PCOMPBIOL-D-24-01213

How distinct sources of nuisance variability in natural images and scenes limit human stereopsis

PLOS Computational Biology

Dear Dr. White,

Thank you for submitting your manuscript to PLOS Computational Biology. After careful consideration, we feel that it has merit but does not fully meet PLOS Computational Biology'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.

​Please submit your revised manuscript within 60 days Feb 17 2025 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 ploscompbiol@plos.org. When you're ready to submit your revision, log on to https://www.editorialmanager.com/pcompbiol/ 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 editor and reviewer(s). You should upload this letter as a separate file labeled 'Response to Reviewers'. This file does not need to include responses to formatting updates and technical items listed in the 'Journal Requirements' section below.

* 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, competing interests statement, or data availability statement, please make these updates within the submission form at the time of resubmission. Guidelines for resubmitting your figure files are available below the reviewer comments at the end of this letter

We look forward to receiving your revised manuscript.

Kind regards,

Christoph Strauch

Academic Editor

PLOS Computational Biology

Marieke van Vugt

Section Editor

PLOS Computational Biology

Additional Editor Comments :

First of all, I want to apologize for the relatively long time it took until we found reviewers - as we can see from the reviews, this isn't due to the quality of the manuscript.

The reviewers and we agree that this is an elegant study that has the potential to advance the field. Overall, the requested methodological details, relatively simple analyses, and further discussion seem doable in a revision. The reviewers further suggested a couple of minor edits/visualizations that the authors may want to follow. Furthermore, scripts needed to reproduce the authors' findings should be made available.

Journal Requirements:

1) Please ensure that the CRediT author contributions listed for every co-author are completed accurately and in full.

At this stage, the following Authors/Authors require contributions: David White, and Johannes Burge. Please ensure that the full contributions of each author are acknowledged in the "Add/Edit/Remove Authors" section of our submission form.

The list of CRediT author contributions may be found here: https://journals.plos.org/ploscompbiol/s/authorship#loc-author-contributions

2) We ask that a manuscript source file is provided at Revision. Please upload your manuscript file as a .doc, .docx, .rtf or .tex. If you are providing a .tex file, please upload it under the item type u2018LaTeX Source Fileu2019 and leave your .pdf version as the item type u2018Manuscriptu2019.

3) Please provide an Author Summary. This should appear in your manuscript between the Abstract (if applicable) and the Introduction, and should be 150-200 words long. The aim should be to make your findings accessible to a wide audience that includes both scientists and non-scientists. Sample summaries can be found on our website under Submission Guidelines:

https://journals.plos.org/ploscompbiol/s/submission-guidelines#loc-parts-of-a-submission

4) Please upload all main figures as separate Figure files in .tif or .eps format. For more information about how to convert and format your figure files please see our guidelines: 

https://journals.plos.org/ploscompbiol/s/figures

5) Some material included in your submission may be copyrighted. According to PLOSu2019s copyright policy, authors who use figures or other material (e.g., graphics, clipart, maps) from another author or copyright holder must demonstrate or obtain permission to publish this material under the Creative Commons Attribution 4.0 International (CC BY 4.0) License used by PLOS journals. Please closely review the details of PLOSu2019s copyright requirements here: PLOS Licenses and Copyright. If you need to request permissions from a copyright holder, you may use PLOS's Copyright Content Permission form.

Please respond directly to this email and provide any known details concerning your material's license terms and permissions required for reuse, even if you have not yet obtained copyright permissions or are unsure of your material's copyright compatibility. Once you have responded and addressed all other outstanding technical requirements, you may resubmit your manuscript within Editorial Manager. 

Potential Copyright Issues:

i) Please confirm (a) that you are the photographer of 1C, or (b) provide written permission from the photographer to publish the photo(s) under our CC BY 4.0 license.

Reviewers' comments:

Reviewer's Responses to Questions

Reviewer #1: Review of "How distinct sources of nuisance variability in natural images and scenes limit human stereopsis"

Reviewer: Michael Landy

This is a really interesting and well-done piece of work that comes to the conclusion that a surprising amount of variability underlying disparity judgments (of patches of natural images) is driven by stimulus variability rather than internal noise, and that the stimulus variability comes from two effectively independent sources (disparity pattern variability and luminance pattern variability). The data are convincing, the methods are perhaps partially drawn from previous work, but pushed to a new level in terms of the conclusions one can draw. All around impressive. But, the Methods section was quite a slog and could be made a bit easier to parse.

Comments, bigger and trivial (page/para/line):

6/3: I would hope the authors can share more than just data (e.g., analysis scripts, etc.).

7 et seq.: I'm still a little confused about the stimuli. The surround is meant to be at zero disparity with the image behind (and the "center" of the image at an uncrossed disparity that is one of the experimental parameters that is varied). The raised-cosine is kind of like a blurry or frosted window through which the scene is viewed. What is the disparity of the window's edge? I assumed it was at zero disparity, in which case more should be seen of the right eye's view at the left edge of the stimulus and vice versa. But, some of the text seems to indicate that a circular patch is cut from both eyes' views centered on the central corresponding point. Which is it, and please clarify that in the manuscript. If it's the latter, then I'm not sure how to think about the scene (mid-gray in front, but what happens at the edges to vignette the scene that's behind it?). I was worried that, e.g., if the standing disparity were 1 deg (and the former version of vignetting was used), there would be NO correspondence. Obviously, that's not what you meant. So, clarify the situation early, please. One confusion is in the definition of v_0, the vergence required to fixate the two center pixels. Maybe I missed it while reading, but I still wasn't sure when I got to this line that the center pixels corresponded.

Eq. 2: Does this pool both images? Using the same scaling?

8/2: Here is where you say the patches are centered on corresponding points, at last.

8/3/5: "central region": You should give its diameter here.

8/3/7: This would be better rephrased by giving the purpose first: "We are interested in the effect of disparity contrast on performance, so we chose patches whose disparity contrast fell either into a "low" range (details) or a "high" range (details), and we report performance below for each group of images separately."

8/4/3: I'm not sure it's cool to cite something as "n.d." with no source or means for a reader to find the citation.

10/2: I'd like a figure panel that shows the stimulus in situ with the background and "reticle" (whatever that is!). What's an "ordinal" direction? (New use of that word for me...)

11/5/2-3: 5x5x2 does not equal 10. Later, comparison disparity is NOT treated as a condition (but rather as an independent variable).

Eq. 4: d' is usual separation/spread, so why is it over variance here??? S/N is usually in energy terms, so separation^2/variance. This equation is neither.

13/1/5: Section 4.6 -> Eq. 6

Eq. 11: This tripped me up, because I stupidly looked at the equation and thought "r^2 is the proportion of variance accounted for, not r!" I was wrong, because this is not linear regression but the correlation between two variables with shared and exclusive variance, and that comes out differently. But, for readers who might be similarly tripped up by leaping to a false analogy, you might warn them. Maybe it's just me, though ;^/

Page 16, et seq.: This feel a bit like teaching rather than developing the math needed here. You are going to do a case that is symmetric and with a criterion at zero, but you develop the math for a general case with two different means (for the two passes) and nonzero criteria. Why bother? Once you assume zero and, in particular, equal criteria, and equal means for the two passes, the picture is symmetric (so that +- and -+ come out the same). It would be nice to save some verbiage by going with what is actually happening here from the get go. This will simplify Eq. 17, for example (using symmetry).

21/1/3: "depthnzz"

22/2/3: inconsistent WITH the expansion

22/2/11-12: twice "subscripts [to] denote" (delete "to")

24/2/1: with-observer -> within-observer

25/2/4: asses -> assess ;^)

25/4/3: Here's where you list 10 conditions the right way

Figure 2A: The Pass 2 timeline is missing the stimulus in the 2nd interval.

32/1/3: parallel to ONE another

32/2/1: nobody -> no one

33/3/6: You should probably show the data (as a function of pooling radius)

Eqs. 48 and 49: What's with all that space in the parentheses in the denominators?

41/3/4: These findings should also probably be shown in a figure.

Fig. 2B: The x-axis label is cut off

What does it mean that the dominant variability stimulus driven? It would be nice to speculate about this. You do ask whether they pool disparities around the center but for a larger area and reject that idea. But what if they use the largest far disparity? The largest near disparity? A randomly chosen small area on the stimulus? Some other strategy that gets knocked around by stimulus details and thus shows variability across stimuli? In other words, what does it mean that different stimuli produce different performance? Maybe the internal noise variance on a given trial depends, somehow, on the stimulus? You assume additive noise that is fixed, so what if it's Weberian based on something about the stimulus or otherwise multiplicative?

Reviewer #2: This manuscript introduces a method for partitioning the variability in a psychophysical task into stimulus-driven components (e.g., disparity and luminance contrast) and internal noise. The primary result, which is consistent with image-computable ideal observer analyses in past work by the same lab on different tasks, is that luminance variability drives perceptual variability substantially. Here, they find that disparity and luminance both contribute to perceptual performance and their contribution is correlated across observers. I think this type of variance decomposition is central to the scientific endeavor when we have “noisy” measurements or “noisy” behavior and the “Law of Total Variance” is a severely underappreciated conceptual framework in perceptual science. The challenge here is partitioning the variance with binary observations governed by truncated cumulative normal distributions. As far as I can tell, the authors got it right and the results are elegant. I have only a few comments.

My only major comment is that I wonder how different the results would be if you just put together a big logistic regression to predict every choice from the different measurable stimulus components and plot the Beta coefficients. It seems like the main advantage of portioning the variability as the authors did is that it naturally communicates the results in terms of effect size (i.e., what fraction of variance is due to what). I was surprised then that this quantity is reported in terms of “threshold contribution” with no units. My understanding is that these are capturing the variances? Why not represent in the “fraction of total variance”?

Optional clarity suggestions: it would be quite useful to see all the analytic solutions in one place (e.g., a table). I think the field would find this paper more useful if they could easily map the variance partitioning multi-pass experiment framework into their variables of interest. Right now, the presentation is thorough, but there’s a lot of jumping back and forth between equations to follow the logic.

Typos:

Equation 4: shouldn’t it be the standard deviation instead of the variance? For the formula in equation 3 to be correct, the denominator of d’ should be the common standard deviation. I believe this is probably a typo.

Section 4.9.3 : Depthnzz

Discussion “see Resutls”

Reviewer #3: This study applies an elegant logic to the task of separating out different components of variability that affect participants in a stereoacuity task using stereo photographs of natural stimuli. My main comments concern the clarity with which the authors describe the psychophysical task and stimuli. In an ideal world, they should be able to present a figure where the reader can see one of the main effects for themselves.

Major

1. Stimulus and task. It would help the reader to be clear what the stimulus was. p10 the Methods say 'In one interval... standard...In the other interval, ....comparison stimulus..' The assumption is that the same scene is presented in both intervals, with only the pedestal disparity changing. It would be helpful to make this explicit. The image in Fig 2A is too small to be helpful.

Also, for Fig 2A, delta is given as -7.50 arcmin. It might be helpful to change delta_{cmp} to something different to help the reader understand the stimulus.

Even more helpful would be for the reader to see two examples of a trial, so that they could see whether they get the main effect themselves. For example, the authors could repeat Fig 1C with a different pedestal disparity, then repeat both rows but now with a low disparity range. This should be considerably easier (Fig 4A). The icing on the cake would be a cross section of the stimulus in each case (x-axis showing pixel number across the image, y-axis showing disparity), times four (standard and comparison for high and low disparity contrast) so that readers instantly understands the stimulus and task. At the moment, a lot of reading of text is required to piece together the stimulus and task.

2. Monocular cues. If the same scene is used for the standard and comparison stimulus then, in theory, the task could be done monocularly (unless the authors sampled the scene with a slightly different lateral position of the scene relative to the camera to avoid this, but I did not find this in the text). This does not affect the authors' analysis (it would just contribute to lowering sigma_L, the luminance-driven variability), but it would be helpful if the authors could comment on this in the text.

3. Disparity range. I found it hard to pick what the authors' model was for the effect of disparity contrast (although they isolate the effect very clearly). For example, Harris and Parker (1995) discussed the efficiency of integrating disparity information across a number of samples in order to compare the mean disparity of dots spanning a range of disparities. Do the authors envisage a similar integration process? The authors point out that the participant could use the central pixel and perform in the same way as if the disparity contrast was zero, in which case no integration would be necessary.

4. A different experiment. It might be helpful for the reader if the authors mentioned in the Discussion what they thought would happen if the experiment had used an entirely different stereo photograph for the comparison stimulus. This would make the stimulus more similar to other experiments using dynamic random dot patterns where the luminance component is changed but the disparity profile remains. It would still be possible to do double pass experiments, provided the stimulus and comparison were repeated exactly on the second pass. As I understand it, the results of that experiment are not necessarily predictable from the sources of variability that the authors have gathered in the current experiment. Discussing this case would help to clarify how abstract a representation of disparity the authors believe participants retain between the standard and comparison (e.g., if just a single estimate of the pedestal value is retained, then changing between different stereo photographs might not matter).

Minor

p8. 'high' 'low' disparities, wrong way round.

p10. 'that proportion comparison...' Difficult sentence that does not effectively explain the basis on which the stimuli were chosen.

Fig 3. Explain the dashed line in the legend.

Fig 11. 'affect of' -> effect of

**********

Have the authors made all data and (if applicable) computational code underlying the findings in their manuscript fully available?

The PLOS Data policy requires authors to make all data and code 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 and code 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 or code —e.g. participant privacy or use of data from a third party—those must be specified.

Reviewer #1: No: 

Reviewer #2: No: Data are available but code is not yet.

Reviewer #3: None

**********

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

Reviewer #3: Yes: Andrew Glennerster

[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.]

Figure resubmission:

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. If there are other versions of figure files still present in your submission file inventory at resubmission, please replace them with the PACE-processed versions.

Reproducibility:

To enhance the reproducibility of your results, we recommend that authors of applicable studies deposit laboratory protocols in protocols.io, where a protocol can be assigned its own identifier (DOI) such that it can be cited independently in the future. Additionally, PLOS ONE offers an option to publish peer-reviewed clinical study protocols. Read more information on sharing protocols at https://plos.org/protocols?utm_medium=editorial-email&utm_source=authorletters&utm_campaign=protocols

Dear Dr White,

We are pleased to inform you that your manuscript 'How distinct sources of nuisance variability in natural images and scenes limit human stereopsis' has been provisionally accepted for publication in PLOS Computational Biology.

Before your manuscript can be formally accepted you will need to complete some formatting changes, which you will receive in a follow up email. A member of our team will be in touch with a set of requests.

Please note that your manuscript will not be scheduled for publication until you have made the required changes, so a swift response is appreciated.

IMPORTANT: The editorial review process is now complete. PLOS will only permit corrections to spelling, formatting or significant scientific errors from this point onwards. Requests for major changes, or any which affect the scientific understanding of your work, will cause delays to the publication date of your manuscript.

Should you, your institution's press office or the journal office choose to press release your paper, you will automatically be opted out of early publication. We ask that you notify us now if you or your institution is planning to press release the article. All press must be co-ordinated with PLOS.

Thank you again for supporting Open Access publishing; we are looking forward to publishing your work in PLOS Computational Biology. 

Best regards,

Christoph Strauch

Academic Editor

PLOS Computational Biology

Marieke van Vugt

Section Editor

PLOS Computational Biology

***********************************************************

The reviewers have suggested only very minor improvements and noted a few small inconsistencies and language errors. All of these can be addressed during the proof stage, provided no further changes are made. Congratulations on this excellent paper!

Reviewer's Responses to Questions

Comments to the Authors:

Please note here if the review is uploaded as an attachment.

Reviewer #1: This is an excellent revision. My comments are only typos/grammar SNAFUs (page/para/line):

6/5/5: chosen by random -> chosen randomly

8/5/7: consistent WITH the predictions

15/2/8: by the product OF the standard deviations

17/6/2-3: "Setting the decision variable..." is not a sentence!

18/1/1: "Hence" doesn't make sense. This is a conclusion from evidence that isn't provided in the previous sentences

Throughout: My standard pet peeve that bucks the tide: "data" is a plural word ;^)

19/1/14: Delete "*B-C*"

29/5/8: "Fig. Fig13B"

Some of the figures (e.g., Fig. 10) have been converted to lo-res pixels. Try to keep things as vector art later

Reviewer #3: Thank you to the authors for their very helpful replies and changes.

The new Fig 2 makes it much easier to understand the stimulus and task.

Fig S1 is also very helpful. As I understand it, the mean disparity of the stimulus is another potential cue (in addition to ‘largest near disparity’ or ‘largest far disparity’) that is not perfectly correlated with the nominal disparity of the stimulus? The disparity of the standard or comparison is defined by the disparity of the central region for that stimulus and that is not necessarily the same as the mean disparity of the stimulus. At least, that is how I understand the stimulus – if that is not the case, then more needs to be added to the section on ‘stimulus vetting’ to explain how this was done. Other studies have assumed that participants attempt to estimate the mean disparity when faced with a task like this. I do not think it would be fair to ask the authors for any more analysis at this stage, but a sentence somewhere, eg the legend for Fig S1B, would be helpful. It should mention that the mean disparity of the stimulus, like the largest disparity, is not perfectly correlated with the nominal disparity used in the figures.

It is now much clearer that the images used for the standard and comparison stimuli were different. In the legend for Fig 2, it says “Each natural image patch was unique across all trials and intervals.” This is not strictly true. It might help the reader to qualify this, e.g: “Over the course of a single pass, a unique natural image patch was used for every trial and interval.”

**********

Have the authors made all data and (if applicable) computational code underlying the findings in their manuscript fully available?

The PLOS Data policy requires authors to make all data and code 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 and code 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 or code —e.g. participant privacy or use of data from a third party—those must be specified.

Reviewer #1: Yes

Reviewer #3: Yes

**********

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: Yes: Michael S Landy

Reviewer #3: Yes: Andrew Glennerster