Turning the Dial on Object Perception

authors point out that " Salt&Pepper " occurs in the rat, where there is no evidence for an ordered map of orien-relatively high density of singularities is always found 5568–5573. These findings are interesting because they show, for the first time, that singularities may be adaptive features whose role is to minimize wire length. (The authors do not say by how much, although presumably it would not be difficult to calculate.) There are some additional interesting implications. It can be shown that previous Perception the Icecube layout of orientation as a final stable state. This can be proved analytically, as Koulakov and Chklovskii do for one particular model, or it can be ob-One of the more challenging aspects of neuroimaging served in simulations that singularities tend to disappear studies of higher cortical function is the isolation of a slowly over time (Wolf and Geisel, 1998). Technically, mental operation of interest. An experimental condition this means that the cost functions for these other models that evokes the cognitive process of working memory, are minimized by arrangements in which singularities semantic retrieval, or spatial representation, for exam-are absent. From the point of view of these previous ple, can be expected to evoke a number of other models (and perhaps the modelers as well), singularities processes as well (e.g., stimulus perception, eye move-are developmental artifacts and represent nonoptimal ments, response preparation, etc.). The " cognitive sub-configurations of the map. That singularities are present traction " method tackles this problem by comparing the in most model maps is because the models either con-neural activity evoked by matched experimental and tain features that prevent the optimum configuration control conditions that putatively differ only in the cogni-from being reached or because it simply takes too long tive process of interest. The assumptions, however, that to compute the optimum solution. In reality, this may cognitive subtraction requires are themselves problem-not be a problem because development, like the models, atic (Friston et al., 1996; Zarahn et al., 1997) and stem may also only incompletely minimize the cost function. in part from the need to evoke the mental operation For example, the factors that bring the critical period to under study in an all-or-nothing fashion. Parametric ma-an end may " freeze " the map in what is, from a computa-nipulations of cognitive processes (in which the level of tional perspective, an intermediate state. This would the operation, as opposed …