Abstract
There is some uncertainty as to the extent to which object representations in late ventral stream visual areas may be translation invariant. Grill-Spector et al. (1999), using faces in an adaptation paradigm, concluded that the representations in LOC were invariant to translation. The stimuli in that study were large ([[gt]]10° of visual angle) and were translated over relatively small distances (5.6°)—about half the stimulus diameter. Alternatively, several groups have now found retinotopic organization in lateral and ventral occipital cortex (Larsson et al 2006, Brewer et al 2005). Consistent with a position-specific representation in late ventral visual areas is the recent single unit work in macaque IT by Op de Beeck et al (2000) and DiCarlo and Maunsell (2003). These investigators reported cells with receptive fields as small as 1 degree of visual angle (the preferred stimuli for these cells was ∼.5°). We investigated whether adaptation effects in the posterior fusiform region of LOC would show position sensitivity when stimuli were translated by more than one stimulus diameter. Subjects viewed a sequence of two object images. All stimuli were presented iso-eccentrically, 4.5° from fixation. The objects subtended 2.3° and the second stimulus was translated either 0°, 2.3°, 4.6°, or 9.0°—zero, one, two, or four stimulus diameters, respectively. The subject's task was to detect whether S2 was a different object than S1—which occurred on 16% of the trials. Objects translated a distance of one stimulus diameter produced a minimal release from adaptation compared to the 0° condition, but both larger translations produced a significant release from adaptation (equal in magnitude for both). This result could explain the apparent conflict among the previously cited studies: translation effects in the posterior fusiform may only be manifested when the extent of the translation is greater than the extent of the object.
This work was supported by NSF BCS 04-20794, 05-31177, & 06-17699 to IB.