Abstract
Using full color photographs of natural scenes, we have previously shown that information about scene category is contained in patterns of fMRI activity in the parahippocampal place area (PPA), the retrosplenial cortex (RSC), the lateral occipital complex (LOC), and primary visual cortex (V1) (Walther et al. J. Neurosc. 2009). If these regions are involved in representing category, then it should be the case that we could decode scene category for any natural scene image that participants can categorize, including simple line drawings. In keeping with this prediction, we found that we can decode scene category from fMRI activity patterns for novel line drawing pictures just as well as from activity for color photographs, in V1 through PPA. Even more remarkably, a decoder trained on fMRI activity elicited by color photographs was able to predict the correct scene categories when tested on activity patterns for line drawings just as often as for color photographs, indicating that the activation pattern elicited by color photographs generalized well to line drawings. Conversely, a decoder trained on activity for line drawings was able to decode activity patterns for photographs just as well as for line drawings, but only in the PPA and V2/VP, suggesting that, in these regions, category information is strongly determined by the edge and line information in a photograph. We conclude that line drawings contain sufficient information about natural scene categories to produce scene-specific fMRI activity patterns all along the visual processing hierarchy. Moreover, the specific encoding of this information appears to be similar to that elicited by photographs as shown by successful decoding of scene categories across the two image types. Our findings suggest that scene structure, which is preserved in line drawings, plays an integral part in representing scene categories.