume here that representation 2 is a simple notational variant of number 1. Both express the
asymmetry between A and B as the main head-dependent asymmetry. Between the firsUsecond and
the third there are significant differences though. No. 3 more explicitly predicts that it is the relation
specification properties that are relevant to higher prosodic organization, and not the properties of
either A or B. Further, 3 conserves the idea of a difference between A and B: A is closer to the head,
and thus less mobile than B. No.1 makes this latter property, the stability of A as compared to B, the
main feature of the organization, by making it the head. The relation properties (R) further specify this
relation, and are in 'specifier' position, also known in generative syntax (e.g., Haegeman, 1994).

Thus, emphasis on different properties of heads contribute to different conceptions of how the A-B-R
constituents should be organized. Further discussions of properties of head-dependent asymmetries and
of the specific properties of heads in the spirit of Dresher & van der Hulst (1994, 1995) can inform
us with regard to the choice between different representations. Also, looking at different conceptions

43 In van der Hulst, 1995ab, this is compared to the speech gesture in spoken language, where manner features also
specify the relation between the active and the passive articulator.
 105

of heads in syntax may be enlightening.

In the current model (van der Hulst, 1995ef), the articulator-place relation and its specifier manner are
represented following model 3 in (7). This is motivated by the fact that manner properties seem even
more stable in the sign than place properties: the manner node never branches (cf. Wilbur, 1993). On
the other hand, the manner node admittedly is not yet studied very well, in the current model it is
something of a black box, as yet (van der Hulst, 1995ab)44. Some of the highly stable properties that
presumably are contained in it, however, like speed or acceleration and repetition, may well be
prosodic properties that are not represented in the sign (see Wilbur, 1993, for remarks in similar spirit).
Other possible properties like movement shape may well be representable in other ways than as
manner features. In other words, the stability of manner properties need not necessarily be expressed
by giving it head status. (Certainly no definite arguments can be made without serious study of manner
features.)

It seems to me that the main property that we wish to emphasize is the head-dependent relation
between articulator and place; the manner properties further specify this relation, and therefore I would
suggest that we put them in 'specifier position', being the highest dependent of (the projection of) the
head. Thus, I would argue that we assign most weight to the phenomenon we are well informed about,
i.e. the asymmetric relation between articulator and place. In this way, I tentatively suggest that the
older version of the model (cf. van der Hulst, 1993) be restored, making place the head of the tree,
articulator its dependent, and manner the specifier.

Considering two-handed signs, we have two articulator nodes, which are in a head-dependent relation,
and it is nothing but very natural to assume that in this case too, there is a node specifying the details
or content of the relation. This relation can be in time, for which case there are the features
[translation] and [reflection], or it can be in space, for which case there is the feature [asymmetric] (or
any future equivalents with more precise information). As I suggested above, I would like to suggest
that these three features are dominated by a single node, which I call 'A-manner'. Further, I suggest
that this node and the two articulator nodes it accompanies follow the organization of the rest of the
sign as proposed above. Thus, I suggest that ASlrong is the head of the articulator node, Aweak its
dependent, and A-manner the specifier. For now, I have nothing to say about the internal organization
of the A-manner node.

Concluding, having incorporated all the adaptations to the model proposed here, we get the
representation in (8).

44 For a large list of possible manner features, see Wilbur (1993:154-5).

106

(8) The model of van der Hulst revised

two-handed sign

~ M
[ ... ]
[ ... 1
[ . .. 1
Po
A-manner
[tn.nslation]
[reflection]
[asymmetric]

Or

PC

6.1.5 Final remarks

The data on symmetry in the NGT lexicon may have other implications that are not discussed here;
this remains open to future research. Also, in this subsection I have examined some implications of
the descriptive data in this thesis for one single phonological model. These implications mayor may
not hold for other phonological models; this is also a topic for future examination.

Admittedly, the proposals made in the previous sections concerning asymmetric and weak: hand place
signs ignore some of the advantages tied to representing asymmetric signs by a place feature [weak:
hand] . I hope that my proposal here can lead to discussion about the relative merits of old and new
proposals, and help uS arrive at a better-motivated representation than the one that I have proposed
here.
 107

The current proposal does have advantages that representations so far did not have. First, a more
principled distinction is made between two main groups of signs which for the largest part corresponds
to the distinction that has been argued for ever since the original study of Stokoe (1960). Second, two
types of problematic categories of signs have been assigned a representation that better fits the
intuitions of many researchers who did not know how to deal with them. Finally, all deviations from
complete symmetry are referred to the articulator node, in line with the proposal for two-handed signs
by van der Hulst (1995e). Either the asymmetry is represented by the presence of a manner feature
at the articulator level, or it is represented by the presence of a non-empty A; node. In the near future,
these advantages need to be compared more carefully to the advantages of earlier proposals by Sandler
(1993) and van der Hulst (1995e), for example.

6.2 Phonetics: articulatory complexity

In chapter 4 I have described which arm and hand movements are physiologically possible, with the
purpose of finding out which of these actually occur in NOT (in chapter 5). The hypothesis or assump-
tion there has been that signs are not randomly chosen from the set of gestures, but that certain types
of gestures are chosen by the language, and not other types. Indeed, we found that in NOT there are
at least four specific types of signs which occur frequently (1111, 2211, 3311, and 6666)45. More in
general, there is a strong tendency for signs to have the same type of asymmetry in the shoulder and
elbow joint pairs, as well as in the radioulnar and wrist pairs.

So far, the assumption has been that everything that is physiologically possible has equal chances of
actually occurring. In other words, the phonology of a language randomly selects a number of
categories to occur in the language from the whole pond of physiologically possible categories. This
assumption will now be modified.

In the study of sound patterns in spoken languages, a framework has been developed in which
researchers try to find phonetic explanations for phonological patterns. On the domain of patterns of
distribution, Maddieson (1991, 1995), Lindblom (1983, 1984, 1990ab, 1991; Lindblom et aI., 1984)
and others have investigated the hypothesis that phoneme inventories in languages are shaped by fac-
tors of 'economy' or 'laziness', to put it bluntly. Ease of perception and ease of production are two
aspects of this general economy factor: the perceiver demands maximal perceptual contrasts in order
to minimize the effort involved in perceiving what is said, and the producer wishes to minimize the
effort involved in production. The crucial part of the hypothesis is that minimization of effort is not

4S It is open for further research whether there are more of such groups, of which accidentally only one or a few
signs occurred in my corpus.
 108

something specific or internal to language, it is independent, being a general phenomenon in biology

(cf. Lindblom, 1983, 1984, 1990; Lindblom et at., 1984; MacConaill & Basmajian, 1969). Being
independent or external, it can be used as an explanatory device to patterns internal to linguistics.

Concluding, in spoken language research it has been claimed that the choice of phonological contrasts
or categories that occur in languages are constrained by factors of economy or ease. It is this
hypothesis that I would like to propose for sign language, too (cf. Lindblom, 1995).

In the study of symmetry in sign language, since the initial studies it has often been remarked that
symmetrical signs are easy both in production and in perception (Battison, 1974, 1978; Friedman,
1976; Frishberg, 1976). Apparently, to these authors it was almost self-evident, as they did not, to my
knowledge, cite any literature on physiology or perception arguing what exactly ease of articulation
or ease of perception means in movements of the arms and hands.

In the rest of this section, I will first argue that there is indeed a sense in which symmetry is easy. In
line with the rest of this thesis, I will only pay attention to the articulation of signs. (Perception will
be briefly discussed in section 6.3.) Then, I will consider possible phonetic explanations for the
patterns found in chapter 5. I will focus on three aspects, listed in (9).

(9) Patterns in NGT for which a phonetic explanation will be considered

out-of-phase movements of the two articulators in sign language are always 180 degrees
out-of-phase
2 in cases where time reflection and time translation are combined, the transformations are
always such that the movements of both articulators start and end at the same time; thus,
the movements of the articulators are synchronized
3 in asymmetric signs, generally only one hand moves

The general idea that symmetry is easy in articulation, and the specific points listed in (9) will now
be discussed in separate paragraphs.

6.2.1 Symmetry and ease of articulation

In hand movements, ease of articulation may figure as the minimization of energy in the action of
muscles, and thus favor smaller movements over larger movements, or maybe movements in which
fewer muscles are involved. Here, however, I will not look at minimization of the absolute amount
of energy spent in muscle actions, but rather at the costs involved in coordination of activity in the
 109

two articulators.

The central idea is that it is easy to program to activate homologous muscles at the same time and to
the same extent, and difficult to perform different actions with homologous muscles. There is evidence
that this is indeed the case from the literature on motor control (e.g, Latash, 1993; Rosenbaum, 1991;
Schmidt, 1988; Kelso, Southard & Goodman, 1979ab).

If one sees symmetry in limb movements as identity of articulation of homologous limbs, then one
can indeed say that symmetrical movements in signs are more easy than asymmetrical movements. In
that way, the strong preference of NGT for symmetric joint pairs, or alternatively the high frequency
of fully symmetric signs (type 1111), is explained by the fact that they are relatively easy to produce.

One critical remark is appropriate at this point. In chapters 4 and 5, the articulatory symmetry has been
considered at the level of individual joints, whereas the above definition of ease of articulation has
been stated with respect to actions of the muscles. This is problematic because it is not the case that
a certain action at a joint is always controlled by the same muscles. This is known as 'context-
conditioned variability' (Bernstein, 1967; Turvey, Fitch & Tuller, 1982): in different contexts, the role
of various muscles may differ in establishing one and the same movement pattern. For example, which
muscles rotate the humerus at the shoulder joint differs depending on the degree of extension of the
humerus. In terms of the joints considered here (cf. chapters 4 and 5), if there the shoulder joints are
in an asymmetric stable state (symmetry type 6), the symmetric action of the elbow (symmetry type
1) might involve different muscles moving the elbow joints in the two cases. I will ignore this
additional complexity here, recognizing that further study of the kinesiology literature is needed to
estimate the relevance of this variability. Thus, I will consider a (moving) type 1 joint pair always as
easier to produce than a type 2 or 6 joint pair, regardless of the (a)symmetry of other joints. In the
fully symmetric sign type 1111 this context-conditioned variability obviously plays no role at all.

Concluding, at the level of joint pairs, symmetry is indeed easy compared to asymmetry. we can say
that symmetry type 1 is easier than the other symmetry types. For now, I will not differentiate between
the various other types, considering them all hard to articulate compared to type 1.

6.2.2 Phase differences

In the survey of the NGT data in chapter 5, we found that time translations as the only type of
asymmetry always occurs in repeated or oscillatory movement, and thus has the effect of a phase
difference. Such a phase difference was found in quite a few signs. Interestingly, we found that the
110

phase difference was always 180 degrees, never 50 or 90, for example. The question arises whether
this constant size of the phase difference is a particular choice of the grammar of NGT, or whether
it reflects more general language-external factors involved in the control of movement in humans.

Indeed, it appears from a large body of literature on motor control that there are two 'stable' phase
differences in all kinds of two-limb oscillatory movements: these movements tend to be or become
either 0 degrees (movement in-phase) or 180 degrees (out-of-phase) (e.g., Turvey, 1990; Kelso & Jeka,
1992; Schmidt, Shaw & Turvey, 1993; Jeka & Kelso, 1995). Thus, the particular realization of the
feature [time translation] as a 180 degree phase difference given that the movement is repeated is not
something that is particular to sign language movements let alone to NGT. It is a general feature of
human movement, and thus needs no further explanation from linguists. It does of course need an
explanation to researchers whose topic is motor control, and for possible explanations I refer to the
above references.

Of the two states, 0 or 180 degree relative phase, the in-phase movement is the only one that remains
as a stable state at higher speeds of movement (Turvey, 1990; Rosenbaum, 1991). It might be
interesting to consider in the future to what extent that also means that out-of-phase movements are
harder to produce than in-phase movements at lower speeds, probably the speeds found in sign
language. If it would indeed appear that out-of-phase movements are more difficult in all cases, than
that would explain the fact that type 3 symmetry is far less frequent in NGT than type 1 symmetry.
This remains a topic for further study.

6.2.3 Synchronization of movements in the two articulators

In chapter 5 we found that in the few signs in which there is a combination of the features [translation]
and [reflection], thus, a non-repeated movement in opposite directions by the two articulators, the
translation is always such that the begin and end of the movement of the two articulators concur.

This specific timing of the movements of the two articulators, such that they move synchronously, can
also be shown to be derived from more general aspects of human behavior than language. In studies
on human motor behavior it has been shown that movements of the two hands tend to occur
synchronically. even if the hands need to perfonn different movements (Kelso, Southard & Goodman,
1979ab). The time that a hand needs to move to a target has been shown to depend mainly on the size
of the target and the distance of the target; this is known as Fitts' law (Kelso et aI., 1979ab). Thus,
one would predict that reaching out to two targets of different 'complexity' at the same time with the
two hands cause the easy target to be reached quicker than the difficult target. This prediction was not
borne out in experiments by Kelso et ai. (l979ab): even if two targets are of different size and
 111

complexity, the hands start and end their movement at the same time. The movement time then is as
long as or longer than the movement required to move to the difficult target.

The particular movement we are looking at, the two articulators moving in opposite directions, seem
to be less complex than the movements described above. Thus, it is probably justified to state that the
synchronization we find in sign movements, too, is a manifestation of a more general characteristic
of human behavior, and not particular to sign language, let alone to NGT.

6.2.4 Movement of only one articulator

Finally, there is the interesting phenomenon that the weak hand does not always move. I have argued
in section 6.1 that whether the weak hand moves or not is fully predictable in all but a few exceptional
cases. The 'weak hand movement rule' (as in (6) above) stated that if the hands need to perform
different motor acts only one hand moves. Recognizing that this rough generalization might need some
modification, we can say that it is a phonetic statement, and not a phonological one. We might expect,
then, that there is a phonetic explanation for this phenomenon that lies outside the realm of grammar.

Actually, we have already seen the some of the ingredients of one such a (tentative) explanation. In
the previous section we saw that the hands tend to move in synchrony even if they have to perform
different motor acts. Also, we saw that this synchrony leads to a longer movement time for the easy
action than would be necessary if it were performed alone. Now if we assume that the signs in which
only one articulator moves would require complex coordination of the two hands, it is likely that such
a two-handed signs would require a longer stretch of time to be articulated. If we then also assume
that sign sentences require a more or less constant rhythm in which complex signs would cause
disruptions, then we might have an explanation for the fact that in asymmetric signs only one hand
tends to move: one of the ways to limit the complexity of a sign and thus the movement time required
for its articulation is to move only one hand, as this simplifies the coordination involved. For such an
explanation to be valid we need to test the different assumptions involved, of course. One of the
aspects of sign language that is ill understood so far is its rhythmic and prosodic structure, both at a
phonological and at a phonetic level; this is one of the aspects of the explanation that requires future
research.

There may also be a more straightforward explanation involved, which places the one-hand-moving
phenomenon even further outside of linguistics: it may be the case that merely the complexity of
coordination of movements of the two articulators in asymmetric signs is responsible for the fact that
such complex movements do not occur in NGT. In that case, it is not specifically the difficulty of
timing that is responsible, rather moving only one hand is the solution to the problem of too much
112

complexity. One sees in signs that this factor cannot be absolute, in the sense that the complexity is
too large to be executed at all; some signs do have such complex movement, even though it is difficult
to do. Thus, there is a pressure towards avoiding complexity in motor programming, not an absolute
prohibition of complex coordinations.

The fact that some fully symmetric signs are allowed to lose the movement of the weak hand, the
process called Weak Freeze in Padden & Perlmutter (1987; cf. p. 26 above), cannot be explained in
the same way: as the sign is fully symmetric, there is no complexity in the motor program
coordinating the two hands. Thus, I do not fully agree with Sandler (1993), who says that it "seems
reasonable to assume that any sign that utilizes two articulators is more complex motorically than signs
utilizing only one" (1993:345). Signs in which the articulators are not performing exactly the same
motor acts are more complex than signs which are fully symmetrical. But in terms of motor
programming, fully symmetric two-handed signs are not more complex than one-handed signs (which
are left completely out of consideration here). It may be the case that such a symmetric two-handed
sign costs more energy than a one-handed sign, and that energy tends to be minimized in human action
in general (cf. Lindblom, 1983a). That seems to me to be a different factor, which might explain the
Weak Drop process. Also, the energy factor might explain part of the weak hand movement rule, as
it would require less energy to move only one hand in a two-handed sign. However, this factor will
not be further discussed here.

No matter which explanation (or combinations of explanations) for the weak hand movement rule is
correct, it seems reasonable to assume that there are strong phonetic pressures regarding reduction of
motor complexity and energy use that are responsible for the generalization expressed in the 'weak
hand movement rule'. Further study of the literature on motor control can shed further light on the
precise character of these phonetic constraints.

6.3 Ease of articulation versus ease of perception

In this thesis, I have completely focused on the articulation of signs and the symmetry therein. It
appeared that from an articulatory point of view, there is quite a lot of symmetry in two-handed signs,
asymmetries being limited in number and in the possible different appearances. These articulatory
symmetries and asymmetries have been explained in terms of ease of articulation: symmetry is easy
to perform. However, there is also a perceptual side to symmetry. Many sign researchers in the past
have argued that symmetry in signs enhances redundancy and therefore is favorable to perception
(Frishberg, 1976; Battison, 1978; Siple, 1978). I will assume without further examination that
 113

symmetry is indeed easy for perception as well, and thus forms another pressure shaping the
lexicon 46 • The question that I will briefly discuss, however, is to what extent symmetry in perception
and symmetry in production do overlap.

If a sign is symmetric from the point of view of perception, there is a plane in which one articulator
can be reflected to give the configuration of the other articulator. We need to ask, now, to which part
of the articulators the sign perceiver directs his attention. In other words, what part of the articulators
need to be symmetric for a sign to be perceived as symmetrical? Most (space-based) models so far
have made the assumption that only the hand matters in a sign, as I discussed above (p. 41). If this
is indeed true, then the configuration of the rest of the articulators would not matter. I will assume for
the moment that these models are indeed right in assuming that what the signer looks at or perceives,
again recognizing that this is something that needs serious investigation instead of mere assumptions.

Considering only the symmetry of the hands, there is no reason to give any symmetry plane a special
status: any mirror is as good as any other in making the sign seem symmetric. Thus, bilateral
symmetry is not tied to merely the sagittal plane. In the corpus of signs studied here, however, it
seems that there are only a few signs in which the hands are bilaterally symmetric but in which the
symmetry plane is not the sagittal plane: these signs have a shifted symmetry plane47 • These are the
signs KROKODIL, NIJLPAARD, and ELASTIEKJE. In all these signs, the symmetry plane is the
transverse (horizontal) plane48 • Further, there are the signs GEHAKT and OMGAAN-MET, in which
there seem to be two symmetry planes: the palm orientation is symmetric with respect to a transverse
plane, whereas the finger orientation is symmetric with respect to the sagittal plane. But for all other
signs in which the hands are symmetric perceptually, the symmetry plane is the sagittal plane. It seems
then, that although reflection in any plane gives rise to the perception of symmetry, the lexicon of
NGT is strongly influenced by factors of ease of articulation.

An alternative interpretation would be that this fact (the rareness of signs with a symmetry plane
different from the sagittal plane) suggests that many models are wrong in assuming that only the hand
matters to perception: apparently signs in which there is a different symmetry plane than the sagittal
plane are disfavored, and this is because they are not completely symmetric in perception nor in

46 It should be remarked that if symmetry is studied in signs with the intention of finding out to what extent this

symmetry aids perception, the symmetry should be studied at the level of phonetics, and not of phonology: what is
perceived of a sign is not directly its phonological representation, but rather its appearance, or phonetic representation.

47 The articulatory characteristic that is tied to these signs with what I call a shifted symmetry plane is the opposite
state of the forearm (or the radioulnar joint): in these signs the strong articulator is (almost) maximally pronated, whereas
the weak articulator is (almost) maximally supinated.

48 It would be more appropriate to talk about a transverse plane, as there is no single transverse plane at a fixed

height like there is a single sagittal plane for the body.

114

production49 • I think it is an interesting topic for future research to find out to which extent the focus
on the hands in many current models is justified, or in other words, to which extent the sign perceiver
is indeed focused on the hands and ignores the position of the arm (see also p. 42 above).

Whatever the main phonetic explanation is (ease of perceptIon or ease of production, or maybe still
other factors), any phonological model should represent the fact that the sagittal plane has a special
status. It seems that this is harder for models which are completely space-based than for models which
are in part body-based. Thus, the predominantly space-based model of Uyechi (1994ab, 1995ab) in
its current form gives no special status to the sagittal plane, although Uyechi (p.c.) recognizes that this
special status is something to be represented. If the model keeps focusing on the hands and remains
so strongly space-based, then an extra 'grounding condition' (Archangeli & Pulleyblank, 1994) is
needed to give the sagittal plane its special status.

Now, back to the original question posed at the outset of this section, to which extent do symmetry
in perception and symmetry in production overlap? Without much argument, I suggest that we can
make the following statement: a joint pair is perceived as symmetric only if the more proximal joint
pairs are symmetric too. If we go on assuming that only the hand matters to perception, then this
comes down to saying that the hands are perceived as symmetrical if and only if the joint pairs
proximal to those involved directly in hand position (radioulnar and wrist), thus, the shoulder and
elbow joint pairs, are symmetric too. If one (or both) of these two joint pairs is asymmetric, then the
hands are not perceived as being symmetric regardless of whether the radioulnar and wrist joints are
symmetric or not.

49 This interpretation hinges on the assumption that ease of perception, caused by symmetry. is a dominant factor

shaping the lexicon. However. if not only the hands but also the arms matter to perception. then it is not clear how ease
of perception and ease of production can be distinguished.
 115

7 Conclusions

In this thesis, I have paid attention to one aspect of two-handed signs that had not yet received much
attention: their symmetry from the point of view of articulation. In this concluding chapter, I will first
review to which extent the research questions posed in section 2.4 have been answered in subsequent
chapters. Then, I will suggest some areas where more research is needed in the future.

1 How can the 'bilateral symmetry of the body' be described for the sign language articulators
on a phonetic (articulatory) level?
I have suggested that complete symmetry of the whole body in a phonetic description comes down
to the two members of all pairs of joints of the body being in the same anatomical state. Thus, a sign
is fully symmetric only if all of the joint pairs are in the same state.

2 In which ways can possible gestures be symmetry-breaking, and which of these types of
symmetry breaking occur in sign languages?
I argued that a gesture may be asymmetrical in one or more joint pairs, and that there are five different
ways in which a joint pair may be asymmetric. A joint pair can be asymmetric all the time of a
gesture, during part(s) of the time, under a time reflection, under a time translation, or under a
combination of a time translation and a time reflection.

Signs which are fully symmetric appear to be more frequent in NGT than asymmetric two-handed
signs. The only time transformation relating the movement of the two hands that was found frequently
in the NGT corpus was a translation, visible as a phase difference. This asymmetry occured mainly
in the shoulder and elbow joint pairs. There were only a few instances of a combination of a time
translation and a time reflection. A time reflection on its own does not seem to be a systematic feature
of any joint pair in the NGT lexicon: it only occurs in two signs which presumably are not lexicalized
forms. Asymmetries of type 2 and 6 (asymmetric during most of the sign or during all of the sign)
occur for all joint pairs, roughly with the same frequency. A strong tendency that we find looking at
different combinations of asymmetries in the different joint pairs of a sign, is that the shoulder and
elbow pairs tend to be of the same symmetry type, and so do the radioulnar and wrist joint pairs.
Another general trait of the types of gestures that we do find in NGT is that only seldomly more than
two different symmetry types are combined; in almost all cases where two symmetry types are
combined one of the two is type 1 (fully symmetric during the whole sign).

Apart from the fully symmetric signs (type 1111) and the so-called 'alternating signs' (type 3311)
which are already mentioned, only two other sign types occur frequently: types 2211 and type 6666.
116

These types cover all the signs that have traditionally been considered ' weak hand place' signs, and
some other asymmetrical signs.

3 Can the study of symmetry in signs lead us to a classification of signs that helps to reveal
phonological patterns of distribution?
The study of symmetry in this thesis has not yet led to a classification of signs. Moreover, it is
doubtful whether a classification based on the level of detail discussed here would be useful from the
point of view of phonology. However, the distinctions proposed here (the different joints involved in
the articulation of a sign, and the five ways in which a joint pair may be asymmetrical) may aid future
research on a classification of two-handed signs, if a classification is considered necessary at all, of
course.

4 What are the implications of the answers to questions 1-3 for current phonological models?
I have suggested that the characteristic aspect of alternating signs, the presence of a time translation
relating the two hands, has a broader application than was traditionally assumed: from an articulatory
point of view, so-called 'shadow signs' just as well are characterized by a time transformation.
Furthermore, there appear to be a few NGT signs in which a translation in time is combined with a
reflection in time. To represent these signs in the phonological model of van der Hulst, I have
proposed to replace the feature [alternating] with the two features [translation] and [reflection], which
indicate the kind of transformation in time that relates the movements of the two hands. I have
suggested that these features are manner features of the articulator node, and that they can be seen as
filling the specifier position of this node.

I also proposed that movement of the weak hand, which seemed to be the only formational property
that could cover most weak hand place signs, is fully predictable on the basis of the amount of
asymmetry in these signs. Thus, there is no need for a feature [weak hand] characterizing a group of
signs, which, apart from the fact that they have no movement of the weak hand, have no property in
common. Rather, as suggested before, part of these signs can be represented in the model as fully
symmetrical; the fact that the weak hand moves can be derived from the specification for direction of
movement, and maybe partly from contact features. Another part of these weak hand place signs is
strongly asymmetrical in the states of the different joints; these typically are signs of symmetry type
6666. For these signs I have proposed to add a new feature [asymmetric], which has the same function
as the two time asymmetry features: it relates the two hands. Therefore, I suggested to put this feature,
too, in the specifier position of the articulator node.

5 To what extent can symmetry be considered as 'easy' in the articulation of signs

 117

Although I have indeed found studies in the field of kinesiology and motor control indicating that for
humans in general (not specifically for sign movements), symmetry is easier than asymmetry.
However, I have not yet found nor discussed why this is actually the case. In that sense, this research
question is still open to investigation.

6 To what extent can the factor 'ease of articulation' be seen to be responsible for the patterns
found in answer to question 2?
I argued in chapter 6 that the human motor system in general favors 180 degree phase differences, and
thus, that this particular value is not a choice of the grammar of NGT. It would indeed be highly
surprising if we would find a sign language which does use phase differences smaller or larger than
180 degrees. Further, I tried to show that if the two hands would need to move in highly asymmetric
signs, the motor complexity just becomes too large to be acceptable; not to the grammar of the
language, but to the human body. One obvious way of reducing this complexity is to move only one
hand, keeping the asymmetric configuration. This, I argued, explains the absence of movement in the
weak hand in many signs.

There are a number of issues left for future research to explore. In the first place, advantages of the
present proposal for adaptations to the van der Hulst model are still to be compared to advantages of
the old model. More specifically, we still need to find out whether there is indeed, as Sandler (1993)
claims, a unified role for the weak hand in weak hand place signs, and thus, whether they do need to
be characterized as a class. In the present proposal, this allegedly natural class of weak hand place
signs has fallen apart.

Second, it would be interesting to investigate in the future whether the more detailed scale of
symmetry in signs proposed in this thesis, which admittedly is not really a clear-cut classification of
signs, might aid the description of Weak Drop. It may be the case that this process which is elusive
to researchers can be characterized in phonological tenns with the help of the revised phonological
model of van der Hulst, in which there is no longer one unified class of signs characterized by the
place feature [weak hand].

Third, further study is necessary to see whether the view of manner features as specifiers suggested
here, and the presence of two different manner nodes, is indeed a viable alternative for the current
model of van der Hulst. More knowledge about the character of manner features in the traditional
sense, being manner of movement, is needed to shed more light on the status of manner features.

In the fourth place, future research may compare the relative influences on the structure of the lexicon
of ease of symmetry for perception on the one hand and ease of symmetry for production on the other
118

hand.

Finally, more extensive study of the literature on motor control and the anatomy and physiology of
the 'upper extremity', the arms and hands, is necessary to determine which formational properties of
signs are actually a matter of grammar, and which ones can be seen as expressions of more general
properties of human behavior.
 119

References
Ann, 1. (1991) Constraining sign language handshapes: toward a phonetically grounded account of
handshapes in Taiwan Sign Language and American Sign Langage. Proceedings of Western
Conference on Linguistics (WECOL), 20.
Ann, J. (1992a) Physiological constraints in Taiwanese Sign Language handshape change. Nordic
Journal of Linguistics, 15.
Ann, J. (1992b) Physiological constraints in two finger handshapes. CLS.
Ann, J. (1993) A linguistic investigation of the relationship between physiology and handshape.
Dissertation, University of Arizona.
Archangeli, D. & D. Pulleyblank (1994) Grounded phonology. Cambrigde, MA: MIT Press.
Baker, C. & D. Cokely (1980) American Sign Language: a teacher's resource text in grammar and
culture. Silver Spring: T.J. Publishers.
Baker-Shenk, C. (1985) The facial behavior of deaf signers: evidence of a complex language.
American Annals of the Deaf, October, pp. 297-304.
Ballering, C. (1995) The phonology of sign languages: an analysis of place. MA thesis, University of
Leiden.
Battison, R. (1974) Phonological deletion in American Sign Language. Sign Language Studies, 5:5-19.
Battison, R. (1978) Lexical borrowing in American Sign Language. Silver Spring, MD: Linstok Press.
Battison, R., H. Markowicz & J. Woodward (1975) A good rule of thumb: variable phonology in
American Sign Language. In Fasold, R. & R. Shuy (eds.) Analyzing variation in language.
Washington, D.C.: Georgetown University Press.
Bernstein, N. (1967) The co-ordination and regulation of movements. Oxford: Pergamon Press.
BJees, M. (1994) Een fonologische analyse van de handvormen van de Nederlandse Gebarentaal. MA
thesis, University of Leiden.
Blevins, J. (1993) The nature of constraints on the nondominant hand in ASL. In Coulter, G. (ed.)
Phonetics and phonology, vol. 3: current issues in ASL phonology. San Diego, CA: Academic
Press. Pp. 43-62.
Brennan, M . (1995) Talk presented at the 1995 Linguistics Institute, Albuquerque, NM.
Brentari, D. (1990) Theoretical foundations of American Sign Language phonology. Dissertation,
University of Chicago.
Brentari, D. (1995) A prosidic account of two-handed signs. Ms., University of California, Davis.
Brentari, D. & J. Goldsmith (1993) Secondary licensing and the non-dominant hand in ASL
phonology. In Coulter, G. (ed.) Phonetics and phonology, vol. 3: current issues in ASL phonology.
San Diego, CA: Academic Press.
Clements, G. (1991) Place of articulation in consonants and vowels: a unified theory. In Laks, B. &
A. Rialland (eds.) L'architecture et la geometrie des representations phonologiques. Paris: Editions
de C.N.R.S.
Coerts, J. (1992) Non-manual grammatical markers: an analysis of interrogatives, negations and
topicalisations in Sign Language of the Netherlands. Dissertation, University of Amsterdam.
Cohen, Namir & Schlesinger (1977) A new dictionary of sign language. Employing the Eshkol-
Wachmann movement notation system. The Hague: Mouton.
Corina, D. & W. Sandler (1993) On the nature of phonological structure in sign language. Phonology,
} 10:165-207.
Ebbinghaus, H. & J. Hessmann (1995) Signs and words. Accounting for spoken language elements
in German Sign Language. To appear in International Review of Sign Linguistics, 1.
Edmondson, W. (1986) Autosegmental phonology: the syllable and the sign. In Tervoort, B. (ed.)
Signs of life. Proceedings of the second European congress on sign language research. Amsterdam:
University of Amsterdam; pub!, Inst. General Ling. 50. Pp. 79-94.
Edmondson, W. (1990) Segments in signed languages: do they exist and does it matter? In
 120

Edmondson, W. & F. Karlson (eds.) SLR '87: papers from the fourth international symposium on
sign language research. Hamburg: Signum. Pp. 66-74.
Ewen, C. (1994) Dependency relations in phonology. In Goldsmith, J. (ed.) The handbook of
phonological theory. Oxford: Blackwell.
Friedman, L. (1976) Phonology of a soundless language: phonological structure of the American Sign
Language. Dissertation, University of California, Berkeley.
Frishberg, N. (1975) Arbitrariness and iconicity: historical change in american Sign Language.
Language, 51:696-719.
Frishberg, N. (1976) Some aspects of the historical development of signs in American Sign Language.
Dissertation, University of California, San Diego.
Frishberg, N. (1985) [In SLR '83]
Geertz, C. (1973) Person, time, and conduct in Bali. In Geertz, C. XX, XX. Pp. 360-411.
Goldsmith, J. (1991) Autosegmental and metrical phonology. Oxford: Blackwell.
Gould, SJ. (1987) Time's arrow, time's cycle. Cambridge, MA: Harvard University Press.
Greftegreff, I. (1992) Orientation in indexical signs in Norwegian Sign Language. Nordic Journal of
Linguistics, 15: 159-182.
Haegeman, L. (1994) Introduction to government and binding theory. 2nd ed. Oxford: Blackwell.
Hollerbach, J. & C. Atkeson (1981) Characterization of joint-interpolated arm movements. In
Heuer, H. & C. Fromm (eds.) Generation and modulation of action patterns. Berlin: Springer
Verlag. Pp. 41-54.
Hulst, H. van der (1989) Atoms of segmental structure: components, gestures and dependency.
Phonology, 6:253-284.
Hulst, H. van der (1993) Units in the analysis of signs. Phonology, 10:209-241.
Hulst, H. van der (1994a) Dependency relations in two-handed signs. Ms., University of Leiden.
Hulst, H. van der (1994b) An introduction to radical CV phonology. In Shore, S. & M. Vilkuna (eds.)
1994 yearbook of the Linguuistic Association of Finland. Helsinki: Suomen Kielitieteellinen
Yhdistys. Pp. 23-56.
Hulst, H. van der (1995a) Dependency relations in the phonological representation of signs. To appear
in H. Bos & T. Schermer (eds.) Proceedings of the 4th European congres on SLR, Munich 1-3
September 1994; Hamburg: Signum Press.
Hulst, H. van der (1995b) Fonologie en gebarentaal. Paper presented to the ATW vrijdagclub,
University of Amsterdam.
Hulst, H. van der (1995c) Radical CV phonology: the categorial gesture. In Katamba, F. & J. Durand
(eds.) Frontiers of Phonology. London: Longman.
Hulst, H. van der (1995d) Radical CV phonology: segments and syllables. Paper presented at Current
Trends in Phonology, Royaumont, June 19-21 1995.
Hulst, H. van der (1995e) On the other hand. To appear in Lingua.
Hulst, H. van der (1995f) The composition of handshapes. To appear in Trondheim Working Papers
in Linguistics.
Hulst, H. van der & W. Sandler (1994) Phonological theories meet sign language: two theories of the
two hands. Toronto Working Papers in Linguistics, 13-1 :43-73.
Jeka, J. & J.A.S. Kelso (1995) ManipUlating symmetry in the coordination dynamics of human
movement. Journal of Experimental Psychology: Human Perception and Performance,
21-2:360-374.
Kelso, J.A.S., D. Southard & D. Goodman (1979a) On the nature of human interlimb coordination.
Science, 203: 1029-1031.
Kelso, J.A.S ., D. Southard & D. Goodman (1979b) On the coordination of two-handed movements.
Journal of Experimental Psychology: Human Perception and Performance, 5-2:229-238.
Kelso, J.A.S. & B. Tuller (1984) A dynamical basis for action systems. In Gazzaniga, M. (ed.),
Handbook of cognitive neuroscience. New York: Plenum. Pp. 321-356.
 121

Kelso, J.A.S. & J. Jeka (1992) Symmetry breaking dynamics of human multilimb coordination.
Journal of Experimental Psychology: Human Perception and Performance, 18-3:645-668.
Kendon, A. (1988) Sign languages of aboriginal Australia. Cambridge: Cambridge University Press.
Kenstowicz, M. (1994) Phonology in generative grammar. Oxford: Blackwell.
Klima, E. & U. Bellugi (1979) The signs of language. Cambridge, MA: Harvard University Press.
Koenen, L., T. Bloem & R. Jansen (1993) Gebarentaal: de taal van doven in Nederland. Amsterdam:
Nijgh & van Ditmar.
Kooij, E. van der (1994) De component beweging in Nederlandse Gebarentaal: een fonologische
analyse. MA thesis, University of Leiden.
Kroeber, A. (1958) Sign language inquiry. International Journal of American Linguistics, 24-1:1-19.
Ladefoged, P. (1983) 'Out of chaos comes order'; physical, biological, and structural patterns in
phonetics. In Broecke, M. van den & A. Cohen (eds.) Proceedings of the tenth international
congress of phonetic sciences. Dordrecht: Foris. Pp. 83-95 .
Ladefoged, P. (1967) Three areas of experimental phonetics. London: Oxford University Press.
Latash, M. (1993) Control of human movement. Champaign, IL: Human Kinetics Publishers.
Liddell, S. (1984) THINK and BELIEVE: sequentiality in American Sign Language. Language,
60:372-399.
Liddell, S. & R. Johnson (1989) American Sign Language: the phonological base. Sign Language
Studies, 64:195-278.
Lindblom, B. (1983) Economy of speech gestures. In MacNeilage, P. (ed.) The production of speech.
New York: Springer Verlag. Pp. 217-245.
Lindblom, B. (1984) Can the models of evolutionary biology be applied to phonetic problems? In
Broecke, M. van den & A. Cohen (eds.) Proceedings of the tenth international congress ofphonetic
sciences. Dordrecht: Foris. Pp. 67-81.
Lindblom, B. (1990a) On the notion of 'possible speech sound'. Journal of Phonetics, 18:135-152.
Lindblom, B. (1990b) Explaining phonetic variation: a sketch of the H&H theory. In Hardcastle, W.
& A. Marchal (eds.) Speech production and speech modelling. Dordrecht: Kluwer.
Lindblom, B. (1991) The status of phonetic gestures. In Mattingly, I. & M. Studdert-Kennedy (eds.)
Modularity and the motor theory of speech perception. Hillsdale, NJ: Lawrence Erlbaum Associates.
Pp.7-24.
Lindblom, B. (1995) A view of the future of phonetics. Proceedings of International Congress of
Phonetic Sciences, 1995, Stockholm, vol. 1:462-469.
Lindblom, B., P. MacNeilage & M. Studdert-Kennedy (1983) Self-organizing processes and the
explanation of phonological universals. In Butterworth, B., B. Comrie & V. Dahl (eds.)
Explanations for language universals. The Hague: Mouton.
Luttgens, K. & K. Wells (1982) Kinesiology: scientific basis of human motion. 7th ed. Philadelphia:
Saunders College Publishing.
MacConaill & Basmajian (1969) Muscles and movements: a basis for human kinesiology. Baltimore:
The Williams & Wilkins Company.
Maddieson, 1. (1991) Testing the universality of phonological generalizations with a phonetically
specified segment database: results and limitations. Phonetica, 48:193-206.
Maddieson, I. (1995) Gestural economy. Proceedings of International Congress of Phonetic Sciences,
1995, Stockholm, vol. 4:574-577.
Mandel, M. (1981) Phonotactics and morphophonology in American Sign Language. Dissertation,
University of California, Berkeley.
Mandel, M. (1982) Major information from a 'minor parameter': point of contact in sign language
phonology. BLS, 8:533-549.
Marr, D. (1982) Vision. New York: Freeman.
McCarthy, J. (1988) Feature geometry and dependency: a review. Phonetica, 43:84-108.
Miller, C. (1990) La geometrie des traits en langue des signes. Ms. University of Quebec in Montreal.
122

Miller, C. (1995) Talk presented at the 1995 Linguistics Institute, Albuquerque, NM.
Morasso, P. (1981) Spatial control of ann movements. Experimental Brain Research, 42:223-227.
Nagahara, H. (1988) Toward an explicit phonological representation for American Sign Language. MA
thesis, University of California, Los Angeles.
Newkirk, D. (1981) On the temporal segmentation of movement in American Sign Language. Working
paper, Salk Institute for Biological Studies.
Padden, C. D. Perlmutter (1987) American Sign Language and the architecture of phonological theory.
Natural Language and Linguistic Theory, 5:335-375.
Perlmutter, D. (1990) On the segmental representation of transitional and bidirectional movements in
ASL phonology. In Fischer, S. & P. Siple (eds.) Theoretical issues in sign language research, vol.
1: linguistics. Chicago: Chicago University Press. Pp. 67-80.
Prillwitz, S ., R. Leven, H. Zienert, T. Hanke, J. Henning, et al. (1989) HamNoSys, version 2.0.
Hamburg notation system for sign languages, an introductory guide. Hamburg: Signum Press.
Rosenbaum, D. (1991) Human motor control. San Diego: Academic Press.
Sandler, W. (1989) Phonological representation of the sign: linearity and nonlinearity in American
Sign Language. Dordrecht: Foris.
Sandler, W. (1990) Temporal aspects and ASL phonology. In Fischer, S. & P. Siple (eds.) Theoretical
issues in sign language research, vol. 1: linguistics. Chicago: Chicago University Press. Pp. 7-35.
Sandler, W. (1993) Hand in hand: the role of the nondominant hand in ASL. The Linguistic Review,
10:337-390.
Sandler, W. (1995a) One phonology or two? Sign language and phonological theory. Glot
International, 1-3:3-8.
Sandler, W. (1995b) Markedness in American Sign Language handshapes: a componential analysis.
In Hulst, H. van der & J. van de Weijer (eds.) Proceedings of the HIL phonology conference. The
Hague: Holland Academic Graphics. Pp. 369-399.
Sandler, W. (1995c) Phonological features and feature classes: the case of movements in sign
languages. To appear in Lingua.
Schermer, G., C. Fortgens, R. Harder & E. de Nobel, eds. (1991) De Nederlandse Gebarentaal.
Twello: Van Tricht.
Schmidt, R. (1988) Motor control and learning: a behavioral emphasis. 2nd ed. Champaign, IL:
Human Kinetics Publishers.
Schmidt, R. , B. Shaw & M. Turvey (1993) Coupling dynamics in interlimb coordination. Journal of
Experimental Psychology: Human Perception and Performance, 19-2:397-415.
Siple, P. (1978) Visual constraints for sign language communication. Sign Language Studies,
19:95-110.
Stack, K. (1995) Weak drop and the representation of 2-handed signs. Ms., University of California,
Los Angeles, and talk presented at the 1995 Linguistic Institute, Albuquerque, NM.
Stewart, I. (1975) Concepts of modern mathematics. Harmondsworth: Penguin Books.
Stewart, I. & M. Golubitsky (1992) Fearful symmetry: is God a geometer? London: Penguin Books.
Stokoe, W. (1960) Sign language structure. 2nd ed., 1978. Silver Spring, MD: Linstok Press.
Stokoe, W., D. Casterline & C. Croneberg (1965) A dictionary of American Sign Language on
linguistic principles. 2nd ed., 1976. Silver Spring, MD: Linstok Press.
Supalla, T. (1986) The classifier system in American Sign Language. In Craig, C. (ed.) Noun classes
and categorization. Amsterdam: John Benjarnins. Pp. 181-214.
Turvey, M. (1990) Coordination. American Psychologist, 45-8:938-953 .
Turvey, M., H. Fitch & B. Tuller (1982) The Bernstein perspective: I. The problems of degrees of
freedom and context-conditioned variability. In Kelso, J.A.S. (ed.) Human motor behavior: an
introduction. Hillsdale, NJ: Lawrence Erlbaum Associates. Pp. 239-252.
Uyechi, L. (1994a) Visual phonology: another look at movement in American Sign Language. Talk
presented at University of Konstanz.
 123

Uyechi, L. (l994b) Generalizing the articulators in American Sign Language. Ms., Stanford University.
Uyechi, L. (1995a) The geometry of visual phonology. Dissertation, Stanford University.
Uyechi, L. (l995b) The symmetry of two-handed signs in American Sign Language. Ms., Stanford
University.
West, Jr., La Mont (1960) The sign language, an analysis. Volumes I and II. Dissertation, Indiana
University.
Weyl, H. (1952) Symmetry. Princeton, NJ: Princeton University Press.
Wilbur, R. (1990) Why syllables? What the notion means for ASL research. In Fischer, S. & P. Siple
(eds.) Theoretical issues in sign language research, vol. 1: linguistics. Chicago: University of
Chicago Press. Pp. 81-108.
Wilbur, R. (1993) Syllables and segments: hold the movement and move the holds! In Coulter, G.
(ed.) Phonetics and phonology, vol. 3: current issues in ASI phonology. San Diego, CA: Academic
Press. Pp. 135-168.
Wilcox, S. (1992) The phonetics of jingerspelling. Amsterdam: John Benjamins.
Winter, D. (1979) Biomechanics of human movement. New York: John Wiley & Sons.
124

Appendix A: ranges of motion and reference axes for movements at each

joint
Shoulder joint

90°
 125

Appendix A (cont.)
126

Appendix A (cont.)

Elbow joint
 127

Appendix A (cont.)

Radioulnar joint

NEUTRAL

SUPINATION PRONATION
~o __________~w.~________~

Wrist joint

RADIAL ULNAR
FLEXION FLEXION

900------------r-~--r_----------- 900
128

Appendix A (cont.)
 129

Appendix B: corpus of all two-handed signs (n=320)

AANDACHT attention CAFE cafe GEGEVENS data
AANKLEDEN to dress CAMPING/TENT camping GEHAKT minced meat
AANPASSEN to adapt site/tent GEHOORZAAM obedient
AANRANDEN to assault CHIRURG surgeon GELIJK-HEBBEN to be right
AARDAPPEL potato COMBINATIE combination GELIJK equal
AARZELEN to hesitate COMMUNICATIE GENEN genes
ABONNEE subscriber communication GESPANNEN tensed
ADRES address COMPUTER computer GESPREK conversation
ACCEPTEREN to accept CONTACT contact GEVOELlG-l sensitive
ADVOCAAT attorney CONTRACT contract GEVOELlG-2 sensitive
AFBLlJVEN to keep off DAMP vapor GEZOND healthy
AGRESSIE agression DANSEN to dance GIDS guide
APART apart DECEMBER December GIERIG stingy
APRIL April DEFECT defective GRADEN degrees
AUGUSTUS August DEKEN blanket GROEIEN to grow
AUTO car DEMONSTRATIE GROEN/GROENTE
AUTORIJDEN to drive a car demonstration green/vegetables
A YOND evening DEUR door GROEP group
BABY baby DICHT closed GUL generous
BAD bath DIJK dike HAAR-WASSEN to wash
BAKKER/OVEN baker/oven DIRECTEUR managing one's hair
BALLET ballet director HALF half
BANAAN-l banana DISCUSSIE discussion HANDEN-WASSEN to wash
BANAAN-2 banana DOEN to do one's hands
BANK-l bank DONKER dark HANDIG handy
BANK-2 bench DOORGAAN to continue HARD hard
BEER-l bear DROOG dry HATEN to hate
BEER-2 bear DRUKlHAAST busy/hurry HELPEN to help
BEGELEIDEN to guide EGEL hedgehog HERBERG inn
BEGRAFENIS funeral ELASTIEKIRISICO rubber HERFST fall
BELEEFD polite band/risk HOED hat
BESLISSEN to decide ELASTIEKJE/SLAPPELING HOND dog
BETALEN to pay small rubber band/weakling HUILEN to cry
BEWIJS proof EMANCIPATIE emancipation HUIS house
BEZWAAR objection ENE-OOG-IN -ANDERE-OO IN-DE-WAR confused
BINNEN inside G-UIT in-through-one-eye-out- INDUSTRIE industry
BLIJ-l happy the-other INHOUD content
BLIJ-2 happy ENE-OOR-IN-ANDERE-OOR INTERNATIONAAL
BLUT broke -UIT in-through-one-ear-out- international
BOEK book the-other INV ALIDE disabled
BOOM tree FABRIEK factory INVLOED influence
BOOT boat FAILLIET bankrupt ISOLEREN to isolate
BORD plate FEEST party JAMMER a pity
BOS forest FIETSEN to ride a bike JAREN/ZATERDAG
BOTER butter FIGUUR figure years/Saturday
BOUWEN to build FOUT wrong JARIG bithday
BRAND fire FRUSTRATIE frustration JAS coat
BRIEF letter GEBAREN to sign JURK dress
BROEK trousers GEBEUREN to happen KAAS-I cheese
BROOD-I bread GEBRUIKEN to use KAAS-2 cheese
BROOD-2 bread GEDRAG behavior KAMPIOEN champion
BUS bus GEDULD patience KANGOEROE kangaroo
130

KANTOOR office with (somenone) SCHOON clean

KAPOT broken ONDERBROEK underpants SCHOPPEN to kick
KATER hangover ONDERSTEUNEN to support SEPTEMBER September
KERK church ONDERTITELING subtitles SLA lettuce
KEUKEN kitchen ONDERZOEK research SNEEUW snow
KIP chicken ONGELUK accident SOCIAAL social
KLAAR ready ONHANDIG clumsy SOK sock
KLEDING clothes ONTSPANNEN relaxed SPANNEND exciting
KLETSEN to chat ONZEKER insecure SPELEN to play
KOE cow OORLOG war SPOOKIGEEST phantomlgoast
KOFFIE coffee OP-SLOT locked SPORT sport
KOK cook OPEN open STAAN to stand
KOKEN to cook OPLOSSING solution STAKING strike
KOPIEREN to copy OPRUIMEN to clear away STEMPEL stamp
KOUD cold OPSCHEPPER bragger STIER bull
KRANT newspaper OVERLEDEN deceased STOEL chair
KROKODIL crocodile OVERSTUUR upset SYSTEEM system
LAATSTE last PAARS purple TAFEL table
LATEN-ZIEN to show PADDESTOEL mushroom TAFELDEKKEN to lay the
LAWAAI noise PAPIER paper table
LENTE spring PARACHUTE parachute TAXI cab
LESGEVEN to teach PAUZE break TECHNIEK/MACHINE
LEVEN to live PERSONEEL staff engineering/machine
LEZEN to read PESTEN to tease TEGENOVERGESTELD
LICHT light of weight PIZZA pizza opposite
LIJDEN to suffer POLmEK politics TENTOONSTELLING
LIJST list PRINCIPE principle exhibition
LUCHTBALLON hot air PSYCHOLOGIE psychology THEATER theatre
balloon RAMPIBOM disasterlbomb THEE tea
MAAKT-NIET-UIT doesn't REAGEREN-OP to react upon THERAPIE therapy
matter RECHT-HEBBEN-OP to be TIJD time
MAATREGEL measure entitled to TOEGEVEN to admit
MAMMOET mammoth RECHT(SPRAAK)IRECHTEN TOESTEMMEN to consent
MANIER manner justice/law (school) TOLK interpreter
MASSAGE massage REGELEN to arrange TREIN train
MELK milk REGELS rules TROTS proud
MES knife REIZEN to travel TRUI sweater
MIMIEK mimicry RELATIE relation(ship) TUNNEL tunnel
MINISTER minister RESTAURANT restaurant TUSSEN between
NAAR-BED-GAAN to sleep RESULTAAT result UITSTEL delay
(with someone) ROBOT robot (postponement)
NADOEN to mirror ROKskirt VAAK often
NAT wet ROL role VERANDEREN to change
NEDERLAND the Netherlands ROMANTISCH romantic VERANTWOORDELUKHEID
NEUKEN to fuck ROMMEL mess responsibility
NIEMAND/LEEG RUILEN to trade VERDELEN to divide
nobody/empty RUZIE row VERGADERING meeting
NIETS nothing SAMEN-MET-EEN-GROEP VERHAAL story
NIJLPAARD hippopotamus together with a group VERHUIZEN to move (house)
NlVEAU level SCHAATSEN to skate VERKEER traffic
NON nun SCHAMEN to be ashamed VERSCHIL difference
OCHTEND morning SCHEIDEN to separate VERSCHILLENDE various
OEFENEN to practice SCHOEN shoe VERTELLEN to tell
OMGAAN-MET to associate SCHOOL school VERTRAGING delay
}' '1f1 ~<)\_\ berr
((;tJ1J1~) f1/~
KJ~V\rA f>:\O f ) J
lU;t\~/..JP\.;(\ ./. 1_ fiflA./!I&'(K'rV
v

~~
JjJ::ri< IX)! Kt~ R f ~OkJL-\ \~\~L---
 131

(retardation)
VERTROUWEN-HEBBEN-IN
to trust
VERWISSELEN to exchange
VERZORGEN to take care of
VIJAND enemy
VLEES meat
VOETBAL soccer
VOORBEELD example
VRIEND friend
VRIJ free
WAAR true
WAARSCHIJNLIJK probably
WACHTEN to wait
WAKKER awake
WALRUS walrus
W ANHOOP desperation
WANNEER when
WARM warm
WEDSTRIJD game
WEGGAAN to leave
WERELD world
WERKEN to work
WINKELfBOODSCHAPPEN
shop/groceries
WINTER winter
WOLKEN clouds
X-WOORDEN [diff. words
with xl
ZEKERIVAST certain/secure
ZENUWACHTIG nervous
ZIEK ill
ZIEKENHUIS hospital
ZIELEPOOT pitiful person
ZITTEN to sit
ZONDAG Sunday
ZWAAR heavy
ZWAK weak
132

Appendix C: corpus of asymmetric signs (n=139)

Each gloss is followed by two sets of codes. The first set indicates the symmetry type for the four
joints considered in this thesis: shoulder, elbow, radioulnar, wrist. The numbers refer to the symmetry
types discussed on page xx. In fifth position there is a letter indicating whether the handshapes of the
two signs are different (d) or the same (s). The second set of codes lists for each of the four joints and
the handshape whether one joint in the set is moving (1), both are moving (2), or none is moving (0).
~ 0 ~~ l
(t D - (
AAROAPPEL 11210 00100 22000
AORES 66660 00100 FABRIEK 3311S 22000
AFBLIJVEN 3113S 20020 FIETSEN 3311S 22000
AGRESSIE 3311S 22000 FOUT 66620 00010
AUTO 3311S 22000 GEBAREN 3311S 22000
BABY 6666S 22000 GEBEUREN 3311S 22000
BAO 3311S 22000 GEBRUIKEN 66660 00011
BALLET 3311S 22000 GEOULD 3311S 22000
BANAAN-2 66660 11110 GEHAKT 3361S 22000
BEER-2 3311S 22000 GELUK-HEBBEN 2211S 11000
BEGELEIDEN 66660 22000 GENEN 1155S 22220
BETALEN 62660 11000 GESPREK 3311S 22000
BEWIJS 1211S 01000 GIDS 11610 22000
BINNEN 6212S 11010 GRAOEN 6666S 11000
BOOM 66210 00100 GROEN/GROENTE 2211S 11000
BORD 62660 00010 HALF 2212S 11010
BOS 6621S 22100 HANDEN-W ASSEN 3311S 22000
BOTER 6622S 01011 HARD 12110 11000
BOUWEN 2211S 11000 HERFST 6666S 22000
BRAND 2252S 11200 HUILEN 3311S 22000
BRIEF 61610 22000 INDUSTRIE 3311S 22000
BROOD-2 2261S 11000 INHOUD 22610 11000
BUS 3311S 22000 INVALIDE 1113S 00020
CAFE 62260 11110 IN-OE-WAR 3311S 22000
CHIRURG 66660 11000 ISOLEREN 6611S 10000
COMBINATIE 1131S 00200 JAMMER 22160 11000
COMMUNICATIE 1311S 02000 KAAS-1 2211S 11000
COMPUTER 66660 11000 KAAS-2 66660 11000
CONTACT 1161S 22002 KEUKEN 1131S 00200
CONTRACT 66260 00100 KIP 6666D 11000
DANS EN 66660 00100 KLETSEN 6666S 22002
DEUR 1112S 00010 KOFFIE 6661S 11000
DISCUSSIE 3313S 22020 KOKEN 3311S 22000
OOORGAAN 2211S 11000 KOPIEREN 22660 11001
EGEL 61260 00101 KROKODIL 2266S 11000
ELASTIEKJElSLAPPELING 6666S LAATSTE 2211S 11000
22000 LATEN-ZIEN 66660 22000
ENE-OOG-IN-ANOERE-OOG-UITI411S LESGEVEN 6666D 11000
02000 ~I LEZEN 6666D 00010
ENE-OOR-IN-ANOERE-OOR-UIT 4411S LIJDEN 3311S 22000
 133

LIJST 6616S 11000 22000

LUCHTBALLON 6666D 22000 VERZORGEN 3311S 22000
MAAKT-NIET-UIT 3313S 22020 VOETBAL 6662D 00010
MELK 3311S 22000 VOORBEELD 6666D 00010
MES 2211 S 11000 VRIEND 1161 S 22000
MIMIEK 3311S 22000 WEDSTRIJD 3311S 22000
NAAR-BED-GAAN 3311S 22000 WEGGAAN 6612S 11010
NIJLPAARD 2266S 11000 WERKEN 2211S 11 000
OMGAAN-MET 6663S 00020 WINKELIBOODSCHAPPEN 1113S
ONDERSTEUNEN 6666D 11000 00020
ONZEKER 3311S 22000 WINTER 3366S 22000
OP-SLOT 1121D 00100 ZEKERIVAST 2211S 11000
OVERLEDEN 1155S 00220 ZIEK 6666D 11000
OVERSTUUR 33118 22000 ZIEKENHUIS 6616S 22000
PAARS 2211S 11000 ZIELEPOOT 2211S 11000
PADDESTOEL 6661D 00100
PARACHUTE 6666D 22000
PIZZA 6266D 11010
POLITIEK 6621D 00100
PRINCIPE 2212S 11010
PSYCHOLOGIE 6611D 11000
RECHT(SPRAAK)IRECHTEN 3311 S
22000
RECHT-HEBBEN-OP 2211S 11000
REGELS 6666D 11000
RELATIE 3361S 22000
RESTAURANT 3311S 22000
ROBOT 3311S 22000
ROL 3311S 22000
ROMMEL 3311S 22000
RUILEN 3311S 22000
SCHAATSEN 3311S 22000
SCHOOL 6666D 11010
SCHOPPEN 6662S 00010
STEMPEL 2261 D 11000
TENTOONSTELLING 3311S 22000
THEATER 3311 S 22000
THEE 6666D 11000
THERAPIE 6666D 22000
TIJD 6612D 00010
TaLK 1161S 00100
TUNNEL 6212S 11010
TUSSEN 2261S 11000
VERANDEREN 1155S 00220
VERDELEN 6662S 00010
VERHAAL 3311S 22000
VERHUIZEN 5111 S 22020
VERKEER 3111S 20000
VERTELLEN 3311S 22000
VERTROUWEN-HEB BEN-IN 1161S
 134

Appendix D: corpus of completely symmetric signs (n=181)

AANDACHT FIGUUR MAMMOET SOK

AANKLEDEN FRUSTRATIE MANIER SPANNEND
AANPASSEN GEDRAG MASSAGE SPELEN
AANRANDEN GEGEVENS MINISTER SPOOKIGEEST
AARZELEN GEHOORZAAM NADOEN SPORT
ABONNEE GELIJK NAT STAAN
ACCEPTEREN GESPANNEN NEDERLAND STAKING
ADVOCAAT GEVOELIG-l NEUKEN STIER
APART GEVOELIG-2 NIEMANDILEEG STOEL
APRIL GEZOND NIETS SYSTEEM
AUGUSTUS GIERIG NIVEAU TAFEL
AUTORIJDEN GROEIEN NON TAFELDEKKEN
AVOND GROEP OCHTEND TAXI
BAKKER/OVEN GUL OEFENEN TECHNIEKIMAC
BANAAN-l HAAR-WASSEN ONDERBROEK HINE
BANK-l HANDIG ONDERTITELING TEGENOVERGE
BANK-2 HATEN ONDERZOEK STELD
BEER-l HELPEN ONGELUK TOEGEVEN
BEGRAFENIS HERBERG ONHANDIG TOESTEMMEN
BELEEFD HOED ONTSPANNEN TREIN
BESLISSEN HOND OORLOG TROTS
BEZWAAR HUIS OPEN TRUI
. BLIJ- l INTERNATIONA OPLOSSING UlTSTEL
BLIJ-2 AL OPRUIMEN VAAK
BLUT INVLOED OPSCHEPPER VERANTWOOR
BOEK J AREN/ZA TERD PAPIER DELIJKHEID
BOOT AG PAUZE VERGADERING
BROEK JARIG PERSONEEL VERSCHIL
BROOD-l JAS PESTEN VERSCHILLEND
CAMPINGITENT JURK RAMPIBOM E
DAMP KAMPIOEN REAGEREN-OP VERTRAGING
DECEMBER KANGOEROE REGELEN VERWISSELEN
DEFECT KANTOOR REIZEN VIJAND
DEKEN KAPOT RESULTAAT VLEES
DEMONSTRATIE KATER ROK VRIJ
DICHT KERK ROMANTISCH WAAR
DIJK KLAAR RUZIE W AARSCHIJNLIJ
DIRECTEUR KLEDING SAMEN-MET-EE K
DOEN KOE N-GROEP WACHTEN
DONKER KOK SCHAMEN WAKKER
DROOG KOUD SCHEIDEN WALRUS
DRUKlHAAST KRANT SCHOEN WANHOOP
ELASTIEKIRISIC LAWAAI SCHOON WANNEER
0 LENTE SEPTEMBER WARM
EMANCIPATIE LEVEN SLA WERELD
FAILLIET LICHT SNEEUW WOLKEN
FEEST MAATREGEL SOCIAAL X-WOORDEN
 135

ZENUW ACHTIG
ZITTEN
ZONDAG
ZWAAR
ZWAK