Purchase this article with an account.
Jack Ryan, Joshua Julian, Russell Epstein; Coding of object size and object category in scene regions. Journal of Vision 2015;15(12):1162. doi: https://doi.org/10.1167/15.12.1162.
Download citation file:
© ARVO (1962-2015); The Authors (2016-present)
Recent work suggests that scene-responsive regions such as the parahippocampal place area (PPA), retrosplenial complex (RSC), and occipital place area (OPA) might play a role in processing of nonscene objects in addition to their role in processing scenes. For example, Konkle & Oliva (2012) reported that these regions respond more strongly to large objects than to small objects. Does this indicate that scene-responsive regions encode objects in terms of their real-world size--or, alternatively, preferentially encode large objects? To investigate this issue, we scanned subjects while they viewed objects drawn from 20 categories, 10 of which were physically large (e.g. stove, copier) and 10 of which were physically small (e.g. binoculars, mug). Objects were shown for 1 s each with a 2 s ISI. Large and small categories were matched in terms of a number of low-level image properties, including retinotopic size, chrominance, luminance, and spatial frequency. Preliminary results replicated the previously reported univariate effect: PPA, RSC, and OPA all responded more strongly to big objects than to small objects. Moreover, this finding was echoed in multivoxel pattern analyses: pattern similarity in these regions was greater for object categories of similar size (e.g. stove-copier) than for object categories of different size (e.g. stove-mug). Notably, we did not observe evidence for coding of object category in these regions when object size was controlled. That is, across scan runs, multi-voxel patterns were no more similar for objects of the same category and size than for objects of different category but the same size. In contrast, object category could be readily decoded in the lateral occipital complex. These results suggest that scene regions code spatial properties of objects useful for navigation but are not centrally involved in object recognition.
Meeting abstract presented at VSS 2015
This PDF is available to Subscribers Only