Patterns of handshape errors in the Auslan productions of young children

Abstract

The aim of this study was to describe the handshape substitution patterns observed in Australian Sign Language (Auslan). The spontaneous signed productions of 44 children aged 3; 0–6; 8 years were documented using a new approach across a 2-year period whilst the children were attending an Auslan/English bilingual–bicultural education program. The 3,166 signs within the corpus contained 3,849 individual handshape tokens; 374 of these handshapes were inaccurate. The majority of these errors involved handshape substitutions. The ASL 1 handshape was used most frequently to replace other handshapes. There were similarities and differences in handshape substitutions seen within this Auslan corpus to error patterns previously described for American Sign Language and British Sign Language. The differences in these Auslan data suggest that there is a need for sign language–specific data when considering how handshapes are acquired and used.

Australian Sign Language (Auslan) is the official sign language of the deaf community in Australia, with an estimated 16,000 users (Australian Bureau of Statistics, 2022). Researchers who are interested in studying Auslan development experience significant challenges in finding subjects due to the small number of users and an extensive geographical spread (Johnston, 2004; Schembri et al., 2002).

Examining Auslan development from birth is further hampered by the low proportion of children who are deaf or hard of hearing (DHH) who are born to fluent signing parents. Estimates vary, but more than 90% of children who are DHH will be born to families where their primary caregivers have typical hearing (Mitchell & Karchmer, 2004; Webster & Safar, 2020). While the hearing parents in these Australian families may speak multiple spoken languages (Ching & Dillon, 2013), they are highly unlikely to already know and use Auslan.

There is limited research focused upon Auslan handshapes; the majority of research has been conducted on larger, more historically established sign language populations such as British Sign Language (BSL) users in the U.K. and American Sign Language (ASL) users in the United States of America.¹ The publication of the first Auslan dictionary was just over three decades ago (Johnston, 1989), and dictionary nomenclature for Auslan handshapes was described fairly recently, in 2007 (Johnston & Schembri, 2007).

To date, Auslan research has primarily focused on the development and analysis of the Auslan Corpus (e.g., Johnston, 2010; Johnston, 2024), the linguistic structures in Auslan, and the use of various components of Auslan narrative discourse, including verb modification (de Beuzeville et al., 2009; Ferrara & Johnston, 2014; Hodge & Johnston, 2014). There is a dearth of research concerning phonological development in Auslan, and no publications have focused on describing handshape errors observed in the spontaneous signs produced by children in everyday contexts. It is unclear how and when the parameters of sign phonology, namely: location, movement, handshape, and hand orientation, are acquired in Auslan.

It may be assumed that all sign languages are acquired in much the same way; that is, studies of acquisition of BSL and ASL can be applied to Auslan with limited supporting data. Indeed, Auslan is derived from BSL and is proposed to be very similar in structure (Johnston & Schembri, 2007). These languages share many lexical signs and handshapes and have similar two-handed alphabets; therefore, they may be expected to be acquired by the child along similar lines. American Sign Language, however, is derived from Langue des Signes Francaise (French Sign Language; Van den Bogaerde et al., 2016) and is unrelated to BSL; therefore, it has less lexical similarity to Auslan and, unlike BSL and Auslan, uses a single-handed alphabet.

The authors posit that just as there are some phonological differences in the acquisition of spoken languages due to the frequency and complexity in motor production of some sounds (e.g., blends and clusters; Bowen, 2023) within the lexicon, there may be similar frequency and complexity effects in Auslan compared with other sign languages. There may be differences in handshape frequency and complexity in Auslan, compared to ASL, for example, that warrant attention. Johnston and Schembri (2007) described the potential for “markedness” and the frequency of handshape use to influence order of acquisition, but the veracity of this theory has not been examined in a large Auslan corpus.

Impact of frequency on handshape development

Handshapes are the shapes made by the hand and fingers akin to individual phonemes in spoken languages. A handshape may be fixed within the sign, or there may be a transition between two or more handshapes as part of producing the full sign. Battison (1974) proposed a set of rules or conditions for handshape use within two-handed signs. Symmetrical signs may use the same handshape on both hands; asymmetrical signs may use different handshapes. In some asymmetrical signs, one hand may act as a base for the other, but this base hand may only use certain handshapes that are easy to produce. The use of the nondominant hand as a “base” can be seen in many Auslan signs, and in many fingerspelled words, because Auslan uses a two-handed alphabet.

Describing the phonology of Auslan is important as it differs from ASL in a number of ways; we cannot simply transfer what we know about ASL to Auslan. Particular handshapes may be used more often than others due to the prevalence of specific handshapes within lexical signs. There may also be an influence of the handshapes used within ASL’s single-handed alphabet. For example, the ASL letter R is used frequently in ASL signs for words that start with “r” (such as READY) and some name signs (e.g., RONALD REAGAN). This ASL handshape is used in Auslan but infrequently, as it is not related to any letter within the Auslan alphabet. Some sign languages, such as Taiwanese Sign Language (TSL) utilize handshapes that appear to be unique to and may only be permissible within that sign language’s phonological inventory (Ann, 1996). Handshapes that exist in TSL or ASL, but do not exist in Auslan, would not need to be considered as part of an Australian child’s phonological development of expressive handshapes. Taking an example from spoken languages, Russian has the initial blend “gd” as a permissible blend, as in “gde” (meaning “where”). This initial blend does not exist in spoken English and therefore would not form part of any assessment of an Australian child’s phonological repertoire.

Boyes Braem (1990) proposed that children learning sign languages respond most strongly to visual cues; therefore, they have a greater chance of making handshape errors for forms that they do not see. If a handshape frequently occurred due to favorite words and concepts within the child’s early play and care routines, then the child would be exposed to that handshape more often and would have more opportunity to see, comprehend, produce, and practice that handshape.

Ann (1992, 1996) was the first to link ease of handshape production and frequency in their discussion of the motor requirements for fingers and hands for handshapes within TSL and ASL. Handshapes within the respective phonological inventories of these languages were coded as “easy” or “hard” to articulate in their 1996 study, and it was observed that “easy” handshapes occurred more often than “hard” handshapes.

Within Ann (2005), it was observed that while many of the “easy” handshapes occurred frequently within the TSL lexicon, not all of the most frequently used handshapes were easy to produce. More recently, Yin et al. (2024) reviewed the 2,723 ASL signs within the ASL-LEX database and found that the more frequent handshapes within the signs of the ASL corpus were indeed easier to produce.

A motoric model of handshape development

In 1990, Boyes Braem proposed a tiered sequence of acquisition of handshapes, based upon analysis of a single 1-hour ASL sample produced by a two-and-a-half-year-old deaf child who had been immersed in ASL from birth with her two deaf parents. There were four suggested levels within this model, partially determined by not only the handshapes present within the language sample but also their theoretical level of motoric complexity. This complexity was linked to the notion of “finger order,” that is, the order by which children can differentiate and manipulate their fingers, with the corresponding ability to use or inhibit particular fingers and finger joints.

Within Boyes Braem’s model, Stage I was composed of two fist shapes, ASL A and S, two pointer shapes, ASL L (Auslan 7) and G (sometimes referred to as 1), ASL 5, C (Auslan bC), and the ASL bO (Auslan gO). Stage II was composed of the ASL B, F, and O. Stage III was composed of eight more complex handshapes in ASL: Y, I, D, P, 3 (Auslan 8), V, U (Auslan H), and W (Auslan 3). Stage IV was composed of five handshapes proposed to be the most complex, namely, ASL 8o, 1h or ASL X (resembling a crooked index finger as in “come here”), ASL 7, and then R and T from the ASL alphabet.

Within this model, the author suggested that an ASL L would be present within a child’s emerging sign language productions prior to an ASL B handshape. They also proposed that an ASL 5 would be used as a substitute more than a B handshape. Boyes Braem’s reasoning behind this proposed hierarchy of substitutions considered the physical challenges involved in making these handshapes. For example, the ASL 5 handshape involves a spread hand, but the B handshape requires most fingers to be kept together, while the thumb remains in an open position, which was suggested to be more difficult. There is, however, very little observational or empirical evidence of this motor hierarchy for ASL and no evidence for any motor hierarchy of difficulty for Auslan.

Handshape complexity and markedness

If underlying motoric constraints are not the only factor that determines frequency, it is possible to consider alternate linguistic models, including the concept of markedness that is applied to spoken languages. In brief terms, markedness theory posits that phonemes contain a possible range of complex features such as voice or aspiration. Sounds that contain these complex features are “marked,” and those that do not contain these complex features are “unmarked” (Morgan, 2006; Rice, 2007). Marked phonemes are more complex, acquired later, and are thought to occur less often than unmarked phonemes, both within and across languages (see Rice, 2007 for an overview).

Markedness can also be applied to sign languages, such as within the Prosodic Model of Sign Language Phonology, proposed by Brentari (1998). This model discusses a branching structure for handshape production whereby branches are added at particular nodes depending on the use of fingers (e.g., isolation of the thumb) and joints (e.g., to bend or flex) to change the shape of the neutral hand. The greater the number of alterations to this neutral handshape, the more branches, and the handshape becomes more complex and marked.

The author’s proposed criteria for marked handshapes in sign languages were that they were acquired later, harder to articulate, and less frequently occurring across languages. Furthermore, marked handshapes would not be used as a base handshape, and be more likely to be omitted within children’s signs, whereas less marked handshapes were proposed to be substituted within their errors (Brentari, 1998). The work of Mertz et al. (2022) investigated the ability of nonsigners to repeat signs of different complexity in line with this model in French Sign Language. Their results supported the notion that more complex signs are harder to produce and, furthermore, that handshape complexity impacts accuracy.

Henner et al. (2013) considered both frequency and complexity when describing “unmarked” handshapes in ASL, in a corpus of 1,000 signs from adult users. They noted that the B handshape, while thought to be more marked in ASL, occurred with the highest frequency. The other most frequent handshapes were the ASL 5, A, and G. The ASL C, proposed to be unmarked and within Stage I of the Boyes Braem (1990) model, did not occur with a high frequency within their corpus. This led the authors to suggest that the ASL C handshape should not be incorporated within this unmarked set, and should, instead, be replaced by the ASL B.

We suggest that handshape complexity and markedness, akin to the complexity of spoken phonemes, go beyond the ease of articulation and are likely to involve more than just the number of fingers or joints that are utilized. Frequency of handshapes within sign languages must also be considered. Brentari (2019) stated that handshapes that are used less often can also be considered more complex due to their rarity, consistent with informational complexity in spoken languages. Markedness as defined by both complexity and frequency would also align with Boyes Braem’s model, whereby Stage I handshapes would be considered to be more lexically frequent and Stage IV handshapes would be lexically infrequent (Boyes Braem, 1990).

In summary, certain handshapes may be used more frequently not only because they are physically easier but also because they are simply present more often within the signs that are salient to the child. It is proposed that if we apply elements of markedness, including frequency to sign languages, in alignment with Brentari (1998, 2019), the young child’s emerging sign phonology would be composed of motorically easier, unmarked, high-frequency handshapes. Furthermore, these same handshapes may also be used (incorrectly) as substitutes for more complex or marked handshapes until motoric proficiency improves and the child’s phonological system develops. While Johnston and Schembri (2007) proposed a set of seven handshapes that would be unmarked in Auslan, specifically the Auslan 6, bC and ASL S, B, 5, 1, and O, their frequency was based upon an ASL model and determined through comparison with the Auslan dictionary. No study to date has examined this theory in a corpus of Auslan signs produced by children.

Handshape substitutions in sign languages

In spoken languages, developmental errors are thought to follow the degree of markedness, for example, when a child is acquiring marked sounds, they will use unmarked sounds as substitutes. Rarely, if ever, will a child use a marked sound to replace an unmarked sound (Watts & Rose, 2020). Similarly, the Boyes Braem model proposed that Stage I and II handshapes may often be used as replacements for other handshapes in sign languages. For example, the ASL 5 would be the most frequent substitution as it is the least phonologically complex. The Stage III handshapes may occasionally be used as substitutions; however, the Stage IV handshapes would never be used as replacements for other handshapes, due to their complexity (Boyes Braem, 1990).

Following analysis of their 1-hour language sample, the author found that the ASL A, S, L, bO, G, 5, and C Stage I handshapes made up 49% of the child’s handshape productions, and the ASL G handshape was the most frequently used. The ASL 5 handshape was only used as a replacement for other handshapes 12% of the time. The most frequently used handshape to replace other handshapes was the ASL bO, used in 31% of all substitutions. The ASL 3 handshape was used more frequently than expected to replace other handshapes. That is, despite the ASL 3 being from Stage III of the model, it was used as a replacement handshape with the same frequency as the ASL G handshape. It accounted for 9% of all substitution errors observed within the sample (Boyes Braem, 1990).

There are a small number of additional studies that describe handshape substitutions. Siedlecki and Bonvillian (1993) used nine parental diary reports about children learning ASL from their deaf parents. Parents kept language diaries from their child’s first signs in ASL to the point where they began producing signed combinations. Children ranged from 5 to 18 months.

In these data, the parents reported that the unmarked ASL 5, A, B, O, G, and C were used most frequently by the children. All six of these handshapes were within Stage I or II of Boyes Braem’s 1990 model, but were not always produced correctly. The marked ASL 3, X, K, V, E, and baby O were used infrequently and were produced at later time points. The diary reports revealed that the ASL F, I, R, T, Y, W, and 8 handshapes were not produced by any of the children (Siedlecki & Bonvillian, 1993, 1997).

When focusing specifically on the patterns of deletion of handshapes, the authors discussed that the children were more likely to delete one hand within a two-handed sign when the sign made contact with the body (Bonvillian & Siedlecki, 1996). Deletion of one handshape was reported to be rare when the hands made contact with each other, only accounting for 6.5% of handshape deletion errors (Siedlecki & Bonvillian, 1993).

In 2000, the same authors revisited their data with regard to handshape substitutions to examine if children used unmarked handshapes to replace marked handshapes. The ASL 5 was frequently used as a replacement for other handshapes; this handshape was used in 30.8% of all handshape substitutions. The children also used the ASL G, A, B, and O handshapes as replacements for other handshapes. These five handshape substitutions accounted for 93.3% of all substitutions within the study (Bonvillian & Siedlecki, 2000).

Marentette and Mayberry (2000) studied the productions of a single child of deaf parents learning ASL between the ages of 12 and 25 months, gathered by video recording seven 1-hour play sessions. Overall, Marentette and Mayberry (2000) showed that their single case study child preferred to use unmarked over marked handshapes. However, it must be noted that this child was observed to have a clear preference for three particular handshapes, namely, the ASL 5, G, and A, and often used these three handshapes to replace many other target handshapes (Marentette & Mayberry, 2000). It is not known if this type of favored handshape substitution pattern was a typical pattern we may see in sign language acquisition or was idiosyncratic to this child.

Conlin et al. (2000) conducted a study with three deaf children of Deaf parents, videotaping their ASL productions during play sessions from 8 to 11 months, 7 to 15 months, and 9 to 17 months, respectively. There were numerous handshape errors; the children only produced handshapes correctly 25% of the time. The patterns of handshape errors aligned with the unmarked handshapes within Boyes Braem’s model. That is, consistent with previous ASL research, in 58.5% of errors, the child used an unmarked handshape instead of a marked handshape. In a small proportion of cases (8.2%), a child used a more marked handshape in place of an unmarked handshape.

Cheek et al. (2001) gathered data from five deaf children of deaf parents who were acquiring ASL, videotaping them at 7, 10, and 13 months of age while engaged in free play. The authors predicted, following Boyes Braem’s model, that these children would substitute spread handshapes for other handshapes within their first sign productions. This prediction was supported as around 59% of substitutions in early sign errors used a spread hand (similar to an ASL 5) as a replacement for the target handshape. The substitution of a spread hand occurred even in handshapes that were not meant to utilize a spread pattern. The authors noted that the shapes using a single pointed finger or fist were also used as replacements but not as often (9.3% and 9.9% of substitutions, respectively).

Cheek et al. (2001) found that the deletion of one hand within a two-handed sign was the most frequently observed error. This led the authors to suggest that signs that use a unique handshape on each hand are the most difficult to produce. They report that children utilize this pattern in less than 10% of their early signs, whereas the use of two different handshapes simultaneously occurs in 25% of all lexical ASL signs. No data appear to have been reported for this pattern in Auslan. That is, even though Auslan has a two-handed alphabet, the accuracy of complex two-handed signs has not been described to date.

In 2002, Karnopp followed the language development of a single deaf child of deaf parents learning the sign language of Brazil (Libras) between the ages of 8 and 30 months. This child demonstrated expressive use of 20 handshapes, out of a possible 46 handshapes within Libras. The child used the ASL 1, 5, B, and lax hand first (coded by the author as a variation of the ASL F), using these four handshapes by 13 months. These handshapes were also some of the most frequently used handshapes within the corpus.

Karnopp noted that handshape errors accounted for 11% of all phonological errors produced by the child and that unmarked handshapes were used to replace more marked handshapes within their study. Karnopp (2002) also observed the deletion of one handshape within two-handed signs in their Libras corpus. The author counted this as an error of “location,” describing it as deletion of the nondominant or weak hand. The child in their study acquired this location skill and produced signs on the nondominant hand after the age of 2 years.

Morgan (2006) examined video-recorded sign-language productions of a deaf child of deaf parents learning BSL from 19 to 24 months. A large corpus of over 1,000 signs was gathered across this 5-month period. In 2007, Morgan and colleagues re-examined this sample, determining that 41% of handshapes were produced incorrectly. The ASL 1, A, B, and 5 handshapes were the most accurate and were attempted most frequently by this single child.

The researchers noted that errors were more likely to be present with regard to marked handshapes, compared to unmarked handshapes. Furthermore, the child was noted to use unmarked handshapes as replacements for marked handshapes, although the same handshape was not always used as a replacement in every erroneous production. The ASL1 handshape replaced other extended finger handshapes, such as the ASL I handshape, whereas the ASL 5 was used most often to replace curved handshapes such as the ASL C (Auslan bC) and O (Morgan et al., 2007).

In summary, most research that discussed handshape substitution addressed patterns for ASL. There is limited research regarding error processes and the notion of markedness in other sign languages at the time of writing. Information regarding other sign languages is of interest as ASL uses particular handshapes within its lexicon and single-handed alphabet that may not be common to other sign languages. It is conceivable that other sign languages will report different patterns of frequency for marked and unmarked handshapes within their phonological inventory.

Some aspects of past studies into the development of sign language phonology were aligned with Boyes Braem’s model of handshape acquisition in ASL. These studies, however, shared a key limitation, that is, in most cases, their data focused upon a few cases, or a single-child case study that was convenient to their research needs and thus did not represent the development of a group of children or a population of sign users. Extrapolating their findings to other users of that sign language, or to different sign languages, may be problematic. Data specific to the sign language of interest must be analyzed to determine whether similar substitution patterns are present.

Investigating the patterns of handshape substitutions, specifically whether unmarked handshapes replace marked handshapes, is the focus of the current study. This study aimed to describe the handshape substitution and deletion patterns observed in the spontaneous productions of a group of children aged 3–6 years learning Auslan. The majority of the children were DHH and all attended a bilingual–bicultural educational program in Australia. We hypothesized that there may be differences in the frequency and complexity of handshapes within this Auslan corpus, particularly regarding the ASL B, C (Auslan bC), and F handshapes, when compared to Boyes Braem’s model for ASL. Specifically, this study aimed to describe handshape substitutions in Auslan (which has a two-handed alphabet) and compare them to handshape substitution patterns reported in ASL (which has a one-handed alphabet) and BSL literature. The frequency of occurrence of particular handshapes in relation to their markedness will also be considered.

Methods

This study was conducted within a bilingual–bicultural educational early years program (for children aged 3–6 years) in Australia across a 2-year period. In this program, Auslan and English are used interchangeably throughout the day, regardless of the activities, with a staff member in each classroom modelling spoken English and a deaf staff member modeling Auslan. Spontaneous language samples were regularly collected in spoken English, leading to the first author developing a method for collecting samples in Auslan as part of their PhD. Parents provided written consent for de-identified language data to be used for research. Human research ethics were approved by the relevant university ethics board (ID 2056807.1).

Participants

Data were gathered from 44 children between the ages of 3.0 and 6.8 years at the time of sampling. The mean age at first observation was 4.73 years (range = 3.08–6.75; SD = 1.06). The mean age at last observation was 5.43 years (range = 3.47–6.83; SD = 0.85).

Children were included in the study if they used English and Auslan (n = 2) or any other spoken language and Auslan in the home (n = 8). Children were included in the study if they had hearing loss (n = 42), or typical hearing (n = 2), if they had typical motor and cognitive development, or if they had one or more diagnoses other than hearing loss (n = 26; Appendix A). The only exclusion criteria for this study were the presence of a motor and/or physical disability that prevented the child from producing Auslan signs.

The majority of children (n = 41) had parents with typical hearing. Three children had parents with an identified hearing loss, but only two of these children were exposed to Auslan from birth from their native signing parents. Thus, for 42 children, their age at first exposure to fluent Auslan models occurred upon their enrollment in the bilingual–bicultural education program (mean age at enrollment = 3.72 years; range = 1.17–5.53; SD = 1.01).

Materials

The data were gathered using the Handshape Analysis Recording Tool (HART), a tool developed to be used in clinical and education settings to document expressive language samples in Auslan without requiring research annotation skills.

The HART requires the user to write down the equivalent English word (also known as a “gloss”) and document whether the child used the right handshape, in the correct location, with the correct movement and hand orientation. Any errors are noted via a tick or cross in columns for each of these four phonological parameters. The user is able to add further explanations, including any information on nonmanual features, in a comment box. More detailed description is contained within the previous work of the authors (West & Dettman, 2024).

Inter- and intra-rater reliability of the HART was established with a group of 14 mature professionals working with DHH children who were self-reported fluent users of Auslan and were experienced with gathering spoken language samples (West & Dettman, 2024). Coders were trained in the use of the HART and coded 55 video-recorded Auslan samples. Eleven samples were also randomly repeated to examine intra-rater reliability. Using Gwet’s Agreement Coefficient 1, the HART was seen to have “good” interrater reliability for handshape, movement, and location and “very good” inter-rater reliability for orientation. Intra-rater reliability was “very good” across all four parameters.

Procedure

In the present study, the HART was utilized by a group of six deaf educational professionals who were fluent in Auslan to observe the Auslan productions of the 44 children. Spontaneous, novel Auslan productions were documented at the word level using the HART across the school day in a range of natural environments including mealtimes, free play, art, music, and curriculum sessions. This procedure is similar to the use of language diaries by caregivers in home-based settings that do not involve formal testing and sampling at specified time points. This process yielded a sample of 3,166 individual sign tokens across two academic years.

Data analysis

The data from the HART forms were pooled and entered into a spreadsheet for coding and analysis by the first author. Analysis of the full corpus included determining the frequency of particular word classes such as nouns and verbs. This analysis is reported elsewhere (West et al., 2025). Each sign was coded according to the accuracy of the handshape, location, movement, and orientation parameters. The present study focused on the handshape parameter. Future work will focus on the remaining three parameters: location, movement, and orientation.

If the sign contained transitions between handshapes, or two different handshapes were used by each hand, all handshapes were included. Any handshape omissions, additions, or substitutions were coded as handshape errors for that particular handshape. A matrix of correct handshapes versus substituted handshapes was created to establish the number of times a target handshape was produced correctly and how frequently it was replaced by another handshape. The matrix also permitted analysis of which specific handshape had been used as a replacement in each instance. This study focuses upon the patterns of errors that were noted within the corpus; the frequency and accuracy of specific handshapes are the focus of a separate discussion (West & Dettman, under review).

Results

The substitution patterns with the matrix (Figure 1) are discussed below.

Figure 1.

A subset of the error analysis matrix. The number of times that handshape was used to replace another handshape is represented within the boxes; for example, a 5 handshape was used as a replacement for a 1 handshape four times.

Prevalence of Auslan handshapes

A sample of 3,166 individual Auslan signs was gathered; however, as some signs used more than one handshape, there were 3,849 handshapes available for analysis. The most frequently used handshapes were ASL B = 942 times, ASL 1 = 490 times, ASL S = 359 times, ASL 5 = 312 times, and ASL Bent 5 = 170 times. These five handshapes accounted for 59% of all observed handshapes. There were three handshapes, ASL P, Auslan Old 7 and !, that did not occur at all within this corpus (Figure 2).

Figure 2.

Auslan handshapes within the corpus from most frequent to least frequent.

Frequency comparison with Boyes Braem

Five handshapes that occurred within Stages 1 and II of Boyes Braem’s (1990) model were also within the top 10 most frequently occurring Auslan handshapes (ASL B, 1, S, 5, and O). The Auslan H (ASL U) handshape occurred 120 times and was present in Boyes Braem’s Stage III, and the ASL X handshape occurred 148 times and was present in Boyes Braem’s Stage IV. Two of the most frequently occurring handshapes, ASL Bent 5, and Bent B, were not listed within the Boyes Braem model. Several ASL handshapes within Stage III and IV of the Boyes Braem model (ASL D, 7, and T) are not part of the phonological system of Auslan, so they were not observed. The frequency and accuracy of the most frequent Auslan handshapes compared with Boyes Braem’s model are presented in Table 1.

Table 1.

Auslan handshape data arranged in order of Boyes Braem (1990) proposed stages of acquisition in ASL.

Auslan handshape	Number of occurrences	Ranking within Auslan corpus	% incorrect	% correct
Boyes Braem Stage I
1	490	2	8.37	91.63
S	359	3	10.03	89.97
5	312	4	10.26	89.74
C	100	11	17.00	83.00
gO	98	14	14.29	85.71
A	83	16	8.43	91.57
7	42	20	4.76	95.24
Boyes Braem Stage II
B	942	1	5.20	94.80
0	119	8	4.20	95.80
F	100	12	10.00	90.00
Boyes Braem Stage III or IV
2	108	10	13.89	86.11
H	120	7	25.83	74.17
X	148	6	16.22	83.78
Not present in Boyes Braem model
Bent 5	170	5	7.06	92.94
Bent B	111	9	6.31	93.69

Accuracy of Auslan handshapes

Within the 3,849 handshapes observed in this corpus, only 374 (or 9.71%) were produced incorrectly. The top five most frequent handshapes were highly accurate; ASL B was 94.8% correct; ASL 1 was 91.6% correct; ASL S was 89.9% correct; ASL 5 was 89.7% correct; and ASL Bent 5 was 92.9% correct. Several low-frequency Auslan handshapes were also highly accurate. Three handshapes that occurred less than five times within the entire corpus, ASL ILY, Auslan M, and 9, were 100% correct, and the Auslan Mid handshape was 75% correct. Several of the handshapes that were produced with low accuracy occurred infrequently. The five Auslan handshapes that were produced with the lowest accuracy by percentage are presented in Table 2. Further discussion of the frequency and accuracy of the handshapes within this corpus is contained elsewhere (West & Dettman, under review).

Table 2.

The five most inaccurate handshapes in this Auslan corpus, by percentage and the error patterns that were observed.

Auslan handshape	Number of occurrences	% incorrect	% correct	Error types
R	1	100.00	0.00	Replaced by Irish K
I	31	45.16	54.84	Replaced by 1, Bent 6, B, 6, Y, X, omitted twice
Irish K	13	38.46	61.54	Replaced by S, Bent 5, H, Flat bC, A
8	6	33.33	66.67	Replaced by H, Flat gC
Irish H	10	30.00	70	Replaced by Flat gC; Flat bC; omitted once

Omissions of Auslan handshapes

The 374 handshape errors included 42 instances (11.2%) where a handshape was omitted. The ASL 1 handshape was omitted most often (nine times), closely followed by the ASL B handshape (eight times). The ASL 5, S, and X handshapes were all omitted four times, and the Auslan gO was omitted three times. The Auslan 2, H, and ASL I handshapes were each omitted twice. The remaining instances of deletions only occurred once for ASL F, Auslan flat bC, Bent 2, and Irish H handshapes.

Patterns of substitutions in Auslan handshapes

The remaining 332 handshape errors (88.7%) were handshape substitutions. The substitution patterns that were observed for the top five most frequently occurring handshapes within this corpus are presented in Table 3.

Table 3.

The substitution patterns noted for the five most frequently used Auslan handshapes within this corpus.

	Auslan handshape error	Number of occurrences	% of all handshape substitutions
B	B to 1	11	3.31
	B to 5	8	2.41
	B to Bent 5	7	2.11
	B to S	6	1.81
	B to bC	4	1.20
	B to 2	3	0.90
1	1 to B	5	1.51
	1 to S	5	1.51
	1 to 6	5	1.51
	1 to 5	4	1.20
	1 to 7	4	1.20
	1 to X	4	1.20
	1 to 2	2	0.60
	1 to A	1	0.30
	1 to Bent 2	1	0.30
S	S to 5	7	2.11
	S to 1	5	1.51
	S to B	4	1.20
	S to Bent 5	3	0.90
	S to 0	2	0.60
	S to 7	2	0.60
	S to 6	2	0.60
	S to C	1	0.30
	S to 2	1	0.30
	S to H	1	0.30
	S to Bent B	1	0.30
	S to gO	1	0.30
	S to X	1	0.30
5	5 to B	7	2.11
	5 to S	5	1.51
	5 to C	2	0.60
	5 to Bent 5	2	0.60
	5 to gO	2	0.60
	5 to 1	1	0.30
	5 to 0	1	0.30
	5 to 2	1	0.30
	5 to H	1	0.30
	5 to Bent B	1	0.30
	5 to flat bC	1	0.30
	5 to 6	1	0.30
	5 to 12	1	0.30
	5 to X	1	0.30
	5 to Irish K	1	0.30
Bent 5	Bent 5 to 5	3	0.90
	Bent 5 to 1	1	0.30
	Bent 5 to B	1	0.30
	Bent 5 to 0	1	0.30
	Bent 5 to 2	1	0.30
	Bent 5 to 6	1	0.30
	Bent 5 to gO	1	0.30
	Bent 5 to Bent 2	1	0.30
	Bent 5 to X	1	0.30
	Bent 5 to M	1	0.30

The ASL 1 handshape was used as a replacement for other handshapes most frequently (68 times), followed by ASL 5 (41 times), ASL S (32 times), ASL B (29 times), and ASL Bent 5 (23 times; Figure 1). In addition to the ASL 1, 5, and S handshapes, other handshapes within Stage I of Boyes Braem’s model were used in substitution patterns as follows. The Auslan 7 (ASL L) handshape was used as a replacement for other handshapes 18 times, the Auslan bC (ASL C) handshape was used as a replacement 17 times, the Auslan gO handshape was used as a replacement 9 times, and the ASL A handshape was used as a replacement 4 times. With Stage II, the ASL O handshape was used as a replacement for other handshapes 13 times, and the ASL F handshape only replaced another handshape once.

The most frequently occurring substitution patterns included the ASL 1 handshape. The ASL 1 was used as a replacement for an ASL B (11 times), was used as a replacement for an Auslan H (ASL U; 10 times), and was used as a replacement for an ASL X (7 times). The ASL 5 was used as a replacement for an ASL S (seven times) and was used as a replacement for an ASL B (eight times). Conversely, the ASL B was also used as a replacement for an ASL 5 (seven times). The ASL Bent 5 was used as a replacement for an ASL B (seven times). There were 102 of 332 handshape substitutions (30.7%) that only occurred once. The ASL ILY, Auslan gC, 3, 9, Mid, and Irish H handshapes were never used to replace other handshapes within this corpus.

Discussion

Past research has focused on ASL and BSL, but due to differences in lexicons, handshapes, and alphabets, these past studies may not be relevant to the Australian context. Examining differences in Auslan compared to previous research is a first step toward considering developmental sequences that may exist within Auslan. Whether a handshape is marked or unmarked or used frequently (within the alphabet or lexicon) are factors that could affect its accuracy in Auslan compared to other sign languages.

Comparisons of Auslan handshapes with ASL data

There were differences in the Auslan handshape substitution patterns compared to Boyes Braem’s model of ASL acquisition. In this corpus, the ASL B was the most frequent, accounting for around a quarter of all handshapes. It was produced with high accuracy (94.8%), suggesting that it was an easy handshape for the children to articulate. The ASL B was also used as a replacement for other handshapes 29 times, accounting for 8.73% of handshape substitutions. In contrast to the current findings, the ASL B handshape was considered to be in Stage II, not Stage I of Boyes Braem’s model. However, this high frequency was consistent with Henner et al.’s (2013) description of ASL that led to their proposal that the ASL B handshape should be considered unmarked and frequently used.

The ASL 1 handshape was the second most frequently used handshape after the ASL B. While it was produced accurately 91% of the time, it was also the most frequent handshape within substitution patterns. The ASL 1 was used as a replacement for other handshapes 68 times, replacing a range of other handshapes, including other unmarked handshapes such as the ASL B in this corpus.

The frequency with which the ASL B and 1 handshapes were used is perhaps not surprising; the ASL B is thought to be used in more than a quarter of all lexical Auslan signs (Johnston & Schembri, 2007). The ASL B is frequently used as a base hand within Auslan lexical signs and within several letters of the Auslan alphabet such as H, J, M, N, and R. The ASL 1 handshape is used in a range of lexical signs but is also used to produce vowels in the Auslan alphabet. It should be noted that an ASL 1 would also be used to denote pronouns in Auslan, although there were limited occurrences of pronouns within this sample.

Within this corpus, the ASL 1 was often used by the children as a replacement handshape (68 times), accounting for one-fifth of all substitution errors. This may possibly be related to its ease of production, in addition to the increased salience of the ASL 1, and its use as a natural gesture, as pointing behavior occurs from a young age (Colonnesi et al., 2010).

The finding that the ASL 1 was the most frequent replacement handshape in Auslan was not aligned with Boyes Braem’s ASL model. Boyes Braem, instead, had suggested that the ASL 5 handshape would be most frequently used within substitution patterns. In contrast, the ASL 5 handshape was the second most frequent replacement handshape, and it only made up 12% of all substitutions within this Auslan corpus.

It should be noted that the ASL 5 was used as a replacement handshape more often than the ASL B. However, a reciprocal pattern was noted whereby the ASL 5 would replace the ASL B, but the ASL B was also used to replace the ASL 5 with similar frequency.

The current study demonstrated that the ASL A, Auslan bC (ASL C), 7 (ASL L), and gO handshapes had different prevalence and substitution patterns to Boyes Braem’s Stage I handshapes. First, they were not the most frequent handshapes. The Auslan bC (ASL C) handshape was the 11th most frequently used by the Australian children, the Auslan gO handshape was 14th most frequently used, the ASL A handshape was 16th most frequently used, and the Auslan 7 (ASL L) handshape was the 20th most frequently used. Second, they were not always used as replacement handshapes within substitution patterns. The Auslan 7 and bC were used 18 and 17 times, respectively, to replace other handshapes. The Auslan gO was used to replace other handshapes nine times; however, the ASL A was only used four times across all 332 substitution errors to replace other handshapes.

This infrequent use of the ASL A handshape was unexpected, given that it was one of the three preferred ASL handshapes in the single-child case study described by Marentette and Mayberry (2000), and the ASL A handshape was also reported by Bonvillian and Siedlecki (2000) as a frequent replacement for other handshapes. There may be frequency differences between Auslan and ASL, whereby the ASL A correlates to a vowel and is repeatedly used within lexical signs (Henner et al., 2013), whereas the ASL S is used more commonly as a fist shape in Auslan. To this point, the ASL S was the third most frequently occurring handshape in this Auslan corpus (359 occurrences; 9.32% of all handshapes) compared to the ASL A, which made up only 2.1% of the 3,849 handshapes.

The ASL O and F handshapes within Stage II of Boyes Braem’s model also had a different prevalence and use within substitution patterns. The ASL O handshape was the 8th and the ASL F handshape was the 12th most frequently occurring handshape within this study. The ASL O handshape was highly accurate, and only produced incorrectly 4.2% of the time. The ASL O was used 13 times to replace other handshapes; however, the ASL F was used only once within the entire Auslan corpus to replace another handshape.

The current study demonstrated that some patterns of Auslan handshape use did not align with Boyes Braem’s model but were similar to other past ASL research. For example, the ASL F was not used by the children at all in the Siedlecki and Bonvillian (1993, 1997). In the majority of the ASL studies, the ASL F handshape was infrequently used and was not used often to replace other handshapes. Within Henner et al.’s (2013) data on the frequency of occurrence of handshapes in ASL, the ASL F was, in fact, less frequent than many of the handshapes proposed within Stage III of Boyes Braem’s model.

There were nonetheless some similarities in the prevalence and substitution patterns in the Auslan corpus and the Boyes Braem model. Boyes Braem described that handshapes within Stages I and II were more likely to replace handshapes in Stages III and IV, and this was also seen in the Auslan corpus. For example, the second most frequent substitution error observed in the Auslan corpus was the ASL 1 (Boyes Braem Stage I), replacing the Auslan H (ASL U; Boyes Braem Stage III).

Comparisons of Auslan handshapes with BSL data

Past research regarding the frequency of BSL handshapes used within a large sign BSL corpus from a single-child case study were described by Morgan (2006) and Morgan et al. (2007), but limited data were available on typical BSL substitution patterns.

Morgan (2006) described differences in BSL compared to Boyes Braem’s model with regard to which handshapes were marked. The ASL C (Auslan bC) and F handshapes, for instance, which were part of Stages I and II of Boyes Braem’s model and therefore would be considered unmarked in ASL, were suggested to be marked in BSL. The Auslan data within the present study supported Morgan’s proposal and were aligned with their data, which is unsurprising given that these sign languages are from the same sign family (Johnston & Schembri, 2007). That is, both the Auslan bC (ASL C) and ASL F occurred 100 times within this corpus, each representing only 2.59% of the handshapes used, and could be described as marked. While they were produced with high accuracy (83% and 90% respectively), they were less likely to be used as replacements for other handshapes when compared to the other unmarked Auslan handshapes.

Additionally, consistent with Morgan’s BSL corpus, handshape substitutions did not always occur as a direct 1:1 replacement of one handshape for another every time in these Auslan data (Figure 3). Rather, there was a subset of handshapes that were often used in place of other handshapes.

Figure 3.

Example of an inconsistent replacement pattern seen for the Auslan B handshape.

In order, the ASL 1, 5, S, B, and Bent B handshapes were used most often to replace other handshapes. They were also used to replace each other, despite four of these handshapes being unmarked and being within the first two stages of Boyes Braem’s model.

Marked and unmarked handshapes in Auslan

It was not possible to match these Auslan data against all of Rice’s (2007) criteria due to the absence of developmental normative data for Auslan (e.g., that an unmarked handshape would be acquired earlier in a developmental sequence) and overall high accuracy within the corpus (e.g., that an unmarked handshape would be able to be produced more accurately than a marked handshape). Thus, the notion that unmarked handshapes would be used to replace other handshapes was applied. The proposed criteria for considering a handshape to be unmarked in this study were therefore: high frequency, high accuracy, and used to replace other handshapes in substitution patterns.

Considering these criteria, these Auslan data suggested some differences with regard to markedness and frequency compared to the ASL handshapes in Boyes Braem’s model. It appeared that ASL A, and F, and Auslan bC (ASL C), and 7 (ASL L) were less frequent and possibly marked in Auslan, whereas the ASL B was unmarked and highly frequent. Therefore, the unmarked Auslan handshapes derived from the current study data would be the ASL B, 1, 5, and S. The Auslan handshapes that are clearly marked would be the ASL R, P, and ILY and Auslan Old 7, 8, 9, Mid, M, and !.

Auslan handshape omissions

Rice (2007) proposed that marked phonemes are more salient than unmarked sounds, and unmarked sounds are more likely to be deleted in phonological processes in spoken languages; for example, weak syllables will often be deleted. Similarly, one of the proposed criteria for markedness in sign languages is that these unmarked handshapes are more likely to be omitted (Brentari, 1998).

In these data, the ASL 1 and B handshapes were deleted twice as often as any other handshape, accounting for 40% of all handshape omissions. These were both purported to be unmarked, and other unmarked handshapes, the ASL 5 and S, were also deleted somewhat frequently. The ASL X and I were the only marked handshapes to be deleted more than once. It is possible that the ASL 1 and B were more likely to be deleted due to their lower salience as unmarked handshapes as proposed by Rice (2007). Indeed, there is some limited evidence from this corpus of a process whereby a sign that should involve a transition between multiple handshapes is collapsed to a single handshape. However, more data on omission patterns in Auslan are required to determine if this is a pattern present in the developmental acquisition of signs containing multiple handshapes.

Limitations

These Auslan data were derived from a sample of children who were enrolled within a single bilingual–bicultural program. The age at Auslan enrollment and duration of Auslan exposure were not variables that were controlled in this study, and there were no exclusion criteria. That is, all children enrolled in the program who produced expressive signs were recruited for this 2-year observational study of their spontaneous signs. Their ages ranged from 3.0 to 6.8 years; therefore, these data cannot be directly compared with several of the previous research studies. More data are needed regarding the handshape productions of children younger than 3 years of age, but there are very few native signing children exposed to Auslan from birth who are available for participation in such longitudinal observational studies.

The children’s spontaneous handshape productions were documented using the HART in this study but were not specifically tested using formal tools or checklists, as there are no standardized assessments of Auslan handshapes currently available. These data from the present study represent a first step toward describing the handshape productions of young children learning Auslan. Further work is required to create standardized checklists suitable for receptive and expressive Auslan lexicons and handshapes.

Future directions

The HART was useful to document signs in the present study, but it is not Auslan specific. It may be implemented by sign language users to document the productions of young children learning any sign language. This would enable more data to be gathered regarding handshape productions across a range of sign languages.

Gathering larger repositories of emerging sign is a critical first step toward determining typical language development in sign languages. At present, we are unable to recognize atypical development as there are limited normative data available regarding the trajectory that can be expected for children learning sign languages. Without normative data, clinicians and educators are unable to establish which substitution patterns may be shared across different sign languages and, more importantly, which errors may be considered typical development within that language. The current study determined that some patterns of Auslan handshape substitutions were aligned with ASL and BSL data, but many patterns of Auslan use were quite different from ASL. It also determined that certain handshapes proposed to be frequent and unmarked in ASL are less frequent and may be marked in Auslan. Either the Auslan, ASL, and BSL patterns of use are quite disparate, or the data sets were too small to make strong generalizations. Analysis of larger data sets will allow clinicians and educators to determine whether a child is progressing in their Auslan development as they should.

A repository of sign language samples analyzed according to handshape would inform the development of assessment tools such as a closed set of stimulus words in sign languages (much like an articulation test assesses sound production in spoken languages). Such a test would permit clinicians to identify which children have a need for intervention services and to monitor typical versus atypical errors patterns and ensure that they have resolved.

Conclusion

This novel study has provided the first description of the handshapes used by 44 children acquiring Auslan. Analysis of 3,849 handshapes within this data set suggested that there were differences in the frequency of handshapes and the patterns of errors in Auslan compared to past literature for ASL, in particular Boyes Braem’s model for ASL. The Stage I ASL handshapes in Boyes Braem’s model were not produced as frequently in Auslan and were not used as often to replace other Auslan handshapes in this corpus.

Overall, unmarked Auslan handshapes were produced with higher frequency than marked handshapes and were used to replace other handshapes more often than marked handshapes. The findings of this study demonstrated that the Boyes Braem model may not be an exact match for Auslan and that a new model of unmarked and marked handshapes in Auslan development is needed. The HART may be used to gather these data to create an Auslan-specific handshape acquisition model.

Author contributions

Erin West (Conceptualization, Data curation, Formal analysis, Investigation, Methodology, Project administration, Validation, Visualization, Writing—original draft, Writing—review & editing) and Shani Joy Dettman (Supervision, Writing—review & editing)

Conflicts of interest: None declared.