Predictions

The conventionalisation of classifier form in languages over time is a process that allows signers to rely on convention to depict referents rather than only iconic resemblance, inviting the possibility for forms to become phonologically reorganised under the pressure of articulatory ease in cases where the two are indeed opposing forces. Of course, there are other forces influencing classifier form such as semantics, but pressures of ease function within such influences (there are articulatorily more and less difficult ways of picking out the same semantic property). Given the tendency for signs—including classifiers—to submit to pressures of articulatory ease as a system increases in time-depth and becomes more conventionalised, and given the holistic depictions found in village sign languages in lieu of robust systematic phonological structure (Sandler et al. 2011; Hou 2016), we predict that iconic depiction will triumph in any trade-off against articulatory ease. Our first hypothesis is as follows:

#### **Hypothesis 1 (H1).** *Cena classifiers will exploit handshapes of greater complexity than Libras.*

Since it is unlikely that a systematic level of phonological structure has emerged in a language of such an age and sociocultural profile, and that close-knit communities can tolerate grea<sup>t</sup> variation (Wray and Grace 2007) at higher rates than their national counterparts (Meir and Sandler 2019) over long periods of time (Meir et al. 2012), our second hypothesis is:

#### **Hypothesis 2 (H2).** *Cena classifier handshapes will exhibit greater intersigner variation than Libras.*

Last, we turn to the encoding of motion events. Whilst Cena signers have not ye<sup>t</sup> been grouped into distinct cohorts by researchers, they are not homesigners; most of them grew up with an existing language model. However, this linguistic input is different to that of later-cohort NSL signers, who showed a greater preference for the linear sequencing of motion events relative to first-cohort signers. The vertical linguistic input of later-cohort signers was that of various unconnected homesigners whom the establishment of a school had brought together. Although we believe the majority of Cena signers in our study to be roughly second cohort, their language is not undergoing the intense restructuring of second-cohort NSL, or others in creolisation language contexts (though this may have been partly in play with the establishment of Libras over 150 years ago in its early development

with the foundation of INES). Recall also that linear sequencing of motion events has been observed in children acquiring a sign language. As all signers in our study are well past the acquisition period, and given that Cena does not share its context of emergence with NSL and the subsequent type of restructing that follows, we predict Cena signers will prefer simultaneous encoding of manner and path. We also predict Libras signers will prefer this strategy, in line with data from signers of Spanish Sign Language. Our final hypothesis is as follows:

**Hypothesis 3 (H3).** *Both Cena and Libras signers will exhibit a preference for simultaneous encoding of motion events.*

## *1.3. Language Profiles*
