We develop a theoretical model of how a human learner's brain creates the categories of language (e.g. the sound "k", the word "cat", or the morpheme "-ing") as well as the connections between these categories (e.g. the fact that "-ing" can combine with verbs but not with nouns, and that words can start with the sound sequence "kn" in Dutch but not in English). The model is constructed and verified in a mutually feeding exchange with computer simulations of virtual learners and with laboratory experiments involving human participants. Our theoretical basis is the idea that categories and connections are not innately given to the child, but instead are created by the child during her language acquisition period. This "emergentist" view of category creation is inspired by the homogeneity and plasticity of the human neocortex, which is where language and other higher cognitive processes take place. The model is the first to marry ideas from the literature on emergent categorization in neural networks with the knowledge that linguists have amassed about phenomena that occur in the world's languages. The principal investigator will develop the computer implementation of the linguistic-neural model. A postdoctoral researcher and a PhD student will perform the laboratory experiments. These experiments are inspired by the model's predictions about language perception and production, and will, conversely, inspire the model and the simulations. The model brings "whole-language simulations" within reach. A second PhD student will therefore utilize the categorization model in computer simulations of the acquisition of a well-studied real human language (e.g. French). This will in turn inspire the model and the experiments. The proposed combination of computational modelling and empirical data will lead to a better theory of how humans create categories and connections, both in language and in higher-level cognitive processes such as the handling of concepts.
De mens heeft veel onbewuste kennis over zijn wereld en zijn taal. Je hersenen slaan die kennis op als duizenden 'categorieën' en verbanden daartussen. De onderzoekers analyseren hoe het lerende kind de taalcategorieën construeert uit de hem omringende spraak.