*4.2. Movement*

Recall that the tendency to encode movement path and manner simultaneously was found more commonly in gesturers and earlier cohorts of NSL, whilst sequential encoding

emerged increasingly with later cohorts. A linguistic context different to that of latercohort signers for both Cena and Libras led us to Hypothesis 3, wherein we predicted a preference for simultaneous manner and path encoding in both Cena and Libras. Indeed, we found that 93% of Libras responses encoded path and manner simultaneously, as did 80% of Cena responses. In both languages, results pattern more akin to those of gesturers and first-cohort NSL signers, where temporal iconicity prevails in trade-offs against the potential articulatory difficulty of encoding manner and path together. Even in depicting the girl running in a circle, the majority of Cena and Libras signers chose an upright person classifier with running legs whilst moving their arm in a circle, despite this being articulatorily difficult (Mandel 1979).

However, results diverge from those of hearing gesturers in Senghas et al.'s (2004) study in the mere presence of linear segmentation (in Cena, Libras, and NSL alike), supporting the idea that there is something inherantly linguistic about such a process. We acknowledge possible influence from the fact that the stimuli clips were not designed with an analysis of movement in mind18. As such, clips may not be balanced in their likelihood to elicit either manner or path in a given clip or in the set as a whole. An entity moving along a marked path but in a predictable manner (e.g., a ball rolling in a zig-zag) may be more likely to elicit path than manner, for example. Indeed, we found that in a small number of cases, some tendencies emerged based on the stimulus item itself. As an example, the girl running in a circle was the most likely to elicit simultaneous encoding; 85% (29 out of 34) of responses across both groups depicted this using a person classifier running in a circle. The woman running was the most likely to be represented sequentially, with 27% (9 out of 33) of responses across both groups encoding in such a way. There may be some effect from telicity at play here, whereby telic predicates prefer sequential encoding. The woman running across the screen could be construed as the most goal-oriented of the stimuli we analysed. Conversely, the woman walking elicited the strongest preference for simultaneous encoding across both languages—both in proportion and the overall number of tokens, as almost every signer produced a response using this strategy. Many signers depicted the walking as continuous or aimless, with slow movement and pursed lips—both of which are attested markers of continuative aspect (e.g., Oomen 2016, who describes such marking in Sign Language of the Netherlands). The distribution of strategies in all responses by stimulus can be seen in Tables 11 and 12. Overall, both groups preferred simultaneous encoding, but we saw greater variation among Cena signers for the girl running in a circle and the woman running.

**Table 11.** Distribution of Cena encoding strategies by stimulus.


**Table 12.** Distribution of Libras encoding strategies by stimulus.


An ANOVA analysis of encoding strategy and age uncovered a statistically significant correlation (*p* = 0.03), in that older signers showed a greater tendency to encode sequentially (Figure 17). The signers in the upper age brackets in our study are roughly second cohort, as the first signer of Cena was born in 1949. This shows a pattern distinct from results of Senghas et al. (2004), in that it is older or earlier-cohort signers who display a preference for

sequential encoding. As many homesigners with distinct idiolects come together, segmentation and linearisation may be but one effect of the rapid restructuring deaf community sign languages will undergo in their initial stages of emergence. Such restructuring may be phonological, grammatical, or otherwise. These are not the conditions under which Cena has emerged. Libras on the other hand does have an analogous genesis, but this initial period of restructuring likely took place well over a century ago. As such, our results do not sugges<sup>t</sup> that the linear sequential encoding of motion events is a property that emerging languages in general will acquire as they develop, but perhaps rather a product of some specific environmental criteria that NSL, among its second and third cohorts, seemed to meet, and that Cena and Libras as they exist at this moment in time do not. Concerning the tendency of later-cohort child signers of NSL to reanalyse manner and path as sequential, Senghas et al. (2010) observe that "it is as if [the] children see structure where there is none". Perhaps because there are no child signers in our data, and as such we are not seeing widespread segmentation as a by-product of children restructuring language from an unconventionalised input, temporal iconicity is retained.

**Figure 17.** Violin plot of age and encoding strategy showing frequency along the X axis.

One question that emerged during analysis was whether the domain of non-manual features could encode the manner of motion as the hands encoded the path. In response to a stimulus clip of a plastic bag floating, multiple signers traced a path straight downwards with their hand (contrary to the floating motion in the clip) whilst puffing their cheeks or blowing. This stimulus clip did not form part of our chosen stimuli for the current study, and as such we made no decision on whether this non-manual information was aspectual, adverbial, or could otherwise be subsumed into a model of sequential movement encoding, but remaining open-minded as to what can be considered a viable slot for movement encoding will undoubtedly be pertinent in any future studies.

## **5. Conclusions**

Overall, we see a very similar selection of handshape variants used in entity classifiers and size and shape specifiers across Cena and Libras. Thus, the distribution of complexity scores for classifier handshapes of each language largely resembled one another, by virtue of the same or similar handshapes populating each data set. Without additional data it is difficult to tell whether the similar attested handshapes across both groups are a product of the influence of ease of articulation, or semantic categorisation—perhaps both (one can imagine how a signer might choose an index finger extended handshape over a middle finger extended handshape for an upright long thin object, though both are iconic). What the results do tell us is that in this instance, there is no evidence for Cena signers prioritising

iconic representation at the expense of ease, in lieu of conventionalised classifiers. We do, however, present evidence for another aspect of phonology having a direct influence on the handshapes selected for classifiers: assimilation. This phenomenon highlights the robustness of handshape as a phonological component in signs even in young sign languages that potentially lack systematic phonological organisation. We show how such assimilation may influence or even dominate handshape selection, in this case winning out over other influences from articulatory ease, iconic representation, or semantics that converge in a trade-off for handshape selection. This has implications for our findings on variation. Overall, we saw more handshape variants used by Cena signers in entity classifiers, but no difference between the groups for SaSSes. On the surface, a larger repertoire of handshapes in entity classifiers may sugges<sup>t</sup> that classifier handshape is, for now, less restrained by conventionalisation in Cena. As we have shown, assimilation may account for this discrepancy in number of variants between groups. We conclude that we do indeed see one influence of phonology on the selection of classifier handshapes (as the handshape of WATER spread over a whole phrase by one Cena signer), just not from the sourcewehadanticipated.

Last, our results show a pattern different to that of NSL signers in the encoding of movement (Senghas et al. 2004), in that second- and third-cohort Cena signers prefer to encode manner and path simultaneously, akin to gesturers and signers of urban sign languages, including the Libras signers in our study, as well as first-cohort NSL signers. This does not reflect the preference for sequential linearisation of path and manner found in younger NSL signers, those of comparable cohorts to the Cena signers in our study. A relevant variable that differs between Cena and NSL signers is that, unlike Cena signers of a comparable cohort, the vertical language input<sup>19</sup> that second-cohort NSL signers received was from previously unconnected homesigners. The different preferences between Cena and Libras on one hand and NSL on the other in encoding manner and path could well be a result of the rapid restructuring a language undergoes in response to relatively disorganised vertical input from disparate homesigners, in which some motivation other than temporal iconicity tips the balance of the trade-off between encoding strategies. Cena signers did not receive such vertical input. Such an explanation would account for why later-cohort NSL signers seemed to be the exception in their preferences for linear sequencing, and why Cena and Libras signers patterned akin to all other groups. We also found a significant correlation between age and movement encoding strategy, with older Cena signers segmenting manner and path in motion events at a greater rate than younger Cena signers. Such inter-signer variation (in this case along the axis of age) invites future questions about whether this pertains to other domains such as syntax. We have seen in other studies using the Haifa clips that different preferences in word order and argumen<sup>t</sup> structure disambiguation emerge at different rates among cohorts of signers of emerging sign languages when compared to those of the corresponding national sign language of their country (Meir 2010). Considering the existence of several studies on young sign languages using the Haifa clips for this aim (Meir et al. 2017; Ergin et al. 2018), we hope that word order elicitation along these lines may yield further opportunities to draw direct comparisons between an incipient sign language and a national sign language on the same basis.

**Author Contributions:** Conceptualisation, A.N., D.S. and A.A.-S.; methodology, A.N. and D.S.; data glossing, J.C.N.F., D.S.d.S. and A.A.-S.; data analysis D.S.; writing—original draft preparation, D.S.; writing—review and editing, D.S., A.N. All authors have read and agreed to the published version of the manuscript.

**Funding:** This research received no external funding.

**Institutional Review Board Statement:** This study was conducted with approval from the UCL Research Ethics Committee (Project ID: 19269/001, approved on 2 February 2021).

**Informed Consent Statement:** Informed consent was obtained from all subjects involved in the study. Signed (in lieu of written) informed consent has been obtained from the participant(s) to publish this paper.

**Data Availability Statement:** Supplementary data in the form of video examples of all discussed phenomena and corresponding glosses can be found at: https://github.com/ucjudst/Cena-Data.

**Acknowledgments:** We wish to thank the inhabitants of Várzea Queimada for their time and participation, as well as our team in Brazil: Carlos Douglas Carvalho de Macêdo and Iago Pedro Mendes Pires Veras for Cena images; and Telma Franco, Bruna da Silva Neres, Nádia Fernanda Martins de Araújo, Dona Silvana Barbosa, and Marcilene Barbosa for demographic information. We are very grateful to Gabrielle Hodge and Nick Palfreyman for their insights into this project in its initial stages, and for those of Anne Senghas into the encoding of motion events. We thank the Centre for Sign Linguistics and Deaf Studies, CUHK for the handshape font (available at http://www.cslds.org/v4/resources.php?id=1, accessed 1 September 2021).

**Conflicts of Interest:** The authors declare no conflict of interest.
