Pattern deduction in linguistically attested and unattested grammars

Abstract

A popular hypothesis in linguistics posits that language learners are biologically predisposed to learn structures attested in human language – for example, a hierarchically nested phrase structure, while eschewing hypotheses for linguistically unattested structures – for example, one consisting of non-consecutive, linearly alternating “constituents”. The current study explores the robustness of such a predisposition within a controlled artificial language learning task as well as a non-linguistic, general puzzle-solving task. We find evidence that suggests learners more easily acquire linguistically attested hierarchically structured patterns compared to unattested non-hierarchical ones within the non-linguistic task, but not within the language task. We discuss the puzzling nature of this finding, and some work in progress to further unravel the source of this result.