**4. Discussion**

In this paper, we used the phenomenon of iconic handshape contrasts (Handling vs. Object), a distinction used grammatically by many sign languages around the world, as a lens through which to examine the development of conventionalization in emerging sign languages. Specifically, we examined iconic handshape preferences in deaf homesigners and their hearing communication partners when naming objects used as tools (see Table 7 for a summary). Study 1 revealed that participant characteristics influence preference for handshape type. Among families with homesigners, we detected no universal handshape preference in either the homesigners or the communication partners, though several participants (many of them gesturers) tended to use more Handling forms. Within families with homesigners, we also observed variable levels of shared preferences for handshape type and the degree of conventionalization for handshape type. Among homesigners' families, time using the homesign system seems to be important, because adult homesigners' families demonstrated a higher degree of conventionalization than child homesigners' families. Study 2 found that traditional tools (e.g., scissors, hammer) tended towards higher conventionality than other items that were not traditional tools (e.g., mascara, hat). Proxy measures of lexical frequency did not show any correlation with the degree of conventionalization.


**Table 7.** Results summary.

These findings sugges<sup>t</sup> that there is no universal handshape preference among the different participant groups when naming tools, and that similarity of handshape type varies even within families. Both individuals and groups varied in terms of iconic handshape preference. The fact that a family did or did not have a homesigner or, taken as an isolated factor, communicative familiarity (i.e., families versus unrelated group of people), does not promote similarity or conventionalization of iconic handshape type; however, there is evidence that, among families with a homesigner, the longer the family uses the system, the more conventionalized it becomes (adult homesign families are more conventionalized than child homesign families). It may be that individual preference initially drives this type of iconic handshape preference.

#### *4.1. Variation in Iconic Handshape Preference and Conventionalization*

While more individuals (i.e., homesigners, communication partners, and hearing nonsigning adults) produced more Handling than Object handshapes in their responses, as a group, fewer than half showed an overall Handling preference. In other words, while 25 of the 43 participants (58%) showed an individual Handling preference, only 4 of the 11 groups (36%) showed a clear overall Handling preference. The fact that a majority of the data points were from communication partners (i.e., gesturers) and that responses were overall more likely to use Handling handshapes aligns with the finding that generally gesturers, compared to signers, prefer Handling handshapes when labeling tools (Padden et al. 2015). The next step will be to see whether homesigners (particularly adult homesigners) and communication partners use the same specific handshapes, instead of merely the same type of handshape iconicity. In line with existing research, homesigners' responses exhibit more finger and joint complexity than the responses of hearing non-signers (Brentari et al. 2012; Coppola and Brentari 2014), but explicit comparisons of homesigners with communication partners have ye<sup>t</sup> to be completed.

The fact that conventionalization varied greatly among homesigning families, ranging from 48% to 71% conventionalization, suggests that this phenomenon of patterned iconicity may emerge later. This is supported by evidence that families with adult homesigners were overall more highly conventionalized than families with child/adolescent homesigners. In groups where this was not the case, such as the family of child homesigner 2 and the group of unrelated hearing people (both of whom had some of highest conventionalization rates), something else might be influencing handshape conventionalization. We included the hearing non-signing family and the group of unrelated hearing non-signers in order to pull apart

the influence of having a homesigner in a family versus generally interacting within a family; however, it was somewhat surprising that the unrelated hearing non-signers displayed such high conventionalization. Note that our measure of conventionalization does not take into account the complexity of the handshapes produced. Thus, the hearing people in our study may be producing the most straightforward and least complex handshape forms. In their context as hearing people who do not regularly communicate with a deaf person, they are not burdened with additional iconicity demands (e.g., using patterned iconicity systematically) and may therefore consider fewer, less complex handshape options. Since hearing non-signing gesturers already tend to both produce less finger and joint complexity than homesigners (Brentari et al. 2012; Coppola and Brentari 2014) and prefer Handling handshapes (Padden et al. 2015), this may partly be responsible for the convergence. In other words, the Handling handshape preference may also be generated independently in each individual rather than due to convergence in a group. In addition, it follows that they would exhibit restricted options for types of handshapes, and their productions, therefore, may appear more conventionalized, even if little actual conventionalization has occurred with the apparent similarity merely reflecting similar strategies. In other words, although the unrelated hearing non-signers may appear to be highly conventionalized, they are likely using the simplest forms at their disposal and, given their limited handshape options, just happen to be using the same simple handshape types.

This line of reasoning may also explain the finding that the less experience a communication partner has with a homesign system, the more conventionalized they are; essentially, these communication partners may be using the most straightforward or simplest approach without much true conventionalization happening (e.g., everyone converging on a shared complex handshape versus each person producing the simplest responses, which coincidentally happen to be similar). Previous work by Singleton et al. (1993) comparing the productions of the homesigner called David with those of his hearing sister showed that her gestures more closely resembled those of non-signers rather than those of her homesigning brother. Frequent and prolonged interaction between communication partners and homesigners does not appear to be enough to conventionalize gestures in a homesign system. As David was documented correcting his sister's gesture forms, so has Coppola observed adult homesigners correcting their family members, providing similar evidence of standards of form in homesign systems that communication partners do not always pick up on. The relationship between age and experience (discussed more below) could perhaps be explained by the association of younger age and simpler forms, and less to do with practice using the system. Uniformity due to simple forms may be masquerading as conventionalization.

#### *4.2. Age and Experience as Factors in Conventionalization*

We noticed some trends related to the age of the homesigner, mainly that, overall, adult homesigners used the combined Handling+Object form more than child homesigners, and that families with adult homesigners were significantly more conventionalized than families with child/adolescent homesigners. This tendency for a combined Handling+Object form to be produced by adult but not child homesigners is related to the finding that adult homesigners more often produced multiple signs or gesturers for single responses compared to child/adolescent homesigners, who typically produced just one sign or gesture per response. While the Handling+Object form clearly made up a portion of the adult homesigners' multiple signs/gestures per response, adult homesigners also tended to produce multiple of the same type of handshape in a single response, which child homesigners did not do frequently. Adult homesigners' more common use of a combined Handling+Object form has also been observed in independent groups of adult Nicaraguan signers and child homesigners in Nicaragua. Martin et al. (Forthcoming) found that 10 out of the 11 signing Nicaraguan adults in their study produced a combined Handling+Object form at a similar level to that of the adult homesigners in our study, but it was not robust in the responses produced by child homesigners or hearing gesturers. This suggests that this combined form, which is present in adult homesigners and persists among Nicaraguan signers from the second and third cohorts (that is, among signers who entered the community after 1983), albeit at low levels, requires maturational time to develop and is not a response to a communicative context experienced by homesigners in which they are concerned about not being understood.

We also would like to note that we specifically refer to this response type as a combined form, not a compound form, because we did not rigorously assess each response to determine if they were true compound signs. Published criteria for identifying compound signs are limited, and the classification is commonly determined by intuition or judgments (i.e., Meir et al. 2010) or by comparing two-sign combinations in which the first stem is reduced to stand-alone versions of the same signs (Liddell and Johnson 1986). Note that existing criteria are difficult to implement in emerging languages, especially homesign systems, in which the forms themselves may be in flux. Further, traditional acceptability judgments and intuitions are difficult (though not impossible) to elicit from homesigners. In addition, the data collected in this sample did not allow the opportunity to compare single signs and two-sign combinations. Since this combined category included both simultaneous and sequential Handling+Object forms, we decided to refer to them as combined forms rather than teasing out which could be compounds and which were not. In some instances, participants produced a "sandwich form" such as Handling, Object, then Handling again. For example, Adult Homesigner 1 produced a Handling+Object + Handling form when labeling the rake. This strategy has the advantage of making it clear that this is the name of the item, not describing the action carried out by the item. These combined forms tend to be produced very quickly, with little to no pause between the signs. While this category of signs may be a candidate for compound forms, that analysis is outside of the scope of this paper.

Another age-related trend was that, among the families that showed a general preference for producing Handling handshapes, most were families with child homesigners. We did observe a preference for Handling handshapes among the communication partners of one of the adult homesigners; this family had the greatest number of communication partners (5, as opposed to 2 to 3 communication partners like the other adult homesigners). As hearing gesturers, who tend to produce more Handling responses, these communication partners drive up the mean of the overall use of Handling handshapes within the family. Might regular interactions with communication partners influence homesigners to be less consistent with their handshape preferences? Goldin-Meadow et al. (2015) found that homesigners' first inclination, before elaborating more extensively for their hearing communication partner, was to produce a pattern comparable to the pattern produced by ASL and NSL signers (i.e., more consistent). However, communicating with hearing partners, who sometimes struggle to comprehend homesigners' productions (as demonstrated in Carrigan and Coppola 2017), may make homesigners' systems appear less structured. This finding might provide an alternate explanation to the phenomenon of adult homesigners using the combined Handling+Object form more often than child homesigners; perhaps they are used to having to elaborate on initial signs when communicating with hearing partners, therefore they may be inclined to use a combined Handling+Object form in order to be more clear and to avoid having to repeat or elaborate. However, this explanation does not account for the persistence of this combined form among NSL signers over 30 years after the emergence of the community. Clearly, more research is warranted.

Homesign systems have the potential to conventionalize, albeit more slowly than a shared sign language used by a deaf community, since each system is only used primarily by a single homesigner (Richie et al. 2014). We found that participants' age at time of testing and their years of experience using a homesign system were negatively correlated with conventionalization. The younger a communication partner was and the fewer years they spent using a homesign system, the higher the degree of conventionalization. Given Goldin-Meadow et al. (2015) findings that homesigners' subsequent responses are much more variable than their first responses, it is possible that interacting with

communication partners, who are not using the homesign system as their primary system, may hinder or even deconventionalize certain homesign patterns. It is also possible that, among a relatively small number of family members who interact regularly and whose communication is embedded with specific contexts, the lack of a conventionalized form is less of an issue. Future research should investigate how age and amount of time a person has been using a system influence conventionalization in homesign systems, in addition to other factors such as communicative closeness and the amount of time that communication partners actually spend interacting with a homesigner.

#### *4.3. Item-Specific Biases and Proxy Measures*

Although conventionalization varied within and across family groups, we did find that certain types of items were, overall, more highly conventionalized than others. Specifically, tools (e.g., scissors, hammer) had higher conventionalization than makeup (e.g., mascara, lipstick) or items of clothing. Similar findings of tools having higher convergence than other categories of items such as animals or food for homesigners from Nebaj, Guatemala, have also been reported (Horton 2018). This observation led us to consider word or gesture frequency; since not everyone uses or converses about makeup, perhaps that is why makeup items had generally lower conventionalization than more widely used items (e.g., scissors, spoons, socks). Unfortunately, it was not feasible to obtain a homesign frequency measure for each item for each family, so we used word frequency measures from ASL, English, and Spanish (all of which were correlated with each other) as a proxy for the frequency of use of such items in homesign systems. We found that conventionalization was not correlated with any measure of word frequency (or ASL iconicity); however, this does not necessarily mean word usage is not at all related to homesign conventionalization. Instead, it is possible that none of the proxy measures used here are actually representative of the frequency of such items in homesign systems. In order to truly understand the relationship between conventionalization and frequency, future studies should obtain iconicity ratings and a measure of homesign frequency from those using the homesign system and directly compare the conventionalization and usage in the system.

#### *4.4. More Questions and Future Directions*

In future work, we will study whether homesigners and communication partners use the same specific handshapes, in addition to the same type of handshape iconicity (i.e., Handling or Object). This would sugges<sup>t</sup> that there is something shared among homesigners and communication partners that is at the level of phonetic form. Previous research (Brentari et al. 2012) showed that homesigners' responses show higher selected finger complexity than gesturers' responses; however, communication partners' productions have not been analyzed. We do not ye<sup>t</sup> know if the productions of communication partners align more with the complexity levels of homesigners or with those of gesturers. Very preliminary analyses of the current data sugges<sup>t</sup> that families that share a general iconic handshape type do not always produce the same specific handshape, for example, in response to the stimulus item eliciting 'saw', family members may all produce different versions of an Object handshape (e.g., B-handshape, H-handshape, or 1-handshape which all resemble a saw). To illustrate, Adult Homesigner 1 and his brother used the same general handshape type (i.e., Handling, Object, Handling+Object) for 19 items but only used the exact same specific handshape for 7 of those items (37%). In contrast, two unrelated hearing non-signers also used the same iconic handshape type for 19 items but used the same exact handshape for 12 of those items (63%). Further analysis will investigate whether the reported conventionalization of iconic handshape preferences is related to the conventionalization of specific handshapes.

Another question for future work is: Do conventionalization and similar handshape preferences actually improve comprehension between homesigners and communication partners? Previous research suggests that communication partners are not very good at comprehending homesign utterances that are shown to them without any context. Younger family members (e.g., siblings, especially younger ones) scored better on comprehension than older family members (e.g., parents). Deaf native ASL signers, who were not familiar with the homesign systems but did have lifelong experience perceiving and communicating in the visual modality, were actually the best at comprehending the homesign descriptions (Carrigan and Coppola 2017). Perhaps using similar handshape patterns could facilitate homesign comprehension and should be investigated further. It would also be useful to see how consistent personal handshape preferences are by gathering longitudinal data using these stimuli.
