Cortical Maps: Activity-Dependent Development

The term “cortical map” refers to the existence of a nonrandom relationship between the position of a neuron in the cerebral cortex and the value of some property that can be assigned to it on the basis of physiological or anatomical tests. Typically the property in question is a receptive field parameter of a sensory neuron, e.g., the position in space of the receptive field of a visual sensory neuron, or the color of a stimulus that activates it, but it might also be a projective field, e.g., the position in the body of a muscle or group of muscles activated by neurons in motor cortex, or, more speculatively, some aspect of knowledge or behavior coded for, or produced, by activity in single neurons or groups of neighboring neurons (perhaps as measured by fMRI experiments). A further qualification is that map properties often remain constant in value with depth in the cortex and vary only with lateral (equivalently tangential) position. This property, termed “columnar organization,” (Hubel and Wiesel 1977; Mountcastle 1998) means that the cortex is normally treated as a two-dimensional array for modeling purposes. “Activity-dependent development” in the context of “cortical maps” refers to a class of computational neuroscience model that seeks to explain the development of cortical maps as the outcome of mathematically specified rules governing axonal growth and synapse formation in which neural activity plays a primary role. All of these models use Hebbian learning rules to achieve a mapping between cells in one or more input layers representing either cell classes in the retina(s) or the lateral geniculate nucleus (LGN) layers and connections with cells in a single output layer that represents the cortex. The interest of the models lies not so much in what can be computed with them, but in the fact that they are often able to replicate, sometimes in a very detailed way, the geometrical properties of different maps and their spatial relationships (Erwin et al. 1995; Swindale 1996, 2003; Goodhill 2007).