Purchase this article with an account.
Heeyoung Choo, Bardia Nikrahei, Jack Nasar, Dirk Walther; Neural decoding of architectural styles from scene-specific brain regions. Journal of Vision 2015;15(12):520. doi: 10.1167/15.12.520.
Download citation file:
© ARVO (1962-2015); The Authors (2016-present)
The human visual cortex can elicit neural activity patterns that are distinctive for basic-level scene categories (e.g., highways) as well as for superordinate level scene categories (e.g., man-made). Can the human brain also elicit category-specific neural activity patterns for scenes at a subordinate level? Using functional magnetic resonance imaging (fMRI) we recorded brain activity of participants viewing scene categories at a basic level - mountains, pastures, highways, and playgrounds, and scene categories at a subordinate level - buildings in byzantine, renaissance, modern, and deconstructive architectural styles, and buildings designed by four well-known architects of which categorization is likely to be guided more by participants’ knowledge than coherent perceptual structure in buildings. Using multi-voxel pattern analysis, we could decode viewed buildings of different architectural styles significantly better than chance from the parahippocampal place area (PPA), retrosplenial cortex (RSC), occipital place area (OPA), lateral occipital complex (LOC). Decoding of buildings by different architects was also successful in the PPA and OPA. Consistent with previous findings, we could successfully decode viewed scene categories. Comparison of error pattern between a simple V1 model and decoding from fMRI data showed that categorical information about architectural styles in the PPA cannot be explained solely by high low-level visual similarity. On the other hand, our simple V1 model was able to discriminate between buildings by different architects to some extent. Our results suggest that the PPA can maintain categorical representation at multiple levels, including subordinate categories, such as architectural styles of buildings, relying on complex structural characteristics of scenes rather than low-level visual similarity. Top-down influences such as domain knowledge or geographical familiarity may thus capitalize on these fine-grained categorical activation patterns in the PPA.
Meeting abstract presented at VSS 2015
This PDF is available to Subscribers Only