Compositional structure can emerge without generational transmission

Abstract

Experimental work in the field of language evolution has shown that novel signal systems become more structured over time. In a recent paper, Kirby, Tamariz, Cornish, and Smith (2015) argued that compositional languages can emerge only when languages are transmitted across multiple generations. In the current paper, we show that compositional languages can emerge in a closed community within a single generation. We conducted a communication experiment in which we tested the emergence of linguistic structure in different micro-societies of four participants, who interacted in alternating dyads using an artificial language to refer to novel meanings. Importantly, the communication included two real-world aspects of language acquisition and use, which introduce compressibility pressures: (a) multiple interaction partners and (b) an expanding meaning space. Our results show that languages become significantly more structured over time, with participants converging on shared, stable, and compositional lexicons. These findings indicate that new learners are not necessary for the formation of linguistic structure within a community, and have implications for related fields such as developing sign languages and creoles.

Evolving artificial sign languages in the lab: from improvised gesture to systematic sign

“…These findings are consistent with previous work investigating the independent effects of transmission (Kirby et al, 2008;Smith & Wonnacott, 2010) and interaction (Fay & Ellison, 2013;Theisen et al, 2010), as well as their combination (Kirby et al, , 2015 in the lab. We also suggest that our results are consistent with studies that have found systematicity to emerge without transmission (Nölle et al, 2018;Raviv et al, 2019;Winters et al, 2018). In theses cases, other manipulations (e.g.…”

Compositional Hierarchical Structure Evolves through Cultural Transmission: An Experimental Study

“…It is probable that the inclusion of communication and thus of the explicit goal of arriving at a shared system for communication results in conscious design by language users more than in Experiment 1. Nevertheless, the degree of structure increases as languages are transmitted through a learning bottleneck and thus, similarly to Theisen, Oberlander, and Kirby (2010) and Theisen-White et al (2011) in the graphic modality, results show that a certain degree of structure can emerge during communicative interaction but it is through iterated learning that it accumulates (see also Raviv et al 2018 for a similar cumulative effect with an expanding lexicon and a turnover of communicative partners) .…”

“…Our findings reveal that when pressures to generalise and to be expressive are both present, languages become hierarchically compositional, which maximises the language's learnability without jeopardising its expressivity. Moreover, our study suggests that the trade-off between lernability and expressivity pressures is not only present intergenerationally in cultural transmission but also within a single generation during communicative interaction (Raviv et al 2018;Winters et al 2018). In comparison to an artificial pressure against ambiguity during individual production, the inclusion of communicative interaction facilitates the evolution of linguistic structure.…”

“…Altogether, these results suggest that a coordination pressure in communicative interaction contributes significantly to the emergence of linguistic structure. The possibility of coordination between participants leads to a very early emergence of structure in Experiment 2 (see also Raviv, Meyer, & Lev-Ari 2018;Winters et al 2018). With a shared goal to communicate accurately, participants might prioritise the establishment of conventions with partners to bootstrap communicationeven at the expense of faithful reproduction of the learned language.…”

Larger communities create more systematic languages

“…While most iterated learning studies use the z-scores provided by the Mantel test for the correlation described above [43,44], z-scores were inappropriate for our design since they increase with the number of observations, and our meaning space expanded over rounds. Therefore, we used the raw correlations between meanings and strings as a more accurate measure of systematic structure [47,48].…”