Purchase this article with an account.
Timothy Andrews, David Watson, Tom Hartley; A DATA DRIVEN ANALYSIS REVEALS THE IMPORTANCE OF IMAGE PROPERTIES IN THE NEURAL REPRESENTATION OF SCENES. Journal of Vision 2016;16(12):516. doi: 10.1167/16.12.516.
Download citation file:
© ARVO (1962-2015); The Authors (2016-present)
The neural representation of scenes in human visual cortex has been linked to processing of semantic and categorical properties (e.g. categorization of indoor versus outdoor scenes). However, it is not clear whether patterns of neural response in these regions reflect more fundamental visual principles like those that govern the organization of early visual cortex. One problem is that existing studies have involved comparisons between stimulus categories chosen by the experimenter, potentially obscuring the contribution of more basic visual features. Here, we used a data-driven analysis to select clusters of scenes based solely on their image properties. Although these visually-defined clusters did not correspond to conventional scene categories, we found they elicited distinct and reliable patterns of neural response, and that the relative similarity of the response patterns to different clusters could be predicted by the low-level properties of the images. Local semantic properties of the images failed to explain any additional variance in the neural responses of scene-selective regions beyond that explained by the image properties. However, we did find that participants' behavioural classification of the scenes was better predicted by local semantic properties than by image properties. These results suggest that image properties play an important part in governing patterns of response to scenes in high-level visual cortex and suggest that these patterns are at least partially dissociated from behavioural responses which are better explained in terms of local semantic content.
Meeting abstract presented at VSS 2016
This PDF is available to Subscribers Only