Back

A matter of availability: Neural representations of task-relevant stimulus features are sharper when stimuli are memorized rather than perceived.

Chota, S.; Gayet, S.; Kenemans, J. L.; Olivers, C. N. L.; Van der Stigchel, S.

2022-09-05 neuroscience
10.1101/2022.09.01.506184 bioRxiv
Show abstract

Our visual environment is relatively stable over time and an optimized visual system ought to capitalize on this by not devoting any representational resources to objects that are still present. Subjective experience, however, suggests that externally available (i.e., perceived) information is more strongly represented in neural signals than memorized information. To distinguish between these possibilities, we use EEG multivariate pattern analysis to quantify the strength of representation of task-relevant features (color or spatial frequency) in anticipation of a change-detection task. Perceptual availability was manipulated between experimental blocks by either keeping the stimulus on the screen during a two second delay period (perception) or removing it shortly after its initial presentation for the same time period (memory). We find that task-relevant (i.e., attended) memorized features are more strongly represented than irrelevant features. More importantly, we find significantly weaker representations for available (perceived and attended) features than for unavailable (memorized and attended) features. Contrary to what subjective experience suggests, our findings demonstrate that vividly perceived and attended stimuli elicit weaker neural representations (in terms of detectable multivariate information) than stimuli maintained in visual working memory. We hypothesize that an efficient visual system spends little of its limited resources on the internal representation of information that is externally available anyway.

Matching journals

The top 7 journals account for 50% of the predicted probability mass.

50% of probability mass above

"Similar papers" are the closest papers from that journal in the model's embedding space. They show what the match is built on, but the ranking comes mostly from a classifier over the whole training set, not from these examples alone.