7 8

25 Regions selective for faces, places, and bodies feature prominently in the literature on 26 the human ventral visual pathway. Are selectivities for these categories in fact the most 27 robust response profiles in this pathway, or is their prominence an artifact of biased 28 sampling of the hypothesis space in prior work? Here we use a data-driven structure 29 discovery method that avoids the assumptions built into most prior work by i) giving 30 equal consideration to all possible response profiles over the conditions tested, ii) 31 relaxing implicit anatomical constraints (that important functional profiles should 32 manifest themselves in spatially contiguous voxels arising in similar locations across 33 subjects), and iii) testing for dominant response profiles over images, rather than 34 categories, thus enabling us to discover, rather than presume, the categories respected 35 by the brain. Even with these assumptions relaxed, face, place, and body selectivity 36 emerge as dominant in the ventral stream. 37 38 Introduction 39 40 The ventral visual cortex has been implicated in the recognition of visually presented objects. 41 This region includes focal areas selective to single categories of visual stimuli (Kanwisher, 42 2010), including the fusiform face area or FFA, which responds selectively to faces (Kanwisher 43 et al., 1997), the parahippocampal place area or PPA, responding to spatial layout (Epstein 44 and Kanwisher, 1998), the extrastriate body area, EBA, selective for bodies (Downing et al., 45 2001), and the “visual word form area”, VWFA, selective for familiar letter strings (Cohen et al., 46 2000; Baker et al., 2007). Intriguing as these category-selective regions are, they collectively 47 represent only a small percentage of the ventral visual pathway (Kanwisher, 2010), raising a 48 difficult question: Have we found the most prominent category-specific regions, or just the ones 49 we have thought to look for? 50 51 Studies testing for selectivity for other categories have generally not found it (Downing et al., 52 2006), or have found weaker selectivity than the FFA, PPA, EBA, and VWFA (Chao et al., 53 1999). These prior tests have been limited to specific categories, and have assumed that 54 selectivity should be shared by contiguous voxels and should arise in similar locations across 55 subjects. However, selectivity need not be clustered at the grain of adjacent voxels, but could 56 be sparsely distributed. Moreover, selective regions might not respond exclusively to a 57 category, but might prefer some set of object classes -one that might fit intuitions (e.g., 58 animate objects, (Caramazza and Shelton, 1998), or tools, (Chao and Martin, 2000)) or might 59 not (e.g., cars and birds but not bicycles or fish). Altogether, only a small subset of the large 60 number of possible response profiles has been tested. 61 1 Of course, contiguity of functionally similar neurons on the scale of single voxels is a prerequisite for finding any functional structure with fMRI.