M. W. Spratling and M. H. Johnson
Most of adult neuropsychology is concerned with attributing functions to particular regions of the cerebral cortex. Few, however, address the more fundamental question of how such specializations develop in the first place. Kingsbury and Finlay are to be congratulated on addressing this critical question head on. Like many issues in biology, the answer to the question of how the cortex differentiates into areas and regions is far from simple. These authors suggest that for most, but not all, regions of cortex differentiation is a result of multiple interacting molecular gradients in interaction with thalamic input. In their words, it is a "plaid" of interacting threads rather than a "quilt" of separate panels. In this commentary we expand upon these conclusions in two directions: first, we draw out some of the implications of these views for developmental psychologists, and second we draw attention to the likely importance of functional neural activity in cortical differentiation.
One current thrust in infancy research is based on the assumption that, early in life, the infant's brain is composed of a number of domain specific modules. These modules are often assumed to have a genetic basis, and one type of explanation of some developmental disorders of genetic origin is that one or other of these modules is "lesioned". The data reviewed by Kingsbury and Finlay, however, suggests that, primary sensory cortices apart, it is unlikely that there are cortical regions defined by region-specific gene expression. The implication of this is that cortical specialization for cognitive function is better described in terms of an interacting factors framework (Johnson, 2000). Thus, we believe that the evidence reviewed by Kingsbury and Finlay is more consistent with cognitive models that attempt to explain the gradual emergence of functions (e.g. Munakata, McClelland, Johnson & Siegler, 1997).
The second point we wish to make in our commentary concerns an additional factor involved in cortical differentiation that we believe was somewhat underemphasized by Kingsbury and Finlay, namely, functional activity. There is now considerable evidence for the importance of neuronal activity in shaping subsequent neural circuitry (Greenough, Black & Wallace, 1993; Katz & Shatz, 1996). Some of the mechanisms underlying this activity-dependent shaping have been studied in detail through artificial neural network modelling (Jacobs, 1999). These models show that an initially homogenous neural architecture can be differentiated into functionally specialized structures through the application of simple activity-dependent learning rules. Critical factors for such models are the refinement of synaptic weights in response to structured afferent input and competition between elements. The formation of both small-scale and large-scale structures have been modelled using similar mechanisms. In the former class are those models which simulate the functional specialization of neurons and cortical columns using competition between neuronal elements; for example, the many models of ocular dominance and orientation column formation in V1 (Swindale, 1996). In the latter class are models of the functional specialisation of cortical regions, in which there is competition between separate neural networks resulting in each network learning a different task (Jacobs, 1999; Dailey and Cottrell, 1999). Competition between elements (either neurons or networks) is essential for differentiation. An element which has a small initial bias for a particular function will be likely to win the competition for that function and hence learn to be even better suited to it. Competitive mechanisms thus tend to enhance any initial differences that may exist between elements. These models thus show that small, gradual innate differences across the cortex could give rise to large, sharply bounded structures in adults with a high degree of uniformity across individuals despite the plasticity of the cortex. The same outcome could result whether the initial bias was due to a genetic predetermination to generate specific cortical regions, or if the initial bias resulted from arbitrary variations. In contrast, the origin of the functional activity is crucial to determining the resulting structure, and is thus at least as important as molecular factors.
We conclude that the structural and functional development of the cortex are inextricably intertwined, and a full account of structural differentiation within the cortex will need to take account of activity-dependent processes. Computational modelling provides a valuable tool for evaluating and refining such theories and is likely to play an important role in helping to understand the regional specialization of the cortex.
Dailey, M. N., & Cottrell, G. W. (1999). "Organization of face and object recognition in modular neural network models". Neural Networks, 12, 1053-1073.
Greenough, W. T., Black, J. E., & Wallace, C. S. (1993). "Experience and brain development", In Johnson, M. H. (Ed.) Brain Development and Cognition: A Reader. (pp. 290-322). Oxford: Blackwell.
Jacobs, R. A. (1999). "Computational studies of the development of functionally specialized neural modules". Trends in Cognitive Sciences, 3, 31-38.
Johnson, M. H. (2000). "Functional brain development in infants: elements of an interactive specialization framework". Child Development. In press.
Katz, L. C., & Shatz, C. J. (1996). " Synaptic activity and the construction of cortical circuits". Science, 274, 1133-1138.
Munakata, Y., McClelland, J. L., Johnson, M. H., & Siegler, R. S. (1997). "Rethinking infant knowledge: toward an adaptive process account of successes and failures in object permanence tasks". Psychological Review, 104, 686-713.
Swindale, N. V. (1996). "The development of topology in the visual cortex: a review of models". Network: Computation in Neural Systems, 7(2), 161-247.