Cognitive biases for sequential learning in language: A functional and evolutionary approach

Abstract

The cognitive mechanisms involved in forming the structure of language are the subject of much discussion. The world's languages tend to show a strong pattern for various properties, aptly termed 'language universals'. The present study focuses on the universal known as the head ordering principle (HOP), which also includes the principle of recursive consistency. The HOP is the overwhelming tendency for languages to place modifiers in the same order in relation to the heads of their respective phrases, with recursive consistency specifically concerning constituent types that can be recursive. This type of ordering is said to maximise ease of parsing in working memory. However, a central question regarding language universals such as the head ordering principle is how they evolved in the first place, though the view supported here is that of language evolving to fit general sequence learning biases. To explore this, an artificial non-linguistic grammar learning experiment investigated the learnability of certain grammatical features that obeyed the head ordering principle to various degrees. It was found that a VO verb order interacted with recursive consistency to create the most learnable languages in the experiment. The full output of participants' testing phase was also used to model a transition matrix for language evolution, of which the stationary distribution showed a stronger bias for recursively consistent grammars over recursively inconsistent grammars.