**3. Results**

#### *3.1. Study 1: Participant Characteristics*

3.1.1. Do Homesigners and Communication Partners Tend to Express Iconicity by Using a Handling Handshape or an Object Handshape?

First, we want to clarify that we did not expect to see a strong overall preference for one iconic handshape type over the other, because many other features outside of the participant can influence preferences, primarily item characteristics, which we address in Study 2. While we do report overall preferences (i.e., the proportion of Handling, Object and combined Handling+Object responses across all items for each participant), it is important to keep in mind that features of specific items may also influence handshape preferences and are not captured by looking at overall preference.

There appears to be no universal preference across types of participants (e.g., homesigners or their communication partners) or across participant groups for one of the handshape types. Indeed, individuals varied greatly in whether they showed a preference, as well as which handshape type they preferred when they demonstrated a preference. We proceeded

to undertake more detailed analyses of potential patterns within participant groups. Of the 11 homesigners, 5 showed a Handling preference, 4 of which were child/adolescent homesigners, as determined by binomial distribution tests (Table 4). In order to carry out the binomial distribution tests, we only compared two categories: Handling vs. Object; on trials in which a participant used a combined Handling+Object form, the response was counted as half a case of Handling and half a case of Object. For example, for a participant who produced 12 Handling handshapes, 7 Object handshapes, and 8 combined Handling+Object forms, their responses were simplified as 16 cases of Handling and 11 cases of Object handshape in the binomial test. Of the 24 communication partners, 5 showed a reliable Handling preference, and 2 showed a reliable Object preference. Of the 8 hearing non-signing adults, 1 showed a reliable Handling preference. Overall, the majority of participants did not show a handshape preference. Of the participants who did show a preference, a majority had a preference for Handling handshapes.

**Table 4.** Mean iconic handshape preferences of child/adolescent and adult homesigners, communication partners (CPs) and hearing non-signers (related and unrelated). H+O stands for combined Handling+Object handshape. See Appendix A for individual and group preferences.


3.1.2. Do Members of Families with Homesigners Share This Preference with Each Other?

Next, we looked at group preferences to assess how communicative familiarity and using a homesign system with others might influence handshape preferences. With regard to the homesigning families, 3 families showed a Handling preference, and the other 6 families showed no clear preference (see Appendix A). Additionally, neither of the groups of hearing people who had no regular communication with a homesigning family member (i.e., the all-hearing family and the group of unrelated hearing people) showed a clear preference. Note that, of the homesigning families that showed an overall Handling preference, two were families with child or adolescent homesigners, and one was a family with an adult homesigner.

We also examined the conventionality of iconic handshape preferences; instead of looking specifically at handshape type, we calculated the likelihood of participants within a group producing the same handshape type, at the item level, regardless of whether the handshape was Handling or Object. In order to calculate the conventionalization for each group, we compared each family member's responses to each other in a pairwise fashion. For each item, when two family members produced the same handshape type (e.g., both used Handling), the pair was assigned 1 point. If they produced different responses (e.g., one used Handling and one used Object), they were assigned 0 points. If one family member used a combination form (e.g., one used Handling+Object, while one just used Handling), they were assigned 0.5 points. For each family member pair, point values were totaled, and the percentage of similar response types was calculated. Once the similarity percentages were calculated for each pair of participants in each family group, the percentages were averaged for the entire group. For example, in Adult Homesigner 4's family, the homesigner and brother produced the same handshape type on 60% of items, the homesigner and mother produced the same handshape type for 63% of items, and the brother and mother produced the same handshape type on 56% of items, leading to an average of 60% conventionalization for the family. In other words, the members of Adult

Homesigner 4's family shared handshape preferences with one another, on average, 60% of the time. Overall, the conventionalization of iconic handshape type (i.e., Handling or Object) ranged from 48% to 71% across the groups (Table 5). Members of adult homesign families (n1 = 4) were significantly more conventionalized than child/adolescent homesign families (n2 = 5) (U = 1, *p* < 0.05, Mann–Whitney U test for small Ns).


**Table 5.** Average conventionalization of all family members for each group.

3.1.3. Does a Participant's Age at the Time of Test, Age at Which They Begin Using a Homesign System, or the Number of Years They Have Used the System Affect Handshape Type and Conventionalization?

We assessed the relationships between the communication partners' conventionalization and (i) age at time of testing, (ii) years of experience with a homesign system, and (iii) age of first exposure to a homesign system. We used age two years as the "starting point" for homesign systems and based our calculations of age of exposure to the homesign system and years of experience with a homesign system on that value (see Carrigan and Coppola 2017). For the homesigners, years of experience using a homesign system corresponded to their age minus 2 years; for family members, years of experience using a homesign system corresponded to the family member's age, minus the homesigner's age, minus 2 years (the initial years of homesign development). Similarly, the variable of age of exposure to the homesign system also took this two-year period of initial, early homesign development into account. For example, the mother of a homesigner (aged 26) who was 54 at the time of testing would have been exposed to the homesign system at age 30 (i.e., (54 − 26) + 2 = 30). Younger siblings were assigned a value for age of exposure corresponding to their age (assuming they were born two or more years after the homesigner). Note that the age of the homesigner at the time of testing is a good approximation of how long the family has been co-constructing and using the homesign system.

We found moderate inverse correlations between conventionalization and age at the time of testing (rs = −0.49, *p* < 0.05, Spearman's rho) and between conventionalization and years of experience using a homesign system (rs = −0.42, *p* < 0.05). That is, higher conventionalization was associated with being younger when tested and with fewer years of experience using a homesign system. A weak inverse correlation was found between conventionalization and age of first exposure (rs = −0.22, *p* > 0.05), showing that higher conventionalization was somewhat associated with being exposed to a homesign system from a young age. While these findings do not necessarily contradict one another, they do raise some questions which are considered in the discussion section. Linear regressions revealed that the communication partners' age at testing (t = −2.51, *p* < 0.05; F(1, 22) = 6.35, *p* < 0.05, R<sup>2</sup> = 0.22) and their years of experience interacting with a homesigner (t = −2.33, *p* < 0.05; F(1, 22) = 5.41, *p* < 0.05, R<sup>2</sup> = 0.20) each significantly predicted conventionalization. The communication partners' age of first exposure to homesign did not predict conventionalization.

We also found differences across participants in how they used both Handling+Object handshapes within a single response. Adult homesigners produced this Handling+Object response type more often (M = 27% of items, SD = 0.18) compared to child homesigners (M = 6%, SD = 0.02), communication partners (M = 7%, SD = 0.09), and hearing non-signers (M = 8%, SD = 0.08).

#### *3.2. Study 2: Item Characteristics*

We now turn to the item analysis, to determine which individual objects and which semantic classes of object (clothing, grooming/cosmetic items, utensils, and handheld tools) exhibited a greater degree of conventionalization.

#### 3.2.1. Do Some Stimulus Items Elicit Higher Conventionality in Iconic Handshape Preferences?

In order to examine item-specific biases, we conducted binomial distribution tests and found that many of the responses elicited by certain items were not at chance (50%) and, in fact, were biased towards one or the other iconic handshape class. Of the items not at chance, more items were more likely to have a Handling preference than Object (Table 6), and tools were more likely to exhibit a higher degree of conventionalization. None of the items at chance were classified as traditional tools.

**Table 6.** Binomial tests revealed that overall, 12 items were more likely to elicit a Handling handshape and 4 items were more likely to elicit Object handshapes, while 11 items were statistically at chance.


3.2.2. Which Factors Are or Are Not Associated with Greater Conventionality?

The participant families' degree of conventionalization varies; however, certain items may lend themselves to being consistently gestured or signed with one type of iconic handshape over the other. Participants converged on a handshape type for 22 of 27 items; that is, over 50% of the group members used the same handshape type. Traditional tools all had higher conventionalization (all >70%), while makeup items all showed lower conventionalization (Figure 2). As expected, the items that have higher conventionalization are also the items that are more likely to have a handshape bias.

We also considered lexical frequency as a possible motivation for conventionalization. Because we did not have homesign frequency counts, we used ASL, English, and Spanish word frequencies as proxies. We obtained measures of ASL sign frequency, English word frequency, and Spanish word frequency from Corpus del Español (Davies 2015). While ASL, English, and Spanish word frequencies were all moderately to strongly correlated with each other, none of the word frequency measures were correlated with the degree of conventionalization observed for a particular item. While these functional/semantic group-

ings did reveal some trends toward greater (e.g., traditional tools) or lesser (e.g., makeup items) conventionalization, they were fairly weak and do not seem to be explanatory.

**Figure 2.** Items ordered by overall conventionality of handshape type across groups. Each category of object is shown in a different color. Dots represent the percentage of participants who used the most common handshape type for each item (Handling or Object). Lines represent standard error.
