*3.2. Handshape*

First, we present results for the analysis of classifier handshapes. As a reminder, we coded handshapes that were used to depict that the referent in some verb event were coded as entity classifier handshapes. Handshapes used only in depicting the extension of the object (in other words, not in a verb event) were coded as size and shape specifier handshapes. We recorded 91 tokens of some type of classifier depicting the bottle from Cena signers, with at least one from every participant. Fourteen of these were handling classifiers used in a construction depicting the act of opening a bottle to specify the object; as we believe this is in the process of becoming a conventionalised lexical sign for *bottle*, we exclude this as a classifier variant. This leaves 77 tokens: 44 entity classifier handshapes used in verb events and 33 SaSS handshapes used only to depict the extension of the bottle. We observed five variants of entity classifier handshapes (Table 3) and four SaSS handshapes (Table 4) in the Cena data, displayed below with number of tokens and frequency, as well as the number of participants who used the variant.


**Table 3.** Cena entity classifier handshapes for the bottle stimulus.


#### **Table 3.** *Cont.*

**Table 4.** Cena SaSS classifier handshapes for the bottle stimulus.


For the Libras signers, we recorded 56 tokens of some classifier depicting the bottle, with at least one from every participant. In this case, the breakdown was 28 entity classifier handshape tokens used in some verb event and 27 SaSS handshape tokens used only to depict the extension of the referent. The 3 attested entity classifier handshape variants are shown in Table 5, and the 4 SaSS variants in Table 6.


5

 0.18  3

**Table 5.** Libras entity classifier handshapes for the bottle stimulus.

**Table 6.** Libras SaSS classifier handshapes for the bottle stimulus.


Quantitively, we find more handshape variants in entity classifiers in Cena than in Libras. All handshapes used in entity classifiers in Libras form a subset of those used in the same context in Cena. Four handshape variants were attested in SaSSes in both Cena and Libras, with three of the four handshapes being the same across the two languages. The least frequent handshape in each varied only in its degree of openess, the thumbopposed handshape appearing in Cena entity classifiers, and the slightly open version in Libras entity classifiers. At a glance, the results seem to support our prediction for Hypothesis 2 (that of greater intersigner variation in Cena) when considering entity classifiers, since more handshape variants were attested in Cena than Libras. For size and shape specifier handshapes, our prediction was not borne out as the number of handshapes attested across the languages was the same. An evaluation of Hypothesis 1 (that of greater handshape complexity in Cena classifier handshapes) requires assigning complexity scores and determining whether there is a statistically significant difference in the distribution of scores across the two languages, which follows in Section 4.1.
