LEARNING SPEECH-LIKE SIGNALS FROM A SKEWED CONTINUOUS DISTRIBUTION

摘要

An important issue in studying the evolution of language is whether the cognitive mechanisms used for learning language are domain-general or language-specific. If there are language-specific cognitive adaptations, they must have evolved biologically. Recently, researchers have started to investigate how learning biases shape language by using artificial language experiments, most notably through modelling cultural evolution in the laboratory (Kirby, Cornish, & Smith, 2008). In our contribution, we present the first results of a new line of individual artificial language learning experiments, in which we test whether humans have different learning biases for continuous linguistic signals versus continuous non-linguistic signals; the current results are from the linguistic condition only.