Abstract
Many recent studies of the human visual system have made use of multivariate pattern analyses (MVPA) to "decode" feature representations from fMRI activation patterns in the visual cortex. We have previously claimed that such methods are sensitive to signals originating from irregularities in the arrangement of cortical columns, found on spatial scales of millimeters (Kamitani & Tong, 2005; Swisher et al., 2010). Evidence for this claim comes primarily from studies of the representation of visual orientation. However, this interpretation is complicated by the presence of larger-scale anisotropies in the cortical orientation map, such as the oblique effect and radial bias. Accounting for the role that such anisotropies play in orientation decoding has accordingly been a focus of much previous work (e.g. Mannion et al., 2009; Swisher et al., 2010). Recently, Freeman and colleagues (2011) have argued that a large-scale retinotopic preference for radial orientations accounts for the entirety of the signal detected in conventional fMRI orientation decoding. This study found that removal of a radial bias component greatly impaired the accuracy of orientation decoding. Specifically, they found that regressing out a spatial map derived from retinotopic mapping disrupted both orientation decoding and decoding of retinotopic position to an equal degree. Using paradigms similar to those of Freeman et al (2011), we find that the orientation signal is only modestly impacted by removal of retinotopic radial bias. Across a variety of stimulus configurations, we instead find that removal of independently-measured retinotopic components consistently impairs the accuracy of retinotopic decoding significantly more than that of orientation decoding. The presence of a substantial residual orientation signal after removal of retinotopic components is inconsistent with an orientation signal carried exclusively by retinotopic radial bias, as previously claimed. Instead, our results strongly support a sensitivity to millimeters-scale irregularities in cortical feature maps.
Meeting abstract presented at VSS 2012