There is already some support for this idea from electrophysiological studies in primates. The response properties of anterior inferior temporal neurons change as monkeys learn novel associations between

visual stimuli, suggesting a role for this region in the acquisition of concepts (Albright, 2012). In the present study, we tested this hypothesis in humans by studying acquisition of new conceptual knowledge in patients with SD. The hub-and-spoke model predicts that the ATLs are critical for integrating the various sensory features of an object into a unified, coherent conceptual representation that can be generalised to new exemplars. We tested this prediction by training SD patients to recognise GSK-3 inhibitor novel visual stimuli as members of two categories. Previous research has shown that SD patients are able to apply well-defined rules to classify novel stimuli, when the classification rule is provided by the experimenter (Koenig, Smith, & Grossman, 2006). Here, we tested the patients’ ability to acquire more complex category structures that could not be captured by a simple rule and when no information about the nature of the categories was supplied by the experimenter. The

structure of the two categories (shown in Fig. 1A) was designed such Screening Library datasheet that optimal performance could only be achieved by acquiring integrated representations of the various typical characteristics of each category. When presented en masse as in Fig. 1, it is easy to discern the features associated with each category. Members of Category A usually contained squares while those in B contained

circles, though there were exceptions in both categories. The same was true for the number of shapes (members of A usually contain one shape) and the colour of the background square (usually blue for A). The colour of the internal shapes, though perceptually salient, was not diagnostic of category. This category structure, in which a number of features are associated with each category but no single feature Buspirone HCl is diagnostic, is termed a family resemblance structure and is characteristic of object categories in the real world ( Rosch and Mervis, 1975, Smith and Medin, 1981 and Wittgenstein, 1953). Within such a structure, it is impossible to classify with complete accuracy by learning only about a single feature dimension. Optimum performance instead requires participants to form integrated representations that include second-order statistical information about the feature conjunctions that characterise each category, allowing them, for example, to correctly class an exemplar with two circles as a member of Category B, even if it has a blue background. We predicted that forming such integrated representations is a key function of the ATLs and, therefore, that SD patients would be impaired in learning the categories.