No single topography in a canonical or average brain can capture the fine-scale topographies that are seen in individual subjects. The primary motivation for the development of hyperalignment was to find such common response-tuning functions that are associated with variable cortical topographies. The rows in a data matrix contain the model space coordinates of response-pattern vectors for time points or stimuli. The response profile of a single voxel is modeled as a weighted sum of the response-tuning functions for dimensions (Figure S1E). Modeling voxel response profiles as weighted sums of response-tuning basis functions can capture an unlimited variety

of such profiles. Computational approaches that define voxel response profiles as types (Lashkari et al., 2010), rather than as mixtures of basis functions, cannot model this unlimited variation, making them unsuited for modeling fine-grained structure in response topographies. www.selleckchem.com/products/PLX-4032.html The full set of dimensions models topographies that are more fine grained than those of category-selective areas for faces (FFAs) and houses (PPAs; Figure 5B; Figures S5A and S5B). Category-selective areas are defined by simple contrasts, which are single dimensions in the model space. The single dimension that is defined by the contrast MAPK Inhibitor Library purchase between responses to faces and objects produces individual topographies that correspond well with the outline

of individually defined FFAs (Figure 6A). Category-selective regions can be defined based on group data that is projected into an individual’s native brain space. Group-defined FFAs and PPAs in individual brain spaces correspond well with the regions defined by that subject’s own data (Figure 6B). Thus, category-selective response profiles, their associated topographies, and the outlines of category-selective regions are preserved in the common model and can be extracted with high fidelity. Such category selectivities, to however, do not account for a majority of the variance in VT responses to natural, dynamic stimuli. Moreover, single dimensions that define category-selective regions cannot model the fine-grained

variations in response topographies within the FFA and PPA that are modeled well by weighted sums of model dimensions and afford classification of responses to a wide range of stimulus distinctions (see Figure S2E). Single-neuron response-tuning profiles in monkey inferior temporal cortex (IT) reflect complex object features, and patterns of responses over a population represent object categories and identities (Logothetis and Sheinberg, 1996, Tanaka, 2003, Hung et al., 2005, Tsao et al., 2006, Freiwald et al., 2009, Serre et al., 2007 and Kiani et al., 2007). IT response-tuning profiles show a variety that appears open ended and, to our knowledge, has not been modeled with response-tuning basis functions (with the exception of Freiwald et al. [2009]‘s investigation of response-tuning basis functions for faces).