Apparatus & eye tracking for assessing global motion perception: participants were asked to perform 100 trials of a 2-alternative, forced-choice, left-versus-right, global direction discrimination task at 2 to 4, equi-eccentric, peripheral visual field locations chosen for testing of predictive oculomotor behavior (circles superimposed on Humphrey visual fields in Fig. 1; red: blind-field locations, blue: intact field locations). All blind-field locations were tested in each patient (red circles in Fig. 1). Time limitations restricted our ability to measure performance at every intact-field location (blue circles in Fig. 1), but at least one intact-field location was assessed in each participant. Across intact-field locations tested, we saw normal NDR thresholds that varied from 0.1-0.3 (Table 1). Percent correct and direction range thresholds were measured during in-lab testing, with central fixation enforced using an Eyelink 1000 eye tracker (SR Research, Mississagua, Ontario, Canada). Tracking was binocular for all participants except for CB3, who was tested monocularly because she exhibited convergence issues. As such, she had her dominant (right) eye tracked and the non-dominant eye patched both for motion perception and PFR testing. Stimuli were presented in a gaze-contingent manner in either intact or blind regions of the visual field. Viewing distance to a luminance-calibrated CRT monitor (HP 7217 A, 48.5 x 31.5 cm, 1024x640p, refresh rate 120 Hz) was 42 cm, enforced by a chin/forehead rest. Experiments were conducted using MATLAB (The MathWorks, Natick, MA, USA) and the Psychophysics toolbox (Brainard, 1997; Kleiner et al., 2007; Pelli, 1997). At the start of each trial, subjects were asked to fixate a small target at the center of the CRT monitor. The Eyelink 1000 eye tracker was accurate to within 0.25o, with a sampling frequency of 1000 Hz. Subjects were allowed a fixation window of only ± 1o around the fixation spot. If gaze moved outside this window during stimulus presentation, the trial was aborted, reshuffled and patients received a noxious auditory tone as feedback, reminding them to improve their fixation accuracy. Following accurate fixation of the central spot for 1000ms, a random dot stimulus appeared in a 5o diameter circular aperture, at one of the pre-determined locations in the peripheral visual field (see colored circles in Fig. 1; NDR thresholds in Table 1). Black dots moved on a mid-grey background with a 250 ms lifetime, a speed of 10 deg/s, and with a density of 3 dots/deg2. Stimuli were presented for 500 ms, accompanied by a tone to indicate stimulus onset. Dots moved globally with a variable range of directions, uniformly distributed around the left- or rightward vectors (Das et al., 2014; Huxlin et al., 2009; Saionz et al., 2020). On each trial, subjects were asked to report the stimulus’ global direction of motion by pressing the left or right arrow keys on a keyboard (Fig. 2A). Task difficulty was adjusted using an adaptive staircase (Levitt, 1971), which increased the range of dot directions from 0o to 360o in 40o steps after each set of 3 consecutive, correct responses; direction range was decreased by one 40o step for every incorrect response (Das et al., 2014; Huxlin et al., 2009; Saionz et al., 2020). Auditory feedback was provided on each trial, indicating the correctness of each response. Participants completed 100 trials at each of the four visual field locations. For each session, we fit a Weibull function to the data to generate a direction range threshold representing the direction range at which performance reached 75 % correct. Direction range thresholds were then normalized to the maximum possible range of dot directions (360o), generating a normalized direction range (NDR) threshold, defined as: NDR threshold (%) = (360o-Weibull-fitted direction range threshold)/360o x 100 Apparatus & eye tracking for PFR measurements: stimuli were generated using the Psychophysics toolbox in MATLAB 2015b on a PC computer (Intel i7 CPU, Windows 7, 8 GB RAM, GeForce Ti graphics card). They were presented on a gamma corrected display (BenQ X2411z LED Monitor, resolution: 1920x1080p, refresh rate: 120 Hz, gamma correction: 2.2) which had a dynamic luminance range from 0.5 to 230 cd/m2, at a distance of 95.25 cm in a dark room. Brightness on the display was set to 100 % and contrast to 50 %, and additional visual features of the monitor such as blur reduction and low blue light were turned off. Gamma corrections were verified with measurement by a photometer. Position of the left eye was recorded continuously in all participants except for CB3, who had her right eye tracked (see above). Eye position was recorded at 220 Hz using an infrared eye tracker (USB-220, Arrington Research, Scottsdale, AZ, USA). The accuracy of the Arrington Eye Tracking system was 0.25˚, with a precision of 0.15˚. To minimize potential head movements, participants performed the task using a bite bar. PFR stimulus and task: CB patients performed a centrally-cued saccade task towards peripheral motion apertures (Fig. 2B) as previously described in visually-intact controls (Kwon et al., 2019). In brief, and as schematically illustrated in Fig. 2B, trials were initiated by fixation of a small, dark, fixation spot presented on a gray background. After a variable fixation period of 150-200 ms, a saccade cue appeared at fixation together with four dot-motion apertures in a square configuration (colored circles, Fig. 1). The cue (dark bar, 1˚ in length, extending from fixation) was used to indicate the target aperture to which the participant should saccade. Each target aperture was 5.5˚ in diameter and centered at (±5˚, ±5˚), with the exception of CB1, for whom the apertures were centered at (±3˚, ±5˚). There were 180 dots total in each aperture, with dot luminance set to 0.5 cd/m2 (100 % contrast) and dot velocity fixed at 10 deg/s. Following parameters from our previous study (Kwon et al., 2019), a Gaussian envelope was applied to each dot-motion aperture to create a gradient in dot contrast from the center of the aperture (sigma= 1˚). To avoid stereotyped eye movements, we varied saccade directions across trials. Thus, the spatially cued motion aperture could appear in the intact or blind-field of a given participant, on any given trial. Of particular note, the motion itself or its direction were irrelevant to the task. The motion within the aperture was 100 % coherent and ran along a direction that was tangential to an imaginary line from the fixation point to the aperture. For each aperture, the motion was selected independent of the other apertures in one of the two tangential directions relative to the center out saccade, either clockwise or counter-clockwise relative to the screen center. We first compared eye movements in which the peripheral motion aperture was either present or absent upon saccade offset. Participants were instructed to make a saccade to the peripheral aperture as quickly as possible following the movement cue. A saccadic grace period (i.e., a maximum latency) was allowed for participants to initiate the saccade. In half the trials, selected at random, the stimulus motion remained present in all four apertures for 300 ms following detection of the eye landing within 3.5 visual degrees from the center of an aperture. In the other half of trials, the stimulus was removed as soon as the eye had been detected leaving the fixation window, thus leaving a blank screen through the post-saccadic period. A saccade was labelled “correct” when it fell at least 3.5˚ from the saccade target center within 90 ms of the eye leaving the fixation window. Participants completed as many trials as possible during a 1-1.5 hour session that sampled equally from the four locations and different stimulus conditions. During a typical session, stroke participants correctly completed on average 103.6 ± 54.2 (1 SD) trials in each blind-field and 137.6 ± 66.4 (1 SD) trials in each intact-field. For comparison against a set of eight normal controls, we analyzed eye movement data from a previous study (Kwon et al., 2019). The previous study employed a task with a subset of trials identical to the current study, but also in another half of trials tested an alternative stimulus configuration with the four apertures along the cardinal axes rather than the quadrants, thus a total of eight locations. During a typical session, normal controls completed on average 59.6 ± 13.8 (1 SD) trials in each of the eight fields. Eye movement recordings and PFR analysis: eye position data were collected as participants performed saccades from fixation to the peripheral target. Eye tracking and saccade detection procedures were identical to those previously published (Kwon et al., 2019). We sub-sampled eye position using the ViewPoint Matlab toolbox (Arrington Research) at the display refresh rate (120 Hz) to initiate gaze-contingent task events. For offline detection of saccadic eye movements, we used the full eye position data recorded at 220 hz and applied an automatic procedure that detected deviations in 2D horizontal and vertical eye velocity space (Engbert & Mergenthaler, 2006; Kwon et al., 2019). Only the trials where the saccade was labelled “correct” were included in the PFR analysis. We then focused our analysis by time locking eye velocity traces on intervals 200 ms prior to saccade onset and 200 ms following saccade offset. Details for eye position filtering, smoothing, and saccade detection were as previously described (Kwon et al., 2019). In brief, the 2D eye velocity was computed from smoothed eye position traces and then projected onto the motion vector in the target aperture on each trial. These projected velocity traces were then aligned to saccade onset or offset, and averaged across trials for each participant. To quantify the net target-related eye velocity in each trial, we used a second measure of eye velocity that did not involve any filtering or smoothing of eye position. We computed a vector for the PFR in units of velocity (deg/sec) as the 2D vector difference in the raw (non-smoothed) eye position from 20 to 100 ms after saccade offset normalized by that time interval. Excluding the first 20ms after saccade offset from this analysis interval reduced the influence of saccade related effects to instead focus on post-saccadic smooth movements. Like velocity traces, we projected this 2D vector onto the vector of the target’s motion to produce a single velocity value along the axis of stimulus motion, which we term the ‘open-loop’ PFR (Kwon et al., 2019). To assess the average PFR across trials, we computed each CB patients’ eye movements relative to the target motion direction so that positive average eye velocities meant that the eye was moving along the target motion direction, and negative average eye velocities meant that the eye was moving opposite to the target motion direction. Finally, we considered to what extent the post-saccadic following response tracked target velocity by quantifying the PFR gain: the eye velocity computed from the open-loop PFR normalized to the target velocity, with +1 indicating a perfect match of eye velocity to the target motion, and negative values indicating eye velocity in the opposite direction.

Visual pathways that guide actions do not necessarily mediate conscious perception. Patients with primary visual cortex (V1) damage lose conscious perception but often retain unconscious abilities (e.g. blindsight). Here, we asked if saccade accuracy and post-saccadic following responses (PFRs) that automatically track target motion upon saccade landing are retained when conscious perception is lost. We contrasted these behaviors in the blind and intact fields of 8 chronic V1-stroke patients, and in 8 visually-intact controls. Saccade accuracy was relatively normal in all cases. Stroke patients also had normal PFR in their intact fields, but no PFR in their blind fields. Thus, V1 damage did not spare the unconscious visual processing necessary for automatic, post-saccadic smooth eye movements. Importantly, visual training that recovered motion perception in the blind field did not restore the PFR, suggesting a clear dissociation between pathways mediating perceptual restoration and automatic actions in the V1-damaged visual system.