The relationship between SLI in English and Modern Greek: Insights from computational models of language acquisition

We present a computational modelling approach to the study of SLI in two languages with different typological characteristics, namely English and Modern Greek. Our modelling approach was based on the development of three neural network (connectionist) architectures, each assumed to underlie the acquisition of a core domain of language (inflectional morphology, syntax comprehension, and syntax production). The architectures were exposed to artificial linguistic environments reflecting the characteristics of their target domains in English and Greek. Computational simulations also considered conditions of atypical learning constraints, corresponding to different theoretical proposals for the type of deficit underlying SLI. The simulation results, combined with some shared properties of the three models, point to a unified explanation of the impairment under the connectionist framework.