A thesis submitted in partial satisfaction of the requirements for the degree Master of Science in Linguistics by Kelly M. Stack
For such a little thing, this thesis has taken a great deal of my time and attention. I'd like to express my appreciation to the members of my committee, Susie Curtiss, Pat Keating and Bruce Hayes, for their valuable time and attention to my efforts here. I'd especially like to thank Bruce Hayes for the uncanny timing of his praise and criticism, for his patience, for the gusto with which he approaches the analysis of ASL, and for his generosity in sharing his keen insights.
Finally, a bowling award to my husband, Arthur Ginsberg, whose encouragement and support has been essential and then some.
This paper presents an analysis of the segmental and syllable structure of American Sign Language in which only the static targets of movement are represented at the phonological level. Toward this end, secondary change and primary change are analyzed as different instantiations of a unified phenomenon. Phonotactic constraints on syllable structure are examined to discover an underlying syllabic template. Concatenative morphological processes are shown to reveal the presence of a class of signs that is sub- minimally specified; these signs cannot surface in isolation without undergoing a rule of epenthesis. The framework presented here is shown to account for observations made by other researchers, in a simple and natural way that may be considered preferable because it is akin to analyses of spoken languages.
American Sign Language (ASL) is the native or first language of Deaf(1) people in the continental United States (and parts of Canada). It is a natural language, apparently the product of a creolization process that took place in this country around 1816, with Old French Sign Language as the superstratum and indigenous colonial signed languages as the substratum (Woodward, 1978).
Serious linguistic investigation of ASL is relatively new, having begun in 1965 with the publication of A Dictionary of American Sign Language on Linguistic Principles (Stokoe, Casterline and Croneberg).
In this paper I propose an underlying syllabic structure for monomorphemic signs, including horizontal and vertical dimensions of the syllable geometry. I begin in Section 1 by describing three phonological formatives of signs: location, hand configuration, and palm orientation, proposing that in a sequential framework movement need not be analyzed as a primitive. In Section 2, I define primary change and argue that a putative fourth formative, secondary change, is actually a variation of change that occurs on the first three tiers. Phonotactic laws governing minimum and maximum syllable structure are described in Section 3, as well as a phonological rule which ensures syllabic well-formedness. In Section 4, formal rules and representations are discussed, including a syllabic template. Finally, I compare my analysis with those presented in Perlmutter 1987) and Liddell and Johnson (1986).
In the interests of making this paper accessible to linguists who have not yet familiarized themselves with research on ASL, I have attempted to avoid assuming prior knowledge about ASL, and to explain conventions in the field. One such convention involves the labeling of the hands, each of which is a unique articulator. For example, in signs involving both hands, one hand has relative freedom regarding hand configuration, location and palm orientation. The other hand is far more restricted. The former hand (usually the right hand for right-handed signers) is often called the "strong" hand; the other hand is called the "weak" hand. I use these terms in this paper.
In the further interests of accessibility, and because it is not crucial to my claims, I have remained neutral regarding various proposals of ASL feature systems. I have attempted to present articulatorily descriptive phonological representations, rather than strictly binary feature lists. The reader is warned, therefore, that my phonological representations do not reflect the complexity of the feature system of ASL.
Finally, glossing of signs, though standardized to some degree, is subject to variation. The appendix contains pictures of most of the signs discussed in this paper, alphabetized by the English glosses I have assigned. Unfortunately, however, for those who do not know the language, a drawing of a sign can be just as mystifying as a written description. The appendix will undoubtedly be of greater use to people fluent in ASL than to non-signers, but it is offered to both in interests of clarity.
In 1965, Stokoe, Casterline and Croneberg named three formatives of ASL signs: location, movement and hand configuration. A fourth formative, palm orientation, was later added by Battison (1978). Movement was necessarily a formative in Stokoe's analysis since his model was a simultaneous one -- each sign was a simultaneous bundle of specifications from each of the formatives.
Most current research has abandoned the simultaneous model. Working within a sequential model, Sandler (1986) pointed out that path movement in such a model is actually transition from one location to another. This point of view would seem to dismiss movement as an ASL primitive. Yet Sandler (1987), Liddell and Johnson (1986) and Perlmutter (1987) all view movement as a formative. I believe the features they attach to movement autosegments (contour(2) , plane and quality) can be accounted for in my framework without appeal to movement as a formative.
Thus there are three formatives of monomorphemic signs: location, hand configuration and palm orientation.
Location (LOC) describes the placement of the hand(s) in space or on the body (including placement of the strong hand with regard to the weak, or vice versa). In a complete description, LOC includes not only the location of the hand, but the part of the hand that comes in contact with the location. For example, in THINK, the location of the hand during part of the sign is the ipsilateral side of the forehead, and it is the tip of the index finger that comes in contact with that location.
Another important factor about location is the manner in which the hand comes into contact. Contact may be single-point, as in MONEY, or brushing, as in NEW, two signs which contrast in this regard. In single-point contact, the hand makes contact with a single point at the location in question. Crucially, the location in question may or may not be on the body -- that is, contact may be made with a point in space. The hand may or may not then move to another location, but if it does move, the next location is unpredictable.
In brushing contact, the hand contacts the location with a brushing movement (again, the location may or may not be on the body or in space) and moves on to a final location that is predictable given the trajectory from and point of contact of the previous location. For example, in NEW, the initial location of the strong hand is above and to the ipsilateral side of the heel of the weak hand. The second location is the heel of the weak hand, where brushing contact is made. The final location is above and to the contralateral side of the heel of the weak hand, and is predictable given the previous location specifications and the fact that brushing contact is made.
In this paper, these details will be overlooked in the interests of clarity. Broad, simple terms will be used to describe the location of the hand.
Palm orientation (PO) describes the orientation of the hand in two planes. A detailed phonetic description, as shown in Liddell and Johnson (1985), should include two referents and the part of the hand aligned with each. For example, in the citation form of STARE, the tips of the fingers point toward a spot slightly to the right of and medially distant from the trunk (this is what Liddell and Johnson call the "facing") and the base of the hand is in the horizontal plane ("orientation," according to Liddell and Johnson). In broad terms, in this paper, palm orientation is usually described as being toward the body, away from the body, up, down, etc.
Hand configuration (HC) describes the position of the fingers and thumb. There are 26 hand configurations corresponding to the letters of the English alphabet, but not all of these are used in ASL signs. There are several additional hand configurations corresponding to certain numbers, some of which are used in signs, and there are some hand configurations which appear in signs but do not happen to be symbols for letters or numbers. Several researchers have proposed feature systems for hand configuration (Liddell and Johnson, 1985 and Sandler, 1987) including description of forearm involvement, general handshape, finger extension, laxness, and details of thumb configuration. Conventional notation (which I will generally follow) usually makes use of the English written alphabet and Arabic numerals along with some diacritics to refer to various hand configurations.
Secondary change (more commonly called "secondary movement" or "local movement") usually refers to repeated change in palm orientation, hand configuration or location. This repeated change is uncountable or of varying count. Secondary change can occur during the transition from one LOC to another LOC or while the hand is otherwise motionless.
Liddell and Johnson (1985) defined the following types of "local movement":
Perlmutter (1987) looked at several types of secondary change, including one not noted by Liddell and Johnson (1985):
Other secondary changes I have observed include:
Examining each secondary change in turn, I will show that secondary changes are really repeated (and usually slightly reduced) primary changes. To do this, I must show that the list of secondary changes can be categorized in terms of location, palm orientation and hand configuration. Then I must show that these changes exist in other signs as primary (non-iterated) change.
Before proceeding with the discussion of secondary change, let me clarify the definition of primary change. Primary change is a single change from one location to another, one hand configuration to another, or one palm orientation to another.
For example, in the underlying representation of THINK, there is a primary change from an initial LOC which is proximal to the ipsilateral side of the forehead to a final LOC which is the ipsilateral side of the forehead itself. In the underlying representation of NAME, there is a primary change from an initial LOC which is proximal to the weak hand and a final LOC which is the weak hand itself. NAME (unlike THINK) is a lexically reduplicated sign so in the surface representation, there are two identical primary changes. In fact, lexical reduplication in ASL (Perlmutter, 1986) is a syllable copying operation. The two primary changes in a lexically reduplicated sign are clearly different from the uncountable tautosyllabic changes in a sign involving secondary change.
All secondary changes may be grouped into three categories, corresponding to the three sign formatives discussed in Section 1:
Given, as shown above, that secondary change occurs within the formatives characterizing primary change, we must ask if each secondary (iterated) change has a corresponding primary (non-iterated) change.
Twisting, as in TREE, is the iterative change in palm orientation from prone to supine (or from supine to prone). There are signs which involve only two palm orientation specifications and thus a single (primary) change from one to the other: FINISH (supine to prone) and LIGHT (prone to supine).
Nodding, as in OH-I-SEE, is likewise the iterative change in palm orientation from flexed wrist to extended wrist. This change occurs once in signs specified for two palm orientations such as BEAT (extended to flexed) and SELL (flexed to extended).
Pivoting, as in WHERE, is the iterative change in orientation from ulnar oblique to radial oblique. This change occurs once in signs specified for two orientations such as ALLOW (radial to ulnar) and LECTURE (ulnar to radial).
Figure (2) below illustrates the comparison:
SECONDARY PRIMARY 1st PO 2nd PO CHANGE CHANGE EXAMPLE EXAMPLE Twisting: TREE FINISH toward body away Nodding: OH-I-SEE BEAT away from body down Pivoting: WHERE ALLOW tips away from body tips up
Swinging involves repeated back and forth change between two locations (SING) or vacillation along the trajectory from one location to another (ART). Primary change from one location to another occurs in a majority of ASL signs.
Circling seems to be of two major types: rotary movement with the wrist or elbow as the hub, and description of a circle in a plane. We can find examples of non-iterative circling to compare with secondary circling for both of these. Secondary rotary circling (TRAVEL) can be compared with FACE. Secondary planar circling (COFFEE) compares with non-iterative planar circling in HOUR and AMERICA. Circling may be describable as a third manner of LOC contact, in addition to single-point contact and brushing (see Section 1.1). Like swinging, it may occur in a single area or along a trajectory from one location to another.
Figure (3) illustrates the comparison of location changes:
SECONDARY PRIMARY 1st PO 2nd PO CHANGE CHANGE EXAMPLE EXAMPLE Swinging: SING HIRE distal from body medial from body Circling: COFFEE AMERICA point in horiz. plane circle in plane
Secondary change involving hand configuration is repeated change from one hand configuration to another. For each secondary change involving hand configuration there is a sign with the corresponding primary change. The opposite, however, is not the case: not every primary hand configuration change can be iterated as a secondary movement.
Figure (4) shows some example comparisons between primary and secondary hand configuration change.
SECONDARY PRIMARY 1st PO 2nd PO CHANGE CHANGE EXAMPLE EXAMPLE Hooking: ANALYZE DOUBT V", thumb opp(3) & closed V"" INSECT OVERHEAR V", thumb opp & open V"" DREAM ASK 1", thumb opp & closed 1"" RUN FAST 1", thumb unopp & open 1"" Flattening: GOSSIP BIRD 1, thumb opp & open Flat 1, closed PIG ALMOST B, thumb unopp A STICKY CHOOSE 8, thumb opp & open Flat 8, closed Releasing: SHIRK AWFUL 8, thumb opp & closed 8, open RESPONSIBILITY Rubbing: DIRT MELT closed 6, thumb opp & closed A, unopp Squeezing: ORANGE CHERISH 5", thumb opp & open A, closed
Wiggling is difficult to describe in terms of two distinct end points, but it does occur in a countable non-iterative change in the sign BEAUTIFUL(4). BEAUTIFUL contains a single, non-repeated, five-finger wiggle: there is sequential bending at the first joint of each finger, beginning with the little finger. In signs like COLOR, there is similar sequential bending, but it is repeated an uncountable number of times. Thus, even wiggling occurs as primary change as well as secondary change.
Recall that primary change is described as single, non-iterated change from one LOC, HC or PO specification to another. I have shown that secondary change occurs only within the three formatives that characterize primary change, and that for each secondary change there is a corresponding primary change. I have thus argued that secondary change is iterated primary change from one LOC, HC or PO specification to another, and is not a fourth formative. Given the three formatives and two types of change, we will now consider how these interact to form signs.
Like any language, ASL has a number of phonotactic constraints, including, for example, limits on sequences of hand configurations and rules governing weak hand activity in two-handed signs. In this section, constraints that confirm the claim that secondary change should be analyzed as similar to primary change, and syllable-level constraints will be examined.
There are no uninflected signs consisting of exactly one hand configuration, one location, and one palm orientation.(5) What then is the minimal sign? Two specifications in any of these classifications suffice to form a sign:
Signs specified for two locations only, such as GOOD:
LOC1 LOC2 PO HC
Signs specified for two palm orientations only, such as SMART:
LOC PO1 PO2 HC
Signs specified for two hand configurations only, such as UNDERSTAND:
LOC PO HC1 HC2
Signs specified for any one type of secondary change only, such as:
COFFEE: LOC1 (circle) LOC2 PO HC TREE: LOC PO1 (twist) PO2 HC ORANGE: LOC PO HC1 (squeeze) HC2
The Minimality Condition expresses these facts:
There is a relatively large class of signs that meets the Minimality Condition in a rather minimal way. These signs consist of one HC, one PO, and two LOC specifications. In the signs in this class, the first LOC specification is always proximal to and perpendicular to the surface plane of the second LOC specification. In other words, there is a bare minimum of movement, the first location being just far enough away from the second to tallow noticeable movement. Examples of such signs are THINK, TEACH, NOW, and WANT. I call these Monosegmental Signs.
Included in this class are letters and numbers, which are candidates for underspecification in everything but hand configuration. When letters and numbers are signed in isolation, they are pronounced with a short path movement like those of other Monosegmental Signs.
Using evidence from concatenative morphological processes, I will argue that the first location in the surface form of Monosegmental Signs is not a part of the underlying representation. The fact that the first location appears in surface citation form may instead be viewed as the result of a highly productive phonological process of epenthesis.
When signs in this class undergo concatenative morphological processes, the first location is always deleted; yet this is not the case with other signs undergoing the same processes. This suggests that the first location in a Monosegmental Sign is not actually present prior to the morphology. We will now examine three concatenative processes and compare their effects on Monosegmental and non-Monosegmental Signs.
When two signs are concatenated to form a compound, the first sign in the compound undergoes significant phonological reduction, as described in detail by Liddell and Johnson (1986). For example, when GOOD combines with NIGHT to form GOOD+NIGHT, only the first of the two locations in GOOD is retained:
GOOD NIGHT LOC [chin ] [medial to chest] [above weak hand] [weak hand] PO [to body] [away from body ] HC [^B ] GOOD+NIGHT LOC [chin ] [weak hand ] PO [to body] [away from body ] HC [^B ]
However, when THINK combines with MARRY to become BELIEVE, it is the second of the two locations in THINK that is retained:
THINK MARRY LOC [prox. to forehead] [forehead] [above weak hand] [weak hand] PO [to body ] [away from body ] HC [1 ] [B ] BELIEVE(6) LOC [forehead ] [weak hand] PO [away from body ] HC [1 ] [B ]
Liddell and Johnson (1986) explained this discrepancy by proposing the Contacting Hold Rule, which states that if there are contacting holds in the first sign in a compound, only those segments are retained in the final form. Yet the Contacting Hold Rule does not explain why the second location rather than the first is retained in TODAY (NOW+DAY), since there is no contact at all in the sign NOW.
There is a broader general phenomenon at work in compounding than the contacting hold rule. Signs that lost their first LOC when they are compound-initial all belong to the class of Monosegmental Signs. That is, they are signs that in surface citation form contain only the barest (and predictable) change from one LOC to another LOC, and no other primary or secondary change. These signs consistently lose their initial LOC in other morphological processes as well, as we shall see.
There is a non-productive negation process in ASL which consists of suffixing a PO [away from the body] and a LOC [down and medial to the previous LOC] to a sign. For example, GOOD becomes BAD:
GOOD BAD LOC [chin ] [medial to chest] [chin ] [medial to trunk] PO [to body] [to body] [away from body ] HC [^B ] [^B ]
However, when KNOW becomes DON'T-KNOW, the first LOC in KNOW is absent in DON'T-KNOW:
KNOW DON'T-KNOW LOC [prox. to forehead] [forehead] [forehead] [medial to chin] PO [to body ] [to body ] [away from body] HC [^B ] [^B ]
Again, as in compounding, the signs which lose the first LOC during negation belong to the class of Monosegmental Signs. Other signs retain the first LOC during negation.
The agentive suffix is added to verbs in a productive morphological process to create the obvious meaning. For example, LEARN+AGENTIVE becomes STUDENT:
LEARN AGENTIVE LOC [wk hand ] [above wk hand] [med. to chest] [med. to trunk] PO [to wk hand] [to wk hand ] HC [^5, open ] [^5, closed ] [B ] STUDENT LOC [wk hand ] [above wk hand] [med. to trunk] PO [to wk hand] HC [^5, open ] [^5, closed ] [B ]
Yet when the agentive suffix is added to TEACH, a Monosegmental Sign, the first LOC in TEACH is lost in the resulting TEACHER:
TEACH LOC [prox. to med. to forehead] [med. to forehead] PO [to wk hand ] HC [flat O ] TEACHER LOC [med. to forehead] [med. to trunk] PO [to wk hand ] HC [flat O ] [B ]
TEACH is a Monosegmental Sign. LEARN is not.
If we assume that the first LOC in a Monosegmental Sign is present in the underlying representation, a rule of LOC deletion must operate in each affixational process -- compounding, negation, agentive formation. Moreover, this rule sometimes acts to delete the initial, sometimes the final LOC, and can only be formalized with transformational notation (see Section 5.2).
If we assume, however, that the first LOC is not present in the underlying representation, a more general and simple rule of Location Epenthesis emerges (to be formalized later in Section 4):
Therefore, location epenthesis is available to "rescue" signs that are underlyingly specified for one location, one palm orientation and one hand configuration -- signs that would otherwise violate the Minimality Condition. When morphological processes apply, they apply to the lexical representation, prior to Location Epenthesis.
Having discovered the minimal requirements for a sign to be a sign, it is natural to look for the upper boundary. The investigation now focuses on the syllable -- in particular, the underlying representation of monosyllabic and lexically reduplicated signs (as defined by Perlmutter, 1987). Perlmutter (1987) lists hand configuration changes that cannot occur tautosyllabically. Signs containing such HC changes, therefore, are not monosyllabic. Also excluded from consideration here are inflected signs, compounded signs, and signs that have undergone location epenthesis. This leaves, however, a very large group of signs.
What is the maximal monomorphemic monosyllabic sign?
Signs specified for two locations and two palm orientations, such as INCREASE:
LOC1 LOC2 PO1 PO2 HC
Signs specified for two locations and two hand configurations, such as FORGET:
LOC1 LOC2 PO1 HC HC2
Signs specified for two locations and any one type of secondary movement, such as
DRAW: LOC1 LOC2 (swing) PO HC SENTENCE: LOC1 LOC2 PO1 (twist) PO2 HC DREAM: LOC1 LOC2 PO HC1 (hook) HC2
Signs specified for two palm orientations and two hand configurations:
* LOC PO1 PO2 HC1 HC2
Signs specified for any secondary change and either two palm orientations or two hand configurations:
* LOC1 (2ary) LOC2 PO1 PO2 HC * LOC1 (2ary) LOC2 PO HC1 HC2 * LOC1 LOC2 (LOC 2ary) PO1 PO2 HC * LOC1 LOC2 (LOC 2ary) PO HC1 HC2 * LOC PO1 (2ary) PO2 HC1 HC2 * LOC PO1 PO2 HC1 (2ary) HC2
Recall that secondary change is essentially the same as primary change; it is specified on either the LOC, PO, or HC tier. Then the following generalization emerges:
As the reader has probably already guessed, I will propose an autosegmental analysis of ASL, with three major tiers and a great deal of underspecification. Underspecified slots are filled by spreading.
In standard autosegmental notation, a feature F spreads from one segment X to another segment Y, by the addition of an association line:
F |\ X Y
In the notation I will use, a specification SPEC will spread from one column X to another column Y (in either direction), by the addition of what I will call a temporal domain line:
X Y [SPEC]----
To represent the difference between primary change from one specification to another and secondary change from one specification to another, I will introduce two new temporal domain lines:
I propose the following template in order to formalize the constraints described in the Minimal and Maximal Conditions:
The Underlying Syllabic Template:
A syllable consists of one or two segments, given here as two columns.
The first segment (the left column) in each lexical entry is specified underlyingly for one location, one palm orientation and one hand configuration. The second segment (the right column) may or may not have up to two filled slots. Either one or both of these slots may be left empty.
The top slot in the second segment, if filled, will connect to the first [LOC] by a primary change line. It will therefore have to be a [LOC] specification.
The bottom slot, if filled, may connect to any one of the three specifications in the first column. If the bottom slot is filled with a [LOC], it will connect via a secondary change line to the [LOC] in the first segment. In this case, the top slot may be filled with a [LOC] specification, resulting in a sign such as SNAKE, consisting of two locations related by primary change and a third location related by secondary change. If the bottom slot is filled with a [PO] or [HC], it will connect to its tier-mate in the first segment by either a primary or secondary line, and the top slot may or may not be filled with a second [LOC] related to the first [LOC] via a primary change.
The Minimal and Maximal Conditions emerge from the Syllabic Template in interaction with Location Epenthesis.
If both slots in the second segment remain unfilled, a monosegmental lexical entry will exist. This will be "rescued" by Location Epenthesis if it surfaces without undergoing some morphological process.
If only the top slot in the second segment is filled, a minimal syllable with two locations connected by primary change, one HC and one PO will be formed. There are many signs in ASL that fit this description, including GOOD (see 3.1.(5)).
If only the bottom slot in the second segment is filled, signs containing three minimal types of syllables may be formed. The first type has two PO's and one of everything else (see SMART, 3.1.(6)). The second type has two HC's and one of everything else (see UNDERSTAND, 3.1.(7)). The third type has any type of secondary change and one of everything else (see COFFEE, TREE, ORANGE, 3.1.(8)).
If both the top and bottom slots in the second segment are filled, signs containing three maximal types of syllables may result. The first has two LOC's, two PO's and one HC (see INCREASE, 3.3.(17)). The second has two LOC's and two HC's and one PO (see FORGET, 3.3.(18)). The third has two LOC's and any one type of secondary change and one of everything else (see DRAW, SENTENCE, DREAM, 3.3.(19)).
But syllables with both two PO's and two HC's cannot be formed since there is only one additional slot available for either of these.
And syllables with any secondary change and two palm orientations or two hand configurations, or two secondary changes, cannot be formed for the same reason.
Location epenthesis is a highly-productive phonological process that applies to underlyingly monosegmental signs. It was originally described:
Since my focus has been on underlyingly monosyllabic signs, the above formulation is adequate. But the syllable is the true domain of Location Epenthesis, as will be seen later when it acts on signs that are lexically reduplicated. Thus its formal instantiation below includes syllable association lines, and it is the syllable boundary, not the sign boundary, that is important.
Location Epenthesis: ([S] = syllable boundary) 0 --> [proximal(LOC)] / [S ___ [LOC] ]S
The fact that Location Epenthesis applies to the left side of the syllable brings up a troubling observation about the Syllabic Template. The Template consists of a fully-specified segment followed by a partially-specified segment. Yet when Location Epenthesis applies to Monosegmental Signs, the result is the mirror image of the Syllabic Template. This suggests that although the Template produces correct results for quantity, positioning of segments may be in error.
Location Epenthesis is accounted for more naturally if segment order within the Template is reversed, yet this reversal would result in the necessity for rather messy segmental restructuring rules in concatenative processes. For example, recall the discussion of compounding in Section 220.127.116.11. When GOOD combined with NIGHT to form GOOD+NIGHT, only the first of the two locations in GOOD was retained:
GOOD NIGHT LOC [chin ] [medial to chest] [above weak hand] [weak hand] PO [to body] [away from body ] HC [^B ] GOOD+NIGHT LOC [chin ] [weak hand ] PO [to body] [away from body ] HC [^B ]
If GOOD is represented with an initial partially-specified segment followed by a fully-specified segment, a segment restructuring rule must be proposed at some point in the derivation:
GOOD - underlying representation with mirror image Syllabic Template
LOC [chin] [medial to chest] PO [to body ] HC [^B ]
GOOD after second LOC deletion:
LOC [chin] PO [to body] HC [^B ]
GOOD after concatenation to underlying representation of NIGHT:
GOOD + NIGHT LOC [chin] [weak hand ] PO [to body] [away fr. body] HC [^B ] [^B ]
Some kind of segment restructuring must apply to GOOD in this case. If GOOD is not restructured into a single segment, it would appear to become a geminate, assuming that spreading applies to fill empty slots and each segmental column represents some kind of timing unit. This does not fit the facts; the GOOD in GOOD+NIGHT is not held for an extra beat, which would be the result of gemination. Similar restructuring rules would be needed with other concatenative processes.
Another, weaker, argument against the mirror image of the Syllabic Template involves informal observations of assimilatory processes in connected signing. For example, anticipatory HC assimilation appears to occur in the phrase "GOOD EGG," with the change in hand configuration from GOOD to EGG occurring during the production of GOOD. If GOOD is represented as below, the spreading seems natural:
GOOD EGG LOC [chin ]===[medial to chest] [weak hand ]==[below weak hand] PO [to body]------ --------[to wk hand]-------- HC [^B ]------ --------[U ]--------
But if we use the mirror image of the Syllabic Template:
GOOD EGG LOC [chin]===[medial to chest] [weak hand]===[below weak hand] PO ------[to body ]---------------------[to weak hand ] HC ------[^B ]---------------------[U ]
we are hard pressed to account for the timing of hand configuration change in this example of connected signing. Further research on this phenomenon is needed.
At this stage in the research, neither the Syllabic Template nor its mirror image are ideal. I have chosen to set aside the questions raised by Location Epenthesis, preferring not to propose a rule of restructuring, but this should not be seen as a strong argument in favor of the position of segments in the Syllabic Template as it stands.
Perlmutter (1987) proposes that certain monosyllabic signs are marked in the lexicon for eligibility to be reduplicated, and that the effect is achieved via a syllabic template. He does not elaborate on the exact nature of the template.
For concreteness in subsequent sample derivations, I will propose a rule of lexical reduplication that is simply, "make a copy of everything and concatenate it," and applies only to items lexically marked as eligible. The usual copying of segmental material and mapping is assumed to apply.
S --> S S
(applies only to items lexically marked to undergo reduplication.)
Many ASL feature specifications are absolute, but many are relative, like [proximal(LOC)]. In this case, the location specified is proximal to another, absolute, location. First segment specifications in secondary change are functions of second segment specifications (phonetically reduced), as described below.
Twisting: PO2 = [180 degree rotary(PO1)] Nodding: PO2 = [90 degree planar(PO1)] (referent plane is the palm of the extended hand from heel to finger- tips) Pivoting: PO2 = [45 degree axial(PO1)] (referent axis is the line between the tip of the middle finger and the base of the hand) Circling: LOC2 = [circle(LOC1)] (circle between LOC1 and LOC2) or LOC3 = [circle(LOC1==LOC2)] (circle along a trajectory from one location to the other) Swinging: LOC2 = [30 degree vacillate(LOC1)] (vacillation between two locations) or LOC3 = [30 degree vacillate(LOC1==LOC2)] (vacillation along a trajectory from one location to the other) Hooking: HC2 = [(HC1)"] Flattening: HC2 = [flat(HC1)] Releasing: HC2 = [open(HC1)] Wiggling: HC2 = [wiggle(HC1)]
An example of a derivation involving only spreading follows. Spreading is the mechanism for filling empty slots, and it proceeds from right to left or left to right, as needed.
An example of a derivation involving Location Epenthesis and spreading:
To sum up, syllables in ASL are analyzed here as being composed of two segments, one fully specified and one underspecified. Specifications in the same tier can be related by either a primary change temporal domain line or a secondary change temporal domain line. Slots lacking specification are filled in by spreading. Location Epenthesis applies to monosegmental syllables, turning them into "legal" disegmental syllables. Feature specifications, especially when connected via secondary change temporal domain lines, may be relative rather than absolute. The derivations demonstrate the action and interaction of the rules: Location Epenthesis feeds Spreading, and Lexical Reduplication and Location Epenthesis are unordered with regard to each other.
The foundation for much of current ASL phonological analysis, including my own work in this paper, was laid by the work of Perlmutter, Liddell and Johnson. Perlmutter (1987) argued convincingly that underlying lexical representations of ASL signs are often quite different from the isolation forms, proposed a syllable typology, and discussed the role of moraic phonology in ASL. Liddell and Johnson (1986) proposed a detailed systematic phonetic description of ASL, and produced the standard account of compounding in ASL.
My analysis, though indebted to theirs, differs in crucial ways. In this section I will discuss some of the more important differences.
Perlmutter (1987) proposes two segmental types: P's and M's. The P segment is a "position;" the M segment is a path movement. Although he was not explicit on this issue, it appears the M segments are specified for a beginning location and a movemental direction of some type, such as "move" or "move downward." The P segments are also specified for a location, and a kind of anti-movemental direction, such as "stop" or "stay." Hand configuration has its own tier which links directly to the moraic tier. Palm orientation is not discussed in detail.
Perlmutter also proposes three syllable types: the Movemental Syllable, (P)MP; the Syllabic P; and the Geminate M. The underlying lexical representations for LEARN, UNDERSTAND and DANCE in Perlmutter's analysis are shown below:
Perlmutter observes that when hand configuration change occurs, it only occurs during the M of a (P)MP syllable or the P of a P syllable. He additionally notes that hand configuration change never occurs in a Geminate M syllable. His explanation is that each featural representation on the hand configuration tier is linked to a mora in the lexical representation. In the case of (P)MP syllables, the mora in turn links underlyingly to the M, resulting in change occurring during the M. In Syllabic P syllables, the mora links to the only thing it can link to -- the P, thus change occurs during the P. With Geminate M syllables, Perlmutter invokes the No-Crossing Constraint to explain why two hand configurations are disallowed in Geminate M's:
In my analysis, there are also two segmental types, but they are both made from the same material. The difference between them is that one is fully specified and the other is not. Movement is not a formative. There are three tiers: for location, hand configuration and palm orientation. The Syllabic Template defines the upper and lower bounds for syllable well-formedness. The underlying lexical representations for LEARN, UNDERSTAND and DANCE in my analysis are shown below:
The crucial difference between the two analyses is the status of movement. If, as Perlmutter claims, movement is a primitive, change occurs segment-internally. Then Perlmutter is required to appeal to underlying moraic linkage to explain why change can sometimes occur in P segments but at other times cannot. In my view, movement from one location to another is change from one location to another, and has the same status as change from one hand configuration to another, or change from one palm orientation to another. The fact that all these changes occur at the same time (in between the end points) is entirely unremarkable. All changes occur between segments. This is the nature of spoken-language segments (with the exception of contour segments, which nevertheless contain discreet end points between which change occurs) as depicted in current phonological theory. It should not be surprising that a similar abstract representation would be in use for signed language.
Thus, the timing of hand configuration change in LEARN coincides with path movement because path movement is change from LOC to LOC, as hand configuration change is change from HC to HC. And the fact that hand configuration change can occur apart from LOC change (on a "Syllabic P" syllable) as in UNDERSTAND is unremarkable. But what about signs like DANCE -- what explains their inability to support hand configuration change?
There are two questions to be considered: is Perlmutter's representation of the underlying form correct, and is he correct in using the No-Crossing Constraint to rule out impossible signs? Perlmutter's analysis can to some extent be empirically tested: are Geminate M signs pronounced with exactly four gestures (each M is a back-and-forth movement)? In fact, I have seen pronunciations (Hayes, Nagahara and Stack, 1987) varying from three to four to more gestures, lending support to my claim that such signs involve secondary (or uncountable) change. Finally, as Hayes, Nagahara and Stack (1987) pointed out, the No-Crossing Constraint does not necessarily rule out hand configuration change in Geminate M syllables, because the following mappings could be possible:
Yet Perlmutter's observation that hand configuration change does not occur in Geminate M syllables is valid and must be accounted for. The theory proposed here does in fact provide an adequate account: namely, that such signs are ruled out by the Syllabic Template. In DANCE (as in all "Geminate M" signs), it is the bottom slot in the second segment that is filled with [180 degree rotary(PO2)], since the top slot is reserved for primary change LOC specifications. There is no other slot available for additional HC or PO specifications.
Perlmutter (1987) observes that the timing of secondary change (which he calls "secondary movement") coincides with that of hand configuration change: it occurs only on the M of a (P)MP, the P of a Syllabic P, and is impossible in a Geminate M.
Perlmutter explains this with a well-formedness condition that states that featural representations containing secondary change must be monomoraic in the underlying representation:
My analysis explains the distribution without appeal to the moraic tier. First, with regard to the impossibility of secondary movement in a "Geminate M" syllable, recall that in the framework I have presented, these syllables are already specified for secondary change. The Syllabic Template rules out more than one secondary change per syllable.
A Perlmutter (P)MP syllable is in my framework a syllable with a primary LOC change. If such a syllable contains a secondary change, Perlmutter notes that it coincides with the primary LOC change, and not during the P on either end of the syllable. Recall that segments in my framework are static end-points. Change occurs during transition from one to the other; never segment-internally. This is more obvious when the change is primary, but remains true when secondary change is involved as well.
Finally, the fact that secondary change can occur on what Perlmutter terms a "Syllabic P" and in my analysis is a syllable without primary LOC change, is to be expected in the framework I have presented.
The most comprehensive account to date of ASL compounding is Liddell and Johnson 1986. They propose two morphological and two phonological rules to account for changes undergone by signs when they compound. In the framework they developed, there are two types of autosegments, M's and H's. These attach to articulatory bundles containing very detailed specifications for location, hand configuration and palm orientation. M's represent change, and must always attach to at least two articulatory bundles. H's represent stasis, and may attach to a single articulatory bundle.
The relevant details of Liddell and Johnson's underlying representations of GOOD, NIGHT, THINK and BELIEVE are shown below:
GOOD H M H (segmental tier) |/ \| a b (articulatory tier) NIGHT M H M H |\| |\| c d c d THINK M H |\| e f MARRY M H |\| g h
The Single Sequence Rule applies first in the compounding process. It deletes a sequence of segments if it is adjacent to an identical sequence.
Single Sequence Rule (Liddell & Johnson, 1986) M1 H2 M1 H2 M1 H2 | \| | \| => | \| a b a b a b
In the compound process leading to GOOD+NIGHT and BELIEVE (THINK+MARRY), the Single Sequence Rule applies to NIGHT:
Application of Single Sequence Rule to NIGHT: M H M H M H |\| |\| --> |\| c d c d c d
The next rule in the compounding process is the Contacting Hold Rule. This rule applies to the first sign in a compound, only if the sign contains a segment that makes body contact. It acts to delete everything except the body contact.
Contacting Hold Rule (applies to initial sign in compound) (Liddell & Johnson, 1986) [ (A) H (B) ] | [ c ] (c=contact) [ ] [body ] 1 2 3 => 2
In the examples above, the Contacting Hold Rule acts to remove everything but the first segment in GOOD, and everything but the second segment in THINK:
Application of Contacting Hold Rule to GOOD: H M H H |/|\| --> | a b a Application of Contacting Hold Rule to THINK: M H H |\| --> | e f f
The final two rules are phonological clean-up rules. Movement Epenthesis inserts an M autosegment between two segments with dissimilar articulatory bundles, thus insuring there is a legal change from one set of features to another. The Hold Deletion Rule removes H autosegments but not their associated feature bundles, if they appear between two M autosegments, thus insuring that the transition from one featural bundle to the next is smooth. These two rules yield the following surface representations of GOOD-NIGHT and BELIEVE:
After Single Sequence Rule and Contacting Hold Rule: GOOD+NIGHT --> Movement --> Hold --> Surface Epenthesis Deletion H M H H M M H n/a H M M H | |\| --> | |\| --> --> |/ \/\| a c d a c d a c d THINK+MARRY --> Movement --> Hold --> Surface Epenthesis Deletion H M H H M M H n/a H M M H | |\| --> | |\| --> --> |/ \/\| f g h f g h f g h
The proposal adopted here proceeds from rather different assumptions. Once again, the view of movement as a segment-external change from location to location is a crucial difference between my analysis and that of Liddell and Johnson. Also important is the representation of lexically-reduplicated signs like NIGHT as being underlyingly specified for only one syllable, as per Perlmutter (1987). Finally, the idea that signs like THINK and MARRY are underlyingly monosegmental, surfacing in isolation only after Location Epenthesis has applied provides a somewhat simpler account of compounding. The underlying representations my analysis proposes for GOOD, NIGHT, THINK and MARRY are:
GOOD LOC [chin ]===[medial to waist] PO [to body] HC [^B ] NIGHT LOC [weak hand ] PO [away from body] HC [^B ] THINK LOC [forehead ] PO [toward body] HC [1 ] MARRY LOC [weak hand ] PO [away from body] HC [B ]
The Single Sequence Rule is unnecessary in this framework because the underlying representations do not have reduplicated sequences. (Of course, the price of eliminating the Single Sequence Rule is the addition of a Lexical Reduplication Rule. But the resulting analysis has the advantage of minimal redundancy in the underlying representation.) The phonological clean-up rules are likewise irrelevant in this analysis, because M's and H's are not autosegments in my framework.
This leaves us with the Contacting Hold Rule, which is unnecessary because in signs like THINK the segment that would have been deleted (being predictable) is not present in the underlying representation. As already discussed in Section 3.2, the Contacting Hold Rule is not only unnecessary in an analysis including Location Epenthesis, but also inadequate to account for the facts in the compound NOW+DAY (TODAY), which behaves like a body-contacting sign, but isn't. Instead, a simple Compound Deletion Rule which deletes the second LOC in the first sign in a compound can be used:
Compound Deletion Rule (applies to initial sign in compound) X --> 0 / # A __ (A = segment; X = any other material)
The effect of Compound Deletion is shown below on GOOD:
Application of Compound Deletion to GOOD: LOC [chin ]===[medial to waist] [chin ] PO [to body] --> [to body] HC [^B ] [^B ]
GOOD-NIGHT and BELIEVE after compounding:
BELIEVE LOC [forehead]===[weak hand ] PO --------[away from body] HC [1 ]===[B ] GOOD-NIGHT LOC [chin ]===[weak hand ] PO [to body]===[away from body] HC [^B ] [^B ]
I assume the OCP would take effect to eliminate the second (identical) hand configuration in GOOD-NIGHT.
Note that the palm orientation from THINK is not retained in BELIEVE; the palm orientation from MARRY is retained and spreads leftward from the second segment of BELIEVE. This fact is not accounted for by my rule of Compound Deletion nor by Liddell and Johnson's Contacting Hold Rule. However, were both palm orientations and both hand configurations retained in BELIEVE, a sign violating the Syllabic Template would be formed. In GOOD-NIGHT, this possibility doesn't loom, since GOOD and NIGHT happen to have identical hand configurations.
In this paper I have argued that the underlying syllable structure of ASL lexical entries consists of up to two segments. The first is partially specified; the second is fully specified. The facts of syllable-level phonotactic constraints emerge from this analysis. Moreover, my analysis permits a simple and natural explanation for observations regarding the timing of hand configuration change, secondary change and the compounding process.
The major points of my analysis are:
In a sequential framework there are three formatives: location, palm orientation and hand configuration. Change from one specification to another in any of these categories is called primary change. Secondary change, rather than having the status of phonological primitive, is uncountable iteration of primary change.
Contact is shown to be more abstract than originally proposed, in that it occurs with points in space as well as on the body.
The Syllabic Template describes upper and lower bounds for underlying syllables. Location Epenthesis acts on Monosyllabic Signs to license them at the surface if they have failed to undergo any morphological processes.
Constraints on hand configuration change noted by Perlmutter (1987) are accounted for naturally by the Syllabic Template, since in my analysis all change is timed together. Timing of secondary change, accounted for by Perlmutter with regard to moraic linkage, is likewise explained in this paper in terms of inter-segmental transition rather than segment-internal movement.
The Contacting Hold phenomenon described by Liddell and Johnson turns out to be part of a larger phenomenon involving the contrast between Monosegmental Signs and other signs, and the proposal that Location Epenthesis is a post-lexical operation. Compounding is analyzed as involving only one morphological rule, the Compound Deletion Rule.
Taken together, the proposals herein present a view of the phonology of American Sign Language as one in which segments are static targets rather than movemental in nature. The standard view of human spoken languages is that they are composed of such segments at the abstract psychological level. A similar view of human signed languages is to be therefore preferred since it requires no new theoretical explanations attempting to link the psychological patterns of language to the physical channels of language.
Hayes, Bruce; Hiroyuki Nagahara and Kelly Stack. 1987. Comments on D. Perlmutter, "Prosodic Properties of Lexical and Surface Representations in ASL Phonology." Presented at the UCLA Phonology Symposium, October 18, 1987.
Humphries, Tom; Carol Padden and Terrence J. O'Rourke. 1981. A Basic Course in American Sign Language. Silver Spring, MD: T. J. Publishers.
Klima, Edward S. and Ursula Bellugi. 1979. The signs of language . Cambridge, MA: Harvard University Press.
Liddell, Scott and Robert Johnson. 1986. American Sign Language compound formation processes, lexicalization, and phonological remnants. Natural Language & Linguistic Theory, 4:4.
Perlmutter, David. 1987. Prosodic properties of lexical and surface representations in ASL phonology (draft). Paper presented at the UCLA Phonology Symposium, October 18, 1987.
Sandler, Wendy. 1987. Sequentiality and Simultaneity in American Sign Language Phonology. Doctoral dissertation, University of Texas at Austin.
Stokoe, William C., Jr.; D. Casterline and C. Croneberg. 1965. A dictionary of American Sign Language on linguistic principles. Washington, D.C.: Gallaudet College Press. [Rev. edn., Silver Spring, MD: Linstok Press, 1976.]
Wilbur, Ronnie B. 1987. American Sign Language: linguistic and applied dimensions. Boston, MA: College-Hill Press.
Woodward, James. 1978. Historical bases of American Sign Language. In P. Siple (Ed.), Understanding Language Through Sign Language Research . New York: Academic Press.
Pictures of all cited signs may be viewed in Humphries, Padden, O'Rourke (1981). A listing of sign glosses used in this paper and the page numbers on which they may be found follows:
(1) "Deaf" refers to a linguistic minority which is primarily identified by use of ASL and only secondarily identified by auditory loss. People with impaired hearing who do not use ASL are not considered a part of this minority.
(2) Contour includes arcing and circling paths of movement, which is probably the most difficult phenomenon to describe without resorting to a movement feature or autosegment. Although a detailed discussion of the status of movement is outside the scope of this paper, I believe it will prove fruitful to analyze arcs in terms of brushing contact with points in space. Additionally, Sandler's (1987) claim that circles are composed of complementary arcs seems to be an important step toward eliminating movement as a phonological primitive in the analysis of ASL.
(3) "opp" is short for opposed; in this case the thumb is opposed to the palm. "unopp" is short for unopposed. "closed" indicates that the thumb is touching either the palm or fingers; "open" indicates it is not.
(4) I am grateful to Bruce Hayes for first bringing this comparison to my attention.
(5) There is a morphological process that results in signs containing exactly one specification for each formative, plus an extra-long hold in one place. This is the protractive aspectual inflection, which is at work in signs like STARE. But its uninflected form, LOOK, is similar to that of GOOD; two LOC specifications, one PO specification and one HC specification.
(6) The reader may have noticed that the second sign in the compound, MARRY, also loses its initial LOC segment, and that it fits into the class of Monosegmental Signs. In Liddell and Johnson's analysis, the first segment of MARRY is retained in the final compound. However, the presence or absence of that segment is difficult to determine. In Liddell and Johnson's terms, it is present as a location that the hand passes through without stopping. In my analysis, the hand still travels the same path, but the intermediate location is not represented.