Closed loop motor-sensory dynamics in human vision.

Vision is obtained with a continuous motion of the eyes. The kinematic analysis of eye motion, during any visual or ocular task, typically reveals two (kinematic) components: saccades, which quickly replace the visual content in the retinal fovea, and drifts, which slowly scan the image after each saccade. While the saccadic exchange of regions of interest (ROIs) is commonly considered to be included in motor-sensory closed-loops, it is commonly assumed that drifts function in an open-loop manner, that is, independent of the concurrent visual input. Accordingly, visual perception is assumed to be based on a sequence of open-loop processes, each initiated by a saccade-triggered retinal snapshot. Here we directly challenged this assumption by testing the dependency of drift kinematics on concurrent visual inputs using real-time gaze-contingent-display. Our results demonstrate a dependency of the trajectory on the concurrent visual input, convergence of speed to condition-specific values and maintenance of selected drift-related motor-sensory controlled variables, all strongly indicative of drifts being included in a closed-loop brain-world process, and thus suggesting that vision is inherently a closed-loop process.

Introduction

The visual system usually perceives its environment during egomotion [1-3]. In particular, retinal encoding results from continuous interactions between eye movements and the environment [1, 4–10]. The kinematics of eye movements contain two major motion components: saccades, which quickly (speeds between ~10 to ~1000 deg/s) shift the gaze from one region of interest (ROI) to another [11-13] and drifts, which slowly (speeds in the order of 1 deg/s) scan each ROI [1, 14–16]. These two kinematic components, saccades and drifts, fully characterize the movements during all kinds of visual activities, whether while fixating, pursuing moving targets, reading or exploring a scene. Thus, fixation includes small saccades and drifts, pursuit includes mostly drifts with occasional saccades (when the target disappears), and reading and scene viewing include saccades and drifts. The faster components, saccades, are often divided into macro-saccades and micro-saccades, depending on their amplitudes. Typically, they shift the gaze to targets beyond or within the foveal field (~1–2 visual degrees at the center of the entire visual field [17], respectively). It is currently accepted that all saccades can be characterized along the same kinematic continuum, controlled by the same circuits, and serve the same function–shifting the gaze to a new ROI [15, 18–22]. Hence, here we analyze all saccades in one category.

Two contrasting perceptual schemes might be consistent with this fast-shifts and slow-scanning kinematic pattern that dominates primate vision: an open-loop computational and a closed-loop dynamical scheme. In both schemes, saccades are controlled, at least in part, in a motor-sensory closed-loop manner, in which ROI selection depends on the accumulated visual information [11, 12, 21, 23–26]. The two schemes, however, divide in their assumed acquisition process in between saccades, during what is referred to as drifts. The open-loop computational scheme assumes that at each ROI the visual system acquires one snapshot of the ROI per saccade and computes its internal representation through an open-ended sequence of computations, much like in current computer-vision implementations [27-31]. The dynamical scheme, in contrast, assumes that the acquisition, and hence also the perception, of each ROI is a continuous closed-loop process, whose dynamics are determined by the interactions between the drift motion and the external image [32-36].

As a result, the two schemes entail contrasting predictions. (i) The computational scheme predicts that the ocular drifts will not depend on the concurrent visual input, whereas the dynamical scheme predicts that it will. (ii) The dynamical scheme further predicts that the drifts will exhibit convergence dynamics during perception (that is, their kinematics should gradually approach steady values during the process), as expected from the dynamics of a closed-loop system around its attractors [36]; the computational scheme predicts that drifts dynamics should follow a predetermined or random pattern. (iii) Lastly, with closed-loop perception, adaptive changes are part of a process that attempts to maintain specific dynamic variables (the “controlled variables”) within specific ranges of values (the “working ranges”) such that perception is optimized [36-38]. Open-loop systems do not have this active capacity and depend on a-priori mapping between environmental and internal variables. Thus, for example, closed-loop perception predicts that oculomotor dynamics will change such that retinal outputs will have temporal characteristics that are optimal for neural processing whereas open-loop perception predicts that such oculomotor changes will reflect motor adaptations such as muscle fatigue.

To test the predictions of the two schemes, we measured ocular behavior while modulating the available spatial information by reducing image size and by mimicking tunnel vision. We tracked subjects’ gaze while they were viewing simple shapes of two sizes, and while revealing to them either the entire visual field (“natural viewing”) or only a portion of the image around the center of their gaze (“tunneled viewing”).

Results

Five subjects were asked to identify an image on a screen as one of five options (square, rectangle, circle, triangle or a parallelogram) after viewing it either naturally or through tunneled viewing, during which spatial information was exposed only around the center of their continuously-tracked gaze. Two image sizes (~10.80±0.15 deg2 and ~0.90±0.03 deg2) and two tunneling windows (~2.90±0.15 deg2 and ~0.24±0.03 deg2, respectively) were used, with similar ratio between the image and window sizes (see Methods, and ). Success rates were 100% for natural viewing of both sizes, 94±6% for the tunneled-large shapes and 60±2% for the tunneled-small shapes. Only correct trials were used for the analyses reported here.

As predicted by both perceptual schemes, limiting the available spatial information had a dramatic effect on gaze locations. During natural viewing of large shapes, the gaze was typically directed around the center of the shape, while during tunneled viewing of large shapes the gaze was typically directed to the borders of the shapes (see examples in ). In fact, most (55±7%) of the saccades made by all subjects in all tunneled-large trials were border-saccades, i.e., started and ended near the border (, see Methods).

With tunneled viewing, saccades were directed along the image borders ( and ), without being able to acquire any information from the target location prior to landing (mean saccadic amplitude was significantly larger than window size, 3.45±0.07 vs 2.90 deg; p<0.05, n = 4648, one-tailed t-test; the large tunneling window is overlaid on the rectangle example in ). Moreover, saccadic border-following was evident already at the beginning of each trial and, on average, kept a constant profile along the trial (). This suggests that saccade planning was primarily based on the information collected during the immediately preceding fixational pause(s) rather than on an accumulated estimation of the object’s shape along the trial. Thus, our tunneling results show that target selection can be based on the information actively acquired at the fovea during a single pause without using peripheral cues.

To better understand how this information is collected in each pause, we examined the dependency of ocular kinematics on concurrent visual acquisition within each fixational pause. The exact kinematics of the drifts are not directly accessible to current tracking technologies, due to measurement noise [8, 39]. In many previous studies that address drift kinematics, thus, drift speeds are reported after significant filtering. However, filtering the raw data also removes the fast ocular transitions that are known to be most effective in activating the visual system [40-44]. Since we are interested here in the kinematics that are relevant to visual acquisition, we chose to report the kinematic values computed from the unfiltered data. Since these data are contaminated with measurement noise, we term the computed speed “the upper-bound of the drift speed (UB speed)”. Naturally, the distributions of our UB (upper-bound) speeds (Fig 2A, blue) show higher values than those typically reported in the literature [45]. High drift speed values might have also been reported if the drift movement included small misdetected saccades. To verify that our analysis did not yield higher drift speeds because of a different classification of the ocular movement to saccades and drifts, we replicated the filtering and saccade detection methods reported previously [45]. Importantly, this method results in lower drifts speeds (Fig 2A, black), without significantly changing the classification of the data. Fig 2B–2D show that our classification method, with its matched saccades detection threshold (see Methods and [16, 21, 46–48]), yields similar saccade statistics and does not yield saccade-like components in its drift sections. The saccadic rate and saccades duration are not significantly different between the two methods (p>0.1, two tailed t-tests, Fig 2B and 2C). All our drift traces were inspected to verify the lack of saccade-like components (see Methods). A typical single trial example (Fig 2D) demonstrates the difference between the algorithms (note the single difference in saccade detection at t ~ 2230ms). Given the arbitrariness unavoidably included in these computations, all our results are based on comparisons between conditions, thus are insensitive to the absolute values of the actual kinematic variables.

Fig 2

Bounds of drifts kinematics.

(a) Distribution of the instantaneous drift speeds of all measured fixational pauses of the experiment. Two methods were used: 1.the upper bound (UB) was assessed by derivation of the raw eye position data, and thus using a higher saccadic detection threshold of 16 deg/s for minimal peak velocity and 0.3 deg for minimal amplitude (blue, UB–these are the values reported in this paper). 2. A commonly used published method was assessed by filtering (a third order Savitzky-Golay filter with window size of 3 samples [45]) and thresholding (lower saccadic detection thresholds: 3 deg/s for minimal peak velocity and 0.05 deg for minimal amplitude [45]) our data (black). (b) Distributions of saccadic rates calculated on all experimental data using our UB method (blue) and the typical reported one (black). No significant difference was found (p>0.1, two tailed t-test) (c) Distributions of saccades durations calculated on all experimental data using our UB method (blue) and the typical reported one (black). No significant difference was found (p>0.1, two tailed t-test). (d) A single trial example of saccades detection (red) on the horizontal and vertical recorded data, using our UB methods (upper panel, blue) and the typical reported one (lower panel, black). One mismatch can be seen on 2230ms.

We first examined the dependency of the scanning patterns of drifts on concurrent visual details. Comparing drifts in ROIs that did or did not contain borders revealed the following difference. When challenged by tunneling, drifts scanning depended on the concurrently scanned visual image—the eye tended to drift in a curvier pattern when scanning a border, remaining closer to its starting location. The distributions of curvature indexes (see Methods) differed between border and non-border drifts in both tunneled conditions (p<0.05, Kolmogorov–Smirnov test) (). A difference between the mean values was also found in the natural small condition (p<0.05, two tailed t-test; ), suggesting a similar trend of drift curviness also when the visual system was challenged by small images.

The presence of an image border in the field of view had an immediate effect on additional kinematic variable—the UB-speed of the ocular drifts (see Methods). The drifts UB-speed was significantly lower for border-containing ROIs ( p<0.05 in 3 out of the 4 conditions, two tailed t-tests). The viewing condition (natural or tunneled) had an even stronger effect on the drifts UB-speed. During tunneled viewing the drifts UB-speed was significantly higher than during natural viewing (p<0.05, for both image sizes, two tailed t-tests; 1st and 3rd panels versus 2nd and 4th; see also ).

In addition to a dependency on real-time sensory information, the closed-loop scheme also predicts convergence dynamics [36, 49–52]. Indeed, in all four conditions the ocular drift UB-speed exhibited a converging-like behavior (). The convergence dynamics may differ between border and non-border drifts, especially so for small images (). On average, consistent with previous observations [53], the drift UB-speed was strongly modulated at the beginning of each pause and gradually converged to an asymptotic value during the pause ( and ). The convergence to an asymptotic target speed value was evident also for individual subjects, and these target speeds were typically stable for the entire durations of the fixational pauses (). Importantly, the drift UB-speeds we measured did not depend on the pupil size or on the amplitudes or speeds of the saccades preceding them (r2 < 0.01 for all cases), and these variables were not significantly different across viewing conditions ().

The dependency of ocular kinematics on the concurrent sensory information described so far indicates a closed-loop sensory-motor behavior of the visual system and suggests that the closed-loop processing is done at multiple levels, at least those levels controlling saccades and drifts. What might the visual system try to maintain using these loops, and in particular during natural and tunneled perceptual epochs? To address these questions, we analyzed the averaged kinematics in each condition per subject. For the majority of subjects, both the saccades’ mean rate (Rs, ) and mean target (i.e., for time>50 ms) drift UB-speed per pause (Sp, ) increased in tunneled conditions compared to natural viewing. These differences did not result from the differences in trial length (). Interestingly, the distance travelled during a pause, the drift length or amplitude, (Xp), was hardly affected by tunneling ().

Assuming that visual information is indeed acquired during fixational pauses [23, 54–57], as expected in a closed-loop scheme, and that photoreceptors are activated by illumination changes, the mean rate of visual acquisition (during a pause) should be proportional to Sp [9, 14] and the amount of visual information collected during that pause should be proportional to the integrated distance scanned by the eye (the length of its trajectory) during the pause (Xp).

Importantly, on average, the tunneling-induced changes in Rs and Sp compensated each other, keeping Xp unchanged, for each stimulus size (). Thus, when tunneled, the visual system appeared to increase the ROI sampling rate (Rs) while maintaining Xp and compromising (i.e., loosening) the control of Sp. If the control of Sp was indeed loosened, the trial-to-trial variability of Sp should increase. Indeed, while Sp exhibited relatively small coefficients of variation (CVs) during natural viewing (0.90 and 0.57 for large and small, respectively), its CVs increased significantly when tunneled (1.22 and 1.02 for large and small, respectively; two tailed f-tests, both p < 0.05).

Control variables and closed-loop visual acquisition.

(upper table) Mean ± SEM of visual scanning variables during each viewing condition (NL, natural large; TL, tunneled large; NS, natural small; TS, tunneled small). Sp is the mean instantaneous drift speed during a pause in deg/s, Xp is the distance travelled during a pause (drift length) in deg and Rs is the mean saccadic rate per trial. Light blue arrows indicate values that do not differ statistically (p > 0.15); all other differences were statistically significant (p < 0.05). (a) Coordinated saccade and drift closed-loops. The two loops interact with each other, coordinating the control of the acquisition. (b) Controlled variables. The scheme symbolizes the maintenance of Sp across image size (large and small) and the maintenance of Xp across viewing conditions (natural and tunneled).

Interestingly, a different strategy appeared with size changes. When viewing small sized images, the visual system decreased the ROI sampling rate (consistent with [58]) while maintaining Sp and thus increasing Xp (). A sequential open-loop scheme, in which the visual input affects Rs and Rs affects the drifts variables, is ruled out here; the mean values of Rs did not systematically change along with either Sp or Xp, and the pause-by-pause correlations between each Sp or Xp and its preceding instantaneous Rs (i.e., the inverse of the inter-saccadic-interval) were negligible (R2 < 0.06).

Discussion

Consistent with previous reports [14, 59–64], our results demonstrate clearly that ocular drifts are affected by the visual system. What our results add is that this effect is part of a motor-sensory closed-loop process: drifts kinematics, which affect the visual input, depend on this very same input. We demonstrate this closed-loop behavior via three major observations. First, we showed that the trajectory of the drift within each ROI depends on the concurrent visual sensory data. During fixational pauses in which the drift scanned an ROI that contained an image border its scanning trajectory was controlled to be slower and curvier than when it scanned a non-border ROI (). Second, we showed that the drift speed dynamically converges to a condition-specific target speed; even when starting with a speed that was similar for two conditions, in each condition the visual system gradually changed its drift speed until it converged, on average, to a value that was specific to that specific viewing condition ( see below). Third, we found two potential drift-related controlled variables—drift speed and scanning distance–each affecting the visual information acquired via drift motions (). Controlled variables are a clear signature of closed-loop systems [37, 52, 65].

The dynamics of closed-loop systems are limited by the loop delay, that is, the time it takes for the effect of a signal to travel along the loop. In the drift system this delay can be estimated as < 50 ms [58]. Accordingly, drift oscillations around 10 Hz [61] may reflect fluctuating loop dynamics and may point to the characteristic limit cycle of the loop. Thus, the dynamics that can be controlled in the drift loop are only the slow dynamics and not those determining the fast direction changes carrying out the drift motion (often termed “tremor” [8])–these direction changes are most likely generated by the collection of stochastic processes in this system, including motor-neuron spikes and muscle twitches [66]. It is the slow dynamics of these fast movement events that is under control. Based on our data, we propose that the slow dynamics of the drift movements may be controlled by both the drift and saccade loops (). The data supporting cross-loop control are those demonstrating inter-relationships between the saccadic rate and the drift speed and length (Results, last two paragraphs). The data supporting within-loop control of the drift are those demonstrating within-pause control of the spatial () and temporal () behavior of the drift. As we showed above (Results, last paragraph), these behaviors cannot be regarded as merely reflecting saccadic behavior.

These results do not rule out additional contributions to drifts control, other than those of the concurrent visual input, such as experience- and context-dependent biases [14, 60, 61] or slow motor-control processes [67]. Under the closed-loop framework, these broader and slower contributions would operate within higher-level sensory-motor loops [21, 36].

Measurement noise and head movements

Our measurements are contaminated with measurement noise. With video-based methods, such as the one we used, the major noise sources are inaccurate calibration, slow changes in the pupil size and head movements [68-70]. Importantly, however, pupillary responses to changes in the visual input exhibit significantly slower dynamics than those shown here [71]. Moreover, our results are based on comparisons between conditions, and we used the exact same methods in all conditions. We thus ruled out the possibility that the differences we found stemmed from differences in pupil size () or insufficient calibration (we calibrated the device before each trial–see Methods) between conditions. Head movements most likely cannot account for the rapid kinematics demonstrated in [72, 73]. We remain with a possible contribution of head movements to the slow kinematics; thus, it is possible that part of the differences we measured in the slow kinematics of the drifts are in fact differences in the kinematics of head movements. However, importantly, this possibility has no effect on our major conclusion that visual acquisition is a closed-loop motor-sensory process. Since head movements always contribute to retinal image-motion [1, 59, 73, 74], whether retinal motion was controlled in a closed-loop manner in our experiment via drifts alone or via drifts + head motion makes no difference in this respect.

Another noise source that might be included in the visual loop in our experimental design is the noise introduced into the visual stimulus by our gaze-contingent protocol, due to measurement errors and delays. Importantly, this noise may be generated only in the tunneled viewing conditions, increasing the difficulty of the task in these conditions. Yet, similar to head motion, this noise is included in the retinal motion and thus affecting the behavior of the visual loops (see below).

The exact value of the drift speed is not directly accessible to current tracking technologies [8, 39]. We thus assessed the upper bounds (UB) of the instantaneous drift speed by analyzing the unfiltered tracking data [16]. We also verified that our recording method generates measures of eye speed that match those obtained in previous studies with another method (see and related text). Assuming that the recording noise was similar across the different conditions, comparing ocular kinematics across conditions should yield the same results whether based on the upper bounds or any other method. Indeed, this was the case here (e.g., ). Moreover, we have verified that our method does not include non-drift components in the drift calculations by comparing the saccadic rate and saccades duration between the two methods (). It should be emphasized here, again, that since all our conclusions are based on comparisons of the same measures across conditions, our conclusions are insensitive to the absolute values of the actual kinematic variables.

Drift convergence and post-saccadic enhancement

Saccades are typically followed by a short period of relatively high speed drift, termed post-saccadic enhancement [53, 75]. These periods are associated with periods of enhanced neuronal activity in the brainstem oculomotor nuclei, a phenomenon that was termed pulse-step motor command: a strong initial pulse of action potentials that rapidly moves the eye at saccadic velocities, and a subsequent tonic discharge (step command) that maintains the eye at the saccade endpoint [53, 76]. Our results suggest that this initial boost of drift speed is the beginning of a convergence process, a process through which the oculomotor system converges upon a target drift speed that fits the characteristics of the current condition (e.g., the amount of available visual information, as implied here by the difference in target speeds between natural and tunneled viewing conditions; ). Such an initial boost may be advantageous for a rapid convergence process, as it dictates the direction of convergence–always decreasing drift speed. In any case, our observation that the drift speed converges to condition-dependent and saccade-amplitude-independent speed strongly suggest that the process that stabilizes the drift speed is not merely a passive attenuation of brainstem oculomotor activity.

We showed here that the system converges to its target speed within < 100 ms on average (). Such a rapid convergence can only be achieved in loops with cycle times that are significantly smaller than the convergence time–in the visual system these are low-level (certainly sub cortical; [77]) loops. These loops, thus, probably involve components that are already known to control saccades [78-80] and smooth pursuit eye movements [81]. Which components are shared across these control functions [82] and which are not is an exciting open question. For example, smooth pursuit is hypothesized to be implemented by a feedback loop that minimizes the difference between a target velocity and the actual ocular velocity, where the target velocity is determined by other circuits of the visual system (e.g., [81]). The drift loop implicated by our results () may or may not use the same components proposed for the velocity feedback loop allowing smooth pursuit. If it does, then during natural vision without voluntary pursuit, this loop functions with the target velocity being zero. This of course does not mean that the loop attempts to cancel ocular motion. Rather, it means that the loop attempts to maintain some steady-state with its visual environment. Importantly, this attempt is sufficient for smoothly pursuing consistent external movements [36].

Controlled variables

Our data suggest that under normal conditions the visual system controls its drift speed such as to maintain it within a specific range [see also [49]]. One plausible reason for such a control is to maintain temporal coding within a range suitable for neural processing [8, 21, 52, 83, 84]. When viewing small-size images, the visual system did not compromise this control, possibly for maintaining the reliability of sensory data. However, when challenged with tunneled viewing, which decreases the amount of available spatial information, the visual system compromised the control of drift speed, allowing its increase, for maintaining constant scanning distances under an increased rate of ROI switching (shorter fixational pauses), thus increasing total spatial information (). We assume that part of these modulations result from global changes in the system induced by task difficulty; specifically, such changes might induce the increased rate of ROI switching observed here.

The rational described above is based on the assumption that the loop cannot control two competing variables simultaneously [37, 65]. Imagine that a slowly fluctuating noise is added to the ocular motion due to muscle fatigue. If the loop attempts to maintain one variable (say Xp) constant then the fluctuations of the noisy signal will necessarily induce matched fluctuations in the other related signal (e.g., Sp, if saccadic rate is enforced) and vice versa. The compromised variable functions as a “shock absorber” for the controlled one.

Temporal coding is forced in active sensing. The movements of the eye convert external spatial information (spatial offsets) to temporal information (temporal offsets, i.e., delays) at the output of the retina [8, 9, 85]. The faster the movement of the eye the larger the amount of information coded in a given period, but also the smaller the characteristic temporal delays. Typical mammalian neuronal circuits are limited in their ability to follow or decode temporal delays that are smaller than 1 ms [83, 85, 86]. The temporal inter-receptor delays expected when scanning individual edges are equal to the inter-receptor spacing divided by the drift speed [21]. Thus, assuming foveal densities of up to 166 photoreceptors per deg [87, 88], inter-receptor temporal delays were likely approaching 1 ms on average ().

The dependencies observed here suggest that if the acquisition system indeed functions as a closed loop, one of its controlled variables would likely be the minimal temporal delay among its foveal ganglion cells. If this is the case, inverse relationships between its motor-sensory variables are expected. The most reasonable closed-loop scheme, according to our data, is a one in which the bottom-up arc implements the dependency of Rs on Xp, and the top-down one the dependency of Xp on Rs. Thus, saccadic ROI switching, which is controlled by both global scene- and task-related factors [5, 11, 12, 15, 20, 89, 90] and local ROI-specific factors [15, 21], appears to be controlled in coordination with the control of the ocular drift. We further show here that this coordination is manifested in real time on a cycle-by-cycle basis ().

In our paradigm, the coordinated saccade-drift control maintained, on average, Sp across image sizes in natural viewing, and Xp across viewing conditions (). It should be stressed out that these specific control strategies may be specific to our paradigm, which used simple shapes, and may differ in more realistic environments. Yet, even with these simple shapes, the identification of controlled variables was possible only at the population level, and not with individual subjects. This is not surprising, given the complexity of the human visual system and its need to cope with many changing conditions in life. These results, thus, suggest that Sp and Xp are controlled in most of the individual subjects in the manner described above, as part of a larger set of idiosyncratic controlled variables.

Conclusions

That visual information is acquired continuously along each fixational pause had been demonstrated previously [23, 54–57]. Our results now provide substantial evidence that this acquisition is accomplished by retino-neuro-oculo closed-loops, involving low (certainly subcortical) levels of the visual system. Together with the known closed-loop basis of ROI selection, our results now suggest that vision is inherently a closed-loop process. That is, that all visual processing components are accomplished within brain-environment closed-loops. Being inherently a closed-loop system does not preclude the ability of the visual system to perceive a certain amount of information even when eye movements are prevented, as often occurs in the laboratory. Being a closed-loop system simply entails, in this context, that every movement of the eye is a component of visual perception and that the system continuously attempts to converge to its attractors, getting as close to a given attractor as possible, given the collection of internal and external constrains.

Methods

Subjects

5 healthy subjects with normal vision at the ages 21–28 participated in the experiments (3 females, 2 with right dominant eye, 3 with left dominant eye). All subjects were given detailed explanation about the eye tracker device and the behavioral task, and were paid for their participation (50 NIS, ~15 USD per hour). Informed written consents were obtained from all subjects, in accordance with the approved Declaration of Helsinki for this project. The experimental procedures were approved by the Tel Aviv Sourasky Medical Center Helsinki committee.

Experimental setup

The experiment took place in a darkened and quiet room where subjects sat in front of a high-resolution, fast computer screen (VPixx, 1920x1080, 120Hz). The movements of the dominant eye were recorded using EyeLink II at 100Hz (enabling real time screen manipulation after each sample) while the other eye was blindfolded. Each cycle of sampling + updating was bounded by 10 ms (verified on-line), thus, the maximal delay was 10ms. We have verified that none of the subjects perceived any delay in the screen manipulation by directly asking after each session (1. “Did you feel at any moment that the stimuli is following your eyes? If yes, estimate the delay”. 2. “Did you feel at any moment that your eyes are following the stimuli? If yes, estimate your delay”. All subjects reported no delay at all analyzed trials). Subjects sat 1 meter away from the screen and placed their chin on a chinrest to reduce head movements.

Stimuli and gaze windows

Two kinds of images were created: ‘large’ and ‘small’, and each was associated with a specific gaze window–a horizontal rectangle centered on the subject’s gaze at each sample and through which the image was exposed. The large shapes occupied 10.80±0.15x10.80±0.15 deg (720±10x720±10 pixels), and the large gaze window was 2.90±0.15x1.90±0.15 deg (190±10x130±10 pixels). The small shapes occupied 0.90±0.03x0.90±0.03 deg (60±2x60±2 pixels) with a gaze window of 0.24±0.03x0.16±0.03 deg (13±2x9±2 pixels). The ratio between image and window size was the same for both image sizes ().

Experimental design

We tested the performance of subjects in a five forced choice shapes recognition tasks. In each trial, one out of five filled gray basic shapes against a black background was presented (square, rectangle, circle, triangle and a parallelogram; see ). These images were presented in two forms, large and small, as described above. Subjects were tested during 5 days. During days 1–3 they performed 2 tunneled vision sessions, the first one with large images and the second one with small images. On day 4 they performed two additional tunneled vision sessions with small images. On day 5 they performed 4 sessions of natural viewing, 2 repetitions with each image size: large, small, large, and small (natural trials were only performed after all tunneled trials ended, to avoid familiarity with the shapes). Each tunneled trial lasted up to 30 s, mean trial duration for tunneled large was 9+2 s and for tunneled small 20+4 s (trials with natural viewing lasted 3 s, hence all comparative analyses were further verified using only the first 3 s of all tunneled trials, to control for trial length confounds, ). There were at least 2 repetitions of each shape in each session (10–12 trials per session, only the first two repetitions of each shape were used for analysis), and hence each session lasted up to 12 minutes. Before the beginning of each trial the eye tracker was recalibrated [69]. We used an adaptive calibration method: subjects fixated on a ‘+’ and waited until the gaze report of the eye tracker (marked as ‘X’) stabilized. The error between the two markers was used to correct the eye tracker’s output during the next trial. At the end of each trial subjects reported which of the five shapes was presented, and received a ‘correct/wrong’ feedback. In the tunneled vision sessions, subjects had to identify a shape that was “hidden” on the screen and exposed only through the gaze window (see above). In the natural vision sessions, subjects had to identify the same shapes, naturally viewing them with no constrains.

Eye movement processing

A velocity based algorithm [modified from Bonneh et al. [47]] was used for detecting all saccades and drifts. We used the following threshold parameters for saccades detection: 16 deg/s minimal peak velocity and 0.3 deg minimal amplitude. Each detected saccade was visually examined to verify the quality of saccadic detection, and each fixational drift was visually examined not to include miss-detection of small saccades. Fixation periods between saccades were labeled drifts only if they lasted > 30 ms. Upper bounds of the instantaneous drift speed were computed as the derivative of the raw eye position signal (100Hz) [16]. For and we in addition used a filter (third order Savitzky-Golay filter with window size of 3 samples) and saccadic detection threshold parameters of 3 deg/s for minimal peak velocity and 0.05 deg for minimal amplitude, in order to match and compare previous publications [45].

Borders analysis

Border-following movements during tunneled viewing were those movements in which the border of the shape was visible to the subject during the movement. This was determined by the window size: saccades or drifts pauses that started and ended at less than 1.8 deg (for large), or 0.15 deg (for small) from a border, were classified as “border saccade” or “border drift”, respectively. During natural viewing border-following movements were defined using the same distance criteria.

Curvature index

We defined an index for drift curvature, where Xp equals the length of the drift trajectory and Dp equals the linear distance between its starting and ending points. Hence, c = 0 represents a straight line and c = 1 represents a closed curve.

Statistical analyses

Two-tailed t-tests were used to evaluate the significance of differences in the mean values of kinematic variables (Rs, Sp, Xp, c), all of which exhibited normal distribution for all subjects. N’s for Rs statistics in tunneled-large, natural-large, tunneled-small, natural-small, respectively, were: Subj1 (30,19,30,20); Subj2 (29,20,30,19); Subj3 (27,20,28,18); Subj4 (30,20,30,19); Subj5 (25,20,30,20); N’s for Sp and Xp statistics were (respectively): Subj1 (1756,229,2392,154); Subj2 (534,110,2626,45); Subj3 (723,129,2108,101); Subj4 (409,104,1288,46); Subj5 (1085,203,2775,136).; Distributions of values were also compared using two-sample Kolmogorov-Smirnov tests. Variances were compared via the corresponding coefficients of variation (CV = variance/mean), using two-sample F-tests. Data are expressed as mean ± S.E.M. Shape presentation order was randomized using a uniform distribution. No blinding was done during analysis and none of the data points was excluded.

Demonstrations of tunneled viewing.

Movies of tunneled viewing of large (top) and small (bottom) shapes. The right panels show the entire shape with the tunneling window superimposed and the left panels show what was presented on the screen.

(AVI)

Click here for additional data file.

Demonstrations of the eye trajectories presented in Fig 1.

Movie is slowed down by 2.4. Colors as in Fig 1. In addition, black segments represent an un-classified movement along the trial (e.g., pauses shorter than 40ms).

Fig 1

Eye trajectories.

(a) Examples of eye trajectories in single trials with natural (left) and tunneled (right) viewing of large shapes. Saccades, lighter blue; drifts, darker blue; traces, horizontal and vertical components as a function of time next to each example (movies of these examples are in ). The small yellow rectangle overlaid on the rectangle image shows the window size through which the images were viewed in the tunneled condition. (b) Fractions of border saccades in the two large-shapes conditions for each subject (small dots) and their means (large dots). (c) Distribution of the angles between the orientation of the border scanned during a pause and the direction of the immediately following saccade with tunneled large viewing. Purple, empirical data, Gray, shuffled data (saccade directions were shuffled before angle computation; average of 100 repetitions is depicted). The histograms of the empirical and shuffled data are superimposed in the graph, and statistically different (p<0.05, two-sample Kolmogorov-Smirnov test). Analyses of individual subjects are presented in . (d) Same as (c) for the first, second, third and fourth quarter of each trial. In all four cases, the distributions of the empirical and shuffled data differed significantly (p<0.05, two-sample Kolmogorov-Smirnov tests).

(AVI)

Click here for additional data file.

Control for trial duration differences.

Related to Fig 4. The analyses described in were repeated for the first 3 s of the tunneled conditions, a time period equal to the duration of natural viewing trials. P values represent the probability that the values measured in the relevant tunneled condition were drawn from the same distribution as those measured in the natural viewing conditions (two tailed t-tests for means and two tailed f-tests for variances).

Fig 4

Kinematics of saccades and drifts.

(a) Mean saccadic rates (Rs) in natural (black) and tunneled viewing for large (purple) and small (green) image sizes. Data for each subject is presented (*, p<0.05, two tailed t-tests). (b) Distributions of the mean instantaneous UB-speeds of drifts (target speed values, for time>50ms) per pause (Sp) in the four experimental conditions; data presented as in (a) (*, p<0.05, two tailed t-tests). (c) Distributions of the distances travelled during a pause (drift length; Xp) in the four experimental conditions; data presented as in (a) (*, p<0.05, two tailed t-tests). N’s for Rs statistics in tunneled-large, natural-large, tunneled-small, natural-small, respectively, were: Subj1 (30,19,30,20); Subj2 (29,20,30,19); Subj3 (27,20,28,18); Subj4 (30,20,30,19); Subj5 (25,20,30,20); N’s for Sp and Xp statistics were (respectively): Subj1 (127,497,97,259); Subj2 (92,164,35,491); Subj3 (55,195,41,415); Subj4 (60,115,20,286); Subj5 (117,347,53,446).

(JPG)

Click here for additional data file.

Saccades direction following border scanning.

Related to Fig 1B and 1C. Distributions of the angles between the orientation of the border scanned during a pause and the direction of the immediately following saccade [data shown in purple, shuffled data (saccade directions were shuffled before angle computation; average of 100 repetitions is depicted) in gray]. Data for each is presented. All distributions are statistically different, p<0.05, two-sample Kolmogorov-Smirnov tests.

(JPG)

Click here for additional data file.

Dependencies between kinematic variables.

Related to Fig 3. (a) The mean amplitude of the preceding saccades of all pauses in each of the four experimental conditions; no significant difference was found (p > 0.1, two-tailed t-test); similarly, no significant difference was found for the maximal saccade speed (p > 0.1, two-tailed t-test). (b-d) Each data point represents a single pause (mean pause speed versus (b) the amplitude of the preceding saccade, (c) the maximal speed of the preceding saccade (d) mean pupil size during the pause). R2 < 0.01 in all cases. Colors as in Fig 3. (e) Mean within-pause instantaneous pupil size (f) Mean within-pause instantaneous drift speed (no correlation with the mean within-pause instantaneous pupil size, R2 = 0.02, p = 0.55) (g) Lower bound of the mean within-trial instantaneous drift speed, calculated from the filtered data (a third order Savitzky-Golay filter with window size of 3 samples [45]).

Fig 3

Drifts curvature indices and instantaneous UB-speed.

(a). Normalized distributions of curvature indices of border drifts trajectories (brown) and non-border drifts trajectories, in the four experimental conditions. Natural viewing in black and tunneled viewing in color (large in purple and small in green). The vertical lines depict the mean values of curvature indices per condition. Mean ± SEM of the curvature index and the number of pauses for border and non-border scanning, respectively: 0.48±0.04 vs. 0.49±0.01 for natural-large (n = 192,374 pauses); 0.52±0.01 vs. 0.48±0.01 for tunneled-large (n = 1405,757); 0.65±0.05 vs. 0.55±0.02 for natural-small (n = 23,270); 0.60±0.03 vs. 0.53±0.01 for tunneled-small (n = 66,2723); p-values in each panel are for (i) comparing means using two tailed t-tests and (ii) comparing distributions using two-sample Kolmogorov-Smirnov tests. (b) Mean within-pause instantaneous UB drifts speeds of border drifts and non-border drifts, in the four experimental conditions (colors as in (a)). Mean ± SEM and number of pauses for border and non-border scanning (target speed values, for time>50ms), respectively: 4.64±0.09 vs. 5.14±0.07 for natural-large (n = 143,310 pauses); 6.19±0.05 vs. 6.56±0.07 for tunneled-large (n = 863,457); 3.9±0.2 vs. 4.81±0.07 for natural-small (n = 19,229); 5.9±0.2 vs. 6.20±0.03 for tunneled-small (n = 48,1851); (*, p<0.05, two tailed t-tests).

(JPG)

Click here for additional data file.

Stability of instantaneous drift speed convergence.

Related to Fig 3. (a) Mean within-pause instantaneous drift speeds presented for large (left) and small (right) objects, depicted for 0

(JPG)

Click here for additional data file.

20 Aug 2020

PONE-D-20-13933

Closed loop motor-sensory dynamics in human vision

PLOS ONE

Dear Dr. Ahissar,

Thank you for submitting your manuscript to PLOS ONE. After careful consideration, we feel that it has merit but does not fully meet PLOS ONE’s publication criteria as it currently stands. Therefore, we invite you to submit a revised version of the manuscript that addresses the points raised during the review process.

Both reviewers raised substantive but addressable concerns about the methods, data reliability, clarity of exposition, and some of the interpretations -- although I think we all agree it is a clever and technically impressive contribution to an interesting topic.

Some additional specific comments I have:

(1) I agree with Reviewer 2 that some of the reasoning could be spelled out better. It could be that the reviewers (and editor) lack some of the technical understanding and familiarity with the terminology used by the authors. However, if this is the case then certainly many readers will have similar confusion. For instance, it is stated that “saccades and drifts fully characterize the movements of the eyes during all kinds of visual activities including pursuing moving targets” — but smooth pursuit itself is generally described as having both open and closed-loop components, and other smooth eye movements such as optokinetic response surely depend on the incoming sensory input and thus would not be entirely open-loop.

Relatedly, in Discussion: “…our results demonstrate clearly that ocular drifts are actively controlled by the visual system. What our results add is that this control is part of a motor-sensory closed-loop process” — I guess I’m wondering, how could it be otherwise? It’s not made clear to the general systems neuroscience reader how the visual system could actively control ocular drifts without it being a motor-sensory closed loop process.

(2) I don’t understand figure 2 and associated text (neither, apparently, did Reviewer 1). It is stated that the current data “show higher values than those typically reported in the literature (40).” But then the next sentence starts “To verify that our recordings *did not* yield higher drifts speeds than those previously reported,”  — how can both be true? In the figure, the black curve is stated to be the ‘same distribution’ as previously reported, but how are we to evaluate this claim when the previously reported distribution is not shown? Overall, the details of this comparison with previous work should be spelled out more clearly.

(3) Reviewer 1 also raises a number of concerns about statistical power, measurement noise, and the saccade thresholds, among other methods choices. Some or all of these may need to be addressed not only in the rebuttal but in the manuscript as well.

(4) The journal’s Data Availability requirements (https://journals.plos.org/plosone/s/data-availability) have not yet been fully met. The submission form states “Important: Stating ‘data available on request from the author’ is not sufficient. If your data are only available upon request, select ‘No’ for the first question and explain your exceptional situation in the text box.”

The journal strongly recommends deposition of data in an appropriate public repository. If this cannot be done, a statement of availability upon request can be made, authors must “identify the group to which requests should be submitted (e.g., a named data access committee or named ethics committee).” Also, “The reasons for restrictions on public data deposition must also be specified.”

Minor:

-Fig 3a has no ordinate label.

-Fig 4a: brackets and asterisks are all all over the place.

-The embedded figures are rendered with rather poor resolution and in several places are almost impossible to read. The reason for this becomes clear when zooming into the attached .PNG files, which themselves, although readable, are not particularly high-res. If this can be remedied in subsequent revision(s) — ideally by using a vector graphics format, but at least just notching up the resolution — the authors and editor would be most grateful.

Please submit your revised manuscript by Oct 04 2020 11:59PM. If you will need more time than this to complete your revisions, please reply to this message or contact the journal office at plosone@plos.org. When you're ready to submit your revision, log on to https://www.editorialmanager.com/pone/ and select the 'Submissions Needing Revision' folder to locate your manuscript file.

Please include the following items when submitting your revised manuscript:

A rebuttal letter that responds to each point raised by the academic editor and reviewer(s). You should upload this letter as a separate file labeled 'Response to Reviewers'.

A marked-up copy of your manuscript that highlights changes made to the original version. You should upload this as a separate file labeled 'Revised Manuscript with Track Changes'.

An unmarked version of your revised paper without tracked changes. You should upload this as a separate file labeled 'Manuscript'.

If you would like to make changes to your financial disclosure, please include your updated statement in your cover letter. Guidelines for resubmitting your figure files are available below the reviewer comments at the end of this letter.

If applicable, we recommend that you deposit your laboratory protocols in protocols.io to enhance the reproducibility of your results. Protocols.io assigns your protocol its own identifier (DOI) so that it can be cited independently in the future. For instructions see: http://journals.plos.org/plosone/s/submission-guidelines#loc-laboratory-protocols

We look forward to receiving your revised manuscript.

Kind regards,

Christopher R. Fetsch

Academic Editor

PLOS ONE

Journal Requirements:

When submitting your revision, we need you to address these additional requirements.

1. Please ensure that your manuscript meets PLOS ONE's style requirements, including those for file naming. The PLOS ONE style templates can be found at

https://journals.plos.org/plosone/s/file?id=wjVg/PLOSOne_formatting_sample_main_body.pdf and

https://journals.plos.org/plosone/s/file?id=ba62/PLOSOne_formatting_sample_title_authors_affiliations.pdf

2. We note that you have indicated that data from this study are available upon request. PLOS only allows data to be available upon request if there are legal or ethical restrictions on sharing data publicly. For more information on unacceptable data access restrictions, please see http://journals.plos.org/plosone/s/data-availability#loc-unacceptable-data-access-restrictions.

In your revised cover letter, please address the following prompts:

a) If there are ethical or legal restrictions on sharing a de-identified data set, please explain them in detail (e.g., data contain potentially sensitive information, data are owned by a third-party organization, etc.) and who has imposed them (e.g., an ethics committee). Please also provide contact information for a data access committee, ethics committee, or other institutional body to which data requests may be sent.

b) If there are no restrictions, please upload the minimal anonymized data set necessary to replicate your study findings as either Supporting Information files or to a stable, public repository and provide us with the relevant URLs, DOIs, or accession numbers. For a list of acceptable repositories, please see http://journals.plos.org/plosone/s/data-availability#loc-recommended-repositories.

We will update your Data Availability statement on your behalf to reflect the information you provide.

[Note: HTML markup is below. Please do not edit.]

Reviewers' comments:

Reviewer's Responses to Questions

Comments to the Author

1. Is the manuscript technically sound, and do the data support the conclusions?

The manuscript must describe a technically sound piece of scientific research with data that supports the conclusions. Experiments must have been conducted rigorously, with appropriate controls, replication, and sample sizes. The conclusions must be drawn appropriately based on the data presented.

Reviewer #1: Yes

Reviewer #2: Partly

**********

2. Has the statistical analysis been performed appropriately and rigorously?

Reviewer #1: Yes

Reviewer #2: Yes

**********

3. Have the authors made all data underlying the findings in their manuscript fully available?

The PLOS Data policy requires authors to make all data underlying the findings described in their manuscript fully available without restriction, with rare exception (please refer to the Data Availability Statement in the manuscript PDF file). The data should be provided as part of the manuscript or its supporting information, or deposited to a public repository. For example, in addition to summary statistics, the data points behind means, medians and variance measures should be available. If there are restrictions on publicly sharing data—e.g. participant privacy or use of data from a third party—those must be specified.

Reviewer #1: No

Reviewer #2: Yes

**********

4. Is the manuscript presented in an intelligible fashion and written in standard English?

PLOS ONE does not copyedit accepted manuscripts, so the language in submitted articles must be clear, correct, and unambiguous. Any typographical or grammatical errors should be corrected at revision, so please note any specific errors here.

Reviewer #1: Yes

Reviewer #2: Yes

**********

5. Review Comments to the Author

Please use the space provided to explain your answers to the questions above. You may also include additional comments for the author, including concerns about dual publication, research ethics, or publication ethics. (Please upload your review as an attachment if it exceeds 20,000 characters)

Reviewer #1: The authors demonstrate that drift operates—in part—in a closed-loop manner, dependent on concurrent visual input. I think the work is interesting and meaningful, and I am largely convinced of the conclusions. However, the number of participants and trials is quite low, and the saccade thresholds may have been adjusted in response to the low amount of data, in a way that might include some saccadic movements in the drift dynamics.

The data are available upon request, rather than being made fully available in a public repository.

Only five subjects were included in the study, and there were only ~10 trials per session. This study might be underpowered as a result, particularly for determining whether effects are consistent across people. How were the number of subjects and number of trials determined? Was a power analysis conducted? It is unclear how statements like “most of the individual subjects” (line 357) should be interpreted, when even one subject showing a different pattern would be 20% of the sample, and when a sixth or seventh participant might display markedly different behavior.

Why were the threshold parameters used here so different from those used in previous work? An explanation is not provided. Were these thresholds used because there was insufficient data (as a result of the low number of trials and participants)? The threshold values used seem like they would consider more samples to be drift, and fewer to be saccades. The text describes manual verification of detected saccades, but does not indicate that the drift data was examined for potential saccades that were not detected. The higher values in drift speed may reflect the inadvertent inclusion of saccadic movements. If so, some of the evidence demonstrating that drift is a closed-loop process may be the result of saccadic movements and not drift.

These simple shape stimuli differ substantially from photorealistic images. It seems entirely possible that drift changes in response to increased information (natural vs tunneled viewing) only when the information is very sparse to begin with, and drift may not change under more realistic conditions. A more ecologically-valid task might reveal very different results. The possibility that the results in the current study may be stimulus-dependent is not discussed.

The possibility of open-loop changes to drift (e.g., as a result of fatigue) are disregarded in the discussion, although the current study does not rule out open-loop components (especially in longer tasks or natural non-laboratory viewing tasks).

Lines 125-135: The distributions in Fig 2 do not appear identical, and no statistical test is shown to demonstrate that these distributions match, but they are described as having the same distribution.

Minor comments:

Line 88-89: It is unclear whether the provided values refer to the image sizes for large and small shapes or to the window sizes in those conditions. The image and window sizes are described as having a similar ratio, but it is not clear what that ratio is.

Line 377: How much were subjects paid for their participation?

Line 383: The screen’s refresh rate is not identical to the EyeLink’s sampling rate. It therefore seems peculiar to describe gaze contingencies as “real time”. How long were the delays between gaze samples and screen updates in this setup?

Fig 3B: The border-following movements do not appear to have converged to an asymptotic value in all conditions.

Lines 318-319: Why does it necessarily follow that low-level loops must involve components that are already known to control saccades and smooth pursuit movements? Could these be separate, unshared components?

Reviewer #2: Gruber and Ahissar investigated if the inter-saccadic ocular drifts functioned in an open-loop manner or a closed-loop manner. Using real-time gaze-contingent display, the authors manipulated the spatial extent of incoming visual information and tested if the drift kinematics depended on these concurrent visual inputs.

The reviewer has a few major comments on this study.

1. One of the main findings of this paper is the increased inter-saccadic drift speed observed in tunnel vision conditions. It is hard for the reviewer to understand how this result can be the supporting evidence for the closed-loop process in the drift. Because the saccade works in a closed-loop system, the feedback information obtained here might be used for increasing the gain of the drift. If the ocular drift itself is in a closed-loop system, it should change its kinematics depending on moment-to-moment changes of visual information during the drift. The reviewer cannot find this result from the paper (ex. Drift direction changes as a function of post-saccadic visual input. I do understand that this is technically challenging, though). It would be very helpful for the reviewer (and potentially for readers) if the authors provide a more compelling explanation for this.

2. The second point is related to the first point. The authors used a gaze-contingent display for the tunnel vision condition. Gaze-contingent visual stimulus manipulation is always a tricky business. In the paper, authors keep mentioning that the measurement noise would not be a problem because all the critical tests are from relative ones. However, the drift speed in tunnel vision condition can be severely affected by this measurement noise because the calibration error or other gaze-related error will have a substantial effect on the visual stimulus in this condition. This measurement noise in the gaze could introduce additional noise in visual stimulus, and this noise could contribute, somehow, to the overall gain increase in the drift. At least, authors should explain their eye calibration procedure in more detail (p19, 409-412), and discussed the potential problems that could be induced by the effect of measurement noise on the visual stimulus in tunnel vision condition.

3. Task difficulty problem. Even if the paper’s primary interest is on the mechanical property of the visual-motor system, other potential components should be considered and thoroughly discussed. One distinct part would be task difficulty. As the performance of the participant showed (p5, 90-92), the task seems to be very difficult in tunnel-vision small stimulus conditions. The reviewer thinks this component might have an influence on saccade and drift kinematics (for example, Supplementary figure 2a, saccade amplitude). Any difference due to this factor is more likely to be related to global changes in the system, rather than the evidence for closed-loop process in the ocular drift.

4. Relationships among Rs, Sp, and Xp. The authors suggested that the controlled relationship between these three variables would be the result of the closed-loop process. It was difficult for the reviewer to understand why. For example, why the visual system increases Rs and maintains Xp while compromising the control of Sp??? The increase of Rs and maintenance of Xp make some sense. Still, it was difficult for the reviewer to digest the rationale of CV increase in Sp (for example, Sp increase in tunnel vision condition could be simply because of measurement noise, please see the second comment). Maybe the reviewer is missing something. Please provide a more precise explanation (pages 11 – 12, lines 220 – 240).

Minor comments:

P 15, lines 320-321. The sentence seems to need revision.

Please put the y-axis label in Figure 3a.

**********

6. PLOS authors have the option to publish the peer review history of their article (what does this mean?). If published, this will include your full peer review and any attached files.

If you choose “no”, your identity will remain anonymous but your review may still be made public.

Do you want your identity to be public for this peer review? For information about this choice, including consent withdrawal, please see our Privacy Policy.

Reviewer #1: No

Reviewer #2: No

[NOTE: If reviewer comments were submitted as an attachment file, they will be attached to this email and accessible via the submission site. Please log into your account, locate the manuscript record, and check for the action link "View Attachments". If this link does not appear, there are no attachment files.]

While revising your submission, please upload your figure files to the Preflight Analysis and Conversion Engine (PACE) digital diagnostic tool, https://pacev2.apexcovantage.com/. PACE helps ensure that figures meet PLOS requirements. To use PACE, you must first register as a user. Registration is free. Then, login and navigate to the UPLOAD tab, where you will find detailed instructions on how to use the tool. If you encounter any issues or have any questions when using PACE, please email PLOS at figures@plos.org. Please note that Supporting Information files do not need this step.

5 Sep 2020

PONE-D-20-13933

Closed loop motor-sensory dynamics in human vision

PLOS ONE

Point by Point Reply

Dear Editor and Reviewers. Thank you for your constructive and thoughtful review. Here we provide a point-by-point reply to your comments, with detailed replies including pointers to the revisions we made in the manuscript (using the NEW line numbers). Our replies begin with “Reply:” and are colored in blue. The references cited in this Reply letter appear all together at the end of the letter.

-------------- Editor -----------

Both reviewers raised substantive but addressable concerns about the methods, data reliability, clarity of exposition, and some of the interpretations -- although I think we all agree it is a clever and technically impressive contribution to an interesting topic. Some additional specific comments I have:

(1) I agree with Reviewer 2 that some of the reasoning could be spelled out better. It could be that the reviewers (and editor) lack some of the technical understanding and familiarity with the terminology used by the authors. However, if this is the case then certainly many readers will have similar confusion. For instance, it is stated that “saccades and drifts fully characterize the movements of the eyes during all kinds of visual activities including pursuing moving targets” — but smooth pursuit itself is generally described as having both open and closed-loop components, and other smooth eye movements such as optokinetic response surely depend on the incoming sensory input and thus would not be entirely open-loop.

Reply: We indeed did not do a good job in this respect. First, this sentence in the Introduction may be incorrectly read as mixing kinematics and function. We thus corrected it and added a sentence, such that it now read (lines 41-45):

“These two kinematic components, saccades and drifts, fully characterize the kinematics of eye movements during all kinds of visual activities, whether while fixating, pursuing moving targets, reading or exploring a scene. Thus, fixation includes small saccades and drifts, pursuit includes mostly drifts with occasional saccades (when the target disappears), reading and scene viewing includes saccades and drifts.”

This is to emphasize that we distinguish between kinematics and function. The point is that the kinematics of all visual functions - fixation, pursuit, reading etc. – can be described as collections of saccades and drifts. Indeed, smooth pursuit is generally described as having both open and closed-loop components. The closed-loop component is comparable to the one we describe here for drift control. However, whereas smooth pursuit closed-loop behavior is thus far hypothesized only for the case in which a subject voluntarily pursuits a moving target, the drift closed-loop is hypothesized to function continuously, in all visual activities. To address this difference in the manuscript we have added a paragraph in the Discussion (lines 365-374).

Relatedly, in Discussion: “…our results demonstrate clearly that ocular drifts are actively controlled by the visual system. What our results add is that this control is part of a motor-sensory closed-loop process” — I guess I’m wondering, how could it be otherwise? It’s not made clear to the general systems neuroscience reader how the visual system could actively control ocular drifts without it being a motor-sensory closed loop process.

Reply: We agree – our choice of words was not accurate enough. Indeed, a control process necessitate closed-loop. What we meant is to first show that we support previous findings that the ocular drift is not a completely random process but rather that it is affected by the visual system. Then we add our finding that it is actually controlled via a closed-loop process. Thus, we now changed it to (line 272):

“…our results demonstrate clearly that ocular drifts are affected by the visual system. What our results add is that this effect is part of a motor-sensory closed-loop process: …”

(2) I don’t understand figure 2 and associated text (neither, apparently, did Reviewer 1). It is stated that the current data “show higher values than those typically reported in the literature (40).” But then the next sentence starts “To verify that our recordings *did not* yield higher drifts speeds than those previously reported,” — how can both be true? In the figure, the black curve is stated to be the ‘same distribution’ as previously reported, but how are we to evaluate this claim when the previously reported distribution is not shown? Overall, the details of this comparison with previous work should be spelled out more clearly.

Reply: Thank you, we indeed did not clarify sufficiently what is shown in Figure 2 and its significance. In order to complete the picture, we have expanded Figure 2 with 3 additional panels, and the revised manuscript now better explains the comparison we made. In many of the literature we have cited, drift speeds are usually reported after massive filtering. As we explain in the text, filtering of the raw data removes the fast ocular transitions that are known to be most effective in activating retinal cells. We thus chose to report the kinematic values computed from the unfiltered data. Since these data are contaminated with measurement noise, we term the computed speed “the upper-bound of the drift speed”.

The purpose of Figure 2 is to show that this is the only difference between our method and previously reported methods – the difference is not in the actual kinematics measured, but only in the filtering method (we include in this term also the threshold parameters). Thus, panel (a) shows that if we were to use the previously reported filtering method we would get values of drift speed that are similar to the ones reported in those previous studies. This shows that our recording device recorded signals that are similar to those recorded in those previous studies. Now, following the Reviewer’s concerns, we have added panels (b-d), to show the extent to which our method affects saccade detection. Panels b and c show that our method hardly affects saccadic rates and saccades durations. Panel d shows an example demonstrating a potential difference between the two methods as far as saccade detection is concerned – whereas clear saccades are detected by both methods, ocular speeds of up to ~8 deg/s are sometimes classified as saccades in previous methods and as drift in our method. Importantly, there is currently no independent criterion that would judge in favor of one or the other classification. In the context of the current paper we think that the important factor to be considered is how the visual system is activated. And here we are convinced that with our method we estimate retinal activation better – all available information indicates that retinal speeds of up to 8 deg/s are highly effective in activating neurons in the visual system (e.g., [1-5]).

As we assume that many of the readers would also run into difficulties with the original brief description, we have provided a detailed description in the revised MS (lines 131-155).

(3) Reviewer 1 also raises a number of concerns about statistical power, measurement noise, and the saccade thresholds, among other methods choices. Some or all of these

may need to be addressed not only in the rebuttal but in the manuscript as well.

Reply: We have addressed all the concerns raised by Reviewer 1 – please see below. Our revisions in this respect include extending the explanations and discussions regarding our methodological choices, statistical power, measurement noise and saccade threshold, and a significant expansion of Figure 2 (legend in lines 157-170).

(4) The journal’s Data Availability requirements (https://journals.plos.org/plosone/s/data-availability) have not yet been fully met. The submission form states “Important: Stating ‘data available on request from the author’ is not sufficient. If your data are only available upon request, select ‘No’ for the first question and explain your exceptional situation in the text box.”

The journal strongly recommends deposition of data in an appropriate public repository. If this cannot be done, a statement of availability upon request can be made, but authors must “identify the group to which requests should be submitted (e.g., a named data access committee or named ethics committee).” Also, “The reasons for restrictions on public data deposition must also be specified.”

Reply: The data are now available in a public GitHub repository: https://github.com/lirongruber/Closed-loop-motor-sensory-dynamics-in-human-vision.

Minor:

-Fig 3a has no ordinate label.

Reply: Fixed.

-Fig 4a: brackets and asterisks are all over the place.

Reply: Fixed.

-The embedded figures are rendered with rather poor resolution and in several places are almost impossible to read. The reason for this becomes clear when zooming into the attached .PNG files, which themselves, although readable, are not particularly high-res. If this can be remedied in subsequent revision(s), ideally by using a vector graphics format, but at least just notching up the resolution, the authors and editor would be most grateful.

Reply: We have now used the Preflight Analysis and Conversion Engine (PACE) digital diagnostic tool, which helps ensure that figures meet PLOS requirements. We hope the figures are much clearer now.

While revising your submission, please upload your figure files to the Preflight Analysis and Conversion Engine (PACE) digital diagnostic tool, https://pacev2.apexcovantage.com/. PACE helps ensure that figures meet PLOS requirements. To use PACE, you must first register as a user. Registration is free. Then, login and navigate to the UPLOAD tab, where you will find detailed instructions on how to use the tool. If you encounter any issues or have any questions when using PACE, please email PLOS at figures@plos.org. Please note that Supporting Information files do not need this step.

Reply: Done.

Reviewers' comments:

--------- Reviewer 1 --------------

Reviewer #1: The authors demonstrate that drift operates—in part—in a closed-loop manner, dependent on concurrent visual input. I think the work is interesting and meaningful, and I am largely convinced of the conclusions. However, the number of participants and trials is quite low, and the saccade thresholds may have been adjusted in response to the low amount of data, in a way that might include some saccadic movements in the drift dynamics.

The data are available upon request, rather than being made fully available in a public repository.

Reply: the data is now available in a public GitHub repository: https://github.com/lirongruber/Closed-loop-motor-sensory-dynamics-in-human-vision.

Only five subjects were included in the study, and there were only ~10 trials per session. This study might be underpowered as a result, particularly for determining whether effects are consistent across people. How were the number of subjects and number of trials determined? Was a power analysis conducted? It is unclear how statements like “most of the individual subjects” (line 357) should be interpreted, when even one subject showing a different pattern would be 20% of the sample, and when a sixth or seventh participant might display markedly different behavior.

Reply: Thank you for raising this issue. The basic comparisons in this work are not between subjects, but rather between conditions. We thus made sure that we collect enough data, from all subjects together, to conduct our comparisons. The analysis of individual subjects (Fig. 4) aims at (i) showing that our conclusions are indeed valid also to individual subjects and (ii) point to inter-subject variability, although this is not addressed statistically here. Indeed, Figure 4 shows that saccades and drift dynamics are controlled differently in different conditions for each individual, even though the specific implementation of such control probably differs among individuals (e.g., different size and even polarity of speed changes).

We agree that the phrase “most participant” is not valid here and changed it now to “four out of our five subjects” (lines 221,226).

Why were the threshold parameters used here so different from those used in previous work? An explanation is not provided. Were these thresholds used because there was insufficient data (as a result of the low number of trials and participants)? The threshold values used seem like they would consider more samples to be drift, and fewer to be saccades. The text describes manual verification of detected saccades, but does not indicate that the drift data was examined for potential saccades that were not detected. The higher values in drift speed may reflect the inadvertent inclusion of saccadic movements. If so, some of the evidence demonstrating that drift is a closed-loop process may be the result of saccadic movements and not drift.

Reply: We realize that we indeed did not explain this point clearly enough, and we thank the Reviewer for bringing it up. We have now added 3 panels to Figure 2, and added detailed explanation in the text (lines 131-155). Importantly, saccades thresholds were not adjusted in response to the amount of data. The threshold values reported in the method sections are commonly used in our and other labs (e.g., [6-8]). As we now explain in the text, filtering of the raw data removes the fast ocular transitions that are known to be most effective in activating retinal cells. We thus chose to report the kinematic values computed from the unfiltered data. Since these data are contaminated with measurement noise, we term the computed speed “the upper-bound of the drift speed”. The purpose of Figure 2 is to show that this is the only difference between our method and previously reported methods – the difference is not in the actual kinematics measured, but only in the filtering method (we include in this term also the threshold parameters). Thus, panel (a) shows that if we were to use the previously reported filtering method we would get values of drift speed that are similar to the ones reported in those previous studies. This shows that our recording device recorded signals that are similar to those recorded in those previous studies. Now, following the Reviewer’s concerns, we have added panels (b-d), to show the extent to which our method affects saccade detection. Panels b and c show that our method hardly affects saccadic rates and saccades durations. Panel d shows an example demonstrating a potential difference between the two methods as far as saccade detection is concerned – whereas clear saccades are detected by both methods, ocular speeds of up to ~8 deg/s are sometimes classified as saccades in previous methods and as drift in our method. Importantly, there is currently no independent criterion that would judge in favor of one or the other classification. In the context of the current paper we think that the important factor to be considered is how the visual system is activated. And here we are convinced that with our method we estimate retinal activation better – all available information indicates that retinal speeds of up to 8 deg/s are highly effective in activating neurons in the visual system (e.g., [1-5]).

As we assume that many of the readers would also run into difficulties with the original brief description, we have provided a detailed description in the revised MS (lines 131-155).

As for verification of detected saccades - as we now explain in more details in the Method section (lines 492-495), saccade detection was also manually verified by inspecting the algorithmic classification of each recorded trial, looking for falsely classified saccades (false-positive) as well as mis-detected saccades in the drift periods (false negative).

These simple shape stimuli differ substantially from photorealistic images. It seems entirely possible that drift changes in response to increased information (natural vs tunneled viewing) only when the information is very sparse to begin with, and drift may not change under more realistic conditions. A more ecologically-valid task might reveal very different results. The possibility that the results in the current study may be stimulus-dependent is not discussed.

Reply: We agree with this point. We have added an appropriate discussion (lines 416-419):

“It should be stressed out that these specific control strategies may be specific to our paradigm, which used simple shapes, and may differ in more realistic environments.”

The possibility of open-loop changes to drift (e.g., as a result of fatigue) are disregarded in the discussion, although the current study does not rule out open-loop components (especially in longer tasks or natural non-laboratory viewing tasks).

Reply: We say in the Introduction (lines 76-79):

“Thus, for example, closed-loop perception predicts that oculomotor dynamics will change such that retinal outputs will have temporal characteristics that are optimal for neural processing whereas open-loop perception predicts that such oculomotor changes will reflect motor adaptations such as muscle fatigue.”

According to our interpretation, any “open-loop components” are “controlled out” in a closed-loop scheme, as the loop attempts to maintain its controlled variables. Indeed, the loop may lose resources over time, which may result in slow drifts of kinematics, but we would not term those “open-loop components”. We may miss the point made by the Reviewer here and if so would be happy to learn about it.

Lines 125-135: The distributions in Fig 2 do not appear identical, and no statistical test is shown to demonstrate that these distributions match, but they are described as having the same distribution.

Reply: Again we apologize for the non-detailed explanation. Please see our answer to the Reviewer’s second comment (above), regarding the new version of Figure 2, and the related new text in the manuscript.

Minor comments:

Line 88-89: It is unclear whether the provided values refer to the image sizes for large and small shapes or to the window sizes in those conditions. The image and window sizes are described as having a similar ratio, but it is not clear what that ratio is.

Reply: Fixed (lines 90-94).

Line 377: How much were subjects paid for their participation?

Reply: Fixed (line 442).

Line 383: The screen’s refresh rate is not identical to the EyeLink’s sampling rate. It therefore seems peculiar to describe gaze contingencies as “real time”. How long were the delays between gaze samples and screen updates in this setup?

Reply: Each cycle of [sampling + updating] was bounded by 10ms (verified on-line). Thus, the maximal delay was 10ms. We use the term real time according to its common interpretation in engineering as a process whose delays do not interfere with the on-going functioning of the system. Our way to judge the effect of this delay on the perceptual process was to ask the subjects after each session:1. “Did you feel at any moment that the stimuli is following your eyes? If yes, estimate the delay”. 2. “Did you feel at any moment that your eyes are following the stimuli? If yes, estimate your delay”. All subjects reported no delay at all analyzed trials. This is now added to the Methods (lines 449-455).

Fig 3B: The border-following movements do not appear to have converged to an asymptotic value in all conditions.

Reply: When examining longer periods, we see that the speed indeed converged in all four conditions. However, when looking separately at border and non-border drifts, the dynamic patterns look different, as the Reviewer observed. We now mention this explicitly in the text (lines 190-192).

Lines 318-319: Why does it necessarily follow that low-level loops must involve components that are already known to control saccades and smooth pursuit movements? Could these be separate, unshared components?

Reply: Correct. They are likely to but do not have to. We have changed this sentence to read (line 363):

“These loops, thus, probably involve components that are already known to control saccades [9-11] and smooth pursuit [12] eye movements.”

--------- Reviewer 2 --------------

Reviewer #2: Gruber and Ahissar investigated if the inter-saccadic ocular drifts functioned in an open-loop manner or a closed-loop manner. Using real-time gaze-contingent display, the authors manipulated the spatial extent of incoming visual information and tested if the drift kinematics depended on these concurrent visual inputs.

The reviewer has a few major comments on this study.

1. One of the main findings of this paper is the increased inter-saccadic drift speed observed in tunnel vision conditions. It is hard for the reviewer to understand how this result can be the supporting evidence for the closed-loop process in the drift. Because the saccade works in a closed-loop system, the feedback information obtained here might be used for increasing the gain of the drift. If the ocular drift itself is in a closed-loop system, it should change its kinematics depending on moment-to-moment changes of visual information during the drift. The reviewer cannot find this result from the paper (ex. Drift direction changes as a function of post-saccadic visual input. I do understand that this is technically challenging, though). It would be very helpful for the reviewer (and potentially for readers) if the authors provide a more compelling explanation for this.

Reply: This comment is in place. We now provide a more detailed explanation that also includes what we mean by a closed-loop system; we hope that the Reviewer will find it compelling. The following explanation appears in the MS (lines 286-301) :

“The dynamics of closed-loop systems are limited by the loop delay, that is, the time it takes for the effect of a signal to travel along the loop. In the drift system this delay can be estimated as < 50 ms [13]. Accordingly, drift oscillations around 10 Hz [14] may reflect fluctuating loop dynamics and may point to the characteristic limit cycle of the loop. Thus, the dynamics that can be controlled in the drift loop are only the slow dynamics and not those determining the fast direction changes carrying out the drift motion (often termed “tremor” [15]) – these direction changes are most likely generated by the collection of stochastic processes in this system, including motor-neuron spikes and muscle twitches [16]. It is the slow dynamics of these fast movement events that is under control. Based on our data, we propose that the slow dynamics of the drift movements may be controlled by both the drift and saccade loops (Fig. 5a). The data supporting cross-loop control are those demonstrating inter-relationships between the saccadic rate and the drift speed and length (Results, last two paragraphs). The data supporting within-loop control of the drift are those demonstrating within-pause control of the spatial (Fig. 3a) and temporal (Fig. 3b) behavior of the drift. As we showed above (Results, last paragraph), these behaviors cannot be regarded as merely reflecting saccadic behavior. “

2. The second point is related to the first point. The authors used a gaze-contingent display for the tunnel vision condition. Gaze-contingent visual stimulus manipulation is always a tricky business. In the paper, authors keep mentioning that the measurement noise would not be a problem because all the critical tests are from relative ones. However, the drift speed in tunnel vision condition can be severely affected by this measurement noise because the calibration error or other gaze-related error will have a substantial effect on the visual stimulus in this condition. This measurement noise in the gaze could introduce additional noise in visual stimulus, and this noise could contribute, somehow, to the overall gain increase in the drift. At least, authors should explain their eye calibration procedure in more detail (p19, 409-412), and discussed the potential problems that could be induced by the effect of measurement noise on the visual stimulus in tunnel vision condition.

Reply: We agree. This is indeed a delicate point that requires attention. First, we would like to emphasize that while the visual stimulus might indeed be affected by the measurement noise, it cannot affect it back (we believe that the Reviewer is indeed aware to this fact). Thus, the experimental design does not add any external, artificial, loop to the loops of the visual system. What is true, and we believe that this is what the Reviewer aims at, is that among the signals that potentially affect the loops of the visual system we should also count the noise introduced into the visual stimulus, and that this noise may differ between conditions. We fully agree with this and thus added the following piece to the Discussion (lines 323-329):

“Another noise source that might be included in the visual loop in our experimental design is the noise introduced into the visual stimulus by our gaze-contingent protocol, due to measurement errors and delays. Importantly, this noise may be generated only in the tunneled viewing conditions, increasing the difficulty of the task in these conditions. Yet, similar to head motion, this noise is included in the retinal motion and thus affecting the behavior of the visual loops (see Controlled Variables below). “

3. Task difficulty problem. Even if the paper’s primary interest is on the mechanical property of the visual-motor system, other potential components should be considered and thoroughly discussed. One distinct part would be task difficulty. As the performance of the participant showed (p5, 90-92), the task seems to be very difficult in tunnel-vision small stimulus conditions. The reviewer thinks this component might have an influence on saccade and drift kinematics (for example, Supplementary figure 2a, saccade amplitude). Any difference due to this factor is more likely to be related to global changes in the system, rather than the evidence for closed-loop process in the ocular drift.

Reply: Right. In fact, this is one of the points that we have tried to convey in our discussion of controlled variables, though apparently not in the best way. In our original discussion the effect of task difficulty was hidden within the discussion of what the visual system attempts to maintain when constrained; however, our description of constraints did not explicitly include task difficulty. We have now corrected this, and the relevant Discussion piece is (lines 384-387):

“We assume that part of these modulations result from global changes in the system induced by task difficulty; specifically, such changes might induce the increased rate of ROI switching observed here. “

4. Relationships among Rs, Sp, and Xp. The authors suggested that the controlled relationship between these three variables would be the result of the closed-loop process. It was difficult for the reviewer to understand why. For example, why the visual system increases Rs and maintains Xp while compromising the control of Sp??? The increase of Rs and maintenance of Xp make some sense. Still, it was difficult for the reviewer to digest the rationale of CV increase in Sp (for example, Sp increase in tunnel vision condition could be simply because of measurement noise, please see the second comment). Maybe the reviewer is missing something. Please provide a more precise explanation (pages 11 – 12, lines 220 – 240).

Reply: Our rational here was based on the assumption that a given loop cannot control two competing variables simultaneously. When the rate of saccades increases there is less time between saccades and the drift loop can either maintain the same speed and then shorten the length or maintain the same length and then increase the speed. We assume that the Reviewer was not concerned by this rational but rather asks: ok, but even if the loop had to increase motion speed why did it have to give up its control? We agree that this is a valid question, and in fact we asked ourselves the same question before realizing that, as mentioned above, the loop cannot control two competing variables simultaneously. Imagine that a noise enters the system (for example, the noise justifiably described by the Reviewer) – if the loop attempts to maintain Xp constant than the fluctuations of the noisy input will necessarily induce matched fluctuations in Sp and vice versa. We now added this explanation to the Discussion (lines 388-397):

“The rational described above is based on the assumption that the loop cannot control two competing variables simultaneously [17, 18]. Imagine that a slowly fluctuating noise is added to the ocular motion due to muscle fatigue. If the loop attempts to maintain one variable (say Xp) constant then the fluctuations of the noisy signal will necessarily induce matched fluctuations in the other related signal (e.g., Sp, if saccadic rate is enforced) and vice versa. The compromised variable functions as a “shock absorber” for the controlled one.”

Minor comments:

P 15, lines 320-321. The sentence seems to need revision.

Reply: Fixed

Please put the y-axis label in Figure 3a.

Reply: Fixed

________________________________________

[NOTE: If reviewer comments were submitted as an attachment file, they will be attached to this email and accessible via the submission site. Please log into your account, locate the manuscript record, and check for the action link "View Attachments". If this link does not appear, there are no attachment files.]

==========================

PBP References

1. Hubel D. Cortical unit responses to visual stimuli in nonanesthetized cats. Am J Ophthalmol. 1958;46:110-21.

2. Skottun BC, Grosof DH, De Valois RL. Responses of simple and complex cells to random dot patterns: a quantitative comparison. J Neurophysiol. 1988;59(6):1719-35. PubMed PMID: 3404201.

3. Pack CC, Born RT, Livingstone MS. Two-dimensional substructure of stereo and motion interactions in macaque visual cortex. Neuron. 2003;37(3):525-35. PubMed PMID: 12575958.

4. Grunewald A, Skoumbourdis EK. The integration of multiple stimulus features by V1 neurons. J Neurosci. 2004;24:9185-94.

5. Manookin MB, Patterson SS, Linehan CM. Neural mechanisms mediating motion sensitivity in parasol ganglion cells of the primate retina. Neuron. 2018;97(6):1327-40. e4.

6. Bonneh YS, Donner TH, Sagi D, Fried M, Cooperman A, Heeger DJ, et al. Motion-induced blindness and microsaccades: cause and effect. Journal of vision. 2010;10(14):22.

7. Fried M, Tsitsiashvili E, Bonneh YS, Sterkin A, Wygnanski-Jaffe T, Epstein T, et al. ADHD subjects fail to suppress eye blinks and microsaccades while anticipating visual stimuli but recover with medication. Vision research. 2014.

8. Ahissar E, Arieli A, Fried M, Bonneh Y. On the possible roles of microsaccades and drifts in visual perception. Vision research. 2014;118:25-30.

9. Robinson D. The use of control systems analysis in the neurophysiology of eye movements. Annual review of neuroscience. 1981;4(1):463-503.

10. Fuchs A, Kaneko C, Scudder C. Brainstem control of saccadic eye movements. Annual review of neuroscience. 1985;8(1):307-37.

11. Coe BC, Munoz DP. Mechanisms of saccade suppression revealed in the anti-saccade task. Philosophical Transactions of the Royal Society B: Biological Sciences. 2017;372(1718):20160192.

12. Behling S, Lisberger SG. Different mechanisms for modulation of the initiation and steady-state of smooth pursuit eye movements. Journal of Neurophysiology. 2020;123(3):1265-76.

13. Hafed ZM, Goffart L. Gaze direction as equilibrium: more evidence from spatial and temporal aspects of small-saccade triggering in the rhesus macaque monkey. Journal of Neurophysiology. 2020;123(1):308-22.

14. Herrmann CJ, Metzler R, Engbert R. A self-avoiding walk with neural delays as a model of fixational eye movements. Scientific Reports. 2017;7(1):12958.

15. Rucci M, Ahissar E, Burr D. Temporal coding of visual space. Trends in cognitive sciences. 2018;22(10):883-95.

16. Carpenter RHS. Movements of the Eyes. 2 ed. London: Pion; 1988.

17. Marken RS. You say you had a revolution: Methodological foundations of closed-loop psychology. Review of General Psychology. 2009;13(2):137.

18. Marken RS. Controlled variables: Psychology as the center fielder views it. The American Journal of Psychology. 2001;114(2):259.

Submitted filename: PBP letter.docx

Click here for additional data file.

1 Oct 2020

Closed loop motor-sensory dynamics in human vision

PONE-D-20-13933R1

Dear Dr. Ahissar,

We’re pleased to inform you that your manuscript has been judged scientifically suitable for publication and will be formally accepted for publication once it meets all outstanding technical requirements.

Within one week, you’ll receive an e-mail detailing the required amendments. When these have been addressed, you’ll receive a formal acceptance letter and your manuscript will be scheduled for publication.

An invoice for payment will follow shortly after the formal acceptance. To ensure an efficient process, please log into Editorial Manager at http://www.editorialmanager.com/pone/, click the 'Update My Information' link at the top of the page, and double check that your user information is up-to-date. If you have any billing related questions, please contact our Author Billing department directly at authorbilling@plos.org.

If your institution or institutions have a press office, please notify them about your upcoming paper to help maximize its impact. If they’ll be preparing press materials, please inform our press team as soon as possible -- no later than 48 hours after receiving the formal acceptance. Your manuscript will remain under strict press embargo until 2 pm Eastern Time on the date of publication. For more information, please contact onepress@plos.org.

Kind regards,

Christopher R. Fetsch

Academic Editor

PLOS ONE

Additional Editor Comments (optional):

Reviewers' comments:

Reviewer's Responses to Questions

Comments to the Author

1. If the authors have adequately addressed your comments raised in a previous round of review and you feel that this manuscript is now acceptable for publication, you may indicate that here to bypass the “Comments to the Author” section, enter your conflict of interest statement in the “Confidential to Editor” section, and submit your "Accept" recommendation.

Reviewer #2: All comments have been addressed

**********

2. Is the manuscript technically sound, and do the data support the conclusions?

The manuscript must describe a technically sound piece of scientific research with data that supports the conclusions. Experiments must have been conducted rigorously, with appropriate controls, replication, and sample sizes. The conclusions must be drawn appropriately based on the data presented.

Reviewer #2: Yes

**********

3. Has the statistical analysis been performed appropriately and rigorously?

Reviewer #2: Yes

**********

4. Have the authors made all data underlying the findings in their manuscript fully available?

The PLOS Data policy requires authors to make all data underlying the findings described in their manuscript fully available without restriction, with rare exception (please refer to the Data Availability Statement in the manuscript PDF file). The data should be provided as part of the manuscript or its supporting information, or deposited to a public repository. For example, in addition to summary statistics, the data points behind means, medians and variance measures should be available. If there are restrictions on publicly sharing data—e.g. participant privacy or use of data from a third party—those must be specified.

Reviewer #2: Yes

**********

5. Is the manuscript presented in an intelligible fashion and written in standard English?

PLOS ONE does not copyedit accepted manuscripts, so the language in submitted articles must be clear, correct, and unambiguous. Any typographical or grammatical errors should be corrected at revision, so please note any specific errors here.

Reviewer #2: Yes

**********

6. Review Comments to the Author

Please use the space provided to explain your answers to the questions above. You may also include additional comments for the author, including concerns about dual publication, research ethics, or publication ethics. (Please upload your review as an attachment if it exceeds 20,000 characters)

Reviewer #2: Thank you for addressing all the questions that the reviewer had raised. The new additions and modification of the manuscript indeed helped the reviewer to understand and appreciate the importance of the manuscript better.

**********

7. PLOS authors have the option to publish the peer review history of their article (what does this mean?). If published, this will include your full peer review and any attached files.

If you choose “no”, your identity will remain anonymous but your review may still be made public.

Do you want your identity to be public for this peer review? For information about this choice, including consent withdrawal, please see our Privacy Policy.

Reviewer #2: No

6 Oct 2020

PONE-D-20-13933R1

Closed loop motor-sensory dynamics in human vision

Dear Dr. Ahissar:

I'm pleased to inform you that your manuscript has been deemed suitable for publication in PLOS ONE. Congratulations! Your manuscript is now with our production department.

If your institution or institutions have a press office, please let them know about your upcoming paper now to help maximize its impact. If they'll be preparing press materials, please inform our press team within the next 48 hours. Your manuscript will remain under strict press embargo until 2 pm Eastern Time on the date of publication. For more information please contact onepress@plos.org.

If we can help with anything else, please email us at plosone@plos.org.

Thank you for submitting your work to PLOS ONE and supporting open access.

Kind regards,

PLOS ONE Editorial Office Staff

on behalf of

Dr. Christopher R. Fetsch

Academic Editor

PLOS ONE