Developmental disorders

Introduction: Connectionist models have recently provided a concrete computational platform from which to explore how different initial constraints in the cognitive system can interact with an environment to generate the behaviors we find in normal development (Elman et al., 1996; Mareschal & Thomas, 2000). In this sense, networks embody several principles inherent to Piagetian theory, the major developmental theory of the twentieth century. By extension, these models provide the opportunity to explore how shifts in these initial constraints (or boundary conditions) can result in the emergence of the abnormal behaviors we find in atypical development. Although this field is very new, connectionist models have already been put forward to explain disordered language development in Specific Language Impairment (Hoeffner & McClelland, 1993), Williams Syndrome (Thomas & Karmiloff-Smith, 1999), and developmental dyslexia (Seidenberg and colleagues, see e.g. Harm & Seidenberg, in press); to explain unusual characteristics of perceptual discrimination in autism (Cohen, 1994; Gustafsson, 1997); and to explore the emergence of disordered cortical feature maps using a neurobiologically constrained model (Oliver, Johnson, Karmiloff-Smith, & Pennington, in press). In this entry, we will examine the types of initial constraints that connectionist modelers typically build in to their models, and how variations in these constraints have been proposed as possible accounts of the causes of particular developmental disorders. In particular, we will examine the claim that these constraints are candidates for what will constitute innate knowledge. First, however, we need to consider a current debate concerning whether developmental disorders are a useful tool to explore the (possibly innate) structure of the normal cognitive system. We will find that connectionist approaches are much more consistent with one side of this debate than the other