Location, Location, Location

Location, Location, Location Deborah Cates, Eva Gutiérrez, Sarah Hafer, Ryan Barrett, David Corina

Sign Language Studies, Volume 13, Number 4, Summer 2013, pp. 433-461 (Article) Published by Gallaudet University Press DOI: 10.1353/sls.2013.0014

For additional information about this article http://muse.jhu.edu/journals/sls/summary/v013/13.4.cates.html

Access provided by University College London (UCL) (23 Jun 2014 05:34 GMT)

d eborah c a t e s , e v a g u t i é r r e z , sarah haf e r , r y a n b a r r e t t , and david c o r i n a

Location, Location, Location Abstract This article presents an analysis of the relationship between sign structure and iconicity in American Sign Language. Historically, linguists have been pressured to downplay the role of form-meaning relationships (iconicity) in signed languages. However, recent inquiries into the role of traditional phonological parameters of signs (handshape, location, and movement) in the development of metaphor, event telicity, and c-selection have invited a reanalysis of the role of iconicity in signed language structure.We present an analysis of the distribution of handshape, location, and movement parameters in signs and examine which of these parameters are most often used to encode iconic properties. We examine both iconic and visual-metaphorical relationships to the referents of 767 signs of American Sign Language. The findings suggest that body location is salient in conveying iconic information both for signs with concrete referents (iconic) and for signs with abstract referents (visual-metaphorical). We discuss the implications of these findings for structural and processing models of naturally occurring signed languages.

Th e g oal of this article is to understand the role of body location in iconicity and its contribution to the structure of sign language. In the background section we discuss the prevalence of Deborah Cates is a graduate student in the Department of Linguistics at University of California, Davis. Eva Gutiérrez is a postdoctoral fellow at Deafness Cognition and Language Research Centre, University College London. Sarah Hafer is a fifthgeneration deaf signer and studied linguistics at University of New Mexico. Ryan Barrett is a graduate student at Gallaudet University in the Department of Government and Public Affairs. David Corina is a director of the Laboratory for Cognitive Neurolinguistics at the Center for Mind and Brain and a faculty member in the departments of Linguistics and Psychology at University of California, Davis.

433 Sign Language Studi e s Vol . 13 N o. 4 S um m e r 2 013

434 | Sign L ang uag e Studi e s

iconicity at all levels of signed language structure and review both historical and recent treatments of iconicity in language theory and processing. In the section on methods, we detail the methods we used in an analysis of 767 signs of American Sign Language (ASL), looking at the relationship between iconicity and the phonological parameters of the language. In the remaining sections, we quantify our results and present our findings, discussing them in light of the results of several previous studies on iconicity and signed language processing. We highlight a relatively understudied structural parameter for sign form—location—to show how it is used productively both iconically and in metaphorical extension of meaning, as well as to illustrate the way in which body locations in ASL are used productively to refer to real-world semantic forms and categories. Finally, we point to new findings that suggest that body locations deserve a special status in the neurobiology of signed languages.

Background Saussure’s notion of an arbitrary link between words (signifiers) and their referents (signified) has been widely accepted in modern linguistic theories and is taken to be one of the design principles of human language (Hockett 1960). However, there is a growing realization that this central tenet of linguistics might be challenged by a full accounting of naturally occurring signed languages of deaf people. The visual properties of signed languages provide rich ground for the development of iconic relationships. One popular definition of iconicity in the literature comes from Stein (1973, 706): “a sign or representation that stands for its object by virtue of a resemblance or analogy to it.” This definition has been adopted by signed language researchers and applied to signs that bear a close resemblance to the signified object, action, or characteristic (Frishberg 1975; Klima and Bellugi 1979; Supalla 1978). Early work on sign structure established that signed languages are naturally occurring, full-fledged human languages with recurring compositional elements (e.g., handshape, location, movement) that combine in language-specific and rule-governed ways to systematically impart lexical meaning. This research included the use of formal linguistic devices largely borrowed from spoken language research (see

Location, Location, Location | 435

Sandler and Lillo-Martin 2006 for a recent review). One outcome of this focus was a lack of inquiry into the presence of iconicity in signed languages, possibly from a desire to refute the popular belief at the time that signed languages were somehow primitive forms of communication.While most of this work admits that signed languages have iconic elements, it suppresses in-depth treatments of iconicity. However, recent work on signed languages builds upon this early research detailing the structure of elements in sign language but acknowledges that iconicity plays a role in it and is not just a by-product of language in the manual modality.This research encompasses discussion of semantics and phonology (Stokoe 1991), and, more recently, the relationship between grammatical constructions and real space (Liddell 2003), the encoding of event properties in verb movement (Wilbur 2004), and the tension between iconicity and articulatory constraints on sign production (Demey and van der Kooij 2008). Another avenue of research into iconicity relevant to the present study is that of iconicity in the metaphor of signed languages. Taub (2000) posits the existence of a double-mapping process whereby signers attach a visual image to a metaphor and then encode that image into sign. For example, th i nk - pe net rate (figure 1), “get my point across [to you],” looks similar to dri ll (figure 2) in that both involve the insertion of the index finger of the dominant hand

Fig ure 1. th i nk - pe net rate.

436 | S ign L ang uag e Studi e s

F i g ure 2 . dri l l .

through the space between fingers of the non-dominant hand. dri ll is highly iconic; the dominant hand adopts a configuration that represents the tool, and the motion and action on the non-dominant hand represent the drill penetrating an object. th i nk - pe net rate describes the abstract notion of thought getting through a barrier; in this case, the penetration is metaphorical, but the sign realizes the “barrier” of the other person’s mind and the thought penetrating it as the visual image of something drilling through something else (Taub 2000, 40). Metaphorical mappings and expressions in signed languages may be influenced by their ambient cultures ( Jarque 2005). An example of this is the use of the shoulder location for ASL signs of responsibility such as bo s s (figure 3). English speakers use the metaphor of responsibility as a weight on one’s shoulders, and that metaphor has been encoded in ASL via location. However, restrictions appear to exist on the application of metaphorical extension in sign. Meir (2010) posits a double-mapping constraint whereby metaphorical mapping of meaning onto a sign must be structure-preserving. . In other words, unless the same structural correspondence exists between the iconic sign and its metaphorical mappings, the mapping is blocked. Meir uses the example of eat (figure 4) in Israeli Sign Language (ISL), which, like eat in ASL, encodes the image of putting something in one’s mouth. The structural correspondence between eat and the act


Fig ure 3. bo s s.

of eating encodes only the act of putting food into one’s mouth and not the processes of chewing and swallowing that accompany eating. Therefore, metaphorical uses of “eat,” such as “the acid ate the key,” in English are not acceptable in ISL (or ASL, for that matter) because they encode the digestion and erosion aspects of eating, which are not structurally linked to eat. Here we are concerned with understanding the role of body location in iconicity in sign language. We document how the universal aspects of the human form provide rich ground for the manifestation

Fig ure 4 . eat.


of iconic forms, ranging from physical appendages (e.g., horns, noses) to psychological states that reference body locations (e.g., feelings in the heart or stomach; thoughts in the head).

Methods The following analysis was performed on a database of 767 lexical signs. This database was compiled for the purposes of an independent study to examine the role of location in sign recognition and, as such, is here used to determine, a priori, whether one or more sign parameters (location, handshape, and movement) are productive in the expression of iconicity. Lexical signs for the purpose of this study were coded in dictionary citation form and therefore include lexicalized classifiers (e.g., hai r b ru sh , f une ral) and fully integrated compounds (e.g., b roth e r, si ste r, b reak fast ). Therefore, we excluded non-integrated compounds (where the two root signs are still separate in production), non-lexicalized classifiers (such as ve h i c le ), and signs with widespread regional variants. The exploration of form and meaning in non-lexicalized classifiers is a topic for future research. For each sign, we coded handshape, location, movement, and orientation. Coding Parameters

Handshapes were coded as discrete forms if they varied in the bend or flex of any digit or in the contact between fingertips; as such, handshapes that may typically be classed as allophonic variants (e.g., curved 5 or straight 5) were coded separately. In all, this resulted in forty-one discrete handshape codes. Location was coded using the distinctions made in Johnson and Liddell (1984) for signs with body contact, and a three-by-three grid was used to determine codes for neutral space (figure 5). (For a more recent and detailed version of their system, see Johnson and Liddell 2011a, 2011b, 2011c, 2012.) For signs wherein the non-dominant hand is stationary, such as vote (figure 6), or the non-dominant arm establishes a horizon like tre e (figure 7), the non-dominant hand is coded as the location (for a theoretical discussion of the distinction, see Brentari 1998; Brentari and Goldsmith 1993; Sandler 1989). Signs that begin at one location and end at another are coded with both (e.g., woman, which begins at the chin and ends at the chest). In all, this resulted in seventy-one

Fig ure 5. Neutral Space Locations.

Fig ure 6 . vote.

Fig ure 7. tre e.

439


F i g ure 8 . th i nk (top) c e nt (bottom).

separate location codes. Movements were coded as paths of motion and internal movements, including handshape changes. Paths of motion were coded differently if they moved to or away from a location even if their shape was otherwise the same because this has been shown to be an important distinction in minimal pairs (e.g., think and c e nt ; see figure 8). Like location, multiple discrete movements in signs were coded, though reduplication was noted with a “+” but not separately coded. In all, this resulted in twenty discrete movement codes. See the appendix for a complete list of codes. Although Battison (1978) introduced palm orientation as a fourth phonological parameter, the complexity of determining the exact orientation at any point along a sign’s path of motion, the susceptibility


of palm orientation to articulatory ease led to the decision to ignore palm orientation for this article. A native signer fluent in English grouped the signs into lexical categories, and where there were questions or multiple possibilities, a dictionary was consulted. Generally, if a sign could be a noun or a verb, the noun form was selected, leading to a possible under-representation of iconic movement in the analysis since the process of grammaticalization for nouns reduces the movement. In all, in the present database there are 12 adverbs, 119 adjectives, 421 nouns, and 215 verbs. A random sampling from three ASL dictionaries yielded an expected distribution of 20 adverbs, 134 adjectives, 384 nouns, and 234 verbs, so our database is slightly skewed in favor of nouns. Coding Iconicity and Metaphor

For this study, signs were put into one of three groups: iconic, doublemapped metaphor (DMM), and noniconic, following Taub (2000). Signs that, in part or in whole, resemble a physical referent or mimic an action were labeled iconic, such as de e r (figure 9), which encodes the antlers of a buck. This analysis takes into account that iconicity in a sign does not mean a direct one-to-one mapping of all features of a referent and that certain aspects may be iconic whereas others are arbitrary (Klima and Bellugi 1979; Thompson, Vinson, and Vigliocco 2009).

Fig ure 9 . de e r.


F i g ure 10 . scot land.

Of the signs analyzed for this article, the DMM signs fall into two categories. The first comprises signs that encode a physical referent based on a symbolic or stereotypic representation, such as the traditional plaid clothing of scot land (figure 10). The double mapping here is an association of the country with some cultural aspect of its people (mapping 1) and then an encoding of that image (mapping 2). The second category includes signs that encode an image representing an abstract concept such as thought or time. Signs of thought, such as f org et (figure 11), conceptualize the mind and memory as

F i g ure 11. f org et.


Fig ure 12 . ye ste rday (top) tomorrow (bottom).

a physical object—the head. Signs for time, such as ye ste rday and tomorrow (figure 12), conceptualize time as “past is behind” and “future is ahead.” This category of signs is mapped in a similar way as th i nk - pe net rate, discussed earlier. Any signs not meeting these criteria were not considered in this analysis. Upon completion of this categorization, there were 270 iconic, 252 DMM, and 245 signs not iconically or metaphorically related to their referents. Coding Iconicity at the Level of Individual Parameters

Once the signs were categorized, each parameter (handshape [HS], location [LOC], movement [MOV]) was analyzed for iconicity. For a


F i g ure 13. danc e.

parameter to be considered iconic, it had to have some relationship to the referent or image associated with the sign. In the earlier example of de e r, the sign encodes the antlers on a deer’s head. Therefore, the handshape is iconic because it encodes the shape of the antlers; the location is iconic because it encodes the position of the antlers on the head; the movement, however, is not iconic because it does not represent any type of movement of the antlers. Handshape was considered “iconic” when it in some way resembled a part or the whole of the referent when removed from location and movement, such as the representation of legs in danc e (figure 13); in cases where the nondominant hand is the location, the only hand considered for iconic purposes is the dominant hand. Movement was labeled “iconic” when it bore a relationship to the movement of the referent in question or when the movement served to create an outline of the referent such as the movement in africa (figure 14). Location was marked “iconic” when it encoded either the location of the referent image, such as the antlers on the head in de e r, or when it corresponded to the place of use of an object, such as the cheek in phone (see figure 15). Two independent raters coded 20 percent of the iconic and DMM signs in the database using the guidelines described in this section. The author and at least one rater agreed on 85 percent of the iconicity judgments (86 percent for iconic signs and 84 percent for DMM signs). For each parameter, the agreement percentages were as follows: LOC 84 percent, HS 85 percent, and MOV 86 percent.


Fig ure 14 . af ri ca .

Results and Discussion The goal of this analysis is to determine the relative contributions of the parameters LOC, HS, and MOV in giving rise to iconic or metaphorical form-meaning pairings in ASL. Distribution of Iconic Parameters in Iconic and DMM Signs

Tables 1a, 1b, and 2 present the distribution of iconic parameters in iconic signs. Table 1a shows the total number of iconic signs present in the database with each possible combination of iconic parameters,

Fig ure 15. ph one.

446 | Sign L ang uag e Studi e s Table 1 a . Distribution of Iconic Parameters in Iconic Signs with Examples Iconic Elements of Iconic Signs LOC ✓ ✓ ✓ ✓

HS

MOV ✓

✓ ✓ ✓ ✓

✓ ✓ ✓

Total

Example

25 31 51 86 11 23 42

clothe s cat bridge butte r check fre eway gamble

Table 1b. Iconic Parameters across Lexical Categories for Iconic Signs Parameter

Nouns

Verbs

LOC HS MOV

81.5% 78% 75%

17.5% 21.9% 25%

each with an example from the lexicon. Table 1b shows the breakdown of iconic parameters across lexical categories for iconic signs. Table 2 shows three things: the distribution of the total number of signs with iconicity in each parameter; the breakdown of those signs with only one iconic parameter; and the breakdown of those signs with two or more iconic parameters. Table 2 reveals that although the distribution of iconic parameters is relatively equal across all of the iconic signs, the prevalence of LOC and HS outweigh MOV in those signs with only one iconic parameter. Table 1a reveals that the most robust grouping of parameters is in signs with holistic iconicity (iconic LOC, HS, and MOV). The implications of this are discussed in the section on coding iconicity and metaphor. Table 1b also shows that, for verbs, iconic movement is more prevalent than iconic handshape or location, which is not surprising, given that some verbs are subject to imitative movement. Iconic handshape and location seem to be more likely than iconic movement to map onto nouns with concrete referents, except when the movement encodes the action associated with use of the referent. Tables 3a, 3b, and 4 present the distribution of iconic parameters in DMM signs. Table 3a shows the total number of DMM signs present in the database with each possible combination of iconic parameters,

Location, Location, Location | 447 Table 2 . Distribution of Iconic Elements by Parameter in Iconic Signs with Percentages Percentage of Iconic Elements by Parameter total signs w/iconic LOC total signs w/iconic HS total signs w/iconic MOV Percentage of Signs with

193 of 269 72% 202 of 269 75% 170 of 269 63% Only One Iconic Parameter

only LOC 25 of 59 42% only HS 23 of 59 39% only MOV 11 of 59 19% Percentage of Signs with Two or More Iconic Parameters LOC+MOV LOC+HS MOV+HS LOC+HS+MOV

31 51 42 86

of of of of

210 210 210 210

15% 24% 20% 41%

each with an example from the lexicon. Table 3b shows the breakdown of iconic parameters across lexical categories for DMM signs. Table 4 shows three things: the distribution of the total number of DMM signs with iconicity in each parameter; the breakdown of those DMM signs with only one iconic parameter; and the breakdown of those DMM signs with two or more iconic parameters. Table 3 a . Distribution of Iconic Parameters in DMM Signs with Examples Iconic Elements of DMM Signs LOC

HS ✓ ✓

✓ ✓ ✓ ✓

MOV

Total

Example

✓

36 10 27 51 56 32 40

challe nge e ng ine e ring expe nsive happy inve nt jewish blow-top

✓ ✓

✓ ✓

✓

Table 3 b. Iconic Parameters across Lexical Categories for DMM Signs Parameter LOC HS MOV

Adverbs

Adjectives

Nouns

Verbs

0% 0% 3.7%

29.2% 20% 21.38%

26.97% 37.6% 29.6%

43.8% 42.2% 45.3%

448 | Sign L ang uag e Studi e s Table 4 . Distribution of Iconic Elements by Parameter in DMM Signs with Percentages Percentage of Iconic Elements by Parameter total signs w/DMM LOC 179 of 252 71% total signs w/DMM HS 109 of 252 43% total signs w/DMM MOV 159 of 252 63% Percentage of DMM Signs with Only One Iconic Parameter only LOC 51 of 97 53% only HS 10 of 97 10% only MOV 36 of 97 37% Percentage of DMM Signs with Two or More Iconic Parameters LOC+MOV LOC+HS MOV+HS LOC+HS+MOV

56 32 27 40

of of of of

155 155 155 155

36% 21% 17% 26%

Table 4 reveals that, unlike with iconic signs, the distribution of iconicity across the three parameters is not equal for LOC, HS, and MOV. Both LOC and MOV are more robust in the total distribution and in the breakdown of signs with only one iconic parameter. Table 3 reveals that the two largest clusters of DMM signs have only iconic LOC (51 of 252) or iconic LOC and MOV (56 of 252). The differences in iconic parameter distribution between iconic and DMM signs is discussed in the next section.

Analysis of the Distribution of Iconicity across Parameters Tables 1 through 4 show an interesting difference between the distribution of parameters in iconic and DMM signs. Whereas LOC and MOV are consistently robust across categories, HS is not. This distribution is unexpected, given the previous bias in the literature on handshape iconicity but is not surprising for movement and certainly not for location. Notice that although LOC and MOV are consistent in both iconic and DMM signs, the percentage of iconic handshapes in DMM signs is significantly lower than in iconic signs. This may be because iconic signs have a concrete referent; therefore, there is some aspect for HS to encode, whereas many of the DMMs have abstract referents and therefore no discrete form upon which HS may capitalize. Those DMM signs with an iconic HS are primarily country signs

Location, Location, Location | 449 Table 5. Chi Square Results for Iconic Signs with One Iconic Parameter Test 1

Test 2

parameters LOC HS MOV LOC HS actual distribution 25 23 11 25 23 expected 19.6 19.6 19.6 24 24 distribution p value 0.0536 0.7728

Test 3 LOC 25 18

MOV 11 18

0.0196

Test 4 HS 23 17

MOV 11 17

0.0396

in which the image associated with the referent is a concrete object such as a flag symbol or traditional clothing. What is even more interesting and unexpected is the distribution of parameters in signs where only one parameter is conveying iconic information (see table 5). If all three parameters (HS, LOC, MOV) are equal in their representation of iconicity, then we expect by chance that they would be equally distributed among the iconic signs. However, this is not the case. Once again, LOC is consistently robust across categories, but HS and MOV are not. Using a corrected alpha of .0125, we observe that the distribution of parameters for iconic signs shows a nonsignificant trend toward an unequal distribution (p = .0535), with Test 2 and Test 3 indicating that LOC and MOV (p = .0196) and HS and MOV (p = .036) contribute to this unequal distribution. As already noted, one must use caution in this interpretation because, as determined by the conservative Bonferroni correction, these comparisons are not statistically significant. More robust differences are observed in the distribution of parameters for the DMM signs (p AL = ANGLED LEFT 5 AR = ANGLED RIGHT 5~ 4 B (+) = REDUPLICATED MOV B> STR-T = STRAIGHT TO LOC S STR-A = AWAY FROM LOC 1 ARC = ARC NO CONTACT 1> CIR = CIRCULAR X ALT-C = ALTERNATING CIRCULAR F STR = STRAIGHT NO CONTACT L ALT-S = ALTERNATING STRAIGHT 7 = 7-SHAPED MVMT L~ BR = BRUSH G TW = TWIST U WIG = INTERNAL WIGGLE V HSC = HANDSHAPE CHANGE V~ SQU = SQUIGGLE 3 WAG = BACK AND FORTH AT WRIST 3~ ARC-T = ARC TOWARD LOC 8 ARC-A = ARC AWAY FROM LOC o8 FLEX = SLIGHT BENDING/FLEXING A CROSS = DUAL CROSSING MOVEMENT I TAP = REDUPLICATED TAPPING AT LOC Y R 5‘ BE = BEHIND EAR D BP = BICEP E BSN = BOTTOM SIDE NOSE K

Sample c o money gray grapes talk b know s 1 corn x f l watch g u v vampire 3 selfish 8 brilliant a i y r baby d e k continued

460 | S ign L ang uag e Studi e s Coding Key LOC

( ‘ ) = THUMB ACCENT CK = CHEEK HS CKB = BOTTOM OF CHEEK CKT = TOP OF CHEEK CN = CHIN CNB = BOTTOM OF CHIN CNS = CHIN SIDE ER = EAR ERB = BOTTOM OF EAR ERT = TOP OF EAR FF = FRONT OF FACE FH = FOREHEAD FHB = FOREHEAD BOTTOM FHT = TOP OF FOREHEAD FM = FRONT OF MOUTH HNC = HIGH NEUTRAL CENTER HNS = HIGH NEUTRAL SIDE HRT = HEART IELB = INSIDE ELBOW IFA = INSIDE FOREARM JAW = JAW LG = LEG LGB = LEG BOTTOM LGS = LEG SIDE LGT = LEG TOP LNC = LOW NEUTRAL CENTER LNS = LOW NEUTRAL SIDE MO = MOUTH MOB = BOTTOM OF MOUTH MOS = SIDE OF MOUTH NC = NEUTRAL CENTER NCBH = NEUTRAL CENTER BACK HAND NCC = NEUTRAL CENTER CONTAINER NCI = NEUTRAL CENTER INDEX NCKT = NEAR TOP OF CHEEK NCP = NEUTRAL CENTER PALM NCW = NEUTRAL CENTER WRIST NFH = NEAR FOREHEAD NFHS = NEAR FOREHEAD SIDE NK = NECK NKS = SIDE OF NECK

M T W BULL HL 7 N G> X>

m t w alcohol airplane 7 n print fishing

continued

Location, Location, Location | 461 Coding Key LOC

( ‘ ) = THUMB ACCENT NOS = NOSE NS = NEUTRAL SIDE NSB = TIP OF NOSE OELB = OUTSIDE ELBOW OFA = OUTSIDE FOREARM OH = OVERHEAD SF = SIDE OF FACE SFH = SIDE OF FOREHEAD SH = SIDE OF HEAD SHO = SHOULDER TAP SHT = TIP OF SHOULDER SRT = SIDE OF TRUNK ST = STERNUM STRB = SIDE BOTTOM OF TRUNK STRT = SIDE OF TRUNK TOP STS = STERNUM SIDES TEM = TEMPLE TH = TOP OF HEAD TM = TOP OF MOUTH TMS = TOP SIDE MOUTH TR = TRUNK TRB = BOTTOM OF TRUNK TRT = TOP OF TRUNK TSN = TOP SIDE NOSE UA = UPPER ARM NCPR = neutral center palm radial NCPU = neutral center palm ulnar