Statistical learning in infant language development

Abstract

AbstractChildren learn language over such a short span of time and with such seeming ease, that many have assumed they must master language by means of a language-specific device. Artificial languages provide a useful tool for controlling prior learning and for manipulating specific variables of interest. This approach has resulted in a wealth of findings regarding the learning capabilities of children. Infant artificial language learning has become synonymous with statistical learning because of the emphasis in much of the work on learning statistical regularities. However, not all cases of artificial language learning entail learning statistical structure. For instance, some learning requires generalisation of relational patterns. This article explores statistical learning in language development in infants, phonological learning (discrimination of speech sounds, learning phonotactic regularities, phonological generalization), word segmentation, rudiments of syntax, generalization of sequential word order, category-based abstraction, and bootstrapping from prior learning.