PhD & Post-doc Projects

Segments and rules: a comparative study into the computational mechanisms underlying language acquisition

PhD students: Andreea Geambasu (LUCL), Michelle Spierings (IBL), Raquel Garrido Alhama (UvA ILLC)
Supervisors: Claartje Levelt (LUCL), Carel ten Cate (IBL), Jelle Zuidema (UvA ILLC)

This joint, NWO-funded project brings together researchers from the Leiden University Centre for Linguistics (LUCL), the Institute for Biology Leiden (IBL) and the Institute for Logic Language and Computation (ILLC) of the University of Amsterdam. Together we focus on the properties of statistical- and rule-learning mechanisms in relation to the acquisition and evolution of language. We investigate to what extent these learning mechanisms are unique to humans, or to human language, by comparing the acquisition of grammatical rules in two vocal-learning species: humans (infants) and songbirds (zebra finches).


A series of comparable artificial grammar learning (AGL) experiments will expose both species to auditory input streams containing patterns from which rules can be learned. We will investigate a number of domain-general or domain-specific factors that may influence learning, such as perceptual and memory primitives, or phonological biases. We will also study how these cues are weighted as more or less important at different stages of development and when interacting with other cues. Our colleagues in the ILLC will bring together the human and zebra finch data and develop computational models of artificial language learning that, on the one hand, predict optimal learning behavior and, on the other hand, test how different constraints and biases influence learning.


See also:
Assessing the linguistic abilities of a songbird: syntax detection and use in zebra finches (Taeniopygia guttata)
Geamba┼ču A, Ravignani A, Levelt CC. Preliminary Experiments on Human Sensitivity to Rhythmic Structure in a Grammar with Recursive Self-Similarity. Front Neurosci. 2016 Jun 28;10:281