The role of parvocellular and magnocellular shape maps in the derivation of spatially integrated 3D object representations

Our ability to effectively process visual information necessitates the transformation of sensory input from retinotopic to non-retinotopic representations of scene content. One fundamental question concerns the mechanisms, types of representation, and coordinate systems, that mediate these transformations. The complexity of their interactions makes this a formidable challenge. Davida’s case presents with a highly specific deficit affecting the mapping between representations of sensory input based on retinotopic coordinates and higher-level non-retinotopic reference frames. To account for her performance, the authors outline a theoretical proposal that makes several important claims. Among those are: (1) the existence of a level of intermediate shapecentred representation (ISCRs – or “shape maps”) that mediates the mapping between retinotopic and body-centred/spatiotopic representations of scene content; (2) the parallel derivation of independent ISCRs encoding object shape information via parvocellular (P-cell) and magnocellular (M-cell) channels. Davida’s impairment is assumed to arise from a selective deficit affecting the mapping between the proposed ISCR and body-centred/spatiotopic reference frames in the P-cell channel. This interesting proposal invites further speculation about the possible role of the proposed ISCRs in the derivation of spatially integrated representations of complex 3D object shapes – and their prospective role in object recognition. The question we discuss here is how this proposal might link to other recent work about the structure and functional organization of object shape representations in human vision. A growing body of evidence suggests that object representation in human vision is hierarchical, decompositional, and parts-based (e.g., Behrmann & Kimchi, 2003; Behrmann et al., 2006; Biederman, 1987; Hoffman & Richards, 1984; Leek et al., 2003; 2005; 2009; Reppa & Leek, 2003; Robertson & Lamb, 1991). These multi-level representations comprise elementary local features (e.g., edges and vertices), intermediate-level functional units (e.g., spatially bounded 2D regions approximating visible surface structure – Leek et al., 2005; Reppa et al., 2015; Marr & Nishihara, 1978; Palmer & Rock, 1994), and (on some accounts) higher-order primitives such as volumetric parts (e.g., Biederman, 1987; Marr & Nishihara, 1978). Evidence for this complex decompositional representational structure comes from both studies of neurologically intact, and brain-damaged, individuals. For example, there are case reports of patients with acquired object recognition impairments who have difficulty distinguishing among 3D objects that comprise the same geometric parts arranged in different 3D spatial configurations (e.g., Behrmann et al., 2006; Behrmann & Kimchi, 2003); studies showing complementary patterns of local-global feature representation deficits following unilateral brain lesions (Robertson & Lamb, 1991), and partsbased object identification errors associated with socalled integrative agnosia (e.g., Humphreys & Riddoch, 1987; Leek et al., 2012). Other work has