Given the paucity of data that bears on this question, how do we develop viable theories explicating the features underlying the neural representation of objects? One possibility is to focus on feature codes realized in “category-selective” regions within the ventral-temporal cortex. However, most investigations of these regions—for example, the “fusiform face area” (FFA) associated with the detection and discrimination of faces (Haxby, Hoffman, & Gobbini,
2000; Grill-Spector, Knouf, & Kanwisher,
2004), the “parahippocampal place area” (PPA) associated with scene processing (Epstein, Harris, Stanley, & Kanwisher,
1999), or the lateral occipital complex (LOC) associated with the processing of objects more generally (Grill-Spector, Kourtzi, & Kanwisher,
2001)—emphasize specific object-level experiential factors or input characteristics that lead to their recruitment but never establish the underlying visual properties that form the basis of the nominally category-specific representations. Most studies of the visual properties that lead to the recruitment of these class-specific, functionally defined brain regions have focused on the effects of spatial transformations and of the alteration of simple domain-specific features (Tsao & Livingstone,
2008). For example, images of objects from within a given class often elicit similar neural responses when scaled, rotated, or moved to different locations in the visual field although, in the case of picture-plane inversion or 3-D rotation, there is typically some change in neural activity (Perrett et al.,
1984; Haxby et al.,
1999). To the extent that viable models of neural representation have been developed, they have relied on the statistical analysis of the input space within a restricted object domain. For example, “face spaces,” nominally capturing the featural dimensions of human face representation, can be defined using principal component analysis (PCA) on face images or using parameterized models that are generative for constructing what appear to be realistic new face stimuli (Calder & Young,
2005; Freiwald, Tsao, & Livingstone,
2009). Alternatively, the featural dimensions of representation are sometimes made more explicit as in Kravitz, Peng, and Baker (
2011), who found that the encoding of scenes in the human visual cortex can be understood in terms of an underlying set of intuitive properties, including “open/closed” and “natural/artificial.”