Purchase this article with an account.
Ipek Oruç, Jason Barton; Multi-voxel pattern analysis of face and object exemplar discrimination in occipital cortex.. Journal of Vision 2011;11(11):653. doi: 10.1167/11.11.653.
Download citation file:
© ARVO (1962-2015); The Authors (2016-present)
Background: Face or object areas such as the fusiform face area (FFA), occipital face area (OFA), superior temporal sulcus (STS) and lateral occipital cortex (LOC) are typically determined in fMRI by subtractions showing higher activation for faces or objects than other stimuli. However, contrast analyses do not inform us about the functions of these areas, in particular in discriminating between different exemplars. Objective: We used multi-voxel-pattern analysis (MVPA) to determine whether spatial fMRI patterns discriminated between faces and objects, and between different exemplars of faces and objects in different regions. Methods: We used a localizer to identify the LOC, FFA, OFA and STS. In the experimental block subjects viewed two different faces and two different cars, chosen using an ideal observer analysis to equate physical differences between cars with those between faces. We used MVPA to determine discrimination accuracy based on activity in the top 200 voxels of each region. Results: All regions showed significant accuracy for between-category (car/face) discrimination and both within-category tasks (car1/car2, face1/face2). Between-category was better than within-category discrimination in right and left LOC and left FFA, but equivalent in all other regions. Accuracy for car exemplar discrimination was similar to face exemplar discrimination in all regions. Individual voxel weight magnitudes were positively correlated between car and face tasks in OFA and LOC, negatively correlated in STS, and not correlated in FFA. Conclusion: Better between-category discrimination is more typical of object processing in the LOC, but all core face regions and LOC may process the information necessary to individuate not just faces but also cars. Early processes common to both tasks may use overlapping neural populations giving rise to the positive correlations in OFA and LOC, which are not seen in FFA or STS, in which more specialized processes may recruit relatively distinct neural sub-populations.
This PDF is available to Subscribers Only