*4.1. Handshape*

In order to find out whether the handshapes observed in the data are articulatorily easy or simple, we return to the quantification of complexity formulated by Brentari et al. (2012), which we outlined in Section 1. All handshapes attested within entity classifiers in Cena and their resulting finger and joint complexity scores are shown in Table 7, listed in descending order of frequency. The three most frequent entity handshapes have the lowest possible finger and joint complexity scores. Only the two least frequent handshapes have a finger or joint complexity score above the lowest possible value. Such a distribution upholds the general prediction of phonological markedness (Battison 1978) that there should be an inverse relationship between frequency and complexity; that is, the more complex a handshape, the less frequent we expect it to be. Conversely, we expect the most frequent handshapes to be the least complex. This prediction is borne out in the results.


**Table 7.** Entity classifier handshapes in Cena with finger and joint complexity scores.

1 Though not specified in Brentari et al. (2012), the closed-fist handshape meets all the listed criteria for low complexity: it is among those first acquired by learners (Boyes Braem 1981), one of the most frequently occurring crosslingustically (Rozelle 2003), and has a relatively simple structure in the prosodic model of representation (Brentari 1998, p. 112).

Aside from influence from a preference for ease, we can speculate further about the distribution of the data and the presence of the two least frequent handshapes. If it is the case that over time signers may choose to substitute iconic but difficult handshapes for less

iconic easier ones, this alone would not explain the presence of in the data, which is a departure from both iconicity (having no obvious semantic motivation) and finger and joint simplicity. Looking at the classifier within the phrase provides clues. The handshape was only attested (*n* = 3 from two different participants) in cases where the sign WATER—which is specified for the same handshape, although the thumb is not visible in the first image in the following example—preceded the classifier. An example sequence is shown in Figure 12. This appears to be a case of handshape assimilation. Similar to spoken languages where a feature of a particular sound (such as place of articulation, or voicing) may spread onto its neighbour, sublexical features of a particular sign may also spread onto adjacent signs. In this case, the handshape in WATER remains throughout the following classifier.

**Figure 12.** Handshape assimilation in an entity classifier.

Similarly, the curved handshape only appears as an entity classifier when preceded by a SaSS depicting the bottle's cylindrical shape using the same handshape (*n* = 4 from one participant), as in Figure 13. Of course, a phonological explanation is not the only type possible. Such variants could be motivated by reasons of semantics17, in that the handshapes signers select may be motivated by semantic properties determined by certain experiences (or lack thereof) with objects, or certain semantic properties the signers feel to be salient in the object in the stimulus. To tease this apart, one could elicit depictions of different types of the same object, perhaps forms varying in colour, material, or intended use. This would not only foreground different semantic associations, but ideally also encourage varied lexical items preceding or following the classifiers to further investigate a hypothesis of assimilation. However, when we revisit one specific production, additional evidence for assimilation emerges. In Figure 14, a signer produces an account of the bottle

falling. It begins with the sign WATER, with its handshape remaining over a string of several subsequent signs including a lexicalised sign, two classifiers, and an indexical point.

DRINK, which appears in the middle of this string, is a conventionalised sign with the handshape, ye<sup>t</sup> the presence of the extended index finger in this production is evident. We take this as robust evidence for assimilation, and thus apply the same hypothesis to the

case of the curved handshape in entity classifiers, given its similar distribution only following another classifier with the same handshape. It seems that in entity classifiers

that used and , any constraints on markedness or complexity were violated by virtue of other influences from phonology—assimilation. In the case of WATER, the influence of phonology in pulling sign form away from faithfulness to semantics or iconicity is particularly clear.

**Figure 13.** Handshape assimilation of in an entity classifier.

**Figure 14.** Handshape assimilation of across several signs.

Turning to SaSS handshapes in Cena (Table 8), we see two handshapes with high joint complexity scores, both depicting the cylindrical shape of the bottle. The more frequent curved handshape has only high joint complexity since all the fingers are selected and act in unison. The less frequent thumb-opposed handshape scores highly both in finger and joint complexity. There are many handshapes available to signers to depict curvature: , , to list a few in addition to those in the data. Iconic depiction of curvature using handshape is likely to tip the balance out of favour with articulatory ease. As the quantification models of both Brentari et al. (2012) and Ann (2006) show, extended, closed, and flat handshapes all require less articulatory effort than curved ones. Within this small subcategory of handshapes in the data that depicted curvature, we still see the easiest one prevail—the curved handshape. In the choice of handshape to depict the form of the bottle overall, iconicity may have won the trade-off initially, but within the variants selected for that choice, pressures from ease endure.

**Table 8.** SaSS classifier handshapes in Cena with finger and joint complexity scores.


**Table 9.** Entity classifier handshapes

Next, we consider complexity in the Libras data. Complexity scores for Libras entity handshapes are given in Table 9, and SaSSes in Table 10. All entity classifier handshapes had all fingers or the index finger selected, resulting in the lowest possible finger complexity score. The curved handshape is the only entity classifier handshape to receive a high joint complexity score. Every token of this entity classifier directly followed a SaSS depicting the object's curvature, of the form shown in Figure 13. This was a common strategy among Libras signers, to first depict the object's extension before depicting the verb event: 82% of Libras entity classifiers involved in verb events were directly preceded by a SaSS that depicted the size or form of the bottle, e.g., CL:SaSS(height) CL:TALL-OBJECT-FALL, as opposed to only 30% of Cena entity classifiers. The greater relative consistency with which Libras signers used this ordered construction may have had an effect on the distribution of handshapes with regards to assimilation, considering the evidence for assimilation in the same environment in Cena. Among the SaSSes, handshapes receiving high finger or joint complexity scores were involved in depictions of curvature. As the most frequent SaSS handshape fell into this category, the curved handshape, it seems iconicity and

semantics won this particular trade-off. with

 finger and joint complexity scores.

 in Libras


**Table 10.** SaSS classifier handshapes in Libras with finger and joint complexity scores.


Last, we summarise the distribution of complexity scores (Figures 15 and 16) to return to Hypothesis 1—that of greater complexity in Cena classifier handshapes. Overall, classifier handshapes in Cena do not exhibit greater complexity relative to Libras. The languages showed a very similar distribution of finger complexity, in that handshapes

across both languages in both types of classifiers were overwhelmingly of low finger complexity. In both languages, the depiction of curvature explains the presence of high finger complexity handshapes, which comprised roughly the same small proportion of SaSS handshapes across Cena (9%) and Libras (7%).

**Figure 15.** Finger complexity score by classifier type and language.

**Figure 16.** Joint complexity score by classifier type and language.

For joint complexity, the picture is not so similar, but the two languages still share some tendencies (Figure 16). Whilst entity classifier handshapes were predominantly of low joint complexity across both languages, we see more entity classifier handshapes with higher joint complexity in Libras. SaSS handshapes saw the highest proportion of handshapes with high joint complexity in both languages. The curved handshape was the most populous across both languages, accouting for the large proportion of scores of 3. Handshapes with the highest joint complexity (with a score of 4) were those that are stacked, also depicting the curvature of the referent.

In terms of our expectation to find greater complexity in Cena, there was no statistically significant difference between the two languages for finger or joint complexity, across both types of classifier. A Fisher's exact test was used in lieu of a chi-square, since the data set contains low numbers of observations in some cases. In comparing variance between languages for each type of complexity in each type of classifier, the p-value was greater than 0.05 in all cases. The high complexity scores in both groups shows plainly how the semantic property of curvature affected the distribution of handshape complexity in the trade-off between ease (or simplicity) and iconicity. Moreover, we argue the distribution of medium finger complexity and high joint complexity handshapes in entity classifiers in Cena can be partially accounted for by phonological assimilation. In some sense, this finding leads us even further away from our initial hypothesis, suggesting that in the absence of signs with marked handshapes earlier in the phrase, in other words all else being phonologically equal, like Libras signers Cena signers aim for simple unmarked

handshapes. The prevalence of the curved handshape does not appear to exemplify this idea, seeing as it has a relatively high joint complexity score. However, if we recall that it is crosslinguistically frequent and is considered unmarked, it appears more as a discrepancy between using a model of complexity primarily based on representational simplicity, as opposed to models based on markedness or ease of articulation. That is, measures of complexity based on representational complexity (such as that of Brentari et al. 2012) do not capture certain realities of usage that likely affect handshape distribution in classifiers,

including the pervasiveness of the curved handshape as a manual configuration for grasping objects outside of the linguistic system. The prevalence of such a handshape in the data may seem surprising when considered through the lens of complexity as defined by representational complexity alone, but its ubiquity both as a handshape crosslinguistically, and as a configuration for the non-linguistic manipulation of objects goes far in accounting for this.

Next, we address Hypothesis 2, which predicted greater intersigner variation in Cena. For size and shape specifier handshapes, we see the same number of variants in both languages, with a similarly proportional distribution. The picture is slightly different with entity classifier handshapes. Cena signers produced five variants in contrast to the three from Libras signers, and distribution of these variants patterns differently between the groups. In Cena, three variants accounted for 89% of tokens, with the proportion of each variant being fairly similar. For the Libras data, one variant accounts for over half the tokens (53%), showing greater consistency between signers in their selection of a handshape to represent the referent. Overall, the results do not demonstrate greater variation in Cena for handshapes in SaSSes, but for entity classifier handshapes we see more variants in Cena and more equal weighting between them in terms of proportion. However, considering that two variants (interestingly, the most complex variants) may be accounted for by assimilation, the presence of more variants does not necessarily mean that Cena signers have a larger and less conventionalised repertoire of handshapes available for depiction of whole entities.
