Foveal neurons of the monkey superior colliculus signal trans-saccadic prediction errors
Fig 3
Consistency of trans-saccadic error signaling by foveal SC neurons.
(a, b) Similar to Fig 2b and 2c but now showing the normalized firing rate of the population. Each neuron’s results were normalized before averaging across neurons (“Methods”) and error bars denote SEM across neurons. The example neuron results of Fig 2b and 2c were representative of the whole population. (c, d) ROC analyses (“Methods”; also see ref. [16]) comparing intra-saccadic change trials to control trials across the population. Elevations in the area under the ROC value indicate higher firing rates on intra-saccadic change trials. For both low and high spatial frequency foveated targets, there was an elevation in area under the ROC curve post-saccadically when compared to pre-saccadically (and also surpassing the 0.5 chance level). Note that the values below 0.5 at negative times in d reflect an earlier transient associated with the release of fixation, which is not the focus here. Error bars denote 95% confidence intervals. (e) Reafferent response strength (“Methods”) on trials with an intra-saccadic spatial frequency change vs. reafferent response strength on control trials. In both cases, the foveated image feature was the low spatial frequency texture within the circular patch of the saccade target. There was a systematic elevation of reafferent response strengths across the population when there was a visual feature inconsistency across saccades. (f) Similar results for the case in which the foveated texture had a high spatial frequency. (g, h) Neural modulation indices (“Methods”) for the results in e, f. For both foveated image features, there was a systematic elevation of foveal neuron visual reafferent response strength when the experienced post-saccadic image was inconsistent with that expected from the pre-saccadic target appearance. In each panel, the solid vertical line indicates the mean of the distribution, and the dashed vertical line indicates the median. Dark histograms indicate the neurons that were individually significant within a session (“Methods”). The figure’s underlying data are included in S3 Data.