Review Reports - An Unexpected Spontaneous Motion-In-Depth Pulfrich Phenomenon in Amblyopia

Round 1

Reviewer 1 Report

This is a very nice, clear paper. It measures interocular delay in mild amblyopes, concluding that either the amblyopic or the fellow eye may be delayed. They suggest that this variability reflects the competing effects of low contrast in the AE (tending to delay its information) vs blurring in the AE (tending to accelerate its information relative to the fellow eye). I have only minor comments, and congratulate the authors on a thoughtful paper.

First, Table 1 appears to be missing.

I had some questions about the methods (probably answered in the previous paper, but (i) I am writing this review without internet, so can’t check, and (ii) it would be nice to add the few additional details that make this paper stand alone.

1. Was it only the carrier phase that varied, ie did the Gabor envelope have zero disparity?

2. “the sinusoidal angular speed was 18deg/s” I’m not sure how to interpret this. As an angular velocity (degrees of phase) this seems very low, with 20s for a full revolution. As a speed of motion across the screen, this varies for a cylinder. Is 18deg/s the maximum speed, at the middle of the cylinder? When they later vary the speed, I assume the same definition is being used.

3. How were the Gabors placed on the cylinder? It looks as if their density is higher at the edges, reflecting “foreshortening” there, but their width (sigma) is constant everywhere; is that right?

l. 121 “We can see here that for most amblyopic subjects, this psychometric function is clearly offset from 0°.”

Mmm, “clearly” for “most” may be overstating it. Can you reword?

In all the figures, could I ask you to add on reference lines marking IOPD=0 and Probability=0.5 , to help guide the reader’s eye?

A couple of times you use the phrase “one amblyopic subject who reported a clockwise perception as a function of the interocular phase difference”. This doesn’t make sense, since the whole point is that the perception CW/CCW varied with IOPD. I assume you mean something like “one amblyopic subject who reported a clockwise perception at zero interocular phase difference”. Can you reword?

l. 176 “One can see that the psychometric function shifts to the left and the PSE gets closer to 0° when the number of Gabor patches is increased.” I must say this is very hard to see. Is the shift significant under bootstrap resampling of the data? I’m assuming not, and that this is why stats are given at the population level only. I would suggest rewording to acknowledge the weak nature of this effect.

Effect of number of Gabors/ size/ speed. In each case, the threshold (slope) of the psychometric function decreases. This is much more obvious than the shift in PSE. Could the authors comment on this and give p-values?

“the rPSE could be converted in interocular delay” : in -> into

A couple of times the authors write “It is known that X”. I invite them to consider whether they could delete the words “It is known that” without loss.

Author Response

Reviewer #1

First, Table 1 appears to be missing.

-> We apologize for this oversight, it is now included in the manuscript and as a separate file

1. Was it only the carrier phase that varied, ie did the Gabor envelope have zero disparity?

-> The gabor patches were displaced, now clarified line 100

-> We added (degrees of phase) to make it clear line 107

-> Gabor were randomly placed in each trial (now clarified line 104). The foreshortening comes from the sinusoidal trajectory at the “edges” of the cylinder

l. 121 “We can see here that for most amblyopic subjects, this psychometric function is clearly offset from 0°.”

Mmm, “clearly” for “most” may be overstating it. Can you reword?

-> Reworded, now line 157: “We can see here that for some amblyopic subjects, this psychometric function is offset from 0° “

In all the figures, could I ask you to add on reference lines marking IOPD=0 and Probability=0.5 , to help guide the reader’s eye?

-> Ok, we added those lines in the figures

-> Reworded lines 218, 231 “ the psychometric functions of the perceived direction of rotation as a function of the interocular phase difference are displayed for one amblyopic subject”

-> Reworded “ In this figure, it seems” l221

-> Regarding your, and other reviewers’ comments, we added a third panel representing the slope in the figures. We now provide all regression parameters in Supplementary Table 1. Slopes data are presented lines 231, 256 and 283

“the rPSE could be converted in interocular delay” : in -> into

-> ok

A couple of times the authors write “It is known that X”. I invite them to consider whether they could delete the words “It is known that” without loss.

-> Deleted lines 47, 420

p { margin-bottom: 0.25cm; line-height: 115%; }

Reviewer 2 Report

Review of Reynaud et al – Vision

This study aims to measure the interocular delay found in some individuals with amblyopia that have sufficient stereovision to experience spontaneous Pulfrich effects. They use a psychophysical paradigm that presents Gabor patches rotating in depth following a cylindrical path and modulate the interocular phase of the Gabors to estimate the perceived subjective equality of motion direction (ambiguous motion). Overall, this is a nice paper and an interesting replication of effects. I do have some concerns and questions about the manuscript, and would strongly recommend the authors consider including a computational model that could offer additional, an important, information on the temporal integration of neurons (or filters) described in the manuscript.

L62 – I can’t seem to find table 1 in the text or the supplementary materials and from what I have read, it contains important information to make sense of the data presented here.

L73 – Is 60Hz the monocular refresh rate (i.e., the refresh rate of the monitor is 120Hz?).

L75 – Were the Gabor patches presented at maximum contrast in both eyes?

Figure 2 – The psychometric function fits are fine for most observers, but some (particularly A3, A4, A5) do not asymptote at 0 or 1 ), which results is less than desirable fits. The fits could definitely be improved by adding a lapse rate parameter to the psychometric function fits shown in figure 2. This would also give you a much better estimate of the slope of these psychometric functions. It would also be helpful to report goodness of fit measures of your functions, either in a table or in text. I can’t seem to see the data points of participant A1.

Figure 2 – Are error bars represented in Figure 2a. You mention that within participant variability is small, but is it so small that no error bars would be seen in Figure 2a? I find that representing variability graphically is useful to the reader and Figure 2 would definitely benefit from this.

L125 – I find Figure 2a hard to compare to the data reported in text. I imagine that Table 1 (which is missing) indicates the amblyopic eye of observers, but at the moment, Figure 2 shows mostly negative inter-ocular phase differences without indicating which observers were switched to calculate the rectified Point of Subjective Equality.

L131 – I quite enjoy having the measurement of the IOD in ms as it speaks directly to the measurements being made. That said, as these are estimates calculated from the interocular phase difference, it would be helpful to also report the IOD values as interocular phase differences (degrees) to simplify the comparison of data reported in text to the figures. In fact, I recommend including both measures for all figures and data reported in text.

L161 - Observer 8 is evidently an outlier in this sample. Their psychometric functions and performance on all the tasks completed here differ greatly from all other observers. While I understand there is value to representing all observers, I’m unsure it is best to keep this participant in statistical analyses and interpretation of effects. I see that the authors are cautious, but still define the correlation as statistically significant (but not meaningful) when it most likely is not. I would recommend only reporting the statistical analyses with observer 8 removed, but still show their data to represent variability.

L162 – Does the journal allow for subheadings? The transition between L161 and the new paragraph at L162 is a little rough and could be improved by including a subheading in the results section.

L180. The slopes do appear to converge somewhat, but these effects are quite small. For most observers, it appears that the jump from 100 to 200 Gabors brings the rPSE closer to 0, but the increase from 200 to 400 Gabors has little effect. You report a sign test for your slopes, but you do not report whether the slope of the regression line was statistically significantly different from 0. It would also be beneficial to report goodness of fit values for the regression analysis.

Figure 5a. It appears to me that the largest effects in changing the number of Gabors presented to observers is a change in the slope of the Psychometric function and not a leftward or rightward shift on the x axis (as compared to the psychometric functions measured with ND filters). I understand the psychometric functions shown in Figure 5a are for a single observer, but I’m curious to know if this is similar for other observers as well. If so, do the authors believe it may be worth discussing when interpreting their results?

L200 – It would be helpful if the authors reported whether or not the slope of the regression line was statistically significantly different than 0. The sign test is useful in determining whether or not slopes converge for observers with positive and negative rPSEs, but it does not tell the reader whether the sign of the slope is statistically significantly different from 0.

Figure 7. The results of the speed experiment are interesting. Mainly in regards to some of the main conclusions in terms of the sign of the rPSE. You report in the main experiment that many observers have positive rPSE values, which indicates a slower non-amblyopic eye. However, according to the results of this figure, the sign of the rPSE appears to be dependent on speed. At slow speeds (4.5), all but one observer have negative rPSE values. This increases as the speed increases, and we start to see more observers with positive rPSE values once speed reaches 18 degrees per second. I’m curious to know how the authors interpret these findings, especially because the interpretation of which eye is slower comes back in the discussion.

L230 – I don’t believe it is necessary to state that the methods of a previous study were crude. I think it goes without saying that any initial observation of a phenomena will always be improved upon in future studies. In addition, there are more sensitive approaches to measure processing delay than the psychophysical technique measured here (i.e., EEG / MEG), which the authors have previously used. I would recommend removing the statement about crude methods and changing the beginning of the following sentence to: We developed a sensitive psychophysical method …

L296 – Yes, the number of elements does decrease, but how does the area/density of the stimulus change? The Gabor stimuli should increase in size as their spatial frequency decreases and this will influence any summation mechanisms that are contributing to the absolute reduction in rPSE. I think it is also important to mention that the large effects reported for Gabor spatial frequency come from a single observer (A5). All other observers show largest rPSE < .1. From the effects reported in text, I’m not certain these experiments are measuring different aspects of stimulus dependence in the Pulfrich effect and may instead be measuring how spatial / area summation aids the amblyopic observers in the task.

L304 – The authors of this manuscript have, in a previous manuscript, built a computational model that captured the temporal delays of the amblyopic eye measured with MEG. I’m curious to know why they do not attempt to build a model here that can demonstrate the influence of a slow temporal integration VS a delay in transmission. I am unsure of the necessity of the model here, but I am curious to know whether the authors have considered how these effects could be model and whether would they consider one useful to differentiate the influence of temporal integration and transmission delays.

Author Response

Reviewer #2

-> Thank you for your comments. We didn’t have time to develop a model in the time allowed for the revision. However we will definitely try to do so for our upcoming studies.

L62 – I can’t seem to find table 1 in the text or the supplementary materials and from what I have read, it contains important information to make sense of the data presented here.

-> We apologize for this oversight, it is now included in the manuscript and as a separate file

L73 – Is 60Hz the monocular refresh rate (i.e., the refresh rate of the monitor is 120Hz?).

-> It is a polarized line-interleaved display, so the 2 images to the 2 eyes are presented at the same time but on different scanlines. now clarified line89

L75 – Were the Gabor patches presented at maximum contrast in both eyes?

-> They were presented at 80% contrast, now indicated line 105

-> Fitting is done by nonlinear least-squares regression (Matlab’s nlinfit function, now explicited line 125), not maximimum likelihoood. Hence, adding a lapse rate would present only a small benefit on the estimate of the slopes (that is not our main focus) compared to the uncertainty it would add in the estimate of PSE We now provide all regression and goodness of fits parameters in Supplementary Table 1. We corrected the Figure1 to display subject A1 data.

-> Errorbars are not represented because they are so small they would be barely visible. They are however represented in figure4

-> Indeed these information are in Table1. We now indicate in the methods which are the 2 subjects for which the rPSE was switched line 135

-> We added rPSE scales on the righthand of graphs in figures 4 and 8. We added PSE values in the comparisons line 170

-> Observer 8 is the strongest amblyope we screened who was able to perform the task. Actually their data follow the trend of other subjects but with larger values. So we decided to keep him in the general statistical analysis and we report stats with and without him included when he is the only one driving the effect line 199, 332

L162 – Does the journal allow for subheadings? The transition between L161 and the new paragraph at L162 is a little rough and could be improved by including a subheading in the results section.

-> We now divide the manuscript in 2 parts “Characterizing the spontaneous Pulfrich phenomenon” and “Influence of the stimulus parameters”

-> We now report the coefficient of determination and MSEof the psychometric functions fits and the slope, the coefficient of determination with p-values and the p-values of the slopes for the linear regressions of the rPSE and slope in all conditions in Supplementary Table1. We say that only two slopes are significant line 228

-> Regarding your, and other reviewers’ comments, we added a third panel representing the slope in the figures. We present these results line 231

-> We now provide all regression parameters in Supplementary Table 1, including the p-values of the slopes for the linear regressions of the rPSE and slope. Indeed few are actually significant. Clarified line 255

-> We suggest that the blur-based acceleration might be stronger at high speed:

“The observation that all but one subjects had a negative rPSE at low speed (4.5°/s), but some of them switched their polarity at high speeds (Figure7) could also find an explanation with this blur-based acceleration suggestion. Indeed, an object moving fast might appear more blurred than an object moving slow, thus generating a faster acceleration. It could also explain the decrease of the slope of the psychometric function at high speeds, which in that case could indicate a sort of position uncertainty.” line 365

-> ok, we removed this statement

-> We now say only 3 of these slopes are significant line 255 and provide stats in Supplementary Table1

Indeed, despite the decreasing number of elements, the area covered by the Gabor patches increased with their size. We now raise this point in the discussion line 383

-> We didn’t have time to develop a model in the time allowed for the revision (10days). However we will definitely try to do so for our upcoming studies.

p { margin-bottom: 0.25cm; line-height: 115%; }

Reviewer 3 Report

I enjoyed reading this manuscript. The paper is succinctly and clearly written. The discussion does a phenomenal job of considering and integrating theoretical interpretations for the result. The result described in this paper is an important one for the field.

My primary recommendations for revision are (1) to clarify the introductory matter; and (2) clarify a few points in the methods/analysis that will help convey the details of the experiment and results. I consider these changes to be minor. A third recommendation is to present data from control subjects for the density/size experiments.

Table 1 is either missing from the manuscript or hasn’t been uploaded to the review system available for me to see, so I was unable to fully assess. The text was clear enough that I can guess what is in there, but please make sure it is included in revision.

Points of clarification in the introduction:

(35-37) The paper cited by Pete Thompson [14] does not measure spontaneous Pulfrich effect as far as I understand. "In particular" suggests the information following this sentence, about spontaneous Pulfrich in amblyopia, is an example of what is cited in [14]. I don’t think that is the case, nor do you measure an effect in stereoblind observers which was what I was expecting based on the introduction. Can you clarify this in the introduction?

This is important because I didn’t quite grasp at first why the specific amblyopia participants were selected as described in the methods (59-62). Can you briefly provide an explanation in the text why you decided to use non-stereoblind amblyopes, and what you considered "sufficient" stereo vision - i.e., explicitly why stereoblind participants could not do this task?

(39) “ increased retinocortical transmission time” – is this proposed to be a delay in reaching V1, specifically?

(45-48) The longer delay in what? That paper is about interocular saccadic latency differences, can you explicitly tie the relevance of their findings to the magnitude of a spontaneous Pulfrich effect so the reader can follow this line of reasoning all the way through? Also, why do they doubt this? Can you expand for so the argument you are assessing is clear? Can you be clear about why their findings in strab amblyopia should hold for the aniso amblyopia participants assessed in the current paper?

Methods/analysis:

(59) Did you screen aniso + strab with only aniso being able to do this task, or did you screen aniso only?

(109) Please include a formula or equation for the conversion

(121) “Is clearly offset from 0” - It is to me clear the amblyopia group is more variable, but it is not clear to me that they are all offset from 0 as the offsets are small. Can you report summary descriptive here? Or maybe they are in Table 1?

(123) “All of these delays are significant” - What was this test? The delay is significantly different from 0? The PSE (or the |PSE|) is significantly different from 0? Is this 8 tests, one for each participant? Please report in full.

(128, 132, 182, other spots including appendix) Please report the calculated statistic, not just the p-values or the alpha used to control Type 1 error.

Regarding the density/size experiments: If you can present data from 8 controls, I would be much more convinced that something interesting is happening in amblyopia. While I assume there is no effect for controls and I do think the effect in amblyopia is interesting, it would be better to see what sort of idiosyncratic responses controls have on this task.

(Figure 5A) I know you're interested in the point estimate of the PSE as an index of interocular delay, but it also seems clear that the slope of the psychometric function in Figure 5a steepens with more elements. It looks like number of gabors is changing the slope a lot more than it is changing the PSE. Does this happen for all subjects? Does this tell us something interesting?

(Figure 5B and accompanying text) Only two participants start with a negative rPSE at 100 gabors, so I'm not sure you have enough data to make a strong claim here. Can you report the statistics and slope values for the individuals plotted in Figure 5b? I’m not convinced that both starting with a negative value go significantly towards 0 as # gabors increases.

Small details:

(68) For completeness can you include the Psychtoolbox extension information?

(74) For completeness can you include the brand or model of the passive polarized 30 glasses?

(108) I think NAE is being used here before being defined

(152) In observers with typical vision, correct?

(results) This is a stylistic choice that is really up to the authors, but I doubt 4 significant figures is meaningful for the R-squared and p-values. I would advise using 2.

Author Response

Reviewer #3

-> We apologize for this oversight, it is now included in the manuscript and as a separate file

Points of clarification in the introduction:

-> We replaced “in particular” by “Furthermore” line 41

-> Indeed, the word “sufficient” was misleading. It just happened that none of the stereoblind amblyopes we screened was able to perform the task. We rephrased the methods to clarify them line 66

(39) “ increased retinocortical transmission time” – is this proposed to be a delay in reaching V1, specifically?

-> Indeed, this statement was too strong, we added “or because of prolonged neural integration at retinal or cortical sites”

-> Indeed they report a delay correlated with interocular contrast sensitivity difference in all amblyopic groups, plus an extra delay for strabismic amblyopes. We now restated this sentence as “ Indeed, McKee et al., [25] observed a delay correlated with interocular contrast sensitivity difference in amblyopia” line 52

Methods/analysis:

(59) Did you screen aniso + strab with only aniso being able to do this task, or did you screen aniso only?

-> We screened and tested both kind of amblyopes (data was in the missing table1), we added this information in parenthesis too line 66

(109) Please include a formula or equation for the conversion

-> ok

-> Indeed the IOD values are in Table1. We reworded this sentence according to reviewer #1 ‘s comment too

-> We now indicate that for each participant, the PSE is different from 0

(128, 132, 182, other spots including appendix) Please report the calculated statistic, not just the p-values or the alpha used to control Type 1 error.

-> ok

-> We didn’t have time to run a control experiment in the 10 days period allowed for the revision. We assume these parameters would have a small impact on controls as their PSE are initially very close to 0° however it would be nice to test it as you suggest. We will likely test these effects in a future study.

-> Regarding your, and other reviewers’ comments, we added a third panel representing the slope in the figures. We present these results line231

-> We now provide all regression parameters in Supplementary Table 1. Indeed few are actually significant, we now say it line 228

Small details:

(68) For completeness can you include the Psychtoolbox extension information?

-> ok

(74) For completeness can you include the brand or model of the passive polarized 30 glasses?

-> Ok, the brand is Viewsonic, we don’t know the model

(108) I think NAE is being used here before being defined

-> ok

(152) In observers with typical vision, correct?

-> Yes, now clarified line 192

(results) This is a stylistic choice that is really up to the authors, but I doubt 4 significant figures is meaningful for the R-squared and p-values. I would advise using 2.

-> ok

p { margin-bottom: 0.25cm; line-height: 115%; }

Round 2

Reviewer 3 Report

Thank you for your quick turnaround! I think the revisions are great. Here are my final thoughts:

(abstract) The abstract mentions "mild anisometropic amblyopes" in reference to a previous result (I think this is why I originally thought you were studying aniso only), but now that I see Table 1 I can tell your patients are broader than that, which is more compelling. Can you add information to the abstract indicating this? "We observed a spontaneous Pulfrich phenomenon in anisometropic, strabismic, and mixed amblyopia" (or something like this)

(327-328) Readers scanning for this analysis will expect it in the results section; please move this from discussion to results.

Author Response

Thank you for your constructive comments and suggestions.

We made the modification you suggested in the abstract and added the regression information in the figure caption, as we think it is better to keep the figure in the discussion section. We also added an appendix with the experiment on control subjects you previously suggested (Appendix3, referenced line 380). We are sorry this control experiment took us so much time. All new modifications appear in magenta.

Round 3

Reviewer 3 Report

The additional control data goes a long way to strengthen the claims made in this paper. Looks great!