Lexicalization in Signed Languages: When is an Idiom not an Idiom?

Transcription

1 Lexicalization in Signed Languages: When is an Idiom not an Idiom? TREVOR JOHNSTON Macquarie University LINDSAY FERRARA Macquarie University Copyright 2012 Selected Papers from UK-CLA Meetings Vol 1: The internal and often iconic structure of the signs in a signed language (SL) the rough equivalent of words in spoken languages present interesting challenges for linguistic theory. Specifically, lexicalisation in SLs occurs when a signed unit acquires a clearly identifiable and replicable citation form which is regularly and strongly associated with a meaning. The meaning is unpredictable and/or more specific than that implied by the value (meaning) that each component may contribute to the overall form of the signed unit. SL linguists continue to struggle with the dual nature of sign components they appear to be simultaneously phonemes and morphemes, depending on one s perspective. Cognitive linguistics and construction grammars offer a principled analysis of this situation. It holds that the use of linguistic symbols in patterned ways involves constructions that can be differentiated along two continua: one of size or simplicity (from atomic to complex), and one of lexical specificity (from substantive to schematic or abstract). Idioms, famously, are the most obvious manifestation of this lexico-grammatical continuum. In this presentation, we show how lexicalization in SLs can be best understood as a type of idiomaticity: fully-lexicalized signs (atomic and substantive) are idiomatic in SLs in much the same way as multi-word constructions (complex and substantive) can be idiomatic in spoken languages. Lexical signs are in a sense idioms: the components in a lexical construction do not (just) mean what they should mean based on its components. In SLs there thus appears to be a large role for idiomaticity at the level of lexical constructions, and very little use of idiomaticity on the level of multi-sign (multi-word) constructions, unlike spoken languages. The lexico-grammatical continuum exists in all languages, it is just that idiomaticity occurs at a different level in this continuum in SLs. Keywords: idioms, sign language, lexicalization, constructions, Australian sign language (Auslan) 1. Introduction An idiom in any language is an expression whose meaning is partially unpredictable because of regular and idiosyncratic properties. In the words of

2 Selected Papers from UK-CLA Meetings 230 Fillmore et al. (1988) an idiomatic expression or construction is something a language user could fail to know while knowing everything else in the language (p. 504). It has been suggested by Johnston and Schembri (2010) that lexical signs in signed languages (SLs) are idiomatic expressions in much the same way as some multi-word constructions are idiomatic expressions in spoken languages (SpLs). This paper explores Johnston and Schembri s claim in more detail and offers supporting evidence of why idiomaticity surfaces at the lexical level in SLs but not in SpLs. We show here, using a cognitive grammar approach, that idiomaticity in SLs occurs when partly-lexical depicting signs become fully-lexicalized. In other words, when signs that have gestural and iconic components become conventionalized within a community, they acquire an idiomatic reading. First we discuss the types and characteristics of symbolic units along the lexico-grammar continuum in SpLs and SLs. Then we describe how partlylexical depicting signs integrate gestural and linguistic components through real-space blending. The meanings of these signs are more or less componential and analyzable as part of a particular usage event. However, we go on to argue that as these signs become lexicalized their meaning becomes less predictable and thus idiomatic. We conclude with a discussion of how this relates to the sources of linguistic units and processes of meaning construction in Auslan and other SLs. 2. Symbolic Units along the Lexico-grammar Continuum A basic tenet of cognitive grammar posits that the lexicon forms a continuum with grammar and syntax to represent the progressively more complex and abstract symbolic units that exist within a language (Langacker 2002: 102). These symbolic units are bipolar structures that link a phonological pole and semantic pole via a symbolic association. An example of a symbolic unit in Auslan is illustrated in Figure 1; it is comprised of the semantic unit [AUSTRALIA] linked to the phonological unit [australia]. i Figure 1. The Auslan lexical unit [[AUSTRALIA]/[australia]]

3 231 Johnston and Ferrara Symbolic units can be characterized according to two scales: one of content and one of size (Croft and Cruse 2004; Langacker 2005). The content scale ranges from substantive to schematic and characterizes both phonological and semantic units. Substantive units are concrete and specified while schematic units are abstract and less specified. The size scale ranges from atomic (i.e., singular, small) to complex (i.e., multiple, large) and can also be used to characterize phonological and semantic units. In this way, distinctions can thus be made between simple expressions (in English) like the word [cockatoo] and more complex structures, such as the passive construction [SBJ be-tns VERB-en by OBL]. ii The Auslan lexical unit [[AUSTRALIA]/[australia]] is both atomic and substantive phonologically and semantically. That is, this sign is concrete and fully specified for content and it also cannot be further broken down into smaller phonological or semantic units. Contrast this with the Auslan lexical unit [[TOMATO]/[tomato]] that is phonologically and semantically more complex in size but still substantive in content (Figure 2). Figure 2. The Auslan lexical unit [[TOMATO]/[tomato]] In this sign, the complexity in size results from compounding; this sign derives from the integration of the two Auslan lexical units [[RED]/[red]] and [[BALL]/[ball]]. Complexity also results from lexical units being integrated into larger units such as phrases and clauses. In cognitive and construction grammars, a symbolic unit is therefore any pairing of a form with a meaning, while a construction is a symbolic unit that is itself made up of two or more smaller symbolic units. Strictly speaking therefore a symbolic unit which is not composed of any smaller recognizable symbolic units (e.g., a simple word or morpheme) is not also a construction. Thus in most of the literature construction is only used to refer to units which

4 Selected Papers from UK-CLA Meetings 232 are at the complex end of the lexico-grammar continuum. Nonetheless, in some accounts of construction grammar a minimal symbolic unit, such as a morpheme, is sometimes also described as a construction, e.g. Croft and Cruse (2004: 259). Croft and Cruse (2004) discuss how different constructions with different values of size and content can be mapped onto traditional linguistic labels. For example, constructions that are complex and mostly schematic in size and content (e.g. the passive construction cited above) are traditionally considered a part of syntax, whereas constructions that are complex in size and mostly substantive in content (e.g. the [kick-tns the bucket] construction in English) are traditionally treated as idioms. Croft and Cruse (2004: 255) attempt to exemplify the lexico-grammar continuum in tabular form in one vertical dimension (Table 1). Table 1. Examples of lexico-grammar continuum associated with traditional grammatical labels (adapted from Croft and Cruse 2004) These observations on the lexico-grammar continuum are represented by Langacker (2005) using two axes, one for each of the scales of schematicity and complexity, in a two-dimensional graph (Figure 3). As seen by this representation, what is traditionally known as the lexicon occupies the atomic end of the content scale but can range from simple to complex on the size scale. Grammar, on the other hand, is made up of constructions that fall on the schematic end of the content scale but also range from atomic to complex on the size scale (Langacker ibid.). In Figure 4 we map Croft and Cruse s examples and their matching traditional grammatical labels from Table 1 on to an adapted form of Langacker s two-dimensional scale of content and size.

5 233 Johnston and Ferrara Figure 3. Construction types along the size and content scales (reproduced from Langacker 2005) Figure 4. Example English constructions mapped onto the scales of size and content

6 Selected Papers from UK-CLA Meetings 234 Important for the discussion to follow is the part of the graph comprised of constructions that are more substantive than schematic and more complex than atomic, i.e. idioms. Idioms, as stated previously, are traditionally described as having both regular and idiosyncratic features. For this reason, idioms are relegated to the lexicon in generative grammar frameworks. However, within a cognitive grammar framework, idioms are not descriptively or analytically problematic because they simply represent different types of constructions that vary along the content and size scales of the lexicogrammar continuum (Langacker 1987). Furthermore, in this framework there is no assumption of componentiality and thus the unpredictable semantic structure of these composite structures is also unproblematic. For example, the English construction and idiom [kick-tns the bucket] is complex and mostly substantive except for the unspecified tense of the verb kick. Other English idioms may be fully substantive, e.g. [by and large] and [all of a sudden]. Still other idioms may be more schematic, e.g. [the X-er the Y-er] or [X let alone Y]. We now replace the English constructions with Auslan constructions on the graph of content and size (see Figure 5). Similar to the mapping of the English constructions, the more atomic and substantive Auslan units are placed in the area associated with the lexicon. This includes signs for Australia, green, lock, paper, and tomato. As constructions become more schematic and complex, they move towards the realm of traditional morphology and then syntax. Of interest here is the lack of constructions in Auslan that are complex yet more or less substantive. In other words, there seems to be virtually no multi-sign idioms in Auslan. This also appears to be the case in all known SLs. One of the very rare examples of a potential multi-sign idiom in Auslan (TRAIN GO SORRY the train has left, sorry or its variant PRO-2 MISS TRAIN you missed the train ) is actually borrowed from American Sign Language (ASL). The ASL idiom, in turn, appears to be a modified translation calque of the English to miss the boat. The ASL/Auslan idiom, like its apparently related English idiom, also means to miss out on something by being too late but is used in a more restricted way; namely it is used as response to a request for clarification about an on-going conversation from a new entrant to that conversation to mean you are too late and I am not inclined to repeat or recontextualize. We can see that this phrase has non-compositional meaning, because the overall meaning cannot be based on the signs TRAIN, GO, SORRY or MISS in combination alone. In the rest of this paper we use a cognitive grammar framework to examine if there is a place for idioms, or idiomaticity, in SLs nonetheless. However, before proceeding, we first need to clarify the status of the signs that can be found in any SL. Of course, comprehensive and detailed discussion of the question of sign types in SLs is beyond the scope of this paper. In the following section, we simply wish to summarise the main points as presented in Johnston and Schembri (1999, 2010) and Johnston (2010) regarding lexicalization in SLs insofar as they are needed to understand the ensuing discussion of lexical idioms (see Janzen, 2012, for a related discussion).

7 235 Johnston and Ferrara Figure 5. Example Auslan constructions mapped onto the scales of size and content 3. Sign Types in SLs: Partly- and Fully-lexicalized Signs We need to know if all signs are conventional symbolic units and, if so, if they are all conventional in the same way or to the same degree. In other words, we need clear a notion of what constitutes a token and the types that each may instantiate. On examination it is clear that the signs uttered when communicating in a SL are not all of the same type. Unproblematically, from one point of view, the atomic conventionalised units of a SL can be divided into the two broad classes, just as they can in SpLs: an open class of content (or lexical) signs/words and a closed class of function (or grammatical) signs/words. Discovering the language-specific conventionalized semantic content of these forms is an empirical question, requiring detailed textual analysis and fieldwork. From another point of view, however, there is a further sign/wordlevel distinction that needs to be made for SLs: a distinction between fullylexical signs and partly-lexical signs. In this second-order distinction, a fully-lexical sign is roughly equivalent to the commonsense notion of word generally used to refer to the atomic conventionalized form/meaning pairings found in a language (i.e. the free

8 Selected Papers from UK-CLA Meetings 236 morphemes). Of course, a fully-lexical sign may be either a content sign or a function sign. Fully-lexical signs constitute the listable lexicon of a SL in the strictest sense of the word. Partly-lexical signs, on the other hand, are signs that, though conventionalized at the level of the form and meaning of some of their formational features, do not have associated with them in any usage event a meaning that is additional to or unpredictable from the value of those components given that context. Lexicalization in SLs essentially occurs when a signed symbolic unit acquires a clearly identifiable and replicable citation form which is regularly and strongly associated with a meaning which is more specific than the sign s componential meaning potential, even when cited out of context; cannot be predicted based on these components alone; or is quite unrelated to its componential meaning potential, i.e., it may be arbitrary. Of course, this does not preclude the possibility that a sign may also be lexicalized instantaneously through the assertion, and acceptance, of any form/meaning link within a linguistic community. Finally, there exists a third category of sign type non-lexical signs. This includes gestures (both manual and non-manual) and other communicative bodily acts with are the least conventionalized in terms of form and meaning. They rely on context to be construed as signs in the first instance, let alone to be correctly interpreted. 4. Where have all the Idioms Gone? Given the recognized lack of traditional idioms in SLs, we may well ask if idiomaticity plays a significant role in SLs at all. From a construction grammar point of view and in the light of the distinction between partlylexical and fully-lexical signs just discussed, it would appear that idiomaticity does indeed have a significant role to play in the lexico-grammar of SLs because, as we explain in this section, idioms have not disappeared in SLs; rather, idiomaticity manifests itself in a different area of the language design space than for SpLs, as mapped by Langacker in Figure 3. While we explained above that SLs rarely have idioms in the traditional sense, the position we advance in this paper is that idiomaticity is actually found at the lexical level in SLs. Therefore, following Johnston and Schembri (2010) we propose that fully-lexical signs should be analysed as idioms or, better, lexical idioms in order to distinguish them from traditional multi-word idioms: [ ] there is a sense in which fully-lexicalised signs (constructions which are atomic and substantive) are idiomatic in SLs in much the same way as multiword idioms (constructions which are complex and substantive) can be idiomatic in SpLs and rarely in SLs [ ]. In other words, given the meaningfulness and/or iconicity of sign components, the idiosyncratic meaning of a fully-lexical sign suggests that lexicalization in SLs is similarly idiomatic. Lexical signs are in a sense idioms. (p ) To do this, we first suggest that partly-lexical signs in Auslan (and other SLs) are transparently complex structures. Then we describe how these signs may

9 237 Johnston and Ferrara become fully-lexical and that the resulting semantic structures of these signs are not straightforwardly predictable from component contributions. Next we describe how all fully-lexical signs, no matter what the provenance of the lexicalization in each case, have essentially two faces : the one as a unitary symbolic unit, the other as a construction in its own right. In other words, component symbolic units are latent in fully-lexical signs if lexicalized from partly-lexical signs, or they can be read into the sign s formational components despite the fact that this may bear no relation whatsoever to its overall symbolic value as a fully-lexical sign. 4.1 Partly-lexical Signs are Complex Structures In SpLs, the complexity of idiomatic expressions is manifested as a sequence of multiple linguistic units. For instance, the complex symbolic unit and idiom [paint the town red] is the integration of four atomic units in a particular order. If this assumption about complexity is extended to SLs, then as we have already noted most SLs have very few or no idioms in the traditional sense. However, the position advanced in this paper is that signs of SLs are different from words of SpLs, because they can contain complexity that stems from gestural or iconic sources. Such complexity is sub-atomic in size and involves the arrangement of bundles of formational/articulatory features (e.g. hand configuration, orientation, location, timing sequence, and other nonmanual features) both simultaneously and sequentially across one sign. Thus, what appear to be minimal symbolic units in SLs are actually themselves constructions because they, too, are composed of identifiable symbolic units that link forms to meanings. As we have seen, some are partly-lexical and some are fully-lexical. To accommodate this type of sign internal complexity, the scales of content and size along the lexico-grammar continuum need to be adjusted (Figure 6). Figure 6. Adjusted scales for size and content to accommodate the sub-atomic components of signs

10 Selected Papers from UK-CLA Meetings 238 We adopt Liddell s (2003a) model of real space blending to explain how parts of a sign may be meaningful and integrated to form a single complex sign. Real space blending is a type of mental space blending, or conceptual integration, that involves projecting elements from input mental spaces and elements from the signer s conceptualization of their immediate physical environment (real space), into another mental space, the blend, where parts of the signer, the space around the signer, and the signs themselves can be conceptualized as something else (Liddell 1995, 2003a, 2003b). iii Liddell (2003a) proposes the term depicting verbs for one group of blended signs and describes them as having lexically fixed features combined with additional meaningful, gradient aspects of form (p. 269). We suggest here that it is this group of signs that gives rise to lexical idioms. To present an example, the blend structure of a sign that depicts an instance of a person locking a door with a key is illustrated in Figure 7. Figure 7. Real space blend structure of to turn a small object against a flat surface The signer s right hand depicts a small object (or a person holding a small object) while the signer s left hand depicts a portion of a flat vertical surface, which prompts a conceptualization, via mappings between the input mental space, real space, and the blend, of a person locking a door with a key. iv While the same predication would require a complex expression involving multiple linguistic units in a SpL (consider the English expression [a person locks a door with a key]), a SL like Auslan can partially demonstrate and encode the same meaning with a single sign. Our view that depicting signs are complex stems from this ability to depict multiple blended elements and the relations between them. The description of partly-lexical signs given in section 3 above also applies to the signs Liddell (2003a) calls depicting verbs, because the components acquire meaning through correspondences in a real space blend. Extending Langacker s (1987) position on novel complex utterances in SpLs to partly-

11 239 Johnston and Ferrara lexical depicting signs, he explains that they are necessarily fully analyzable, because the speaker, in putting it together, must attend specifically to each component and its contribution to the desired composite sense (p. 297). This means that the semantic structure of a partly-lexical sign reflects each of the sub-atomic components in addition to the complete blend structure. The blend structure is an essential component of a partly-lexical sign s semantic structure, because it provides instructions for how correspondences are to be made between parts of the sign in real space to elements in the other input mental spaces. For instance, the form of the sign in Figure 7 could, depending the mental space input, depict a person unlocking a door with a key, screwing a nail into a wall, a person twisting a nail out of a bookcase, or turning a knob to turn on/off a light or appliance. We explain above that the meaning constructed by a partly-lexical depicting sign arises through the integration of linguistic and gestural components as they correspond to elements in the input mental spaces. Generally, linguistic status is given to the handshapes of these signs as they conventionally are used to depict certain types of things (Brennan 1992; Engberg-Pedersen 1993; Johnston 1989; Johnston and Schembri 1999; Schembri 1996). For example, in the (Auslan) sign above, the configuration of the right hand is one used to depict small objects (or a person holding a small object) while the configuration of the left hand depicts flat surfaces. Thus, we expect the sign s meaning to include a small object and a flat surface. As another example, in Auslan, a hand with index and middle fingers extended downward, palm facing the signer is conventionally used to depict a person s two legs. Depicting signs with this handshape then may be used to depict, for example, a person jumping, walking, laying down, etc. The gestural, or gradient, features of these signs (often related to the location and movement of the sign) also correspond to elements in the blend. They do so via iconic mappings. In the sign from Figure 7, the relation of the hands to each other as well as the movement of the right hand are the gradient features of this sign that map iconically onto the sign s semantic structure. As such, an Auslan signer would expect the spatial arrangement of the hands to iconically represent the arrangement of elements in the mental space input. That is, if the mental space element being depicted is indeed the surface of a door, then it would be unexpected to see a signer position a flat hand with the palm facing downward. In a similar fashion, we expect the timing and orientation of the sign to predicate a relation of turning, as this is what the signer is depicting. If in fact the intended predication involved the flipping of a switch on a wall, instead of the turning of a key in a door, a signer would have to indicate this by changing features to demonstrate that action. In a sign that depicts a person jumping, the general handshape (index and middle fingers spread and extended) and to some extent orientation (palm towards signer) are linguistically determined, although still motivated by iconicity. As the gestural components of the sign, the timing segments/movement, location features, and orientation features are determined by iconic mappings to partially show the predication. In this case, the movement of the hand up and down along with the fingers bending and straightening is meant to (partially) show a person jumping and bending his

12 Selected Papers from UK-CLA Meetings 240 legs at the knees. Because there are conventions for composing and interpreting the components of partly-lexical signs (guided by symbolic and iconic mappings), it is possible to identify a generalized semantic structure that is instantiated during usage events. Again, continuing with our example sign in Figure 7, a general meaning of this sign could be summarized as to turn a small object against a flat surface. This general and literal meaning is a result of more or less straightforward integration of the components of this sign. This meaning is not idiomatic but rather analyzable and predictable. However, in the next section we show that as partly-lexical depicting signs become conventionalized and entrenched within a signing community and thus become fully-lexical signs they exhibit idiomaticity. 4.2 From Partly- to Fully-lexical Signs: Lexicalization through Idiomaticity Langacker (1987, 2002, 2005) discusses lexicalization in cognitive grammar as a dual-process of entrenchment and conventionalization. Entrenchment is related to automatization, or the development of cognitive routines. With repeated use a novel linguistic structure becomes a pre-packaged assembly and thereby achieves what is called unit status. Once a unit, a language user can access and use the structure as a whole without having to attend specifically to its possibly complex internal composition. It is in this way that it constitutes an automatic cognitive routine. When a linguistic structure has reached unit status for a substantial number of language users, then it may be described as conventional linguistic unit and considered a part of a language s lexicon; it has been lexicalized. Johnston and Schembri (1999) explain that any sign can be lexicalized. And the partly-lexical signs just discussed are no exception. In fact, it has been suggested that the lexicalization of depicting signs is an important source for new signs across SLs (Brennan 1992; Johnston and Schembri 2007; Sandler & Lillo-Martin 2006; Schembri 2003; Schick 1990). In Auslan, there are numerous examples of signs that originated as partly-lexical signs. Take for instance the signs shown in Figure 8; all began as real-space depicting blends that over time lexicalized and came to act as a single meaningful units (Johnston and Schembri 2007: 174). It has been noted that once a partly-lexical depicting sign reaches unit status, its meanings that are originally general and compositional (as described in the above section) are backgrounded to those that are more specific and less predictable (Johnston and Schembri 2010). The lexical idiomatic reading is not only foregrounded it becomes the automatic default reading in most usage events. For example, the sign shown in Figure 7 that depicts turning a small object against a flat surface has been lexicalized in Auslan and now has the more specific meaning to lock or a key. And while the lexical meaning of this sign is motivated by its sub-atomic components, it cannot be predicted. It could just have easily been lexicalized with the meanings unlock, a door, or to screw, etc. In this sense it is idiomatic, not merely compositional. The sign

13 241 Johnston and Ferrara TICKET, shown above in Figure 8, provides another illustration of what we mean. Its use as a partly-lexical depicting sign prompts a literal reading of a smallish flat rectangular object that acquires a more specific meaning in context as part of a real-space depicting blend. Over time this sign has become conventionally associated with a set of more specific readings that include a card, a ticket, a receipt, or a sales docket. And even though it may seem rather obvious for this sign to be associated with these meanings, they are not the only possibilities. This sign could have been lexicalized to mean photograph or envelope or some other small rectangular object. And in fact, in ASL, one of the primary readings of this sign is a cheque/check, which also is congruent with the literal reading of this sign. However, in Auslan there is another sign that means a cheque, just as in ASL there is another sign meaning a ticket. Figure 8. Examples of lexicalized signs that originated as partly-lexical depicting signs (reproduced from Johnston and Schembri 2007) Another way the literal and general meanings of partly-lexical depicting signs become more specific and less predictable is through metaphorical or metonymical extension (Taub 2001; Janzen 2012). Take for example the Auslan sign THRILLED. This sign originated as a partly-lexical sign depicting a person jumping up and down. Then as it became lexicalized, the depiction of a person jumping up and down became associated with jumping up and down with excitement leading to one of its primary lexical readings to be thrilled or delighted. In other instances the same literal (partly-lexical) meaning of a form found in two or more unrelated sign languages has the same or identical idiomatic (fully-lexical) meaning (e.g., MEET in ASL and Auslan, see Figure 8). In both languages, however, the lexical idiom could have been agree, couple, partner, club or any number of non-predictable but metaphorically or metonymically related plausible meanings. Indeed, metonymic associations can easily produce a lexicalization that appears (and to all intents and purposes is) completely arbitrary to users of the language. For example, a depiction of meet/bump-into standing for the place I met you could become the fully-lexical sign the name of the city (in which is found the place I met you), which in turn could become the name of the country (in which is found the city in which is found the place I met you). The steps in the metonymic chain can be lost leaving a fully-lexical sign which is completely opaque and arbitrary. This is indeed what happened with the Auslan sign MELBOURNE which is the old sign for town.

14 Selected Papers from UK-CLA Meetings 242 Above we presented traditional multi-word idiomatic expressions as complex symbolic units whose meanings are unpredictable from their components. Then we justified the existence of lexical idioms in Auslan by explaining that partly-lexical depicting signs have a meaning which is predictable from their components (symbolic units) and the usage environment. If and when they become fully-lexical signs they acquire a default meaning which is not entirely predictable. It may even be arbitrary. In the following section, we address how signers can prompt either the componential meaning or the lexical meaning of these signs. 4.3 The Two Faces of Fully-lexical Signs (Lexical Idioms) From Fully- to Partly-lexical Signs: De-lexicalization In cognitive grammar, semantic structure is conceived of as encyclopedic and involves a structured network of related senses (Evans & Green 2006, Ch. 5). This view entails that the semantic structure of complex expressions includes all component structures, the relations between components and the resulting composite structure (Langacker 1987). Accordingly, the semantic structure of a sign which is a lexical idiom will contain an entrenched, schematized blend structure and related componential meanings alongside the meanings specific to the lexical unit as a whole. What is interesting here is that these signs never lose their componential structure; the componential structure just becomes backgrounded to the lexical associations. Thus, the components of a partly-lexical depicting sign are still retrievable if necessary after lexicalization. When needed a signer can at any time, using Johnston and Schembri s (1999) term, de-lexify (or delexicalize) the sign in context. De-lexicalization is a process of re-activating a sign s underlying component blend structure. In Figure 9 this process is represented superimposed on to the scales of size and content in the lexicogrammar continuum discussed above. Figure 9. The cycle of lexicalization and de-lexicalization

15 243 Johnston and Ferrara Consequently, the sign can be used by a signer to prompt either the componential, token meaning that arises in a usage event as part of a realspace blend or the idiomatic lexical meaning that is now conventionally associated with the sign. Janzen (2012) speaks of signs having a continuum of articulation possibilities from fully compositional to fully lexicalized and non-productive. We suggest here that these dual meanings, literal and lexical, exist because the semantic structure of a lexical idiom never loses its original blend structure. The different meanings simply relate to the salience afforded the blend structure in a given usage event. The lexical meaning will background the original blend structure while profiling the less predictable structure (e.g., the meaning of the sign LOCK as to lock or a key ). However, the blend can be (re)-activated when needed by the signer and will represent the use of the sign as a token symbolization of a specific, context-dependent conceptualization, i.e., a novel partly-lexical depicting sign. Figure 10 shows this type of reactivation. Figure 10. Illustration of a lexical sign's blend structure re-activated Thus, we suggest this sign signals a token depiction of a door being unlocked (if we assume that the reactivation of the blend involves similar inputs to the lexicalized sign) or, say, someone screwing a fastener into a wall, if the mental space inputs are different. In the latter case, the sign is not a result of reactivating the blend structure specific to the sign LOCK but rather a more general real space blend that has not yet developed into a lexicalized sign and just happens to share the same formational properties. In summary, partly-lexical depicting signs are composite structures and as such they are componential and analyzable. Different bundles of articulatory features are mapped onto mental space elements; some of these features are

16 Selected Papers from UK-CLA Meetings 244 regularly used for certain types of elements, the lexical part of the sign, and some are gestural/gradient features which map iconically. However as these partly-lexical signs become fully-lexical signs they are perceived as unitary structures rather than composite structures. Additionally, instances of these signs no longer represent mere tokens but now represent tokens of types (the lemma). Langacker (1987) speaks of this transition and the effect this process has on the meaning of composite structures in English: A novel expression is necessarily fully analyzable, because the speaker, in putting it together, must attend specifically to each component and its contribution to the desired composite sense. Once a complex expressions reaches unit status, however, there is at least the potential that its composite structure itself now a unit may be activated autonomously (i.e. without the supporting activation of the component structures). Certainly this does not happen immediately; the component structures and composite structures may continue to be co-activated indefinitely. It is nevertheless the composite structure that furnishes the primary content of an expression, the role of the component structures being subsidiary and dispensable. With continued use, then, a fixed expression will tend to gravitate towards the negative pole of the analyzability scale (though local reversals are possible), until its compositionality fades away entirely (p. 297). In contradistinction to the phenomenon in English, we maintain that the compositionality of signs never fades away entirely, even in fully-lexical signs. This link is maintained, we suggest, mostly through iconic mappings that motivate both the lexical and gestural inputs to these signs Fully-lexical Signs are Complex Structures too As described above, lexical idioms begin as partly-lexical depicting signs which then become idiomatic with lexicalization. However, not all signs originate as real space blends and, in these fully-lexical signs, there is no underlying blend structure that is associated with the lexical meaning of the sign. In normal usage events they are un-analyzable at a sub-atomic level. As a result, their meaning would not be seen as non-componential but rather simply the associated semantic pole of a lexical unit. For example, the sign JOKE did not arise as part of a real space blend (Figure 11). With respect to its conventional lexical meaning, the formational features cannot normally be divided into meaningful components, gestural or otherwise, like the sign LOCK can. That is, the bundle of features that includes the hand with thumb, index, and middle finger extended, the orientation of the palm to the side, the brushing movement just under the chin with the thumb, and the double arc movement away from the signer, do not together contribute (via correspondences in a real space blend) to the semantic structure of JOKE. They simply serve an articulatory function and constitute the phonological pole of this symbolic unit.

17 245 Johnston and Ferrara Figure 11. JOKE Figure 12. SISTER However, what distinguishes even these fully-lexical signs from atomic monomorphemic symbolic units in SpLs is that just like traditional multi-word idioms of SpLs they also have a possible literal interpretation, by which we mean a meaning that is predictable from, and congruent with, the sign s component symbolic units. Consider the English idiom [he kicked the bucket]. It has a default idiomatic reading (to die), but it also has a latent literal meaning that simply requires a suitable context to activate: He kicked the ladder, he kicked dog, and then he kicked the bucket. He was furious. Similarly, fully-lexical signs are constructions, and not simple atomic symbolic units ( mono-morphemic signs ), because the formational components of the sign are each potential symbolic units in their own right, even if none can be independently articulated. Of course, a suitably robust context is needed in order to force a literal interpretation of a fully-lexical sign that is not transparently derived from a partly-lexical sign or has no apparent link between its form and meaning, i.e. it appears to be completely arbitrary as in SISTER (Figure 12). (Recall, however, that every extant sign was probably iconically or gesturally motivated in some way in its original sense, even if semantic shifts through metaphor or metonymy completely obscure this fact in the current sense.) To force the abandonment of the default ( idiomatic ) lexical reading of this type of fullylexical sign to a contextually congruent literal meaning (e.g., SISTER as the depiction DS:TAP-ON-NOSE-WITH-HOOKED-OBJECT) one has to activate the right context of use. Of course, it may be very difficult to construct such a context of re-interpretation for some extremely common fully-lexical signs, but it is always, in principle, possible. Indeed, SL users do experience signs as having two faces as simultaneously unitary symbolic units and constructions in their own right. For instance, offering folk etymologies for almost any sign is quite irresistible for most signers. In other words, though a sign may actually be opaque with respect to iconicity it is arbitrary signers are often willing to hypothesise a gestural/iconic motivation, and hence origin, for almost all signs, even where historical evidence actually shows these folk etymologies to be quite false. This supports that our claim that even fully-lexical signs which have no apparent depicting origin or association can also be de-lexicalized. The latent presence of these symbolic units in fully-lexical signs is thus psychologically real for most signers.

18 Selected Papers from UK-CLA Meetings Conclusion If our description of some lexical signs in SLs as idioms is correct, what does the existence of lexical idioms tell us about the sources of linguistic units and processes of meaning construction found in these languages? We suggest we are able to make non-lexical signs or parts of signs meaningful because we have a cognitive ability to construct real space blends. This ability, it should be noted, is not limited to users of SLs. Users of SpLs also construct real space blends while talking. For example, imagine a person telling a story about her commute to work one day. She says, And this crazy person came out of nowhere. He was speeding and weaving in and out of traffic. Very dangerous. While saying weaving in and out of traffic she also produces a manual gesture a flat hand with fingers extended, palm towards the side, moves forward and left to right a few times. It is apparent that during this usage event, the speaker constructs a real space blend where her hand is the car and the path is the weaving in and out. However, there is a key difference. In SpLs, gestures are never considered a source for new words, but in SLs, they are considered a source for new signs. Wilcox (2004) explains the reason is because gestures and SLs share articulatory and perceptual systems. For example, the gesture described above would probably be produced as a partly-lexical depicting sign by an Auslan user, as there is a conventional handshape for depicting vehicles. The gesture-sign interface that results from a shared modality and an ability to construct real space blends also explains why the signs discussed in this paper exhibit sub-atomic or internal complexity. Signers are adept at integrating multiple lexical and gestural elements within a single sign. The subsequent lexicalization of these signs provides further evidence that gesture is essential to the emergence of conventional symbolic units in SLs. Finally, the fact that it is often difficult to determine whether a sign belongs to the category of fully-, partly-, or non-lexical sign is precisely due to the prominent role gesture/iconicity and real space blending plays in SL grammar. For example, we mentioned how lexical idioms, which are conventional/fully-lexical signs, can be de-lexicalized in context by reactivating component blend structures prompting literal, non-idiomatic readings. The process of de-lexicalization in discourse gives SL users a powerful tool to modify and qualify conventional meaning non-lexically. It also works to re-strengthen connections to the gestural substrate and highlights the tight integration of gesture and language in Auslan, and most likely other SLs. Notes i Each Auslan sign referred to with an English gloss in this paper can be viewed as a video clip by visiting and typing the gloss in the search box. The sign

19 247 Johnston and Ferrara ii iii iv transcription of the phonological unit in Figure 1 is written in HamNoSys (Hamburg Notation System) from the Institute for German Sign Language, Hamburg University, Germany. See Table 1 for an explanation of the abbreviations and format. For an in-depth discussion of blending as a general cognitive process, see Fauconnier & Turner (2002) and Fauconnier (1997). Vertical brackets signal blended entities. References Brennan, M. (1992). The visual world of BSL: An introduction. In D. Brien (ed.), Dictionary of British Sign Language/ English. London and Boston: Faber and Faber. pp Croft, W. and A. Cruse (2004). Cognitive Linguistics. Cambridge: Cambridge University Press. Engberg-Pedersen, E. (1993). Space in Danish Sign Language: The semantics and Morphosyntax of the Use of Space in a Visual Language. Hamburg: Signum Press. Evans, V. and M. Green (2006). Cognitive Linguistics: An Introduction. Edinburgh: Edinburgh University Press. Fauconnier, G. (1997). Mappings in Thought and Language. Cambridge: Cambridge University Press. Fauconnier, G. and M. Turner (2002). The Way we Think: Conceptual Blending and the Mind's Hidden Complexities. New York: Basic Books. Fillmore, C.J., P. Kay and M.C. O Connor (1988). Regularity and idiomaticity in grammatical constructions: The case of let alone. Language 64: Janzen, T. (2012). Lexicalization and grammaticalization. In R. Pfau, M. Steinbach and B. Woll (eds.), Sign Languages. Berlin: Mouton de Gruyter. pp Johnston, T. (1989). Auslan Dictionary: A Dictionary of the Sign Language of the Australian Deaf Community. Sydney: Deafness Resources Australia. Johnston, T. (2010). Degree, not kind: Non-lexicalized points are symbolic indexicals regardless of whether they occur in the composite utterances of spoken languages or signed languages. 4th Conference of the International Society for Gesture Studies (SGS). European University Viadrina Frankfurt/Oder. Johnston, T. and A. Schembri (1999). On defining lexeme in a signed language. Sign Language and Linguistics 2: Johnston, T. and A. Schembri (2007). Australian Sign Language: An Introduction to Sign Language Linguistics. Cambridge: Cambridge University Press. Johnston, T. and A. Schembri (2010). Variation, lexicalization and grammaticalization in signed languages. Langage et société 131: Langacker, R.W. (1987). Foundations of Cognitive Grammar. Volume 1: Theoretical Prerequisites. Stanford: Stanford University Press.

20 Selected Papers from UK-CLA Meetings 248 Langacker, R.W. (2002). Concept, Image, and Symbol: The Cognitive Basis of Grammar. New York: Mouton de Gruyter.. Langacker, R.W. (2005). Construction Grammars: Cognitive, radical, and less so. In F.J.R.D.M. Ibanez and M.S.P. Cervel (eds.), Cognitive Linguistics: Internal Dynamics and Interdisciplinary Interaction. Berlin: Mouton de Gruyter. pp Liddell, S. (1995). Real, surrogate, and token space: Grammatical consequences in ASL. In K. Emmorey and J. Reilly (eds.), Language, Gesture, and Space. Hillsdale, NJ: Lawrence Erlbaum Associates. pp Liddell, S. (2003a). Grammar, Gesture, and Meaning in American Sign Language. New York: Cambridge University Press. Liddell, S. (2003b). Sources of meaning in ASL classifier predicates. In K. Emmorey (ed.), Perspectives on Classifier Constructions. Mahwah, NJ: Lawrence Erlbaum Associates. pp Sandler, W. and D. Lillo-Martin (2006). Sign Language and Linguistic Universals. New York: Cambridge University Press. Schembri, A. (1996). The Structure and Formation of Signs in Auslan (Australian Sign Language). Sydney: North Rocks. Schembri, A. (2003). Rethinking 'classifiers' in signed language. In K. Emmorey (ed.), Perspectives on Classifier Constructions in Sign Languages. Mahwah, NH: Lawrence Erlbaum Associates. pp Schick, B.S. (1990). Classifier predicates in American Sign Language. International Journal of Sign Linguistics 1: Wilcox, S. (2004). Gesture and language: Cross-linguistic and historical data from signed languages. Gesture 4: