Cortical mechanisms for afterimage formation: evidence from interocular grouping

Whether the retinal process alone or retinal and cortical processes jointly determine afterimage (AI) formation has long been debated. Based on the retinal rebound responses, recent work proposes that afterimage signals are exclusively generated in the retina, although later modified by cortical mechanisms. We tested this notion with the method of “indirect proof”. Each eye was presented with a 2-by-2 checkerboard of horizontal and vertical grating patches. Each corresponding patch of the two checkerboards was perpendicular to each other, which produces binocular rivalry, and can generate percepts ranging from complete interocular grouping to either monocular pattern. The monocular percepts became more frequent with higher contrast. Due to adaptation, the visual system is less sensitive during the AIs than during the inductions with AI-similar contrast. If the retina is the only origin of AIs, comparable contrast appearance would require stronger retinal signals in the AIs than in the inductions, thus leading to more frequent monocular percepts in the AIs than in the inductions. Surprisingly, subjects saw the fully coherent stripes significantly more often in AIs. Our results thus contradict the retinal generation notion, and suggest that in addition to the retina, cortex is directly involved in the generation of AI signals.


12
In case some readers are not familiar with the "indirect proof", here we show a simple example of the 13 "indirect proof" in algebra.

37
As we note, the present study tries to prove the cortical generation notion (2). Thus, we would first assume 38 that the retinal generation notion (3) is true. Generally, AIs appear after adaptation to a stimulus, and adaptation 39 reduces the gain of the visual system. That is, the visual system is not in the same state during AIs as it is during 40 the presentation of a stimulus with AI-similar contrast. Assuming that the retinal generation notion is correct, i.e.

41
the retina is the only origin of AI signals, comparable contrast appearance would require stronger retinal signals 42 in the AIs than in the inductions with AI-similar contrast (4). According to the findings in Experiment 2 that 43 increased inducing contrast caused decreased interocular grouping during the inductions (5), we may reach a 2 prediction (6) that more frequent monocular patterns should be perceived in the AIs than in the inductions with 1 AI-similar contrast. However, this prediction (6) contradicts the empirical observations (1) of subjects actually 2 seeing the fully coherent stripes significantly more often in AIs. As a result, the retinal generation notion (3) is 3 denied. The present study thus suggests that AI formation should also involve cortical processes in addition to 4 the retinal mechanisms (2).

6 7
Supplemental figures and tables 8 9 Figure S1. Candidates for the Butterworth filters and the filtered stimuli. Images were distorted severely for 10 cutoff frequencies lower than 1 cpd. No obvious distortion was found when the stop frequency was between 0.3 11 cpd to 0.7 cpd with a higher cutoff frequency ( ≥ 1 cpd). The central image (framed with dotted lines) was the 12 one selected in the study. The cutoff frequency was 1 cyc/deg. The filtering retained the energy for spatial 13 frequencies lower than 0.5 cyc/deg.  Figure S7. In Experiment 2b, we selected the middle patch in Figure S1 based on our empirical feeling 2 for the AIs. After acquiring the eye movement data, we re-examined whether our selection was hierarchy. We then performed a pixel-by-pixel correlation analysis between this simulated image and 10 each of the 9 possible candidates shown in Figure S1. Specifically, each image array was reshaped 11 into an N by 1 vector. Then we ran a Pearson's correlation analysis between each pair of vectors to 12 obtain a 3-by-3 array of the correlation coefficients for the 9 candidates. For each subject, the arrays 13 for the 15 trials were averaged to show the average correlation coefficients for the 9 candidate patches.     Table S4. Results of 2D Gaussian fitting. We fitted a 2D Gaussian model to the spatial distributions of gaze 12 positions during the full contrast induction phases for each individual subject (marked as 1~12), and to the data 13 pooled from all the subjects (marked as "All"). Here, 0 , 0 showed the positions of the fits (unit: degree),

14
, showed the widths of the fits (unit: degree), and A was the amplitude of the fit.