Recent neurobiological evidence suggests that environmentally derived activity plays a central role in regulating neuronal growth and neuronal connectivity. Artificial neural networks with distributed representations display many features of knowing and learning that are known from biological intelligence. In this article, I advocate artificial neural networks as models for cognition and development. These models and how they work are exemplified in the context of a well-known Piagetian developmental task and school science activity: balance beam problems. I conclude that artificial neural networks, because of their profoundly interactivist nature, are ideal tools for modeling cognitive development and learning in science. © 2000 John Wiley & Sons, Inc. J Res Sci Teach 37: [63][64][65][66][67][68][69][70][71][72][73][74][75][76][77][78][79][80] 2000 How certain can science educators be that their models of learning-quintessential prerequisites for any theory and practice of teaching-are compatible with learning as it occurs in biological systems, including humans? Recent research in the philosophy of mind, computational neuroscience, neurophysiology, cognitive neurobiology, cognitive psychology, and connectionist artificial intelligence suggests that much of classical psychology is incompatible with recent findings in these fields. Thus, [t]he sentential kinematics of folk psychology is but a common sense theory, and almost certainly a false theory, at least as an account of the basic kinematics of cognitive creatures generally. (Churchland, 1989, p. xvi, original emphases) If the models and metaphors of mind used by science educators today-including the mind as information processor or as a mental structure of concepts-are incompatible with research in the neurosciences, how sure can they be that their models of teaching are appropriate to bring about learning and/or development in the students who are in their care?During the past 2 decades there has been relatively little interest in the mechanisms of developmental research; the reasons lie most probably in the metaphors of mind that are used in experimental psychology, cognitive science, and neuropsychology (Elman, Bates, Johnson, Karmiloff-Smith, Parisi, & Plunkett, 1996). Recent developments in neurobiology and computational architectures provide new tools and ideas for looking at questions of cognitive development and for reframing old or unanswered questions about cognitive development in new ways. In recent years, there appears to be some interest in linking educational questions closer JOURNAL OF RESEARCH IN SCIENCE TEACHING VOL. 37, NO. 1, PP. 63-80 (2000) © to research in brain sciences (Anderson, 1992(Anderson, , 1997Bereiter, 1991;Bruer, 1997). The artificial neural network (ANN) framework is a new paradigm in which to explore basic questions about the relations of nature and nurture in the development of cognition (Siegler, 1989). ANN models of knowing and learning provide a new perspective on many questions related to nature an...