Complexity trade-offs and equi-complexity in natural languages: a meta-analysis

Abstract

In linguistics, there is little consensus on how to define, measure, and compare complexity across languages. We propose to take the diversity of viewpoints as a given, and to capture the complexity of a language by a vector of measurements, rather than a single value. We then assess the statistical support for two controversial hypotheses: the trade-off hypothesis and the equi-complexity hypothesis. We furnish meta-analyses of 28 complexity metrics applied to texts written in overall 80 typologically diverse languages. The trade-off hypothesis is partially supported, in the sense that around one third of the significant correlations between measures are negative. The equi-complexity hypothesis, on the other hand, is largely confirmed. While we find evidence for complexity differences in the domains of morphology and syntax, the overall complexity vectors of languages turn out virtually indistinguishable.

Societies of strangers do not speak less complex languages

“…Another advantage of our approach is the use of the two complexity metrics on a comprehensive global sample. Previous empirical studies either focused on one grammatical domain, such as the number of cases (20) or verbal synthesis (21), or they included a wide variety of features corresponding to different coding procedures and interpretations of complexity [see studies based on WALS (23), such as (6,22,26)]. Moreover, we followed a systematic focused approach toward delineating fusion and informativity, which allows us to make principled decisions about the choice of the Grambank features for each metric and avoid including features that did not align with either of the described interpretations.…”

“…Despite our focus on phonologically fused markers rather than morphological inflections, our metric of fusion is still comparable with what other studies measured as morphological complexity. For instance, according to the metric based on a variety of WALS features, Turkish ranked extremely high (0.775) and Vietnamese (Austroasiatic) was the lowest-scoring language (0.141) (26). In our data, we do not quantify fusion scores for Vietnamese but observe a divide between Turkish and some Austroasiatic languages lacking all fusion features: Turkish is assigned 0.44 with the maximally bound language Tariana (Arawakan) reaching as high as 0.7, while the fusion score of Thavung (same Vietic branch as Vietnamese), Hu, Prai, and Rumai Palaung is equal 0.…”

“…Hence, combining several of these dimensions into one metric of grammatical complexity [e.g. (22,26)] may not shed light on the relationship between grammatical structures and the chosen sociodemographic variables reflecting exotericity.…”

Abordagem baseada em Aumento de Dados para Avaliação Automática de Leiturabilidade

“…No entanto, há uma preocupação na escolha do idioma intermediário, i.e., da língua Y, porque afeta diretamente o resultado do processo de RT e, consequentemente, a manutenção do rótulo inicial do texto. Ao mesmo tempo, determinar se um idioma é mais simples ou complexo que outro ainda é uma questão controversa e dependente do contexto; inclusive, estudos recentes apontam que não é possível medir se uma língua é mais complexa que outra (BENTZ et al, 2022).…”

Language evolution in a context of increased socialization: the self-domestication approach

“…This is the equi-complexity hypothesis, which is commonly construed as involving a trade-off principle too, so that, if e.g., a language's syntax is more complex, its morphology will be simpler. Recent statistical analyses using extensive language databases have given support to the equi-complexity principle (Bentz et al, 2022). A common explanation for this is that core features of languages mostly depend on how our brain is configured and works, which is quite similar in all human beings.…”