Back

Fast feature- and category-related parafoveal previewing support natural visual exploration.

Fakche, C.; Hickey, C.; Jensen, O.

2024-05-01 neuroscience
10.1101/2024.04.29.591663 bioRxiv
Show abstract

Studies on vision tend to prevent or control eye movements, while humans naturally saccade every [~]250 ms. As the oculomotor system takes [~]100 ms to initiate and execute a saccade, this leaves only [~]150 ms to identify the fixated object and select the next saccade goal. This is very little time, suggesting that vision relies on parafoveal processing before and after the eye movement. However, evidence of high-level parafoveal access is sparse. The purpose of our study was to use magnetoencephalography (MEG) combined with eye-tracking and multivariate pattern analysis to identify the neuronal dynamics of parafoveal processing which support natural visual exploration. We demonstrated that future saccade goals in the parafovea could be decoded at the feature and category level peaking at [~]90 ms and [~]160 ms respectively. Simultaneously, decoding of fixated objects at the feature and category level peaked at [~]70 ms and [~]145 ms respectively. Also decoding feature and category specific neuronal information related to past parafoveal objects were sustained for [~]230 ms after saccading away from them. The feature and category of objects in the parafovea could only be decoded if they were in the saccade goal. In sum, we provide insight on the neuronal mechanism of pre-saccadic attention by demonstrating that feature and category specific information of foveal and parafoveal objects can be extracted in succession within a [~]150 ms time-interval and may serve to plan the next saccade. This information is maintained also after fixations and may support integration across the full visual scene. Our study provides novel insight on the temporal dynamics of foveal and parafoveal processing at the feature and semantic levels during natural visual exploration.

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.