The goal of this research is to examine the bases of age-dependent learning effects in two domains, spoken language learning in humans and song learning in zebra finches. Learning in both domains involves a critical or sensitive period during which the species-typical behavior is acquired, followed by a loss of plasticity that limits further learning. The main hypothesis to be explored is that these sensitive/critical period phenomena arise from changes in learning capacity that can be understood in terms of the computational principles that govern such PDP networks and their biological correlates. The critical/sensitive period is hypothesized to reflect the operation of a network with intrinsic capacities to encode certain types of stimuli and learn from experience, modeled using a statistical learning procedure. Learning in such systems is constrained: what the model can learn at a given point in time depends on the state of the network, i.e. what has been learned to that point, the nature of the new information being acquired, and the quality of the input. Thus, a conjunction of experiential and constitutional factors constrain how such models learn and give rise to the changes in plasticity associated with critical/sensitive period effects.
The specific aims of the project are to develop computational models that examine changes in plasticity associated with learning and their relevance to critical period effects in both language and song domains. We will first investigate the computational bases of changes in plasticity in PDP networks, examining effects of the timing, amount and quality of experience. We will then use this computational framework to develop models that address age-dependent learning effects in language and song learning. The language research focuses on the bases of the differences in plasticity that are observed in lexical vs. phonological learning. In the song research we will develop models using the same principles that are further constrained by facts about the neural circuitry to explain a parallel set of age-dependent behavioral phenomena. Together the studies will advance the understanding of the bases of age-dependent learning phenomena in both species and the extent to which they reflect common vs. species-specific factors.
Showing the most recent 10 out of 98 publications
