Abstract
Previous studies have shown that only representations of currently task relevant goals can be decoded from brain activity. However, it is still unclear how and where representations of future goals are instantiated in the brain. Here, we measured fMRI of 24 human participants while on each trial we presented two real-world objects from different categories serving as targets for two consecutive visual search tasks. We manipulated the relevance of the objects with a cue that indicated which object to look for first (current), and which second (prospective). Before each search there was an eight second retention interval. We used multi-voxel pattern analysis to decode the dynamical changes in representational space of the object categories in object-selective cortex throughout the trial, as a function of current versus prospective task relevance. As predicted, we observed better category decoding for the currently relevant than for the prospectively relevant category right before the first search. However, even during search for the current target we could successfully decode the future target. When we trained the classifier on the currently relevant category and tested on the prospectively relevant category or vice versa, classification was below-chance during both searches. This indicates that current and future object categories are represented in opposite corners of the representational space. Indeed, representational similarity analyses confirmed that as a trial unfolds, object representations move from object category space (e.g. a cow) into relevance space (e.g. current target), where current and prospective targets of the same category are represented by opposite representational patterns. Taken together, our results demonstrate how the brain shields current from future targets and vice versa.
Meeting abstract presented at VSS 2018