Neural network models of strategy development in children

This paper presents new models of strategy development in young children. Traditional approaches to strategy development: (a) have relied on top-down mechanisms; (b) have not incorporated situational influences; (c) have not been biologically motivated. Neural networks models are described which address the limitations of these traditional approaches. The development of the models was influenced by: (a) Grossberg's modular approach to neural networks; (b) the general theoretical framework of Siegler; (c) empirical research on external representation and memory. Empirically, in situations requiring memory for the arrangement of objects, younger children frequently use simple external memory strategies (e.g. pointing at objects) whereas older children tend toward more advanced strategies (e.g. inter-object orienting) to aid remembering. Computer simulations with the novelty bias model show that strategy novelty and accuracy history can account for selection and evolution of strategies from simple to advanced and for accuracy patterns associated with strategies of different degrees of sophistication. Simulations with a second model, the components model, which overcomes some of the limitations of the first model, demonstrate comparable behavioural evolution using accuracy feedback and selective encoding mechanisms. The models suggest alternatives to the traditional approaches to strategy development, and suggest that the strategy evolution mechanisms involved may apply to a variety of situations. Copyright (C) 1996 Elsevier Science Ltd.