On the Conventionalization of Mouth Actions in Australian Sign Language

Transcription

1 569334LAS / Language and SpeechJohnston et al. research-article2015 Article On the Conventionalization of Mouth Actions in Australian Sign Language Language and Speech Language and Speech 2016, Vol. 59(1) 3 42 The Author(s) 2015 Reprints and permissions: sagepub.co.uk/journalspermissions.nav DOI: / las.sagepub.com Trevor Johnston and Jane van Roekel Macquarie University, Australia Adam Schembri La Trobe University, Australia Abstract This study investigates the conventionalization of mouth actions in Australian Sign Language. Signed languages were once thought of as simply manual languages because the hands produce the signs which individually and in groups are the symbolic units most easily equated with the words, phrases and clauses of spoken languages. However, it has long been acknowledged that non-manual activity, such as movements of the body, head and the face play a very important role. In this context, mouth actions that occur while communicating in signed languages have posed a number of questions for linguists: are the silent mouthings of spoken language words simply borrowings from the respective majority community spoken language(s)? Are those mouth actions that are not silent mouthings of spoken words conventionalized linguistic units proper to each signed language, culturally linked semi-conventional gestural units shared by signers with members of the majority speaking community, or even gestures and expressions common to all humans? We use a corpus-based approach to gather evidence of the extent of the use of mouth actions in naturalistic Australian Sign Language making comparisons with other signed languages where data is available and the form/meaning pairings that these mouth actions instantiate. Keywords Mouthings, mouth gestures, non-manuals, constructed action, Auslan Corresponding author: Trevor Johnston, Department of Linguistics, Macquarie University, Building C5A, Room 546, North Ryde, Balaclava Road, New South Wales, 2109, Australia. trevor.johnston@mq.edu.au

2 4 Language and Speech 59(1) 1 Introduction In this study, we report on the distribution of different types of mouth actions in Australian Sign Language (Auslan), making comparisons with other signed languages where possible. We describe the characteristics of mouth actions in Auslan and quantify the alignment of the attested form/ meaning pairings found in a recently created linguistic corpus of the language, that is, a dataset which is representative of native and near-native signer usage across a number of text-types, linguistically annotated, and machine readable (searchable). We do this in order to gauge the extent and degree of language-specific conventionalization of these actions. Mouth actions that occur while communicating in signed languages have posed a number of questions for linguists: are the silent mouthings of spoken language words simply borrowings from the respective majority community spoken language(s)? Are those mouth actions that are not silent mouthings of spoken words conventionalized linguistic units proper to each signed language, culturally linked semiconventional gestural units shared by signers with members of the majority speaking community, or even gestures and expressions common to all humans? Our thesis is that, though there are pairings that suggest some conventionalization of usage, much of the contribution mouth actions make to individual signs or signed utterances may not be, unlike what has been claimed in the literature, specific to any particular signed language or even signed languages generally. Rather, many mouth actions appear to be either part of the oral and facial expression repertoires used by all humans moderated by local cultural norms and, for hearing people, constrained by the use of speech or are co-articulated mouth patterns derived from the relevant ambient spoken language. The relevance of this study for our understanding of language is threefold. First, it demonstrates the importance of complementing intuitions and elicited datasets in signed language research with corpus-based usage data: these cast doubt on earlier claims of high degrees of linguistic conventionalization, even obligatoriness, in this area. Second, it underlines the complex relationship between signed and spoken languages in many signing communities: the important role of mouthing must be acknowledged. Third, there is a deep synergy between language in spoken and signed modalities: both use oral and manual gestures. 2 Background Signed languages were once thought of as essentially manual languages because the hands produced the basic symbolic units of these languages (i.e., the signs) and these were readily equated with the words of languages and the phrases and clauses that words are the constituents of with which linguists were already familiar. Very quickly in the linguistic study of signed languages, however, it was realized that non-manual activity which involves movements of the body, head and the face (especially involving the upper face or eyebrows) play an important role in understanding the grammar of these languages (Baker & Padden, 1978; Liddell, 1978). It has also long been known that there are many distinct signed languages. Most deaf signed language using communities around the world use historically unrelated languages that are, for the most part, mutually unintelligible. Despite these differences, signed languages across the world nonetheless appear to have many features in common (Newport & Supalla, 2000; Woll, Sutton- Spence, & Elton, 2001), partly because their visual-spatial modality allows room for meaningful uses of space, simultaneity and, in particular, iconicity (e.g., Cuxac, 1999, 2000; Meir, Padden, Aronoff, & Sandler, 2013; Perniss, Thompson, & Vigliocco, 2010; Taub, 2001), and partly because the sociolinguistic situation of deaf signing communities around the world is similar (e.g., Lucas, 2006; Quinto-Pozos, 2007; Van Herreweghe & Vermeerbergen, 2004). Importantly, the spatial, temporal and visual representational affordances of gesture-based face-to-face languages appear to

3 Johnston et al. 5 have been spontaneously and independently exploited in similar ways time and time again in different signed languages. For example, the use of tilts, shakes and nods of the head, the raising and lowering of the eyebrows, and the widening or narrowing of the aperture of the eyes have been observed repeatedly and their functions in the grammars of signed languages have been described in varying detail with different amounts of empirical support. That is, despite strong claims that some of these are formal syntactic markers of relative clauses, content questions, conditionals and so forth, debate has continued not only on their functions but also on whether they are syntactic or prosodic, and their degree of conventionalization or grammaticalization in particular languages (see Pfau & Quer, 2010). 1 The mouth is a prominent site of non-manual activity and movements of the mouth are an obvious accompaniment to manual signing. The linguistic status of mouth actions in signed languages, like other non-manuals, is also a question of debate (e.g., Vinson, Thompson, Skinner, Fox, & Vigliocco, 2010). 2.1 Previous research Non-manual activity that is centred on the mouth (mouth actions) has been documented for a number of signed languages and appears to be a characteristic of all those studied to date. The least problematic in terms of categorization and identification is the class of mouth actions that are transparently complete or partial silent articulations of the spoken words of the ambient spoken language. The remaining mouth actions have the common characteristic in that they are not related to the articulation of spoken words, yet, at the same time, they appear to fulfil a variety of functions which are not well understood. Since the publication of a major overview of mouth actions in signed languages, edited by Boyes Braem and Sutton-Spence (2001), these two general types have usually been referred to as mouthings and mouth gestures respectively. 2 An early issue of interest was the proportion of signs found to co-occur with mouth actions, especially mouthings, in various signed languages (e.g., Schermer, 1990). This was partly in response to well-known early claims that American Sign Language (ASL) used very limited amounts of mouthing when compared to other signed languages (Baker-Shenk, 1983; Boyes Braem, 2001; Padden, 1980). Another area of debate, besides frequency, concerned the status of mouthings: were they an integral part of a signed language or were they marginal, stemming from language contact or borrowing? Addressing both these questions involved describing what mouthing and mouth gestures did and thus categorizing them into general types. More recent publications by Fontana (2008), Lewin and Schembri (2011), Bank, Crasborn, and van Hout (2011, 2013, submitted), and Sandler (2012) on mouth actions in Italian, British, Netherlands, and Israeli signed languages respectively give a good overview of the research, particularly that since Mouthings Briefly, research has shown that mouthings frequently accompany manual signs in many signed languages, including ASL, 3 and there is some evidence though the datasets have never been very large or varied that the rate varies according to text-type. They occur very frequently with fingerspellings. Mouthings have been shown to occur more with nouns and morphologically simple verbs than with morphologically complex signs such as indicating verbs (also known as agreement and spatial verbs) or with depicting signs (also known as classifier signs), both of which may have aspects of their handshape, movement and/or location modified meaningfully. Mouthing has been shown to add meaning to some signs by indicating a more specific reading of a sign, for example, the Auslan sign spouse with the English mouthings wife or husband (Johnston & Schembri,

4 6 Language and Speech 59(1) 2007). Indeed, for Schermer (2001) in these type of environments mouthing is often obligatory in Netherlands Sign Language (NGT). A mouthing can even add independent semantic information. Vogt-Svendsen (2001) cites the example of the Swedish word for red being mouthed along with the Swedish Sign Language sign pull-over to mean a red pull-over. Although mouthings are often closely temporally aligned with their co-articulated sign, they may be stretched, reduced or repeated to maintain an alignment with the duration and rhythm of the manual sign, especially if the sign has itself been modified (Fontana, 2008). Finally, the mouthing itself may spread regressively or progressively to adjacent signs (Bank et al., 2013; Crasborn, van der Kooij, Waters, Woll, & Mesch, 2008). 2.3 Mouth gestures Mouth gestures are all other communicative mouth actions that are not mouthings. They have been described as mouth actions that do not derive from spoken language (Boyes Braem & Sutton- Spence, 2001). However, this needs qualification. It has yet to be shown that signers and speakers in the same speech community and/or culture do not share any of the same or similar mouth action repertoires. Though some research has been conducted on facial expressions that occur when speaking, as correlates of intonation and stress or audio-visual prosody (Krahmer & Swerts, 2009), little or no research has been conducted on mouth gestures made in face-to-face communication especially at moments when speakers are not actually speaking but perhaps gesturing. Thus the extent to which the mouth gestures are actually sourced from or related to the predominantly spoken language community in which signers are invariably embedded is as yet unknown. Some mouth gestures look like voiceless syllables. These syllabic mouth gestures are usually described using letters, for example, pah, ap, tham, woof, pow. The manual signs they co-occur with have been called multi-channel signs because the mouth gestures add specific semantic components to the sign, for example, Lawson (1983) claimed that pah means success in British Sign Language (BSL) and is obligatorily produced with a sign meaning success, at last or finally. Woll (2001) proposed an alternative theory in which these mouth movements echo hand movements, for example, when a manual sign includes an abrupt closing of the hand, the mouth will parallel this by snapping shut. Woll terms this echo phonology as the mouth echoes the movement of the hand. Under this theory, these mouth gestures are devoid of semantic content but match classes of manual movement such as opening, closing and twisting, for example, pah accompanies signs where the handshape opens or the hands separate (often twisting apart). Other mouth gestures are thought to act as morphemes with an adverbial or adjectival function (Liddell, 1980). The classic example, given in Liddell for ASL and supported by other ASL researchers (McIntire & Reilly, 1988), and identified for other signed languages (Crasborn et al., 2008) including BSL (Lewin & Schembri, 2011) and Auslan (Johnston & Schembri, 2007), is the protruding tongue (often written th) to mean something like carelessly. Other modifying adverbial mouth gestures identified for some signed languages have included ee for effort (showing the teeth in a smile-like action) and mm (similar to a lip pout) for effortlessness. Adjectivals have included puffed-cheeks for large/long things and suckedin-cheeks for small things. Woll (2001) categorized the syllabic type of mouth gestures echoes and this second type as adverbials. She called a third sub-type enactions (henceforth, we will use our preferred term, enactments). In enactments, the mouth gesture is part of the representation by the signer of an event a type of role-playing or mimesis in which the mouth was involved one way or another. In an early study of Auslan (Johnston, 1992), it was observed that some mouth gestures appear to have a prosodic role. The upper face (eyebrows, eyes) and the lower face (mouth, tongue,

5 Johnston et al. 7 cheeks) appeared to make different contributions to visual prosody in Auslan the former analogous to pitch, the second analogous to stress. In other studies, mouthings that spread over two or more signs ( stretched mouthings) have been noted for their stress-like qualities in Swiss German Sign Language (Boyes Braem, 2001) and for their effect of binding constituents together into a prosodic word when spreading from host to clitic (Sandler, 1999c) or into even larger prosodic units such as noun phrases or verb phrases (Boyes Braem, 2001). The prosodic role of some mouth gestures, not just stretched mouthings, will once again be raised here, later in this paper. Since 2001, researchers have continued descriptive work on mouth actions as well as investigating mouth gestures as part of non-manual grammatical signals. 4 For a study of spreading behaviour in mouthings in three signed languages, Crasborn et al. (2008) called mouthings M-type mouth actions; adverbials mouth gestures were called A-type (and now also included adjectival mouth gestures as some modified nouns), and echoes became part of a slightly broader E-type category (for semantically empty following Woll (2001)). Enactments were discriminated into two subtypes those involving only the mouth in which the mouth represents itself doing the action described by the sign, such as bite, laugh or lick (these were called 4-type for mouth for mouth ): and W-type (for whole of face ) in which the mouth is simply part of a large whole-of-face expression, for example, an open mouth with wide eyes for surprise. Crasborn et al. (2008) note that W-type mouth gestures are not part of any potential mouth-based semiotic system because any mouth action form is linked to the whole face and consequently its interpretation is a function of the enactment and not a function of a possible conventional value attributable to the mouth form. 2.4 Gestural versus linguistic More recently, some researchers have focussed on the question of the gestural nature of some mouth actions in that they have the same kind of relationship to the conventional manual signs of a signed language as do manual gestures to the conventional spoken words of a spoken language (Pizzuto, 2003). Fontana (2008), drawing on the work of Kendon (2004, 2008) and McNeill (2000), makes the radical suggestion that all mouth actions can be analysed this way. A similar but more conservative observation is made in Dachkovsky and Sandler (2009) and Sandler (2009) who identify a category of gestural iconic mouth actions to distinguish them from syllabic E-type (or lexical ) mouth components and from the conventional adverbial and adjectival A-type modifiers already identified by other researchers (which they do not question). The iconic category comes from the work of McNeill (1992) who, drawing on the work of earlier 20th century researchers (e.g., Efron, 1972; Kendon, 1980), explored how gesticulation played an important role in both clarifying meaning and structuring discourse in spoken language. He divided gestures into four categories: deictic, iconic, metaphoric and beat gestures. Iconic and metaphoric gestures portray visually some aspect of the referent. Iconic gestures are usually made for concrete objects and events/actions with a representation that physically resembles them (objects typically through an image and events/actions typically through a mimetic enactment). Metaphoric gestures are usually employed to refer to ideas and abstract expressions. Sandler found that some mouth gestures in Israel Sign Language are mimetic replicas (oral mimes) and others act as iconic gestures (symbolic icons), for example, a round bowling ball was represented by puffed cheeks, a zigzag shape by a sideways mouth movement, and the whoosh of an object swung through the air by an exhaled faa. Different signers have varying versions of these iconic mouth gestures so there is no standardized or conventionalized form. These iconic mouth gestures fit in well with McNeill s (1992) classification of gestures versus linguistic forms. It should be remembered that Sandler (2012) agrees with most signed language researchers that, despite their gestural origins, many other mouth gestures and facial signals are

6 8 Language and Speech 59(1) Mouth actions Mouth Mouthings Crasborn et al (2008) W-type 4-type A-type E-type M-type Non-linguistic Linguistic Fontana (2008) Mouth gestures Gestural Mouthings Sandler (2009) Mimetic replicas Gestural Symbolic icons Adverbial/ adjectival Lexical Linguistic Mouthings Figure 1. Three possible categorizations of mouth actions. fully language-dedicated gestures that are an integral part of the lexis and syntax of signed languages, that is, they are linguistic. These three understandings of the linguistic (or gestural) status of mouth actions in signed languages are illustrated in Figure 1. 3 This study We document the distribution of types of mouth actions in Auslan making comparisons with other signed languages where possible and describe their overall characteristics in order to test the language-specific conventionalization of these actions Methodology Fifty video texts were selected from the Auslan corpus for analysis. All signs and all mouth actions were examined and all mouthings and mouth gestures were identified, categorized and annotated using ELAN multi-media annotation software. 6 Annotations were exported into a database spreadsheet for further processing. We conducted a statistical analysis of this data using Rbrul, with a multivariate logistic regression model being used to identify predictors of mouthing and mouth gestures in the data (Johnson, 2009). Like the programme GoldVarb, developed by Rand and Sankoff (1991), Rbrul can quantitatively evaluate the influence of multiple factors on variation. 3.2 Data The data in this study have been drawn from the Auslan corpus of native or near-native signers (for further details, see Johnston & Schembri, 2006). For this study, 50 video clips were selected from the corpus, representing 38 individuals and three text types (monologue, dialogue and elicited). The total duration of the data was five hours and 58 minutes, representing 17,002 manual sign tokens. The signed texts ranged from 1:32 to 38:30 minutes in length. The 50 video clips consisted of 25 monologues (these were narratives, all of which involved retellings of two Aesop s fables); 10 dialogic texts (free conversation or responses to a series of interview questions); and 15 sessions of 40 elicited picture descriptions.

7 Johnston et al. 9 Figure 2. The ELAN tiers used to annotate mouth actions [SSNc2a00:00:29.000]. 3.3 Annotations The 50 texts had been previously annotated according to guidelines detailed in the Auslan Corpus Annotation Guidelines 7 and in Johnston (2010). They were selected because they had previously received the most detailed annotations of all the 459 corpus annotation files, and would thus require the least amount of additional work, besides the study-specific annotations, to complete. That is, the majority had been almost completely annotated for glosses and translations, as well as for clause boundaries and periods of enactment or constructed action (a term for all forms of enactment or mimetic behaviour in signed languages; compare with constructed dialogue from Tannen, 1986), and tagged for the relationships between clauses (e.g., embedded, dependent), the part of speech of clause constituents, and the argument roles of signs within the clause (e.g., agent versus patient, etc.). Part of speech tagging, particularly relevant to this study, cannot be carried out without first identifying clauses. Some files already had mouthings identified. We thus only had to complete the incomplete files before then adding the study-specific annotations: namely, annotations for the state of the mouth as: (a) a mouthing; (b) a mouth gesture (further specified for sub-type as explained below); or (c) a period of no mouth action. Annotations were added to two glossing and two grammatical class tiers. A tier for each hand is required because it is possible to articulate two signs at the same time (one on each hand) so each may need to be glossed separately. When this happens, the grammatical class of each sign/hand may also be different and thus two tiers need to be available. Annotations for mouth gestures were made on the mouth gesture form tier (called MouthGestF), and on the mouth gesture meaning tier (called MouthGestM). The annotations for mouthings were made on the mouthing form tier (called Mouthing) (Figure 2). The annotation schema for mouthings is summarized in Table 1. Mouth gestures were divided into types that were based on Crasborn et al. (2008) but further sub-categorized, as illustrated in Figure 3. We did this because our first attempt to annotate mouth actions using the existing categories appeared inadequate we believed finer discrimination was necessary in order to properly identify and account for each type. For example, A-types were divided into two-subtypes: the adverbials (and adjectivals), as described in Crasborn et al. (2008), and prosodic mouth gestures. We did this because the latter did not appear to modify the meaning of co-occurring signs like other traditional A-types. They appeared to have an emphatic role, not unlike stress in speech. (For possible manual correlates of stress in signed languages generally, see Wilbur s (1999) study of stress in ASL.) Their most important characteristic appears to be a tensed posture of the mouth which is held, even if relatively briefly, without changing dynamically, rather than any precise form of the mouth posture itself, for example, the wide or ee mouth gesture used with a pointing sign, as in the two signs on the left of the bottom row in Table 2. This sub-type was identified so that they could be

8 10 Language and Speech 59(1) Table 1. The annotation schema for mouthings. M-type (mouthing) Annotation Examples Complete articulation complete-word race, rabbit, village, far Initial segment i(nitial) v(illage), sa(me), diff(erent), sh(eep) Medial segment (me)di(al) (no)th(ing), (re)mem(ber), (b)e(st) Final segment (fi)nal (success)ful, (fin)ish, (im)prove. (to)day Initial and final segment only in(i)tial f(ini)sh, d(ea)f, s(uc)cesful suppressed articulation* (suppressed) (lady), (have) unreadable* Unreadable anticipatory spreading mouthing-regr ID gloss: pt:pro1sg explain (regressive mouthing) Mouthing: explain-regr explain delayed spreading (progressive mouthing) mouthing-prog ID gloss: finish pt:pro1sg Mouthing: finish finish-prog * A suppressed mouthing annotation is used in a few instances where the annotators are convinced there is underlying movement congruent with articulating a word associated with a sign, but the mouth does not actually open, for example, the y of lady when signing lady. They are identified to distinguish them from mouth gestures, for example, an ee-like mouth gesture. Where annotators were certain a word was being mouthed there are articulatory motions but were simply unable to lip-read it, it is annotated as unreadable. Mouth actions Mouth Crasborn et al (2008) W-type 4-type A-type E-type M-type This study No action Spontaneous Editorial Constructed action Congruent Adverbial expressive Mouth for mouth Prosodic Adverbial/ adjectival Syllabic/ echo Mouthings Figure 3. Types of mouth actions annotated in this study. compared and contrasted to the traditional category of adverbial mouth gestures: were they really morphemes being added to signs (as the A-type category is meant to capture) or actually part of the prosody? A-type mouth gestures were annotated using a coding schema adapted from Sutton-Spence and Day (2001) in their study of BSL. The schema attempts to capture objectively and consistently major features of mouth gestures (mouth-open/mouth-closed, teeth-visible/teeth-invisible, puffedcheeks/no-puffed-cheeks, tongue/no-tongue, narrow/wide, etc.). The adaptation was adequate for our purposes and had the advantage of allowing for an easy and direct comparison with the BSL data (see Appendix). Some typical examples of the most general mouth gestures are illustrated in Table 2. We usually refer to the mouth gestures in this paper by the short-hand glosses, but the codes indicate the range of related forms covered by a single gloss. Five sub-types of W-type mouth gestures were also identified. First, mouth actions that were part of involuntary or spontaneous expressions (or indexes) of the state of mind of the signer (e.g., amused, confused, concerned) were categorized as spontaneous W-type.These were not coded for the present study because we do not consider them to be candidates for inclusion in

9 Johnston et al. 11 Table 2. Mouth gesture (MG) form codes and glosses used for typical exemplars. blow air moves inwards or outwards through the lips which may be pursed or rounded CN8, CN17, ON16-18 bottom-lip-out bottom lip is pushed forward, out or up CN3, CN20, ON11, ON14 down the corners of the mouth are pulled down, mouth can be open or closed, lips can be pressed together, tense or relaxed CN4, CN22, ON4, ON9, ON15 lip-curl top lip is pulled up on one or both sides, as in a sneer CN1, ON5, ON10 lips-out lips pushed forward, as in a pout or shh CN11-14, CN16, ON16 lips-pressed ( mm ) lips are pressed together but the mouth corners are relaxed CN5, CN6, CN21, CN23, open mouth is open ON1-3 puff puffed cheeks CP1-8 slightly-open mouth is slightly open ON6, ON12 sucked-in cheeks are sucked inwards CN24 tongue ( th ) tongue pokes out or is visibly forward all OT codes & CN19 trill ( brrr ) lips vibrate CN7, CN9-10, CN13-15, CN18, CP5, wide ( eee ) the corners of the mouth are pulled wide, mouth can be open or closed, lips can be pressed together, tense or relaxed CN2, ON7, ON8, ON13, ON14

10 12 Language and Speech 59(1) any semiotic system that is potentially language-specific. Second, expressions or subtle metacomments about what the signer is signing (i.e., the signer s attitude) were categorized as editorial W-type expressions. An example would be the use of lip-curl (Table 2) together with furrowed brow, squinting of the eyes and a backward, recoiling movement of the head while producing he say after signing you fat (i.e., you fat, he say) to convey disapproval of the overly direct and thus inappropriate comment being quoted. They were identified in order to gauge their overall role in Auslan and, especially, how they differed from the A- or E-type mouth gestures. Third, mouth gestures that were part of enactments were categorized as constructed action W-types. They involved the whole face and the meaning was dependent on this context. If the constructed action involved the production of one of the common recurring mouth gestures (as described above), the appropriate gloss for that was also added to the annotation. The fourth subtype was congruent W-types. These were expressions which carry very little modifying communicative load and were, in fact, little more than a congruent default expression, that is, they simply matched the semantics of the lexical sign, such as smiling while signing happy, or looking sombre when signing funeral. These whole face affective expressions often appeared when particular semantically appropriate lexical signs were cited. This phenomenon has led some researchers to claim they are obligatory conventional non-manual components. However, though it would be decidedly odd to sign happy in isolation with a negative or sad expression (just as it would be odd to read out the English headword depression from an English dictionary in a joyous manner), this does not mean the expression is lexically specified merely because of a sympathetic collocation of the congruent expression. Systematic annotation of text quickly revealed that signs of this type can be signed with a neutral expression. We identified these separately in order to quantify and compare them to any similar mouth gestures that do not involve the whole face (if any) and to test if they really are obligatory. The fifth subtype, adverbial expressive W-types, were clearly intended to modify and add meaning to the manual sign(s), unlike congruent W-types, for example, signing dance while performing it in a lively or energetic fashion with an overall facial expression of enjoyment. However, they were not classified along with other A-types because they were not restricted to the mouth and were strongly enacting, that is, one way or another the signer was engaging in constructed action. The annotation schema we used for mouth gestures is summarized in Table The time alignment of mouth action annotations. When a mouth gesture is clearly articulated across 50% of the duration of an adjacent sign, it is annotated separately on each co-articulated sign it spreads over with cont (for continued ). When a mouthing begins before its target sign it is marked as regr (for regressive ) on the sign during which it begins; where a mouthing persists beyond its target sign it is marked as prog (for progressive ). Again, a 50% spreading criteria is applied. (The target of a mouthing can usually be easily identified but mouth gestures need not have an unambiguous single target, so they are simply marked as continued from the sign in which they begin.) Only significant spreading is annotated in our dataset the exact onset and offset time of mouth gestures is not a focus of this study (other studies have explored this; see Bank et al., 2013; Crasborn et al., 2008; Sandler, 1999b). 4 Results 4.1 Overall distribution About one in five signs were not accompanied by any noteworthy mouth activity, which is not unlike other signed languages described to date (Table 4).

11 Johnston et al. 13 Table 3. The annotation schema for mouth gestures. Mouth gesture MouthGestF tier begins with MouthGestM tier contains E-type (echo or empty) syll:gloss (= syllable) various meanings as needed temporary tag tier: -im (imagistic), -mi (mimetic), -me (metaphorical) A-type (modifying) prosodic gloss/code(h) (h = held) meaning glosses: activity, emphasis (see Table 2) prosodic (non-specific) No annotation Tag tier: -mh (mouthing held) adverbial gloss/code (see Table 2) meaning glosses: large-amount, careless, unpleasant, smooth, ease, effort, smallamount temporary tag tier: -im (imagistic), -mi (mimetic), -me (metaphorical) 4-type (mouth for mouth) cmo (= Congruent enactment Mouth Only) W-type (whole-of-face) spontaneous no annotation editorial comment no further annotation or various meanings as needed ca (Constructed Action) ca: (= Constructed Action) no further annotation or various descriptions as needed ca using an A-type ca:gloss/code (Table 2) add after the ca: the A-type mouth gesture gloss/code congruent cwf (= Congruent Whole Face) meaning glosses: expression, enactment, emphasis adverbial expressive ca:adv (= Adverbial) expression Spreading mouth gesture annotation-cont on all subsequent co-articulated manual sign(s) Table 4. Mouth action rates in Auslan compared with other signed languages. Auslan BSL NGT SSL Mouth actions 77% 71% 65% 90% No mouth actions 23% 29% 35% 10% *Data are taken from Crasborn et al. (2008). The Auslan data also show that about half of all manual signs were accompanied by mouthing, and that mouth gestures are relatively infrequent, especially 4-type, A-type and E-type (Figure 4). Approximately 73.6% of all mouth actions (i.e., excluding sign tokens with no mouth action) are, in fact, mouthings, which is higher than in previous descriptions of other signed languages, although similar to the recent corpus-based study of NGT (NGT-2 in Table 5) which also included several text-types. Some of the differences in this data may be due to the finer discrimination of sub-types of the major categories in the Auslan data, particularly the identification and annotation of mouthings. In our experience, mouthings whether complete or incomplete are often not noticed on first viewing,

12 14 Language and Speech 59(1) Figure 4. Rate of mouth actions with manual signs. or are often mistaken for mouth gestures, by researchers for whom Auslan is a second language. They may only know the standard mouthing (see below), or may not realize that some signers actually have larger English vocabularies than they credit them with. They are thus not primed to see a mouthing. Equally important are the single genre and small datasets that have been used. For example, the large number of 4-types in the three European signed languages is almost certainly due to the fact that these narratives involve a wolf, or a dog and a bone and present multiple opportunities for this kind of representation. This bias was no longer present in the NGT-2 data. 4.2 Distribution by sign type Studies of lexicology (Brentari & Padden, 2001; Cormier, Schembri, & Tyrone, 2008; Johnston & Schembri, 1999), lexical frequency (Fenlon, Schembri, Rentelis, Vinson, & Cormier, 2014; Johnston, 2012; McKee & Kennedy, 2006; Morford & MacFarlane, 2003), and lexicalization and grammaticalization (Janzen, 2012; Johnston & Ferrara, 2012) in signed languages have revealed that they consist of different types of signs (e.g., core lexicon, productive lexicon and gestural elements). If we are to understand the grammatical organization of these languages, we need to ensure we properly distinguish between the types of signs in our analyses, and not be misled by over-simplistic or ad-hoc glosses. Briefly, signed languages use not only fully conventional lexical signs (more or less the equivalent to the conventional words of spoken languages) 8 and gestures, but also signs that appear to be blends of gestures and lexical signs (pointing signs, indicating verbs, depicting signs). As the data in Figure 5 show, different sign types have very different rates of co-articulated mouth actions, even before we consider what grammatical role tokens of each type may play in any given clause or utterance.

13 Johnston et al. 15 Table 5. Each mouth action type as a % of all mouth actions compared to other signed languages. Auslan BSL NGT SSL HKSL* NGT-2* M-type 73.6% 51% 39% 57% 35% 80% A-type** 8.3% 21% 30% 14% 17.5% N/A E-type 0.5% 2% 8% 7% 9.5% N/A 4-type 0.5% 6% 6% 6% 2.5% N/A W-type 17.1% 20% 17% 16% 36% N/A * NGT-2 data from Bank et al. (2011, 2013); HKSL data from Siu Wai-yan (2007). **Our W-type (adverbial expressives) are included for comparison because they appear most likely to have been coded as A-type by the other researchers. Figure 5. Mouth actions by sign type (ranked by decreasing % of mouthing). Fingerspelled signs are clearly the type of sign most strongly associated with mouthing, but several individuals from the corpus have very low rates, for example, one mouths with only 28% of her fingerspellings. Depicting signs and gestures were the least likely type of sign to have a co-occurring mouthing, though they were the most likely to co-occur with a mouth gesture, by a considerable degree. Multivariate logistic regression analysis, using Rbrul, showed that sign type was a significant predictor of the use of mouthing (p < 0.001). The results (Table 6) are presented as log odds, with positive values indicating an increased likelihood to favour a particular behaviour, while negative values indicate an increased likelihood to disfavour this behaviour. Fingerspelled items (3.339), lexical signs (1.146) and possessive signs (0.896) all strongly favoured mouthing, while pronominals ( 0.128), locatives ( 0.481), buoys ( 0.548), determiners ( 0.692), depicting signs ( 1.282) and gestures ( 2.249) all disfavoured mouthing. An additional Rbrul run (Table 7) showed the sign type was also a significant predictor of the use of mouth gestures (p < 0.001). Gestures (2.313) and depicting signs (1.517) strongly favoured

14 16 Language and Speech 59(1) Table 6. Sign type and mouthing. Application value: Presence of mouthing All factor groups significant at p <.05 Factor group Factor Log odds Tokens % with mouthing Sign type Fingerspelling Lexical signs , Pointing signs: Possessives Pointing signs: Pronouns Pointing signs: Locatives Buoys Pointing signs: Determiners Depicting signs Gestures Table 7. Sign type and mouth gestures. Application value: Presence of mouth gestures All factor groups significant at p <.05 Factor group Factor Log odds Tokens % with mouth gestures Sign type Gestures Depicting signs Buoys Pointing signs: Pronouns Lexical signs , Pointing signs: Locatives Pointing signs: Possessives Pointing signs: Determiners Fingerspelling the use of mouth gestures, as did buoys to a lesser extent (0.636). Pronouns ( 0.053), lexical signs ( 0.078), locatives ( 0.097) and possessives ( 0.326) disfavoured mouth gestures to a certain extent, whereas determiners ( 0.901) and especially fingerspelling ( 3.011) strongly disfavoured them. The type of sign least likely to have any kind of mouth action (mouthing or mouth gesture) was determiners, closely followed by locatives. Figure 6 shows there are considerable differences in mouth action rates depending on the primary functional role of the pointing signs. On the whole, non-possessive pointing signs (pronouns, locatives and determiners in Figure 5) had the least mouth activity, and none were mouthed in the majority of cases. Possessive pointing signs had relatively high rates of mouthing compared to other pointing signs. First and second person points were much more likely to be mouthed than locatives or determiners. A Rbrul analysis (Table 8) found that the type of pointing sign was a significant predictor of the use of mouthing (p < 0.001), although many of the types were represented only by small numbers

15 Johnston et al. 17 Figure 6. Mouth actions with pointing signs (ranked by decreasing % of mouthing). Table 8. Pointing signs and mouthing. Application value: Presence of mouthing All factor groups significant at p <.05 Factor group Factor Log odds Tokens % with mouthing Pointing sign type 1sg possessives sg possessives pl possessives pl pronouns sg pronouns sg pronouns pl pronouns Locatives sg possessives Determiners sg pronouns Buoys of tokens. First singular (1.943) and second singular (1.481) possessives strongly favoured mouthing. Other pointing signs which favoured the use of mouthing included third person plural (0.820) and first person plural (0.715) possessives, together with second person (0.456) and first person singular (0.075) pronominals. The following forms disfavoured the use of mouthing: third person plural pronouns ( 0.053), locative pointing signs ( 0.246), third person singular possessive ( 0.314), determiners ( 0.615), third person singular pronouns ( 0.652) and buoys ( 2.093).

16 18 Language and Speech 59(1) Figure 7. Mouth action rates by grammatical class (ranked by decreasing % of mouthing). 4.3 Distribution by grammatical class Mouth action rates in Auslan varied according to the grammatical class of the accompanying manual sign (Figure 7). Lexicalized nouns showed more mouthing than other classes of sign and even depicting signs functioning as nominals were more likely to co-occur with mouthing than depicting verbs. Verbs, plain or indicating, were more likely to co-occur with mouth gestures than mouthing. Apart from nouns, some of the highest mouthing rates are with numbers and function words (prepositions, auxiliaries, conjunctions and wh-question words and wh-relativizers). These observations were confirmed by Rbrul analysis (Table 9), with grammatical class a significant predictor of mouthing (p < 0.001). We found that numerals (2.283), nouns (1.804) and prepositions (1.582) strongly favoured mouthing, as did relativizers (1.294), auxiliaries (1.066), adjectives (0.938), conjunctions (0.919) and question signs (0.381). Adverbs only slightly favoured mouthing (0.076). Plain and indicating verbs ( 0.298), negators ( 0.360), depicting nouns ( 0.375), pronouns ( 0.984), locatives ( 1.067), determiners ( 1.259), interjections ( 2.116) and depicting verbs ( 3.207) all disfavoured mouthing. Grammatical class was also a significant predictor of the use of mouth gestures (p < 0.001) (Table 10). Depicting verbs most strongly favoured mouth gestures (2.538) followed by interjections (2.054), buoys (1.868), negators (1.366), plain and indicating verbs (1.108), depicting nouns (1.101), locatives (0.257), adverbs (0.192) and pronouns (0.097). Question signs were neutral (0.000), while the following signs disfavoured mouth gestures: adjectives ( 0.251), relativizers ( 0.296), determiners ( 0.467), conjunctions ( 0.910), nouns ( 1.602), prepositions ( 1.785), numerals ( 2.455) and auxiliaries ( 2.813). This distribution of mouthing over signs of different grammatical classes observed in the data was, once again, comparable to other signed languages for which data is available (Table 11). The only true outlier in Table 11 is Hong Kong Sign Language (HKSL). Three of the signed languages from communities in which English is the dominant spoken and written language

17 Johnston et al. 19 Table 9. Grammatical class and mouthing. Application value: Presence of mouthing All factor groups significant at p <.05 Factor group Factor Log odds Tokens % with mouthing Grammatical class Numerals Plain nouns Prepositions Wh-relativizers Auxiliaries Adjectives Conjunctions Wh-question signs Adverbs Plain and indicating verbs Buoys Negators Depicting nouns Pronouns Locatives Determiners Interjections Depicting verbs (Auslan, BSL and ASL) are somewhat similar in distribution with Auslan and BSL, unsurprisingly, very similar indeed. The fourth English-related signed language, Irish Sign Language, had no category of adverbials included in the analysis and hence the data are difficult to compare. Given the distribution of other categories and given that the adjectival category actually could include many adverbials, it too may be fairly similar to the other western and English-related signed languages in the final analysis. 4.4 Standard mouthings and lexical frequency Our data confirms the long made observation that some signs appear to be consistently mouthed with the same word, called a standard mouthing (Bank et al., 2011). 9 Of 1314 fully lexical sign types in the dataset, 442 had a token count of five or greater. Considering only these in order to reduce the bias of signs with very few tokens we found that 36 had only one associated mouthing: more, help, easy, ram, how, father, now, ball, english, after, dog, first, melbourne, should, than, time, train, with, world, access, five, love, waste, week, begin, city, class, deaf-club, few, government, happy, new, new-zealand, problem, return, very. It was much more common, therefore, for lexical sign types to co-occur with more than one type of mouthing. Nonetheless, one of the co-occurring mouthings was usually more frequent that any of the others and could thus also be regarded as the standard mouthing. Not surprisingly these mouthings are usually of words semantically related to the core meaning of the sign (and its standard mouthing), for example, beer, pub, bar with beer. Seven of the fully lexical sign types with multiple mouthings actually had 10 or more unique attested mouthings (same, go, say, group, real, area, look). Most of these mouthings represented wordings semantically related to the core meaning of the sign (e.g., true, truth,

18 20 Language and Speech 59(1) Table 10. Grammatical class and mouth gestures. Application value: Presence of mouth gestures All factor groups significant at p <.05 Factor group Factor Log odds Tokens % with mouth gestures Grammatical class Depicting verbs Interjections Buoys Negators Plain and indicating verbs Depicting nouns Locatives Adverbs Pronouns Wh-question signs Adjectives Wh-relativizers Determiners Conjunctions Plain nouns Prepositions Numerals Auxiliaries Table 11. Distribution of mouthing by grammatical class of manual sign. Auslan BSL NGT SSL ASL* HKSL IrishSL* Noun 35.8% 40.3% 39.4% 42.5% 27.1% 84% 32% Verb 26.9% 24.4% 30.6% 20.9% 15.4% 9% 23% Adverbial 8.4% 6.4% 11.1% 10.1% 20.5% N/A N/A Adjective 8.8% 12.3% 5.6% 11.2% 26.7% 2.5% 27% Other 20.1% 16.7% 13.4% 15.2% 10.3% 4.5% 18% *ASL and IrishSL figures calculated from data reported in Nadolske and Rosenstock (2007) and Militzer (2010). real, really, actually, sure, honestly with real; or same, like, too, also, but, mean, with same), but they also included co-articulations in which the mouthing added additional information for which there was no separate articulated sign, for example, have with group, meaning there was/existed a group ; or I m with say, meaning I said I m or people with look, meaning people looked (c.f., Vogt-Svendsen, 2001). The data also reveal that only a handful of fully lexical signs were always mouthed, that is, almost all, even those with only one associated mouthing, had tokens that also had either no mouthing whatsoever or were accompanied with a mouth gesture. Several signs were produced with mouthing at rates much higher than other signs of the same grammatical class, for example, have and finish. Unlike most other verbs, have is almost always produced with a mouthing (92%), almost always with have and by had when not; and finish is

19 Johnston et al. 21 Table 12. Lexical frequency and standard mouthings. Application value: Lexical frequency as a continuous variable All factor groups significant at p <.05 Factor group Factor Log odds Tokens Mean Mouthing status Standard mouthing Non-standard mouthing No mouthing Table 13. Type of mouthing. Degree of articulation Tokens Full articulation 9197 Initial segment 289 Medial segment 16 Final segment 24 Initial and final segment only 24 Suppressed articulation 6 Unreadable 81 Total 9639 accompanied by a mouthing (as finish ) in 85% of tokens. These mouthing rates are quite high and are comparable to the identical signs have and finish in BSL which Sutton-Spence and Day (2001) report as near obligatory. Of signs with a token count of five or greater in the dataset, 26 fully lexical signs never cooccurred with mouthing (sprint, coincidence, suppose, tease, surprise, object, injection, disappear, amerindian, excellent, learn, do-there, eliminate, age-years(one), attend, go-trace, maybe, automatic, collect, delicious, demote, meet-see, patience, push-aside, run, until). In order to conduct a Rbrul analysis relating the use of standard mouthings and lexical frequency, we eliminated signs that occurred five times or less, reasoning that a sign occurring at such low rates in the data could not be said to have a standard mouthing due to the lack of sufficient tokens. We identified a subset of lexical signs that appeared to occur with the same English mouthing 50% of the time or more, and we categorized this subset as those with standard mouthings. We then ran a Rbrul analysis, with lexical frequency as a continuous dependent variable, and mouthing status as a fixed effect, and participant as a random effect (Table 12). We found that lexical frequency was a significant predictor of the use of standard mouthing (p < 0.001), so that standard mouthings were favoured with higher frequency (98.593) compared to non-standard mouthing ( ) and no mouthing ( ). 4.5 Characteristics of different types of mouth actions M-types. Insofar as can be discerned from lip-reading alone, most mouthings appeared to be fully articulated (over 95%) (Table 13). Incomplete mouthings were overwhelmingly initial segments only; see NGT (Bank et al., 2013). A small number of mouthing tokens (approximately 20)

20 22 Language and Speech 59(1) Table 14. E-type mouth gestures in the dataset (N = 63). Gloss MouthGestF Tokens Gloss MouthGestF Tokens coincidence pah 11 well-known pah 1 real ap 7 dss(h):animal-running pah 1 finish.finally pah 5 eliminate tham 1 automatic woof 4 sleep pah 1 have-not pooh 3 dss(b):object-passes pah 1 ban ap 3 speed-dust boom 1 peculiar pah-pah 2 dsm(b):animal-over-line pah 1 learn-lesson pah 2 dsm(b):animal-stops pah 1 vanish ap 2 diminish tham 1 zoom-off pow 1 afterglow am 1 relieved pah 1 found-out pah 1 unusual pah 1 learn oom_inhale 1 real alam 1 eliminate alam 1 empty thap 1 dss(h):animal-running pah 1 witness pah 1 too-late pah 1 wipe-up alam 1 spasmodic ap 1 cohhlear-implant pah 1 dss(4):assembly-line pah 1 were not accompanied by any manual sign yet they were clearly not redundant. They provided essential disambiguating or logico-cohesive information to the utterance. They were conjunctions, prepositions or adverbial including but, or, for, just, maybe ; sentence modifiers like I don t know, I think ; or other interactives like no, yes, not true E-types. Empty, echo or syllabic mouth gestures were rare in the data, but in all cases they had the minimal effect of making the meaning of the co-occurring manual sign more emphatic. A full list of types and tokens is given in Table 14. The number of different syllabic mouth patterns in this dataset is actually quite small, at 12 forms, and in every case some discernible meaning is associated with the mouth gesture in addition to the underlying emphasis (Table 15). These forms also have consistent, though admittedly general, meanings across usage contexts, that is, they are not semantically empty as such, as suggested in the literature. The underlying emphasis appears to stem from the parallel reinforcement of the form of the manual sign by the mouth gesture: the syllabic opening and closing of the mouth aligns itself with opening and closing of handshapes, the striking of surfaces or the sudden change in hand orientation (twisting, turning and bending). These mouth gestures thus generally conform to the characterization of echo phonology proposed by Woll (2001). One syllabic mouth gesture pah did not display in one instance the type of echoing predicted or attributed to it. In this instance, pah was aligned with the rapidly closing handshape in sleep, which is the opposite of the opening movement it normally aligns with and echoes. Nonetheless, in all instances pah has to be said to be optionally added because, again in contrast to some of the earlier literature (Sutton-Spence & Woll, 1999), 10 signs in this class of multi-channel signs, such as coincidence, finish or real, can occur without this mouth gesture: of 19 instances of coincidence, 11 occur with pah, one with A-type lips-pressed, two with W-type expressions and four with no mouth action at all; of 32 tokens of finish, five with pah, and 27 with mouthing; and of 179 tokens of real only seven tokens occur with ap (137 tokens are mouthed).

21 Johnston et al. 23 Table 15. E-type mouth gestures by associated meanings in context. Mouth gesture form Specific contextual meanings # echo phonology # echo metaphorical # NOT echo # tokens total pah suddenly, quickly ap suddenly, abruptly woof without-impediment, automatic alam all-gone, disappear from view abruptly pooh nothing, negative, remove, blow away tham all-gone, disappear from view pah-pah peculiar, strange am sudden, complete boom all-gone, complete, energetic oom_inhale close-shave, sharp, risky, dangerous pow sudden, energetic thap all-gone, disappear from view abruptly Totals Syllabic mouth gestures are attested across a variety of signed languages. According to Crasborn et al. (2008) they accompany 1.3% of signs in BSL, 4.9% in NGT and 6.1% in Swedish Sign Language (SSL). In our dataset, they occurred with only 0.4% of signs, a much lower rate than the BSL data Adverbial A-types. Only a very small percentage of A-types are potential candidates for conventional mouth-based adverbials (i.e., with their own dedicated or language-specific semantic meaning) that are applied as a modifier to the co-occurring manual sign. The data suggest that combinations of manual signs and these mouth gestures (with may occur with other non-manual elements) can produce specific meanings in particular usage contexts, but that overall the semantic component of the mouth gesture is very broad. Only a very small set of recurring semantic descriptors was needed to capture the apparent contribution of these mouth gestures large-amount, careless, unpleasant, smooth, ease, effort and small-amount (Table 16) Prosodic A-types. The majority of A-types appear to be prosodic in character, according to our definitions. These mouth gestures tended to accentuate an action qua action, that is, as an unfolding process, not unlike Sandler s protracted action category of mouth gestures (2009). (She, however, analyses protracted action expressions as conventional adverbial modifiers proper to Israeli Sign Language, not mouth gestures.) We judged that only two broad labels are needed to capture the effect of these prosodic mouth gestures: emphasis and activity (Table 17). emphasis was the broader default reading accounting for 71% of tokens, activity accounted for the force of the remaining 29%. The actual mouth gesture used with each token varied considerably (and idiosyncratically) when collated according to our form-based coding system. Nonetheless one underlying feature a tensed mouth did unite most of them (all but trill, puff and blow). This supports earlier descriptions of Auslan in which a facial grimace was said to indicate emphasis or stress (Johnston & Schembri, 2007). If we use the broader general form descriptors for A-type mouth gestures (which seem more appropriate and manageable given this variability; see Table 2), we can see that the tensed mouth gestures account for 78% of the prosodic mouth gestures. down, wide, lips-pressed and trill account for 62% of emphasis, and wide is the most frequent for activity.

22 24 Language and Speech 59(1) Table 16. Form/meaning pairings for adverbial A-type mouth gestures* (N = 187). MouthGestF MouthGestM Specific contextual meanings Tokens tongue careless carelessly, easily, with ease, without regard, petulantly, with 109 deliberate careless enjoyment, reckless, slipshod, insouciant trill ease easily, unimpeded, with enjoyment 16 blow small-amount little remaining, blown away, 12 blow smooth smooth, unimpeded, quickly, ongoing 11 trill large-amount large amount, a lot of, unimpeded, energetic, powerful, 8 engine/machine-powered tongue unpleasant unpleasant, distasteful, bad 7 puff large-amount large amount, a lot of, powerful 6 bottom-lipout careless careless, easily, without regard, petulantly, with deliberate 3 careless enjoyment, reckless, slipshod, insouciant lips-out ease easily, without regard, petulantly, with enjoyment 2 lips-out small-amount small amount, trivial, insignificant, nothing to worry about 2 lips-pressed ease easily but deliberately, enjoyable 2 lips-out large-amount large amount 1 puff careless careless 1 tongue smooth smooth 1 wide effort effort 1 lips-pressed effort effort 1 lips-out unclear N/A 1 sucked-in small-amount small amount 1 down careless careless 1 slightly-open effort effort, concentration 1 * Again tokens with more than one possible descriptor were aggregated with a single descriptor that was the most salient. Table 17. Form/meaning pairings for prosodic A-type mouth gestures.* MG form (= emphasis) Tokens MG form (= activity) Tokens down 88 wide 66 wide 78 trill 28 lips-pressed 65 slightly-open 22 trill 63 lips-pressed 22 slightly-open 38 down 19 lip-curl 34 bottom-lip-out 13 puff 31 puff 8 bottom-lip-out 27 lips-out 8 open 23 blow 4 blow 16 open 3 lips-out 13 lip-curl 1 Totals 476 Totals 194 * A number of mouth gestures need to be translated from the formal codes using more than one of these descriptors (e.g., blow, lips-out ). In these cases, the tokens were aggregated into the most salient descriptor with the same general meaning.

23 Johnston et al. 25 Table 18. Mouth for mouth (N = 62). ID-gloss Tokens ID-gloss Tokens ID-gloss Tokens graze 11 eat 5 hearing 2 capture 10 shout 3 catch 1 yell 10 laugh 2 dss(5):animal-teeth 1 amerindian 6 chew 2 bite 1 speech 5 angry 2 chatter-box types. The number of mouth for mouth gestures was extremely low (Table 18). The token frequencies are unremarkably linked to the narratives chosen for the re-tells or the elicitation materials. Insofar as these are all mimetic of the mouth action involved in the process designated by the co-articulated manual sign, they are also iconic W-types. From the data presented in Figure 4 one can calculate that 71.4% of mouth gestures alone are of the W-type. Of these we find that 42.6% are of the constructed action sub-type. There are very few of the other types. Indeed, constructed actions represent 30.4% of all mouth gesture types (i.e., mouth actions excluding mouthings). These were further coded on the MouthGestM tier as enactments (global enactments of the subject of the constructed action, including their facial expression) or expressions (presentations of just the face expression of the subject during the constructed action). A large majority were of the second type. Many enactments and expressions used mouth actions found in the A-type mouth gesture repertoire to convey similar meanings, for example, tongue for carelessly. As a result, at minimum 10% of all putative adverbial A-type mouth gestures in the dataset that clearly modify a co-articulated manual sign in some way are actually part of enactments, that is, constructed action W-types. 4.6 Iconicity The gestural nature of some mouth actions has been recognized by all researchers. The very category of W-types is intended to identify a large proportion of these. Indeed, most of the mimetic replicas and some of the symbolic iconics that Sandler (2009) identifies in Israel Sign Language fall into this category. However, some mouth actions that do not include the whole face, and thus fall into the A-type and E-type categories as coded in this and other cited studies, also appear to be drawing on mimetic replicas or symbolic iconics the form of these mouth actions are related to their meanings. Consider the A-type puffed cheeks mouth action meaning large-amount, or the E-types in which the movement of the hands (which the mouth echoes ) is itself iconic, for example, when the fingers close to represent something getting smaller and smaller in disappear. So, in order to assess the amount of gestural iconicity that may be present in our mouth action data, and evaluate it in the light of Sandler s (2009) observations, we identified those A- and E-type tokens in our data where iconic motivation is present and identified their type (i.e., as imagistic, mimetic or metaphoric). We could then combine these token counts with W-types token count. Among the 63 identified E-types, 18 were examples of echo phonology in which the form was an iconic metaphor. Of the A-types, at least 82 (45%) of the 187 traditional adverbial A-types (i.e., those not in the sub-group prosodic) were iconic in some way (usually metaphoric). Consequently, approximately 40% (100 of 250 tokens) of A- and E-type mouth gestures appear to be symbolic iconics. The majority fell into two main groups: those involving the puffing of the cheeks with or without the expulsion of air or the vibrating of the lips and were associated with the

24 26 Language and Speech 59(1) Table 19. Spreading data. Mouth action Tokens Percent with spreading Auslan bsl ngt-1 ssl Auslan bsl ngt ssl ngt-2 M-type (mouthing) % 25% 20% 12.5% 12.5% Mouth gesture % N/A N/A N/A N/A E-type % 10% 5.3% 11.1% N/A A-type % N/A N/A N/A N/A 4-type % N/A N/A N/A N/A W-type % N/A N/A N/A N/A intensification/augmentation; and those involving the pursing of the lips or narrowing of the mouth, with or without the sucking in of the cheeks and the inhalation or expulsion of air, and were associated with diminution/attenuation on the one hand or care/precision on the other. The forms the remaining prosodic A-types assume appear to be directly related to the stress or emphasis they impart and are thus also non-arbitrary. Indeed, we suggest these prosodic mouth actions are expressions deaf or hearing people spontaneously make when doing or uttering with stress or imitating someone doing so. They may be considered mimetic replicas of an action produced with stress or emphasis, if not actually indexic reflexes (the actual manifestation of stressed production). If considered in this light, then in as many as 80% of all A- and E-types, the form of the mouth action maps onto the meaning in an iconically motivated way (imagistically, mimetically) which may also be metaphorical. Some of these may even be indexic. When these are combined with W-types, fully 96% of all mouth gestures thus appear to be iconically motivated. 4.7 Spreading Spreading behaviour in mouthings in the Auslan data appears less frequent than that reported for other signed languages, despite the fact that the criteria we used for annotating spreading was the same as the other projects (i.e., that the mouth action must spread over at least 50% of the duration of the adjacent sign) (Table 19). Once again it is possible that different annotation practices may partially explain this (see Table 5). The spreading behaviour of mouth gestures when considered alone is, however, comparable to some of the spreading rates reported for two of the other signed languages. By nature, mouth gestures are not as strongly aligned with single signs, having a significant prosodic nature. Nonetheless, the spreading behaviour itself is similar to that which has been described with other signed languages, in terms of direction and source and target. There were examples of all of the types of spreading mentioned in the literature: the most common with mouthings were progressive (88%) but some were regressive (12%) (compare with 94.1% and 5.2% respectively for NGT- 2); of these, the majority were from content or lexical signs to function signs or to pointing or depicting signs, for example, 49% spread over an adjacent pointing sign (compare with 58.3% for NGT-2). The next most frequent were spreadings between lexical signs and the fingerspellings that represent the English equivalent of the sign s intended meaning (the target and source in these examples is uncertain because fingerspellings appeared on either side of the lexical sign). 4.8 Variation Individual variation. There was enormous variation between individuals in the dataset in their use of mouth actions. The rate of mouthing by individuals ranged from a low of 5% to a high of

25 Johnston et al. 27 Figure 8. Mouthing rates for individual signers. Table 20. Mouth action rates across text-types. No mouth action M-type E-type A-type 4-type W-type Monologue (narratives, retell) Dialogue (conversation, interview) Elicited (sentence elicitation, picture stimulus) %. Mouth gesture rates ranged from a low of 2% to a high of 43% (Figure 8). The mouthing rates for the six individuals in the BSL, NGT and SSL datasets, which are based on two signers from each language (Crasborn et al., 2008), have been included in Figure 8 for comparison. (Because only mouthing rates were reported in Crasborn et al., the rates for mouth gestures and no mouth actions are absent and there are white/empty bars above these data points in the graph.) Overall there is a tendency for signers to use mouth gestures less frequently as mouthing rates increase. This is not surprising: the more one mouths, the less opportunity one has to perform a mouth gesture Text-type variation. Mouth action rates vary across text-types (Table 20). Although Sutton- Spence and Day (2001) found rates over 70% for some registers in BSL, their narrative samples had rates around 44%. The fact that the individuals and the number of individuals represented in each text-type count is different may account for some of the range, or lack of range, in variation. However, 11 individuals in the dataset have examples of production in at least two text-types, and one has an example from each of the three text-types (Figure 9). These data give us confidence that the

26 28 Language and Speech 59(1) Figure 9. Mouthing rates in different text-types with same participant. Table 21. Monologues and mouth action types. Application value: Use of mouth action types All factor groups significant at p <.05 Factor group Factor Log odds Tokens % with mouth action type Monologues 4-type E-type A-type W-type No mouth action Mouthing differences in Table 20 are not driven merely by individual differences, which we still expect to be significant. The data appear to support the notion that rates do vary with text-type. Nine of the 11 individuals show a difference in mouthing rates in the same direction (namely more mouthings in dialogues, and more W-type in the monologues) between each of two or more text-types. In a Rbrul analysis, text type was found to be a significant predictor of mouth action types (p < 0.001). Monologues strongly favoured 4-type (1.349) and E-type (0.976) mouth gestures, while weakly disfavouring A-type ( 0.111) and W-type ( 0.299) mouth gestures, and particularly disfavouring mouthing ( 1.095) (Table 21). In contrast, dialogues favoured mouthing (0.801), a lack of any mouth actions (0.600), W-type (0.460) mouth gestures, and disfavoured A-type ( 0.455), E-type ( 0.653) and 4-type ( 0.752) mouth gestures (Table 22) Multivariate analysis of social factors. We conducted multivariate logistical regression analyses using Rbrul on the occurrence of mouthing and mouth gestures with the following social factors as

27 Johnston et al. 29 Table 22. Dialogues and mouth action types. Application value: Use of mouth action types All factor groups significant at p <.05 Factor group Factor Log odds Tokens % with mouth action type Dialogues Mouthing No mouth action W-type A-type E-type type independent variables: signer s sex, age, nativeness and region, with participant as a random effect in a mixed effect model. Although participant variation was significant, we found none of the social factors were predictors of the use of either mouthing or mouth gestures. 5 Discussion 5.1 Methodology Type and size of datasets. At 17,002 manual sign tokens, the Auslan dataset used in this study is much larger and varied than previous studies of BSL, NGT, SSL and HKSL already cited, or even the next largest study (at 11,905) recently completed which we have referred to as the NGT-2 dataset (Bank et al., submitted). Previous researchers called for further research with significantly larger numbers of signers to help distinguish between variation between individual signers and differences between signed languages (see Figure 8). Our dataset enables us to see quite a wide distribution in the rate of mouth actions for individuals, despite the fact that we have deliberately restricted our corpus to native or near-native signers. One cannot fail to see from Figure 8 that if we were to randomly select two individuals from our dataset to represent Auslan mouthing rates (or rather had only selected two individuals in the first instance), the results would most likely have been quite misleading. Similarly, the spreading data for NGT-1 are quite different to the spreading data for NGT-2. It is relevant in this context to note that it has been taken as axiomatic by many signed language researchers that almost all of the symbolic communicative behaviour of signing deaf people is linguistic in the narrow sense. However, this is actually simply a working assumption, in some cases driven by linguistic theories which assume the autonomy of language from other aspects of cognition and communication. If gesture plays a significant role in face-to-face communication (spoken or signed), then some symbolic behaviour may not be linguistic in the sense of being part of a highly conventionalized, rule-governed system in which most of the forms in either primary modality are actually language-specific. Possibilities in wording and morpho-syntactic coding are often highly constrained by the very nature of linguistic systems, that is, some constructional schemas are obligatory in certain contexts and thus many aspects of linguistic symbolic behaviour can be sampled from relatively small numbers of users precisely because of this. However, if the substantive symbols are not actually linguistic in the sense we have described, it is unlikely that any single individual, or small sample of individuals, will provide data upon which can be generalized core constructional schemas of the language. There is reason to believe that some aspects of signing behaviour (like mouth actions) fall into this category.

28 30 Language and Speech 59(1) It is for this very reason we have adopted a corpus-based approach in this study (and in other areas of the description of Auslan). Nonetheless, one may reasonably contend that no datasets yet presented, including our own, are sufficiently representative of native signer usage for our observations to be unlikely to be revised in the future. This needs to be borne in mind Coding and categorization. In a first parse of the data, A-type mouth gesture forms were annotated using the glosses as shown in Table 2. When this first round of data were analysed, our results regarding form/meaning pairings were no different than reported above and discussed below. Given the strong claims made in the literature on the conventionalization of some mouth gestures, we considered that our coding and categorization was perhaps too general to capture this, so we then used the Sutton-Spence and Day (2001) schema, which we adapted (see Appendix), and returned to the primary data and re-coded the entire dataset. The finer coding did not alter the outcome: there remained the range of variation in forms that we originally observed, that is, the more discrete and detailed features did not align in any discernable pattern, and the relative importance of the co-text and context of utterance in describing the meaning of these forms was also unchanged. 11 Categorizing mouth gestures is neither simple nor straightforward. Mouth gestures often overlap, for example, if the mouth is widened, with the lips closed, the lips will often be pressed together, or different lip configurations can occur with a slightly open mouth. This is evident from an examination of the form-based codes we used. Moreover, decisions about how to label the meaning of mouth gestures are even more subjective and thus potentially open to re-interpretation by other researchers, even if reviewed by a native signer (as they were). Inconsistencies could also arise if an annotator classifies a mouth action as a non-w-type and fails to notice that the whole face is actually involved or that the mouth action is performed during a period of constructed action. Again, one needs to keep this in mind. 5.2 Language contact and the conventionalization of mouth actions Relationship of English mouthings to Auslan signs. The majority of manual sign tokens (56.7%) co-occur with mouthing (Figure 4). Given mouthing rates range from 6% to 84% (Figure 8), can it still be suggested that mouthing is an obligatory formational component of some signs? The strongest claim has been made for fingerspelled signs. Only some 6% of fingerspellings have no mouthing (or mouth gesture) (Figure 5). Several individuals in the dataset, however, have very low rates of mouthing with fingerspelling, for example, MVS who is from a multigenerational deaf family, mouths only four of 14 fingerspellings (a rate of only 28%). MVS is not difficult to read for a native signer and her signing style should certainly not be considered atypical. 12 The next strongest claim has been made for the mouthing co-occurring with so-called ambiguous signs, such as Auslan spouse or NGT sibling. 13 True, all five instances of spouse in the dataset were indeed mouthed wife or husband and it is clear that signers intend to make this semantic distinction. They certainly never mouth spouse. (In fact, some Auslan signers would not know what this English word means.) However, sampling the much larger Auslan corpus reveals that, of 37 tokens of spouse, 23 are mouthed (as wife, husband, marry/ied ) and 14 have no mouthing (in all of these senses). 14 The fact that fully one-third of spouse tokens have no mouthing suggests that the use of mouthing is not obligatory and not specified in the signer s mental lexicon. In the dataset, gender or grammatical function is often simply inferred from context without any formal (including mouthed) marker: for example, poss1sg spouse (unmouthed, signed by a man) means my wife; or pro1sg spouse pro3sg (unmouthed) means I married him/her.

29 Johnston et al. 31 Nonetheless, basic core culturally shared distinctions are made and expected. For Auslan signs, this does mean a default mouthing of wife, husband or even, marry when signing spouse, even if it is not obligatory. Their interlocutors, who are often second language learners of Auslan and speakers of English as a first language, give and demand clarification. A similar default coupling occurs with vocabulary items that name basic level cognitive categories that one can expect to recur and overlap cross-linguistically and cross-culturally, for example, tree, flower, boy, girl, dog, cat, mother, father, man, woman and so on (Taylor, 1991). The fact that signers and speakers share essentially the same culture, reinforces the likelihood of overlap and convergence. The fact that Auslan and English are in different modalities means that double coding (mouthed in English while manually signed in Auslan) can occur without any expressive or receptive conflict. Indicative mouthing rates for these types of concepts from the dataset include: tree (87%), flower (78%), boy (77%), girl (85%), dog (95%), cat (72%), mother (84%), father (100%), man (78%), woman (80%), lady (91%). Moreover, despite the fact that many signs are associated with a standard mouthing, as shown earlier, across the lexicon the relationship of manual signs to mouthing is simply not one-to-one: it is one to many. Most signs can co-occur with different mouthings (and sometimes mouth gestures) that specify, limit or extend the core meaning of the sign (Bank et al., 2011; Johnston & Schembri, 1999). The reported findings thus confirm that mouthing is not obligatory in the sense that they are not part of the lexical specification of any sign, a position long-held by Auslan researchers (Johnston & Schembri, 1999) and other researchers (e.g., Ebbinghaus & Hessmann, 2001). Rather, mouthings represent the parallel or simultaneous expression of elements from a second language, and thus the phenomenon is best described as code-blending (Bank et al., submitted; Emmorey, Borinstein, & Thompson 2005; van den Bogaerde & Baker, 2005). Data on slips of the hand and tongue by deaf and hearing BSL signers, suggesting that the mouthing and manual components have separate mental representations, supports this position (Vinson et al., 2010). Mouthings are obligatory only in the sense that if an Auslan signer wishes to make the semantic distinction afforded by the mouthing in a situation where context does not render this inferable and continue to use just the intended manual sign then they must use that mouthing. Otherwise, they must choose another way of incorporating English (e.g., by an interpolated fingerspelling) or instead use circumlocution in Auslan. Researchers have suggested that mouthing is primarily or solely a language contact phenomenon or that formal or oralist education has had an impact on mouthing rates (Bank et al., 2011, submitted; Hohenberger & Happ, 2001; Nadolske & Rosenstock, 2007). With respect to the former, though we do not have data on our corpus participants signing to non-native or hearing signers for comparison, our data are taken from sets of two native (or near-native) signers (and each knew the other was a fluent signer) with a third native signer researcher (known to be native to the other two). Our data do show that mouthing is present to a large degree in most texts, so immediate contact in the context of utterance with non-native or hearing signers cannot be driving this behaviour. With respect to the impact of formal education for deaf children, reports of deaf signers from communities who have no formal education and are illiterate show that they use spoken language related mouthings in their sign production (Fontana, 2008; Nyst, 2007). For example, Fontana observes that for home signers in Okinawa and Italian Sign Language signers in Sicily, speech is part of the communicative economy of deaf individuals and deaf communities everywhere. In the life experience of all deaf people, there will be inevitable exposure to some signs and gestures that are invariably accompanied by the same mouthing when performed by hearing people (signers or non-signers). Deaf people cannot help but observe this: observing and internalizing mouth actions (including mouthing) is an inevitable part of being a deaf person in a hearing world.

30 32 Language and Speech 59(1) Consequently, in Auslan, as in NGT (Bank et al., 2013) and presumably many other signed languages, mouthing is ubiquitous. 15 As Bank et al. have also remarked, this has implications for the independence of signed from spoken languages in developed urban communities with long histories of deaf education and widespread basic literacy. We would agree, at minimum, that monolingual speakers of these languages do not, in fact, exist. All signers display some degree of bilingualism and thus language contact Form/meaning pairings in Auslan. The majority of mouth gestures (71.4%) were categorized as W-type (Figure 4) and are indexic expressions, or mimetic and gestural enactments of one kind or another. The small number of 4-type mouth gestures is also mimetic. Of the A-type mouth gestures, we have determined that fully two in three appear to fulfil a prosodic function, with little evidence of language specific codification. Of the remaining traditional non-prosodic (adverbial) A-types, we consider some 45% to be iconic. We also identified the use of A-types within W-types (during constructed action) and this may be relevant to the question of conventionalization. It would appear that many adverbial A-type mouth gestures represent expressive or iconic/mimetic mouth enactments that can also be used with some signs when no overt constructed action is actually being performed, rather than being examples in which constructed action has co-opted adverbial A-types. It is not clear to what extent we can consider the above mouth gesture types as specific to any signed language, and it is likely that many form part of a more general semiotic system. The remaining non-iconic A-types vary in form and effect. It was difficult to identify recurring form groups and the meanings they conveyed except in the most general of terms. They seemed to have more subtle, global meanings, to be read in combination with other aspects of the delivery, rather than a specific, defined meaning. There has only ever been one good potential candidate in this category: th. It has long been identified as the archetypical A-type modifying mouth gesture morpheme meaning something like carelessly (or with ease which is a suggested second distinctive meaning proposed for BSL; see Lewin & Schembri, 2011). However, the use of this mouth-centred facial gesture does not appear to be specific to signed languages. Studies, some that predate the work of Liddell (1980), on expressive and communicative human facial expressions have documented the meanings and functions of tongue protrusion (th) in great detail (e.g., Rozin & Cohen, 2013; Smith, Chase, & Lieblich, 1974). Few signed language researchers, apart from Fontana (2008), appear to be aware of this work on humans and primates. Data from human children and adults show that tongue protrusion is common and is used to index and communicate distaste, displeasure or disgust; attention or care; recklessness or carelessness; unencumberedness or ease; and disdain or rejection. Some of these meanings clearly overlap the values of carelessly or with ease that are attributed to this mouth gesture in signed languages. However, the apparent specificity of these meanings to these signed languages has much to do with how one describes the meanings of various facial expression (Elliot & Jacobs, 2013). The research of Smith et al. (1974), for example, shows that there are really only two underlying functions of tongue protrusion: (a) indexing distaste/displeasure/disgust; and (b) expressing social exclusion (i.e., a desire to be uninterrupted or a rejection of social engagement), with the former probably lying at the basis of the latter (one can readily appreciate how physical rejection could extend metaphorically to social rejection). When an appropriate general label is given to the latter (which Smith et al. call social exclusion ), one can see how its semantic force would vary according to what co-occurs at the time of the tongue protrusion (i.e., the context) and the positive or negative overall facial expression it is made with. For example, signing drive + negative face + th = recklessly, carelessly, that is, drive in such a way as to reject the concerns or attitudes of others, namely carelessly ; drive + positive or neutral face + th = carefree, easily, that is, drive in such a

31 Johnston et al. 33 way as to reject the concerns or attitudes of others as unnecessary because it is not difficult and you are competent, namely with ease ; cut + neutral face + squint + th = cut with one s full attention, that is, keeping (social) interruptions at bay, namely carefully, attentively or precisely. Signers are able to produce a specific conventional meaning with a manual sign while they produce a mouth gesture. This provides signers with greater control on the context in which inferences using the social exclusion gesture th could be constrained. Speakers, on the other hand, need their mouths to speak so do not have the opportunity to do this. 16 The conventionality lies not in the th mouth gesture but in the conventional signs that with which they co-occur (and the wider signed context). If one does not know a conventional lexical manual sign meaning drive is being produced (driving is an act one would normally do with care and attention), one lacks the intended and assumed inferential context. The fact that a similar mouth gesture has been identified in quite a few signed languages leaves little doubt it is not language-specific. Finally, the fact that no human cultures so far investigated appear to lack it suggests it is not specific to language, as such. This leaves very few other mouth gestures that may be considered candidates as conventional language-specific mouth gestures. The E-type mouth gesture pah is one possible example. First of all, like the example of th above, one needs to discount one meaning associated with it as overly specific and derived from an association with a frequently, but not exclusively, co-occurring lexical sign: finish (or alternatively glosses as success or finally in both Auslan and BSL). On the other hand, one may argue that it has begun to be interpreted as a semantic component ( suddenly ) which can be added to any sign, regardless of whether or not it echoes that sign s movement parameter. For example, pah occurs with sleep to mean fell swiftly into a deep sleep which is not a standard pairing (the mouth gesture opens, whereas the handshape closes). It also occurs with depicting signs which do not have a standard form and thus no standard mouth gesture. (They are productive signs with both conventional components, like handshape, and nonce components, like placement and movement.) There may even be no formational parameter which is echoed in the pah. A second E-type candidate may be woof because in the study dataset it always occurred with automatic. However, there are two reasons why it clearly would be incorrect to attribute automaticity to the meaning of woof. First, across the entire Auslan corpus there are attested examples of automatic also co-occurring with faa, alam and fee, all with the same effect as woof, with the mouth gesture type trill, and the mouthing own (probably an abbreviation for on its own ). Second, it also occurs with no mouthing or mouth gesture. So though the strongest claims to linguistic status have been made for some E- and A-types of mouth gestures, it would appear that their language-specific conventionalization is, at best, incipient. Moreover, as a group they only occur in small numbers and may be considered to be a marginal phenomenon. Overall, the data support earlier suggestions that few Auslan mouth gestures have very clearly specifiable language-specific forms and are obligatory in any meaningful sense (Johnston & Schembri, 2007). Rather than being a part of the phonological specification of some signs or separable formal morphemes added to signs and phrases, they are on the whole contextual symbolic units (gestures of one kind or another) and/or are prosodic. 6 Conclusion Mouth actions, other facial expressions, head and body movements and other aspects of sign articulation (e.g., speed and stress) all contribute in various complex ways to meaning creation in Auslan (Johnston & Schembri, 2007; see also references found in Notes 1 and 2). However,

32 34 Language and Speech 59(1) positing a conventional codified system for mouth actions as part of this dynamic appears not to be supported by the usage evidence, at least for Auslan. The strongest candidates for conventional status were the so-called lexically specified mouth actions (mouthings and echo phonology). But in both cases, no specifiable mouth action appears to obligatorily co-occur with particular signs, and exceptions and variation abound. Mouthings appear to be examples of language blending, even if constant and intimate language contact does mean many signs have a standard paired mouthing. Mouth gestures do add meaning to co-articulated signs but these meanings are relatively general and the actual mouth forms employed vary from person to person. Overall, we have seen that there are often no mouth actions at all in environments in the data where it would be assumed or predicted to occur had the mouth action the linguistic role attributed to it, for example, mouthings with fingerspelling or with lexical specification, or mouth gestures of a certain form with classes of signs with particular morpho-phonological characteristics. These facts do not suggest highly codified usage. Mouth gestures also appear to have parallels in face-to-face spoken language production, although there has been virtually no research on non-manual gesture use in spoken language. Although infrequent (because speakers are most of the time busy speaking so the mouth is not available to gesture), observation of day-to-day spoken language interactions suggests a hearing person s manual gesture can be accompanied by a mouth gesture which echoes the dynamics of the former, for example, a fist opening to a five, or fingers-spread, handshape (from a plosive-like to an open vowel-like syllable) to express rapidly throwing something, something suddenly opening or exploding, a spitting or expelling action, or something splattering on a surface. In this context, it would seem that the definition often given of mouth gestures as mouth actions that are unlike mouthing in that they do not derive from spoken language (Boyes Braem & Sutton- Spence, 2001) as if they arose spontaneously within signed languages and were sign language dedicated needs qualification. Most mouth gestures that occur in signed languages appear to be part of the repertoire of facial expressions used by members of the wider culture in which deaf signers find themselves. Of course, some mouth gestures may be negatively sanctioned in some cultures and, thus, be suppressed and rarely produced, yet they would remain comprehensible to anyone in such a culture given the appropriate context. So mouth gestures also are often sourced from spoken language insofar as speakers constitute the vast majority of any human community. Everyone, signers included, imitate the faces of the people they interact with on a daily and face-to-face level. It is true that many mouth actions are actually the articulation of spoken words (mouthings as far as a deaf person is concerned), but many are not. Most of the data do not suggest that all mouth gestures are autonomous to signed languages or that they have their own unique meaning and function in these languages. We would conclude that mouth actions in Auslan are thus significantly less conventional and more gestural than has been proposed in the literature for a number of other signed languages. We believe that other signed languages may not be very different, but certainty in this regard requires analysing much larger and representative datasets of naturalistic data than has hitherto been the case. We believe mouth actions exist along a continuum of indexicality, iconicity and conventionality (Figure 10). Highly conventional (i.e., linguistic) mouth gestures, should they exist in some signed languages, would only account for a small proportion of all mouth actions in any signed language. Mouthings are linguistic in the sense that they are instances of code blending. Mouth actions work as part of signed languages which we believe, like all human languages, are best described as heterogeneous semiotic systems (de Beuzeville, Johnston, & Schembri, 2009; Johnston, 2013; Schembri, 2001). Mouth actions are additions that support other facets of the utterance to create a synthesized meaning in context. These mouth actions, both the mouth gestures and

33 Johnston et al. 35 Figure 10. A continuum of conventionalization in mouth actions. the mouthings, add meaning to the primary conventional semiotic units of the linguistic system, the symbolic constructions of the language (the individual signs and multi-sign strings). In this sense, mouth actions as a whole perform a role not unlike that often attributed to co-speech gesture: adding meaning which is formally un-stated or under-stated in the wording (i.e., the linguistic coding). However, it would be a mistake to equate mouth actions with co-speech gesture if by this we imply this is the primary or sole locus of gesturing within signed languages (see Sandler, 2009). We do not believe that it is only on the mouth that gesture-in-sign resides ( co-sign gesture as it were) because gestural elements can also be found in other levels of the lexico-grammar of these languages (de Beuzeville et al., 2009; Janzen, 2012; Johnston & Schembri, 2010; Liddell, 2003; Wilcox, 2004). They are not only manifested in mouth actions. Acknowledgements In addition to the annotations created by the three authors of this paper, we also wish to acknowledge the work of students and research assistants who have also contributed to annotating the Auslan corpus. In chronological order since 2005 (most recent first): Lori Whynot, Donovan Cresdee, Christopher Hansford, Ben Hatchard, Michael Gray, Gabrielle Hodge, Lindsay Ferrara, Julia Allen, Gerry Shearim, Karin Banna, Louise de Beuzeville, Dani Fried and Della Goswell. This study grew out of a pilot BA honours research project by Jane van Roekel, supervised by Trevor Johnston. Funding Research for this study was supported by the Australian Research Council, grant #DP to Professor Trevor Johnston. Notes 1. See, for example: Aarons, Bahan, Kegl, and Neidle, 1992; Baker-Shenk, 1983; Dachkovsky and Sandler, 2009; Elliot and Jacobs, 2013; Johnston, 1992; Liddell, 1980; Neidle, Kegl, Maclaughlin, Bahan, and Lee, 2000; Nespor and Sandler, 1999; Pfau and Quer, 2010; Sandler, 1999a, 1999b; Sandler and Lillo-Martin, 2006; Tang, 2006; de Vos, van der Kooij, and Crasborn, 2009; Wilbur, 2009; Wilbur and Patschke, 1999; Zeshan, For earlier research on this topic see Anderson and Reilly (1998) for ASL; Engberg-Pedersen (1993) for Danish Sign Language; Ebbinghaus and Hessmann (1990) for German Sign Language; Schermer (1985) for Netherlands Sign Language; Schroeder (1985) for Norwegian Sign Language; Bergman (1984) for Swedish Sign Language; Vogt-Svendsen (1983) for Norwegian Sign Language; Lawson (1983) for British Sign Language; and Liddell (1978) for American Sign Language. The Boyes Braem and Sutton- Spence volume included papers relating to British Sign Language, Finnish Sign Language, German Sign Language, Indo-Pakistani Sign Language, Italian Sign Language, Netherlands Sign Language, Norwegian Sign Language, Swedish Sign Language and Swiss German Sign Language. Some of these papers are cited individually elsewhere in this study.