Abstract
Object perception requires the neural representation of both object identity independent of viewing conditions, and the representation of viewing-dependent properties such as location. The seminal view is that such representations emerge respectively in the ventral and dorsal streams. Recent research in monkeys (Hong et al., 2016) has questioned this division observing both types of information in high-level ventral visual cortex when objects are shown in cluttered scenes. However, how such representations emerge over time, how they emerge in the processing cascade of the ventral visual stream, and whether they depend on attention remains unknown. To investigate this, we conducted three experiments in humans using EEG and fMRI, and analyzed data with multivariate pattern classification methods. Experiment 1 (EEG, N=27) investigated the emergence of object category and location representations in time, dependent on clutter level of the background. We found that object location information emerged later when objects were presented in cluttered versus uncluttered scenes. Further analysis of object location and category information showed that the brain carried out the same processing steps with and without scene clutter, but shifted in time. Experiment 2 (fMRI, N=14) corroborated the previous result in space. In particular, when objects were presented in cluttered versus uncluttered scenes, location information emerged in higher regions of the ventral stream hierarchy. Experiment 3 (EEG, N =25) investigated whether disentangling object-specific information from cluttered scenes might require top-down attention. The emergence of location representations depended strongly on attention for cluttered scenes but was independent of attention in uncluttered scenes. In contrast, category representations were affected by attention independent of whether the background was cluttered or not. Together, our results provide a new perspective on the role of ventral visual cortex in object perception and show how attention impacts the underlying neural processing.
Acknowledgement: German Research Foundation CI-241/1-1