Diane Brentari

Purdue University

brentari@omni.cc.purdue.edu

The Uses of Orientation in Agreement Morphology

1. INTRODUCTION

Orientation is the least well understood of the basic phonological parameters of sign languages. Phonetically and phonologically, handshape, location, and movement have much more straightforward ways of fitting into theoretical concepts developed for spoken language than does orientation, and it is likely that for this reason it has been given less attention in the phonological and morphological literature on sign. If handshape in sign languages can be considered an active articulator and location considered a passive articulator, they can be compared with their analogs in spoken languages quite easily (see van der Hulst and Smith, 1996 for first mention of this analogy); however there is no analog in spoken languages for orientation. The active and passive articulators of spoken languages-typically, the tongue and palate-have far fewer degrees of freedom to allow for much variation in the orientation of one towards the other. In sign languages, in contrast, the hands can rotate in space and can assume a much wider range of orientation variation.

This paper proposes a model for representing orientation phonologically based on evidence of use as a morphological entity. It is well attested that having morphological status is one of the criteria for establishing a single phonological feature, such as tone (Goldsmith, 1976) or palatalization (Itô and Mester, 1989), as an autosegmental tier. Here I will extend that line of reasoning to argue that, because orientation is used morphologically in the ways described here, our phonological representations must be able to both clearly identify these orientation properties in underlying structure and be able to handle morphophonemic alternation which involve them. This paper is part of a larger project in which three cases of morphological use of orientation are described, and ways of representing and transcribing each are proposed. In this paper, I will be dealing the use of orientation in agreement (Askins and Perlmutter 1995; Meir 1998.

I argue that morphological orientation phenomena will be better illuminated by using a phonological model where orientation is represented two ways: first, it is represented as a relation between articulator features and a place of articulation (Liddell and Johnson, 1989; Greftegreff, 1992; Uyechi, 1995; Crasborn and van der Kooij 1997; Brentari 1998); second, it is represented as a small set of movement features. This contrasts with models in which orientation is represented only one way-namely, by using palm or finger tip orientation features alone which can change in value.

In order to understand how orientation is represented, it is necessary to understand the basic principle by which all of the structure is organized for all of the parameters, then we will address the specific issues related to orientation. In the Prosodic Model of sign language phonology (Brentari 1998), orientation has two possible ways of being represented. One is as inherent features, which have one set of specifications per word (i.e., selected fingers of handshape, major body place of location, etc). Inherent feature specifications typically represent static properties of signs, while prosodic features represent movement. The second type of specification possible for orientation is movement-based and is a part of the prosodic branch of structure. Typically there are two specifications per word for this type of feature (i.e., aperture of handshape, setting within a location, etc.). For monomorphemic items, one of these specifications is based on the features given in the inherent branch of structure or those 'filled in' according to a fundamental signing position (which is a position with elbows bent, palms facing each other) similar to the posture taken when signing TO-SIGN-GERMAN SIGN LANGUAGE or TO-SIGN-ITALIAN SIGN LANGUAGE. The other specification is a feature describing a type of movement, either a shape or direction of movement, or that of a setting, orientation or handshape change, where one static position changes to another, thereby creating the movement. I will argue that for morphological purposes of orientation, it is the inherent features that are relevant.

Figure 1 shows a schematic structure of the total representation within the Prosodic Model. In the Prosodic Model of sign language phonology. Inherent features (IF) and prosodic features (PF), are associated to the root.

Figure 2 shows the parts of the structure and related features involved in expressing each type of orientation; Figure 2a shows inherent orientation; Figure 2b shows prosodic orientation. Inherent orientation is a relation between features of the Articulator and the Place of Articulation. The relevant Articulator features are located at the nonmanual node and at the hand node. The entire structure of Place of Articulation is also involved in the orientation relation.

The Articulator can be expressed nonmanually or manually. This is important, since we have seen in work on body shift (Kegl, 1985; Engberg-Pedersen 1993) and on eye-gaze (Bahan 1996), that morphological agreement expressed in these ways, as well as by the orientation of the hand. If it is a manual expression of orientation, then the hand class node is responsible for the specification of a handpart; this handpart plays a role in the orientation relation. The hand class node has eight possible features specified there, shown below in Figure 3. Place of articulation is first divided into planes: the x-vertical, y-horizontal, and z-midsagittal planes. These three planes compartmentalize the so-called neutral space, and have been discussed at length in Uyechi (1995) and Brentari (1998). The body is also located in an x-plane, and the body is divided first into 4 major articulatory regions- head, torso, arm, and hand. Each region are further divided into 8 sub-regions where signs occur.

It is expected that the eight places within each region will be language particular. Taken together the combinations of plane, region and sub-region specification designates the place of articulation, which is the other part of the orientation relation; that is, the handpart or nonmanual articulator is facing or oriented towards the place of articulation.

The prosodic features [supination], [pronation], [flexion], [extension], [adduction], [abduction] are types of movements. In the prosodic model, they are movements from the fundamental signing position, with elbows bend at a 90 degree, palms oriented towards each other in the mid-sagittal plane. The prosodic orientation features play no role in the discussion of agreement to follow, and they will not be mentioned again. The inherent orientation relation, as just described, is schematically represented in Figure 4.

2. THE ROLE OF ORIENTATION IN THE EXPRESSION OF AGREEMENT

Johnson and Liddell (1984), Sandler (1989), Brentari (1988, 1998), Diane Lillo-Martin (1995), Engberg-Pedersen (1990, 1993), Askins and Perlmutter (1995), Bahan (1996), and Meir (1998) have all discussed the alternations that occur in person agreement verbs. Below, I will use the Prosodic Model to provide a unified account of agreement expressed either as hand-to-place and body-to-place orientation.

Agreement morphemes have traditionally been expressed as loci in space (Liddell, 1984). Engberg-Pedersen discusses a 'calendar plane' in her work, and it is from this and other work on place of articulation as planes rather than points (Uyechi 1995), that I take the morphemic representation of person agreement to be planes rather than points in space (i.e., 2-dimensional mental constructs rather than one-dimensional mental constructs). There are two arguments for a planar representation of agreement morphemes. First, we need a vertical plane-which can be further specified with respect to body part-when considering how surrogate space works, a concept developed by Liddell (1995). He argues that, when using an agreement verb, the expression of agreement may be tailored to fit the dimensions of the object. This type of morphemic use of space can be accommodated within the grammar if an x-plane (similar to the one which functions for the signer's body) can be used for other structures involving person agreement. Second, Brentari (1988) and Askins and Perlmutter (1995) have argued that path features are involved in person agreement. These features are further defined in Brentari (1998) as [tracing] and [direction]. A movement will have a [direction] feature if it describes a line that occurs at a 90 degree angle to a plane; a movement will have a [tracing] feature if it describes a line that occurs in a plane itself. This has consequences for how contact will be realized.

2.1. Path, Place of Articulation, and handpart features

The syntactic facts about agreement important for this paper are as follows. Like most languages, ASL has a class of predicates that can subcategorize for only one syntactic argument, a subject (e.g., SIT, SORRY, DISAPPEAR), and other classes of predicates that can subcategorize for more than one argument. For example, READ, RESPECT, ADMIT, and ADOPT subcategorize for a subject and an object. GIVE is the typical verb that can subcategorize for three syntactic arguments in many languages, including ASL; see Padden (1983) for an analysis of these signs. Phonologically (and independent from their syntactic properties), some ASL verb stems which subcategorize for two syntactic arguments and include a [direction] feature in the verb stem allow for place agreement affixation. Traditionally, signs which subcategorize for an object but do not realize this object via place affixation (1) are called 'plain' verbs (Padden, 1983); other signs can exhibit either object agreement or both subject and object agreement (2). In more recent work, Bahan (1996), Neidle, et al. (2000) have argued that agreement may be expressed nonmanually be either head tilt or eye gaze or both, and that signs in (1) and (2) may include non-manual agreement markers in their forms.

Now, let us address the phonological facts. The phonological representation of the singular person agreement morphemes are planes; specifically, x-planes, located the place of articulation branch of structure. First person corresponds to the x-plane proximal to the signer's body; non-first person corresponds to an x-plane which forms a tangent to the arc which describes the distal signing area. Non-first person plural object marker (i.e., they) is a plane described by the arc itself. There may also be a first person plural (i.e., we) (see Mathur, 2000). The schema of this is shown in (3a); the phonological structures are given in (3b).

The actual location of the agreement x-planes (distal or proximal, with a specific location index) is specified at the setting node in the Prosodic branch of structure.

The features of the person agreement markers are affixed in different parts of structure. The specification for the x-plane is affixed to the place of articulation node. The [distal] or [proximal] features will be affixed at the setting node. If there is a [direction] path feature in the input, which generates 2 timing slots (see details in Brentari, 1998, chapter 4), the sign will affix these distal or proximal features to the timing slots according to an Alignment Constraint (8). If there is no path feature, the agreement will be expressed by orientation alone (e.g. STARE).

As stated above, the orientation relation is expressed by a relationship between some Articulator features and a place of articulation; furthermore, when a sign includes a directional path movement, this is also important for the realization of agreement. Brentari (1988) provided a phonological principle to characterize the possible realization of agreement as the Direction of Transfer Principle (4), which handles the starting point and ending point of signs containing a path feature:

Askins and Perlmutter observed that some signs-- such as those in (2a) exhibit different orientation when agreement changes (1ANSWER3; 3ANSWER1), while other exhibit the same orientation regardless of the agreement morphology (1TELL3; 3TELL1). They analyze this difference as follows. A verb, such as ANSWER, which changes its orientation according to its agreement morphology, has no underlying orientation features, and the agreement morphology is expressed via orientation. A verb, such as TELL, which does not change its orientation according to its agreement morphology, has underlying orientation features, and thus does not express agreement via orientation.

The full range of features involved in orientation or the manner in which the setting features [distal] or [proximal] "spread" to orientation is not fully worked out in Askins and Perlmutter (1995). Here I would like to provide such an analysis of the precise features and manner of spreading that takes place in forms, such as ANSWER, but which is blocked in forms, such as TELL. I would argue that what happens is that default specifications for inherent orientation are filled at the hand node in the absence of any other competing feature, according to the Direction of Transfer Principle, given in (5); (Brentari, 1998).

The principle in (5) would fill in ['back of hand'] ] (given as [4] in Figure 3) in the ANSWER case, but not in the TELL case. The two setting features associate to the timing slots generated by the [direction] feature right-to-left, according to the Alignment Constraint argued for in Brentari (1998), and given in (6)..

The representation for ANSWER before and after agreement affixation is given in (7a-b); equivalent forms for TELL are given in (8a-b). In (7b) and (8b), the features that express person agreement are given in bold. In the ANSWER case, which allows both object and subject affixation, we see that the x-plane is inserted in the place of articulation branch of structure for 3ANSWER1, both the [distal] and [proximal] features are affixed as setting features, and since ANSWER is a typical (forwards) agreement verb, rather than a backwards verb, such as INVITE, the setting feature for the subject occurs first and the setting feature for the object occurs second.

(7a) ANSWER

(7b) 3sgANSWER1sg (The person agreement affix's features are given in bold)

TELL is a sign which accepts only object agreement, and it also has a handpart specified in the input ['fingertips'] in (8a); therefore, in the form 3sgTELL1sg (8b) there is no ['back of hand'] specification at the hand node. It is blocked, following Askins and Perlmutter's proposal. The x-plane is inserted as it did in ANSWER. At the setting node, only the object marker (1sg) [proximal] is affixed, since TELL is specified in the lexicon as taking only object, but not subject, agreement. Because of this, the inherent features for Place (given as IF) fill in that space at the setting node, ordered first, since in typical (forward) agreement verbs, objects associate to the second timing slot.

(8a) TELL

Both ANSWER and TELL include a place specification for the x-plane and setting specifications for subject/object as appropriate. The difference between them occurs at the hand node. Since no feature is specified in the input of ANSWER at the hand node, the back of the hand will be filled in as a part of the agreement morpheme, and this will allow the form to express agreement via orientation, as well as place of articulation. In contrast, TELL has ['fingertip'] specified at the hand node, and this will prohibit expression of agreement via orientation.

(8b) _3sgTELL_1sg (The person agreement affix's features are given in bold)

A further observation about agreement forms is made in Meir (1995), who analyzes forms in Israeli Sign Language (ISL) and argues that 'facing' (i.e., fingertip orientation), rather than palm orientation, is the structural feature of orientation that changes in some agreement forms; she further argues that an object reference must be employed in the direction of transfer principle for Israeli Sign Language. The Prosodic Model can predict which forms will take facing features by observing the plane of articulation of the stem. Not all sign stems are articulated in an x-plane, like ANSWER and TELL; for example, many Type-1 2-handed signs are articulated in a z-plane (the mid-sagittal plane). Consider the form CONTROL and _3sgCONTROL 1sg in this context (9a-b). CONTROL is a symmetrical sign (the hands are facing each other's equivalent handpart), which has an alternating [tracing] path movement.

In the stem CONTROL (9a), we see the repeated straight tracing movement generating 4 timing slots on the timing tier, co-indexed as shown. The [alternating] feature below the H2 node insures that the dominant hand will execute the shown xi xj xi xj segmental sequence, but the nondominant hand will execute a xj xi xj xi sequence. As for most 2-handed signs in which the handshapes are the same (Type 1 signs), the plane of articulation of CONTROL is the z-plane (i.e., the mid-sagittal plane). There is no handpart specified at the hand node.

(9a) CONTROL

(9b) _3sgCONTROL_1sg (the person agreement affix's features are given in bold)

In the agreement form _3sgCONTROL_1sg (9b), the person agreement affixes for 3sg and 1sg are indicated as before. The x-plane and the setting features [distal] and [proximal] are inserted as before, but notice that at the hand node ['base of hand'] rather than ['back of hand'] is specified. The is not stipulated, but is predicted by the system; a z-plane or y-plane stem will insert ['base of hand'], as given in (5), achieving the correct result with forms with fingertip orientation to the object, or with forms, such as HIT, where the fingers are not extended and the handpart oriented toward the object is the back of fingers.

In summary, the Prosodic model can accommodate the range of facts covered by both the Askins and Perlmutter (1995) and the Meir (1995) observations. The important point is that by representing orientation as a relation, the component parts of it (handpart and place of articulation) can be manipulated to achieve the correct result, and that the part of the hand involved in the orientation relation for agreement is predictable under the Prosodic account.

2.2 Nonmanual expression of agreement

Thus far we have focused on place of articulation and handpart as the expression of person agreement in the case of signs with path movement. Let us now turn to the case of agreement using nonmanual features. Recall that the basic orientation relation is that which exists between the Articulator node and the Place of articulation node. In the Prosodic Model, the Articulator node has a nonmanual and a manual branch of structure (see Figure 4). In this section, I will argue that the torso, eyes, or head can express the orientation relation, either redundantly with the handpart specified at the hand node, or, in cases like TELL, as the only expression of agreement via orientation. Lillo-Martin (1995) has argued that the body itself is a logophoric pronoun; Neidle et al.(2000) have argued that head tilt or eye gaze can serve as agreement markers; both have claimed that non-manual expressions of agreement are optional. I will argue that the part of the orientation relation expressed in the Articulator branch of structure can spread to or be expressed by other nodes within this structure.

My argument is based on other cases of assimilation within the Articulator branch of structure-assimilation of dominant hand features to the nondominant features in 2-handed signs hand (Brentari and Goldsmith 1993, Brentari 1998), and assimilation of manual features to the nonmanual tier.

Historically, 2-handed signs with different handshapes (Type 3 signs) have often changed the handshape on the non-dominant hand to match that of the dominant hand (e.g., WORLD, INSTITUTE, WHISKEY, THREAD). In the Prosodic Model, such cases are handled by deletion of features at the H2 node and by the addition of an association line from the H1 node to the H2. This diachronic operation is expressed formally as in (10).

(10) Diachronic change resulting in assimilation of handshape and movement features in WORLD

Type 3->Type 2 sign	Type 2->Type 1 sign
(H2 features deleted, assimilation of handshape features)	(Association line added assimilation of movement features)

An assimilation analysis can also be used to analyze the spreading of manual features to the nonmanual tier. In the Prosodic Model, the manual and non-manual branch of the Articulator node were established on independent empirical evidence in stems. for example, the mouth opening in AGHAST ('jaw-drop'), can be analyzed as assimilation from the opening of the two hands, and tongue wag in REALLY-FINE can be analyzed as assimilation from the wiggling of the fingers. Moreover, it is argued in Brentari (1998) that movements of more distal joints (i.e., handshape changes articulated by finger joints) sometimes spread to more proximal joints (path movements articulated by the elbow joint), thereby making such movements more perceptually salient. This may also be the case here; that is, the addition of nonmanual expression of handpart features of the hand node make the agreement morphology more salient. Using this analysis to account for orientation features in agreement is a natural extension of the model.

(11) Assimilation of hand node features to the non-manual node

At this early stage of knowledge about the phonological behavior of nonmanuals, I leave the features without any hierarchical organization, and leave further investigation of the relationship between manual and nonmanual phonological behavior for future research.

3. CONCLUSION

This phonological analysis of agreement morphology has demonstrated the value of considering orientation a relationship between the Articulator and a Place of Articulation rather than a set of absolute features. The Prosodic Model account proposed here shows :

1. exactly how a redundant value for orientation can be filled in for forms which have not handpart features in their input
2. how focusing on the plane of articulation of the stem can predict the particular default value inserted in cases where agreement is expressed both by agreement loci and orientation features.
3. how a nonmanual node, which is the site for expressing static nonmanual behaviors discussed as agreement morphemes, can be considered a daughter of the Articulator. This branch of structure can participate in assimilation of handpart features, or may be the sole expression of an agreement morpheme.

REFERENCES

Askins, D., and D. Perlmutter. (1995). Allomorphy Explained through Phonological Representation: Person and Number Inflection of American Sign Language. paper presented at the annual German Linguistic Society meeting, Göttingen, Germany.

Brentari, D. (1988). Backwards Verbs in ASL: Agreement Re-opened. Proceedings from the Chicago Linguistic Society 24, Vol. 2, Parasession on Agreement in Grammatical Theory, 16-27.

Brentari, D. (1998). A Prosodic Model of Sign Language Phonology. Cambridge, MA: MIT.

Brentari, D. and C. Padden. (in press). A lexicon with multiple origins: Native and foreign vocabulary in American Sign Language. In D. Brentari (Ed.) Foreign Vocabulary in Sign Languages: Cross-linguistic Analysis of Word-formation Processes. Mahwah, NJ: Lawrence Erlbaum and Associates.

Crasborn, O., and E. van der Kooij. (1997). Relative orientation in sign language phonology. In J. Coerts and H. de Hoop, eds., Linguistics in the Netherlands 1997, 37-48. Amsterdam: John Benjamins.

Engberg-Pedersen, E. (1993) Space in Danish Sign Language. Signum Verlag, Hamburg, Germany.

Goldsmith, J. (l976). Autosegmental phonology. Cambridge, MA: MIT dissertation.

Greftegreff, I. (1992). Orientation in diexical signs in Norwegian Sign Language. Nordic Journal of Linguistics, 15, 159-182.

van der Hulst, H. and A. Mills. (1996) "Issues in Sign Linguistics: Phonetics, Phonology and Morphosyntax," Lingua 98, 3-17.

Itô, J. and R. A. Mester. (1989). Feature predictability and underspecification: Palatal prosody in Japanese mimetics. Language 65.258-293.

Johnson, R. and S. Liddell. (1984). Structural diversity in the American Sign Language lexicon. CLS 20, vol. 2: Parasession on lexical semantics, ed. by. D. Testen, V. Mishra, and J. Drogo, 173-186. Chicago: Chicago Linguistic Society.

Liddell, S. (1990) "Four Functions of a Locus: Re-examining the Structure of Space in ASL, in C. Lucas (1990) Sign Language Research: Theoretical issues, 176-198. Gallaudet University Press, Washington, DC.

Liddell, S. (1995). Real, surrogate, and token space. In K. Emmorey and J. Reilly, Language, Gesture, and Space, 19-4. Hillsdale, NJ: Lawrence Erlbaum.

Lillo-Martin, D. (1995). The point of view predicate in American Sign Language. In K. Emmorey and J. Reilly (eds.) Language, Gesture, Space, 155-170. Mahwah, NJ: LEA.

Mathur, G. (2000). Two kinds of constraints on verb agreement in signed languages. Paper presented at the annual meeting of the Linguistic Society of America, Chicago, IL

Meir, I. (1995) "Syntactic-Semantic Interaction in Israeli Sign Language Verbs: The Case of Backwards Verbs," in H. Bos and T. Schermer, eds. (1995) Sign Language Research 1994., 105-119. Signum Verlag: Hamburg.

Meir, I. (1998). Thematic Structure and verb agreement in Israeli Sign Language. Doctoral dissertation. The Hebrew University of Jerusalem.

Neidle, C., J. Kegl, D. MacLaughlin, Benjamin Bahan, and R. G. Lee. (2000). The syntax of American Sign Language: Functional categories and hierarchical structure. Cambridge, MA: MIT Press.

Sandler, W. (l989). Sequentiality and simultaneity in American Sign Language phonology. Dordrecht: Foris.

Uyechi, L. (1995) The Geometry of Visual Phonology, Doctoral dissertation, Stanford University, Stanford, California, Published 1996, CSLI, Stanford, California.

Wallin, L. (1994) Polysynthetic signs in Swedish Sign Language, Doctoral dissertation, University of Stockholm, Stockholm, Sweden.

List of workshop papers