Experimental setup.

(A) Illustration of one block of the experimental and control conditions, respectively. During each block in the experimental condition, a peripheral saccade target was shown, which participants were instructed to fixate. The target disappeared before it could be foveated, and once fixation was achieved, a new target appeared, until the block was over (duration: 11 seconds). The timing of target appearance and disappearance from the experimental condition was recorded and used in the control condition, where targets appeared at fixation. (B) Depiction of the 4 stimuli used in the experiment, which were matched in either visual shape (horizontal / vertical) or semantic category (animal / instrument). Each stimulus appeared equally often in both conditions. (C) Histogram of the gaze distance from the stimulus right after stimulus disappearance including all trials. The blocks in which the stimulus edge may have appeared in the participants’ fovea during at least one saccade were excluded.

Foveal feedback can be decoded from V1.

(A) Average decoding accuracy over all pairwise comparisons for control and experimental conditions and for cross-decoding from experimental to control condition. (B) Average decoding accuracies for all early visual areas as a function of eccentricity for both experimental and control conditions. Error bars represent standard error of the mean. Note that the graphs have different scales of decoding accuracy. The central eccentricities (1-5 dva) were measured using a retinotopic localizer, the outer ones (6-10 dva) were inferred from structural data using Neuropythy Melcher and Colby (2008).

Foveal feedback is sensitive to stimulus shape, not semantic category.

(A) Schematic depiction of comparisons to assess the information content of the neural representations. Comparisons across category assess similarity of representations in terms of visual stimulus properties, with lower decoding accuracies indicating coding for visual information. Similarly, comparing across shape assesses categorical information. Comparing across both serves as a baseline, showing how high decoding accuracy is between maximally different stimuli. (B) Decoding accuracies for all comparisons using data from foveal regions of V1 and from the lateral occipital area (LO). Note that the graphs have different y-axes of decoding accuracy.

Correlation of foveal decoding with ROI activation.

(A) Cortical masks used in the ROI analyses. These masks were generated using Neurosynth (Yarkoni et al., 2011) with the keyword “eye movements”. For later analyses, only the 100 most activating voxels were selected in each area. (B) Results of the ROI analyses comparing neural activation as a function of foveal and peripheral decoding in three key areas related to eye movements. Activation in the IPS was significantly higher in association with foveal decoding compared to peripheral decoding.

Decoding stimulus content from all early foveal areas.

Decoding accuracies for all comparisons using data from foveal regions of V1, V2, and V3. These graphs show that the results from V1 outlined in this publication generalize to other regions of the early visual cortex.