Professional Documents
Culture Documents
5REHUW(-RKQVRQ6FRWW./LGGHOO
6LJQ/DQJXDJH6WXGLHV9ROXPH1XPEHU:LQWHUSS$UWLFOH
3XEOLVKHGE\*DOODXGHW8QLYHUVLW\3UHVV
'2,VOV
)RUDGGLWLRQDOLQIRUPDWLRQDERXWWKLVDUWLFOH
KWWSVPXVHMKXHGXDUWLFOH
Access provided by UFMS - Fundao Universidade Federal do Mato Grosso do Sul (23 Aug 2016 10:53 GMT)
Toward a Phonetic
Representation of Signs:
Sequentiality and Contrast
This is the first of several papers that will describe the authors complete approach to the phonetic representation of signed languages. One paper will appear in each successive number of Sign Language Studies until the series is
completed. Future papers will deal with the topics of segmentation, hand configuration, placement, orientation, representation of analogical space, and phonological alternations.
The idea that sign languages exist as real human languages and
that they have a structural design like that found in vocally produced
languages is quite recent. Stokoes (1960) proposal that American Sign
Language (ASL) is a real language was considered preposterous by many
linguists because sign languages had never before been treated as fully
realized natural languages. People were used to thinking of them as simple systems of communication or as representations of spoken languages.
In addition, the field of linguistics was wedded to the idea that all languages must be produced vocally. From this perspective, because sign
languages are produced by the hands and body and perceived visually,
they would have to be something other than languages.
Over the course of the next two decades, however, numerous linguists came to accept Stokoes ideas and became involved in describing
Robert E. Johnson is Professor of Linguistics at Gallaudet University, Washington,
D.C. Scott K. Liddell is Adjunct Professor at the University of Utah.
241
Sign Language Studies Vol. 11 No. 2 Winter 2010
the structure of ASL and other sign languages. The idea that sign languages have grammars similar to those of vocally produced languages
gradually gained nearly universal acceptance during the late 1970s and
the early 1980s. Today it would be difficult to find a professional linguist
arguing that sign languages are not real languages.
The most obvious and immediate difference between spoken and
signed languages, of course, lies in the manner of their production and
perception. This difference, often described as a modality difference, is
manifested in the physical activities required to create and perceive
the language signal. Speakers of a vocally produced language coordinate different articulatory systems within the vocal tract that produce
groups of sounds recognizable as words. Signers coordinate activities
of the hands, arms, torso, face, and head to produce groups of visible
physical gestures recognizable as signs. Similarly, different sets of receptors are required to understand the signals. On the one hand, variations in sound waves must be recognized by the auditory system,
and, on the other, variations in light waves must be recognized by the
visual system.
Since the sign language signal is so different from that of a vocally
produced language signal, one cannot simply impose on signs phonetic
structures known to exist in vocally produced languages. Rather, analysts must attempt to understand the phonetic structure of signs on
their own terms. There have been a number of such attempts. In this
article, which is the first in a series of articles proposing a phonetic theory of signed languages, we discuss Stokoes original proposal regarding the underlying physical structure of signs, as well as some of the
subsequent attempts to conceive of signs as being composed of sequential segments. We conclude that while the evidence for segmenting signs is compelling, each of the proposed systems of segmentation
has significant problems, and in subsequent articles in the series we argue for a new way of conceiving of signs as sequentially organized
segments. The new proposal differs in the numbers and types of segments that compose signs. It also differs in that we concentrate on representing segments phonetically rather than trying to conceive of sign
structure at a more abstract phonemic level. Finally, this series of articles taken together proposes a usable transcription system for representing signs phoneticallyone that can be applied to a wide variety
of signed languages and refined or rejected on the basis of the description of naturally occurring linguistic events.
Stokoes System
Cheremes and Claims of Simultaneity
Stokoe (1960) was the first to propose that signs have an internal structure equivalent to the phonemic level in vocally produced languages.
In order to mark the modality difference, he coined the term chereme
as the sign language equivalent of a phoneme, modeling his system after the phonetic and phonemic system of the American Structural
School (Sapir 1925, 1933; Bloomfield 1933; Swadesh 1934; Pike 1947;
Trager and Bloch 1941).1 In Stokoes system every sign is divided into
three parts or aspects: the place of articulation (tabula or tab, commonly
called placement), the active hand (designator or dez, commonly called
handshape), and the action of producing the sign (signation or sig, commonly called movement). Sign language cheremes, like phonemes in vocally produced languages, are said to have the property of distinguishing
one (signed) word from another.
In Stokoes cheremic system cheremes combine simultaneously
rather than sequentially in forming signs. In his model every simple sign
is composed of one location (tab), one oriented handshape (dez), and
one or more movements (sig). This schematic structure was applied to
all of the signs in the Dictionary of American Sign Language (Stokoe, Casterline, and Croneberg 1965). Although the dictionary claims that the
three aspects of a sign combine simultaneously to form a sign, movements could be specified sequentially. For example, consider the sign
CHICAGO, illustrated in figure 1. In producing this sign the hand first
makes a rightward movement, then a downward movement.
In Stokoes cheremic system CHICAGO is represented as C>V
(1965, 50), where the location in space in front of the body is represented by the symbol , the handshape is a C, and the two sequential movements are > (rightward) and subsequently V (downward).2,3
The fact that Stokoe was attempting an emic level of representation is evident in his use of the symbol representing the torso [] (tab).
Although the signs GUILTY, PLEASE, ASSERTIVE, DISGUSTING, and RUSSIA are all represented as being produced at the torso location [] , each
Figure 1.
CHICAGO:
C>v
a reader seeing the notation []Cx>x would know that a C hand configuration contacts the torso (x), moves to the right (>), then contacts
the torso again (x), the notation does not specify which part of the
torso is to be contacted. Since the system does not identify the conditions under which specific locations on the torso might be used, a
person unfamiliar with the sign CONGRESS would be unable to reproduce it solely on the basis of the notation.
The second aspect of signs identified by Stokoe et al. is what acts,
that is, the designator, or dez, now commonly referred to as handshape,
for which Stokoe proposes nineteen symbols. Just as the location symbols do not identify phonetically precise places of articulation for the
hand, the dez symbols do not identify phonetically precise hand configurations. For example, the contrastive hand configurations A, S, and
T are all represented as instances (allochers) of the A chereme. Similarly,
the contrastive hand configurations G, 1, and D are all represented as
instances of the G chereme. Because Stokoe does not specify the complementary distribution of these allochers, there is no principled way
to know which variant to use with which sign. Thus, again, a person
reading a notation for an unknown sign with an A chereme would be
unable to reproduce the signs hand configuration from its notation
alone.5
There is one more significant aspect of the representation of dez
symbols that needs to be mentioned here. The nineteen symbols constituting the dez system are presented without any reference as to how
the hands are oriented. In actual practice Stokoe frequently adds subscripts or other diacritics that describe certain perceptual details of
hand orientation. This is evident in the comparison of the Stokoe notations for BLACK and THINK, illustrated in example 1. BLACK (1a) begins with the fingertip oriented toward the left (G<), while in THINK
(1b), it is oriented toward the signer (Gt ).6
1.
uG x>
<
uG x
t
a. BLACK
b. THINK
(Stokoe, Casterline, and Croneberg 1965, 126, 128)
the number of possible dez symbols jumps well into the hundreds.
This argues for separating orientation from hand configuration in order to reduce the inventory of primitives from which signs may be
composed. Following Battison (1974), Frishberg (1975), Woodward
(1973), and others, it is now common to cite four independent manual aspects of sign formation: hand configuration, location, movement, and orientation.7
The third and final aspect of sign structure proposed by Stokoe is
signation (sig), the action of producing a sign, which is now commonly referred to as movement. While location and hand configuration
features are limited to one per sign, Stokoe allows multiple movements,
which may be simultaneous, sequential, or both. The sign BOARD-OFTRUSTEES, for example, is produced with the same sequence of directional movements as CONGRESS (contact, rightward movement, contact).
The hand initially contacts the far side of the upper chest with a B hand
configuration and the near side of the chest with a T hand configuration.
The movement to the right and the closing movement to a T hand
configuration would need to be represented as two simultaneous
movements between the initial and final contacting movements.
Stokoe also represents some signs without hand configuration changes
with simultaneous movements between other sequential movements,
as in KING, shown in example 2. The sign is produced by first contacting (x) the upper-left side of the chest with the radial side of a K hand
configuration. The hand then moves downward (v) and to the right
(>) simultaneously, followed by another contact (x) at the lower right
side of the torso.
2. [K
x>x
v
about how far to move the hand or what to do with it when it arrives at its destination.
Four sig symbols (s, b, w, n) involve wrist movement rather than
movement along a path. The first two rotate the wrist in one direction or the other, the third represents a continuous oscillating rotation
of the wrist, and the fourth causes the wrist to flex.
The three sig symbols, ] (opening action), # (closing action), and
e (wiggling action) produce changes of hand configuration. In
Stokoes system each of the three signs in example 3 is considered to
be a sign produced with a single A handshape. Each is also represented
with an opening movement, which produces the final hand configuration specified in the square brackets after the symbol ] because
each of the signs ends with a different hand configuration. In this way,
then, he uses movement notations to account for sequential details of
hand configuration.
3.
] [5]
A=<
a. GAMBLE
] [4]
Af
b. SEND
] [G]
uAf
c. UNDERSTAND
sequences of movements, of locations, of handshapes, and of orientations (though orientation was not seen as a fundamental aspect of sign
structure). This makes it possible to claim that signs with more than
one phonetic handshape, orientation, or location have only a single
handshape, orientation, or location because all such sequences are characterized using special movement symbols. In this way Stokoe (and by
extension, those who have adopted the central assumptions of his system) perpetuate the notion that all signs have a single handshape, location, and orientation. The outcome of this practice is that sequences
of handshape, location, and orientation become covert because they
are always interpreted as movements. We demonstrate later that this
practice is largely responsible for the mistaken notion that signs are not
organized as sequences of phonetic or phonological segments.
Problems with the Cheremic System
Figure 2. Minimal pairs of words are drawn from sets of words that have distinct
meanings and are composed of identical sequences of segments except in one position.
The notion that the contrastive features that constitute these signs
are made up in the same way as are vowels and consonants in spoken
languages is somehow correct. However, the idea that two signs that
differ by a single chereme constitute a minimal pair is not. In parallel
to the ISL examples MOTHER and NOON, cited earlier, the ASL signs
APPLE and ONION are frequently cited as an example of a minimal pair.
Both signs are made with an X handshape and a twisting movement,
but APPLE is produced at the cheek, while ONION is produced beside
the eye. In this approach, the oppositional potential of simultaneous
features is taken to be equivalent to the contrastive potential of sequential phones. Figure 3 demonstrates that the two types of minimal
pairs are not equivalent. In vocally produced languages (figure 3a) the
contrastive elements are segments, while in Stokoes representations
the contrastive elements are the descriptive features of a single simultaneous bundle, which is structurally equivalent to a single segment in
spoken languages.
Figure 3 illustrates the schematic geometric differences between
minimal pairs in speech and a putative minimal pair of signs. Minimal
pairs in speech are discovered when sequences of phones differ by a
eye
X
handshape
jaw
twist
single sequential element, as in figure 3a. Thus, [pat] pot and [tat] tot
constitute minimal pairs in English. In figure 3b, where the concept
is applied to APPLE and ONION, the concept of sequentiality is omitted because of the assumption of simultaneity in Stokoes system of
representation. The analysis simply applies the concept of substitutability to the three simultaneous cheremes. Since the two signs differ only in location (cheek and beside the eye), the two signs are
claimed to constitute a minimal pair. The problem here is that substitutability (Saussures paradigmatic opposition) is not sufficient to demonstrate contrast (Saussures syntagmatic contrast).
It is even more problematic when one attempts to apply the concept of minimal pairs to a sign such as BRAVE, for which the Stokoe
notation is shown in example 4. Because the concept of sequence of
segments is integral to the definition of minimal pairs in speech and
because that concept is ignored in applying it to signs, an analyst
would have difficulty in applying the concept to BRAVE.
4.
BRAVE:
[ j5j5xf#
In BRAVE the tab, dez, and sig symbols[ ] (chest), j5j5 (two
5-hand configurations with forearm involvement [ j]), and X (contact)do indeed represent simultaneous articulation because the sign
begins with two 5-hand configurations in fingertip contact with the
torso. Subsequent to this simultaneous beginning, however, a second
simultaneous event involves the original handshapes, the torso, and
two different and subsequent movements: ^ (movement away from)
and # (closing action). Given the assumptions of the system, it is inescapable that Stokoes own representation of BRAVE involves a sequence of two simultaneous articulatory events. As to how the
concept of minimal pairs might apply to BRAVE and a hypothetical sign
differing from BRAVE only in hand configuration, the most straightforward answer is that the concept is impossible to apply. For example,
substituting some other hand configuration for the 5-hand configuration in BRAVE would mean that the new hand configuration occurred
in two simultaneous bundles: first with [ ] and X, and next with [ ], f,
and #. Recall that minimal pairs in speech can be identified only when
If one takes their claim about duality of patterning to mean simply that
all languages combine meaningless elements to form morphemes and
words and that meaningful words combine to form an unlimited
number of possible utterances, then their claim is valid because both
types of languages certainly have that. Once again, however, this requires changing the definition of duality of patterning to eliminate the
notion of syntagmatic contrast at the phonological level. That is, at the
phonological level, the definition of duality of patterning involves sequences of meaningless segments that compose morphemes or words.
The Stokoe model as described earlier has no segments and therefore
no sequences of segments. If, on the other hand, signs can be shown
to consist of sequences of segments, then the concept of minimal pairs
could be applied without the need to redefine the concept. Such a
demonstration would also propose that the signs described earlier as
minimal pairs would not meet the generally applied definition.
Problems with the Putative Simultaneity in Stokoes System
Recall Stokoes proposal that a simultaneous bundle of cheremes attaches to meaning as a single unit. Stokoe represents some signs as
consisting of three simultaneous cheremes. This fact is represented in
figure 4 as vertically arranged groups of features. This is in line with
current practice for the representation of simultaneous phenomena
within a linguistic event. For example, THINK (figure 4a) is represented
with the following three simultaneous cheremes: forehead, G hand-
shape, and contacting movement. GUESS (figure 4b) illustrates another type of simultaneity. It is represented by forehead, C handshape,
and two simultaneous movements: a leftward movement and a simultaneous closing movement. Thus, even though the sign utilizes
two movements, all of its aspects are nevertheless conceived of as
simultaneous.
The concept of a simultaneous bundle of symbols is no longer possible when we consider signs such as CHICAGO (figure 4c), which again
has two movements (rightward and downward), although they are sequential rather than simultaneous. In producing CHICAGO the hand
first moves rightward and then downward. The existence of signs such
as this one argues against the core notion of the Stokoe system, which
claims that a signs features constitute a single simultaneous bundle because the sign actually contains sequences of such features.
If CHICAGO were a unique exception to the notion of representing signs with a simultaneous bundle of symbols, it might be possible
to preserve the fundamentals of a simultaneous system and treat it as
a sign requiring an exceptional sequential representation. But CHICAGO
is not exceptional. Rather, it is representative of a large class of signs
produced with two sequential movements. In addition, Stokoe represents signs such as CONGRESS (figure 4d) with a sequence of three sequential movements: contact, rightward movement, and contact.
Other signs, such as BRAVE (figure 4e), are also represented with three
movement symbols: contact, outward movement, and closing movement. Unlike CONGRESS, however, contact occurs first and is followed
by two simultaneous movements: movement away from the body and
a
THINK
Location
forehead
Handshape
Movement
contact
b
GUESS
forehead
C
left
close
c
CHICAGO
d
CONGRESS
space
space
right down
e
BRAVE
trunk
5
contact away
close
ciated with the one-handed GOOD, and the orientation of the S handshape of the right hand is different from its orientation when ENOUGH
is produced outside the compound.
Because Stokoes system does not provide a way to explain these
differences through changes in the structure of one or both signs,
Klima and Bellugi are forced to talk impressionistically about the re-
GOOD and ENOUGH and the compound sign GOOD^ENOUGH (figure 9.9, from
Klima and Bellugi [1979, 212]).
Figure 5.
Reprinted by permission of the publisher from THE SIGNS OF LANGUAGE by Edward Klima
and Ursula Bellugi, p. 212, Cambridge, Mass.: Harvard University Press, Copyright 1979 by
the President and fellows of Harvard College.
duced and weakened movement of the initial sign (1979, 216) and
about a smoothing of transition between signs (1979, 218). They
also describe the rhythm of the compound as being different from that
of a phrase. In compounds, the signs seemed closer together (1979,
211) than the same two signs in a phrase. They describe the initial sign
as losing stress and repetition, mentioning that one-handed signs tend
to reduce to a single brief contact or stopa duration of a few millisecondsas if representing just the onset of the sign (1979, 216).
The problem posed by this analysis is that, in Stokoes conception
of signs, no part of the cheremic structure is identifiable as the onset
of the sign. All of the aspects are viewed as being simultaneous, so it
is impossible to talk about a beginning or an ending in cheremic terms.
As a result, Klima and Bellugis description of compound formation
relies on impressionistic terms like stress, reduced, and weakened, as well as the notion of seeming closer together. Thus,
although the cheremic system of notation seemed adequate for representing signs in a dictionary, its lack of utility in solving phonological or morphological problems demonstrates that the simultaneous
cheremes model of sign structure has significant problems.10
Liddell (1982, 1984) provides the first arguments that simultaneous feature bundles compose phonological segments and that sequences of
such segments compose signs. We summarize those arguments in the
following sections.
Measuring Movements and Holds. Liddell uses timing data to demonstrate that ordinary signs are divisible into sequences of what he terms
movements (M) (i.e., periods of time when the hand is moving) and
holds (H) (i.e., periods of time when the hand is not moving). He
measured M and H periods in signs both in narratives and when
signed individually from a list. He observed that in both conditions
the hand spends as much time not moving as it does moving and that
in both conditions a consistent pattern of M and H sequences
emerged. Liddell argues that these timing data demonstrate that
whether produced quickly as part of a narrative or more slowly as
signs on a listthe signs have a fundamental and consistent sequential structure.
Noncontacting Contact Signs
Many signs tend to be produced with contact between the active hand
and the body when they are articulated in isolation or in careful discourse. The same signs, when produced in more rapid discourse, often approach the place at which they normally make contact but do
not actually contact. Liddell called these noncontacting contact signs,
the behavior of which was a mystery in the cheremic system. For example, Stokoe represented KNOW as uBtX (forehead location, B hand
configuration facing the signer, and a contacting movement). In producing the contacting form of this sign, the hand moves upward
toward the forehead. As it approaches the forehead it takes on the flat
B hand configuration with the four fingers bent over at the first joint
and the tips of the fingers oriented toward the forehead. The hand
continues its motion toward the forehead until the fingertips make
contact. This contact is then maintained for a noticeable length of
time (i.e., in Liddells terms, the sign ends with a hold). Representing
the sign in Stokoes system with the movement chereme X (contact)
implies that it is the contact with the forehead that is significant, not
the movement to get the hand into contact with the forehead. In addition, the notation does not carry any implication that maintenance
of the contact is a part of the signs structure. Stokoe observes that
KNOW can also be produced without contacting the forehead (1978,
29), though it still moves toward the forehead. He describes this form
of the sign as being produced in the space in front of the signer with
an upward movement (Bt^).
This constitutes a puzzle because, according to Stokoe, in producing the contacting form of KNOW, the hand moves toward the forehead for the purpose of making contact. If the contact chereme is
eliminated, there appears to be no reason for the hand to move at all
since it no longer has an instruction to contact anything. Stokoe replaces X (contactual action, touch) with ^ (upward movement) and replaces the location u (forehead) with (space in front of signer). Now
the sign is described as merely moving upward in space with no target other than a general upward movement and constitutes the claim
that the noncontacting form of KNOW has an entirely different structure, one that is not associated with the forehead. This does not reflect the actual behavior of the hand in the noncontacting form of
KNOW. Simply moving the hand vertically upward will not produce
a sign recognizable as KNOW. In actuality, the hand movement is still
directed toward a specific goalthe forehead. Even though the fingertips do not make contact with the forehead, the hand nevertheless
moves to a position near it and comes to a stop, maintaining this position for a noticeable length of time. Thus, the movement toward the
forehead and the following hold are properties common to both the
contacting and the noncontacting forms of KNOW. Drawing on this
observation, Liddell argues that both forms have the structure MH,
and in both cases the movement is toward the forehead. The difference between the two is where the hand is when it stops its move-
ment toward the forehead. In one case the hand is in contact with the
forehead, and in the other it is not. Viewed in these terms, the noncontacting form does not lose the forehead as the place that guides the
movement of the hand, and it does not lose its movement toward the
forehead. Rather, the difference between the two turns out to be just
what the prose descriptions of the two forms of KNOW states. In one
case the hand moves toward the forehead and makes contact, and in
the other it moves toward the forehead and stops without making
contact. Such a view is impossible in a purely simultaneous model of
sign structure. Liddell argues that if the two versions of the sign in fact
share the first two aspects of activity (where the hand starts and its
movement toward the forehead) and only the last part is different
(whether the hand makes contact or not at the forehead), then it must
have significant sequential ordering.
Compounds
Drawing on these same observations about compounds, Liddell demonstrates that the first sign of compounds such as BELIEVE, historically a
compound formed from THINK (uGtx) and MARRY (CCg), is reduced
to one H segment. Specifically, he proposes that if the first sign has a
contacting H segment, that segment will be retained, and the remainder of the sign will not be produced. Liddell proposes that these facts
about THINK in a compound cannot be explained using a model in
which all aspects of sign structure are represented as a simultaneous
bundle. The first element of THINK^MARRY has a structure different
from that of THINK produced alone because the initial movement
toward the forehead is not present in the compound form. There is
no initial movement toward the forehead when the structure of THINK
is understood to be uGtx. This is because the contacting movement x
is the only movement in the sign. Moreover, there is no final part to
be retained in the compound. Clearly, Liddell argues, in order to keep
only one sequential part of a sign, the sign must have sequential parts.
Sequences of Segments
Liddell proposes that signs composed of a single circling activity would
consist of a single M segment. He concludes that other signs, such as
STARE and BE-SITTING, which are produced without any hand move-
Nonmanual Evidence
Liddell (1984) observed that the production of some signs also involves
nonmanual articulatory sequences that are very tightly coordinated
with manual articulatory sequences. GIVE-IN, for example, begins with
a B hand held with the palm in contact with the chest and the fingertips pointing across the body. The hand then moves slightly down and
away from the chest and comes to a stop. By the time the hand reaches
this position, the wrist has rotated enough so that, from its lower position, the palm it is still facing the chest. This makes GIVE-IN a sign
with the structure HMH. In addition to these manual productions, the
nonmanual aspects of this sign are also sequential. The sign begins
with the lips pressed tightly together, followed by a rapid parting of
the lips as the hand moves away from the chest. The mouth is then
held in an open position during the final H. Regardless of the duration of the initial hold in GIVE-IN, the lip press is maintained throughout that hold. Similarly, regardless of the duration of the final hold,
the mouth will be held open throughout that interval of time. During the movement between the holds, a transition from lip press to
mouth open takes place. Although from this perspective lip press
can be identified as a feature of the production of GIVE-IN, a Stokoe
representation would provide no means of making this observation.
This is because lip press is not a feature of the sign as a whole but
rather is of only the initial hold. Similarly, mouth open is a feature
of the final hold only. The initial and final Hs serve an important function in that the manual and nonmanual features are aligned during
those segments.
Indicating Verbs
Liddell also observes that indicating verbs (Liddell 2003) must be specified for more than one location.13 For example, if GIVExy begins its
movement toward x (the addressee) and ends its movement directed
toward y (a woman standing next to the addressee), the sign is translated as you give (it) to her. The initial location of the sign identifies the addressee as the giver, and the final location identifies the
woman as the recipient. There is no way to describe the movement
of this sign by identifying only a single location for it. Suppose that the
addressee were the location of the sign. What possible movement
could ensure that the sign would end by pointing at the woman next
to the addressee? Similarly, if the woman was understood to be the recipient of the giving, and a location in her direction was specified as
the single location of the sign, what possible way would there be to
ensure that the hand begins directed toward the addressee? This leads
Liddell to conclude that such verbs entail a distinctive and unmistakably critical sequence of places, with movements between them.
Nonpath Movements and Local Movements
Stokoe proposed two types of movement: those that cause the hand
to move along a path and those that do not. Stokoes movement f, for
example, moves the hand on a path in a direction away from the
signer, while the movement ] simply causes the handshape to close.
In WHITE (example 5) the 5 handshape begins with the thumb and fingertips in contact with chest. As the hand moves away from the chest
the hand closes to an O handshape. In example 6 the two-handed sign
BRAVE begins with the thumb and fingertips of a 5 hand in contact
with the fronts of their respective shoulders. The hands then move
outward and end their movements in what the Stokoe system identifies as a (cheremic) A handshape, though the hands are actually in an
S handshape.
5.
] [O]
[5tf
WHITE
6.
] [A]
[55t
BRAVE
These two signs illustrate that the closing action symbol alone is
insufficient as a mechanism for producing the correct final handshape.
That is to say, there is more than one way to close the 5 hand. Stokoe
added the second handshape inside square brackets to represent this
fact. However, specifying a second hand is inconsistent with a system
that claims that signs have only one handshape and with calling it a
closing movement rather than a second handshape. Thus, Liddell
argues that these signs provide evidence of the need to specify hand
configuration sequences, a fact easily accommodated in a system that
admits sequences of segments.
A third type of movementlocal movements (often called hand-internal movements)also argues for the segmentation of signs. These
involve a type of repeated, secondary activity of the hands that may
co-occur with path movements. One such local movement is wiggling,
in which the fingers repeatedly wiggle due to muscle contractions in
the joint closest to the palm. Consider VERY-LONG-AGO. In Liddells
treatment VERY-LONG-AGO is an MH sign with the movement beginning well ahead of the shoulder with a 5-hand configuration with the
palm facing across the body. The fingers wiggle during this movement. When the hand approaches the shoulder, it stops its movement,
and the fingers also stop wiggling. The nonwiggling hand briefly
maintains a nonmoving configuration (H). The problem for a simultaneous representation is that wiggling cannot be assigned as a property of the sign. Rather, wiggling is a property only of the movement
toward the shoulder since there is no wiggling during the H. However, if the M and H have their own properties, then it is straightforward to assign wiggling as a property of one but not the other. As long
as each segment has its own features, it is straightforward to restrict
wiggling to the segment in which it actually occurs.
The Revised MH Model
The initial MH model (Liddell 1984) was considerably expanded and
revised in Liddell and Johnson (1989). It maintained Ms and Hs as the
two basic types of segments that compose signs but separated them as
timing units from features that configure not only the shape, orientation, and location of the hand but nonmanual signals as well. This
The goal of minimal specification of features in an emic-level system of representation is also a characteristic of the syllabic model proposed in Wilbur (1993):
Predictable phonetic detail and redundant feature specifications
should not be included in a truly phonemic-level representation.
(Wilbur 1993, 137)
The appropriate approach, then, is to posit the least amount of internal structure in lexical entries until such time that it is conclusively
shown that additional structure must be added to them. (Wilbur
1993, 149)
Because we are seeking to develop a system that will enable a description of and a solution to ordinary morphological or phonological
problems and since none of these proposals deal with phonetic-level
representations, we do not review them here. However, there remains
one proposal regarding segmentation to review here. It suggests that
movements need not be transcribed in the representation of ASL signs
and so is directly pertinent to our analysis.
A Movement-Less Model
C
delayed release
falling tone
hi
lo
Phonetic Transcriptions
Vocally Produced Languages
At a minimum the phonetic system should capture the sound distinctions needed to distinguish one word from another. It is possible that
the phonetic transcriptions of multiple instances of the word play uttered by several speakers of different genders and ages could all be represented as [phleI]. This does not mean that all of the speakers
pronounced the word in exactly the same way. Rather, it means that
the analyst recognized the same sequence of phonetic units each time
one of the speakers produced the word. The three units are the initial aspirated consonant [ph], the consonant [l], and the diphthong [eI].
The actual productions of these instances of play could have differed
in the intonation of the voice, the pitch of the sound, the length of
the vowel, the degree of aspiration of the initial consonant, and so on.
That is, even though a womans or a childs voice is typically at a
higher pitch than a mans, the phonetic representation [phleI] does not
attempt to represent that fact. What the analyst attempts to do is identify linguistically significant phonetic units within the sound stream,
while ignoring other aspects of the sound stream that are the result of
linguistically insignificant, idiosyncratic phenomena.
Transcribing the word play phonetically requires making a decision
about whether the initial sound falls into the category of sounds represented phonetically as [ph] or the category of sounds represented
phonetically as [p]. A typical American speakers production of play
would be represented phonetically as [phleI]. Anyone who has listened
to a large number of speakers knows that there are many individual
differences in pronunciation. Yet, those differences are generally not
great enough to change phonetic categories. It is possible, however,
to find a speaker who pronounces play in a way that requires a different phonetic representation. We have observed a television personality, for example, whose production of play would have to be
represented phonetically as [pleI]. That is, he does not aspirate the initial p to a degree that warrants its categorization as aspirated.
We emphasize here that phonetic transcription symbols are symbols for abstract categories of sounds, but the categories are not as abstract as phonemic categories. The primary difference between
phonetic and phonemic categories is that phonemic categories are abstractions across (abstract) phonetic categories. That is, for English,
the abstract phonetic categories [ph] and [p] both fall within the still
more abstract phonemic category /p/. For other languages these two
abstract phonetic categories are potentially capable of distinguishing
one word from another and could therefore fall into different phonemic categories.
Signed Languages
Notes
1. Stokoe, Casterline, and Croneberg (1965, xxix) describe the motivation for the term chereme as follows: (CARE-eem, the first syllable from a
Homeric Greek word meaning handy). The origin of the term is described
differently in Maher (1996, 67), who states that it comes from the Greek
word chirologiathe eighteenth- and nineteenth-century term for fingerspelling or signing.
2. By convention, Stokoe omits the symbol for the location in front of
the body (). His dictionary represents CHICAGO as C >V. We have added
the tab symbol here for clarity.
3. Although hand orientation is critical and although it changes during
the course of the sign, it is not described as an independent aspect in Stokoes
notation. If it is represented, it tends to be shown as a part of the handshape
(dez) or as a specific movement, such as p (supinating movement).
4. The native ASL two-handed sign is opposed to the one-handed sign
RUSSIA, borrowed from Russian Sign Language.
5. The use of such categories is often defended by saying that Stokoe was
not trying to do phoneticsthat he was interested only in emic representationand so such details are not relevant. It is not the case, however, that
representing phonemes of a spoken language makes the phonetics of the spoken language irrelevant. In fact, phonemes are used together with a set of
References
Baker, C., and C. Padden. 1978. American Sign Language: A Look at Its Story,
Structure, and Community. Silver Spring, Md.: TJ Publishers.
Baker, C., and D. Cokely. 1980. American Sign Language: A Teachers Resource
Text on Grammar and Culture. Washington, D.C.: Clerc Books.
Battison, R. 1974. Phonological Deletion in American Sign Language. Sign
Language Studies 5: 119.
. 1978. Lexical Borrowing in American Sign Language. Silver Spring,
Md.: Linstok.
Bloomfield, L. 1933. Language. New York: Holt.
Brentari, D. 1998. A Prosodic Model of Sign Language Phonology. Cambridge,
Mass.: MIT Press.
Fridman-Mintz, B. 2006. Tense and Aspect Inflections in Mexican Sign Language Verbs. PhD diss., Georgetown University.
Friedman, L. A. 1975. Space, Time, and Person Reference in American Sign
Language. Language 51: 94061.
Frishberg, N. 1975. Arbitrariness and Iconicity. Language 51: 696719.
Hayes, B. 1993. Against Movement: Comments on Liddells Article. In Current Issues in ASL Phonology, ed. G. R. Coulter. San Diego: Academic Press.
Hockett, C. 1960. A Course in Modern Linguistics. New York: Macmillan.
Johnson, R. E. 1996. The Continuative Aspect Inflection in ASL: Evidence
for Variable Phonological Feature Values. Plenary paper presented to the
Theoretical Issues in Sign Language Linguistics Conference, Montreal,
Quebec, Canada.
Klima, E., and U. Bellugi. 1979. The Signs of Language. Cambridge, Mass.:
Harvard University Press.
Liddell, S. K. 1977. An Investigation into the Syntactic Structure of American Sign Language. PhD diss., University of CaliforniaSan Diego.
. 1982. THINK and BELIEVE: Sequentiality in American Sign Language
Signs. Linguistic Society of America, University of Maryland, Summer
Session.
. 1984. THINK and BELIEVE: Sequentiality in American Sign Language.
Language 60: 37299.
. 1985. Compound Formation Rules in American Sign Language. In
SLR 83, Proceedings of the III. International Symposium on Sign Language Research . Silver Spring, Md.: Linstok Press.
. 1988. Structures for Representing Handshape and Local Movement
at the Phonemic Level. Unpublished manuscript, Gallaudet University,
Washington, D.C.
. 1996. Numeral Incorporating Roots and Non-incorporating Prefixes in American Sign Language. Sign Language Studies 92: 20126.