3.4 A proposal for a feature set for SLN
3.4.3 Orientation
3.4.3.2 Arguments for specifying only relative orientation
I want to expand the argument made by Greftegreff, by putting forward the hypothesis that it is always relative orientation only that is specified in the representation of the sign. In most cases, one specification suffices, relating a side of the hand to the end of a path movement (i.e. the second setting specification).68 In a small set of signs, a second value is needed that specifies the side of the hand that faces the major location. There are two types of arguments in favor of this proposal. In the first place, specifying both relative and absolute aspects of orientation leads to redundancy in the phonological representation. This is especially so if contact features would also be present in the representation (see van der Kooij 1996, 1997 for discussion). To properly characterize the phonotactic patterns of the language, one should strive towards eliminating such redundancy, and opt for either one of the two types of orientation features, but not both. In the second place, there are several arguments in favor of relative orientation. These have to do with variation in
68 In a subset of signs, including PEOPLE and DEAD, the specification of only one value is sufficient provided that it relates to the major location rather than the end setting. The second degree of freedom is then the predictable result of the simplest articulation of the specified location and orientation parameters. The question then arises whether it is predictable whether the specified orientation value faces the end setting or the major location, or whether this information needs to be stored in the lexicon. One way to test the claim that only one orientation value is specified might come from coarticulation studies. As the non-specified degree of freedom is predicted to be the result of the easiest articulation of the various aspects of the sign that are specified, one would predict that in different phonological contexts, for example a distinction in the height of the location preceding the target sign, different realizations of the non-specified degree of freedom would occur.
handshapes, variation in orientation, and morphosyntactic phenomena. I will discuss these three issues in turn.
The handshape variations described by Greftegreff (1992) for indexical signs appear to occur widely in other signs in SLN, too. For example, the SLN sign PEOPLE in the citation form is produced with a B-bent handshape (see Figure 3.49i below). However, one frequently encounters variants where the handshape is quite different, looking more like a curved B-hand (all fingers slightly bent at all joints, or like a handshape in between B and B-bent (i.e. more than 0 and less than 90 degree flexion at the MCP joints). The same goes for many signs with B or 1 handshapes in which the fingertips touch the location, such as I and SHOW. It seems, then, that it is not an absolute or concrete specification of handshape and hand orientation that is called for in the representation of such signs, but rather a more abstract representation that leaves open more degrees of freedom in the actual surface form. For orientation, this representation would have to contain the relationship between the fingertip of the hand and the place of articulation. For the handshape, it seems that in many signs what needs to be specified is knowledge of the selected fingers. The finger configuration is what varies in different utterances, so this should be left open in these pointing signs, as far as the perceptual specification is concerned. From a traditional phonological point of view, one would say that underspecification plays a crucial role in this proposal. If we take the view of the Functional Phonology model, one would say that the perceptual target that is specified concerns a minute part of the articulator: the finger tip. In such a case the production grammar that generates the articulatory specification has to specify the state of all joints in the upper extremity, including the finger joints. Like the state of the wrist and other arm joints, the state of the finger joints is predicted to vary in different contexts, such that the easiest articulation is used. Variation in finger configuration will be discussed more extensively in §3.4.4.
Another source of variation is the actual, or phonetic, absolute orientation of the hand. Here, too, for many signs one finds an overwhelming number of variants from utterance to utterance. For example, the sign IDEA (index finger extended, moving from ipsilateral temple outwards) may be produced with many different palm orientations. In the ‘dictionary form’ (KOMVA 1993) the palm orientation is diagonally downwards-contralateral. However, both in connected signing and in productions in isolation the palm orientation may vary from almost straight down to contralateral to backwards. What remains constant under all these utterances is the relation between the tip (or pad) of the index finger and the place of articulation (the ipsilateral temple). Thus, while the absolute orientation values differ for each variant, the relative orientation is constant for all variants. I interpret this as another source of support for representing signs like IDEA with a relative orientation value, and not an absolute one.
A third source of evidence for the importance of relative orientation comes from morphosyntactic agreement. In agreement verbs, the beginning and end locations of the path movement and the orientation of the hand may differ depending on the arguments of the verb (Bos 1993, Meir 1995). As pointed out by Meir, it is not the
absolute direction in space nor the direction of the movement that marks the relation between subject and object. Rather, it is the fact that the palm of the hand faces the object of the verb, and consequently, that the back of the hand faces the subject. Absolute orientation is not a useful notion in the description of verb agreement.
Concluding, then, there are three arguments favoring the use of relative orientation as opposed to absolute orientation. The arguments from variation in the production of signs appear very powerful. Although they are based on informal observation, it is clear that variation in absolute phonetic orientation is omnipresent. As the Functional Phonology model for SLN describes the relation between the lexical specification and the surface form, we further need to know the contexts in which each variant occurs. The question also arises how exactly the actual handshapes (more appropriately, the actual finger positions) and orientations are derived from the other values in the representation together with the context. In other words, we need knowledge on the phonetic implementation of the phonological specification of signs.