Overall assessment:
Although I find the premise & data interesting, I can't help but feel that this submission would need major revision to be a worthy inclusion in PRS-B. Most of the issues I take are with respect to statistics and their true interpretation. I am enough of a fan of the study overall to recommend a major revision rather than an outright rejection, as I am hopeful the authors should be able to address my concerns.

Major remarks:
- The topic is interesting and should be one that most readers of PRS-B would appreciate. I also commend the authors on the efforts put into their thorough characterization of anatid visual characteristics.
- Although I am not averse to the publication of "negative" or "null" results, I do have an aversion to seeing situations where weak results seem over-advertised as strong. In reviewing Figure 3 along with tables S3 through S7, seems that the explanatory variables do a poor job of truly explaining variance in binocular field shape. Models with the proposed explanatory variables do not out-compete null models in a meaningful way. The title of the submission feels far, far, far too definitive in its description of the patterns uncovered. And even if the patterns in the data were more compelling, calling something "the primary evolutionary driver" is almost an impossible statement to justify -- you have no way of knowing if it is THE evolutionary driver.
- One important analysis that is missing from the manuscript is an assessment of the phylogenetic signal within foraging category. It seems clear enough from Figure 2 that foraging categories are not evenly or randomly distributed across the tree -- there seem to be strong clustering patterns among monophyletic groups. This is an extremely important detail, especially when it comes to later interpreting statistical models showcased in S3 - S7. Since foraging category ("Primary forage") is used as a potential explanatory variable, an under-appreciated aspect of these PGLS models is that phylogenetic relatedness is likely embedded in foraging category in addition to being used to inform the error structure of the models. I do not take issue with this approach per se, but this potential influence absolutely needs to be addressed, as it affects how all models should be interpreted. It's of course not possible for me to say anything definitively without knowing how foraging category and phylogeny interrelate, but my sense is that the overriding explanatory factor for all analyses and all data sets in this study is actually phylogenetic relatedness, but that this is being masked. Moreover, phylogenetic relatedness might also inform the patterns in body mass or diet. So, understanding the relationship between the explanatory variables and phylogenetic relatedness itself is essential to fully understanding what is going on here.
- There are a large number of models run on this data set -- to the point of data-dredging. Accordingly, it is also hard to know if the significance of results are true or if they are more likely due to random chance. I'd recommend revising the statistical framework. Some options include: a) a more careful selection of proposed models, 2) some way to correct for multiple-hypothesis testing (akin to a Bonferroni correction, but one that would be appropriate for the current framework), or perhaps 3) a more integrated approach that assesses variance in all dependent variables simultaneously (note: the models in tables S6 and S7 are not what I mean here).
- I don't have a great sense of how well the models actually explain variance in the dependent variables. Log-likelihoods and AICs are great for model-comparative purposes, but they do not tell us anything about goodness of fit.
- I also wonder how much of all this is outlier-driven. In Figure 3, the authors acknowledge the three extreme outliers in the PCA plot, which I very much commend. But the issue then becomes that the reader is left to wonder if these outliers are dictating the results. I therefore strongly recommend the authors entertain the idea of performing sensitivity analyses. The sensiPhy package in R (Paterno et al. 2018) provides great tools for these kinds of analyses. The idea here would be to see if similar results can be achieved with iterative removals of one species at a time.

Minor comments:
- Line 55 of the abstract (sentence starting with "Although foraging..") is an incomplete sentence.
- Figure S3 does not seem to show grey shading, despite the description in the caption.
- This is just a personal opinion, and I hope it is not too rude, but I do not find the figures to be especially compelling towards supporting the assertions that are made in the text. To put it somewhat bluntly, I don't look at Figures 2 or 3 and then say "Oh yes, foraging ecology is the primary evolutionary driver of binocular vision in Anatidae."

I previously reviewed a version of this manuscript several months ago. I'd like to commend the authors for the steps they took to incorporate my comments & those from other reviewers. My main concerns had been about the overstating of results, nearly all of which rely on association of phenomena rather than direct experimentation and/or formal assessment of causality. I find that this version of the manuscript is appropriately toned down and does a much better job of presenting the ideas, results, and conclusions in a fair manner. It is more credible.

I am happy to recommend publication. This work will serve as a fine addition to PRS-B and should be of interest to its readership.There is one (literally just one) thing I would request be changed before doing so:

Line 50 (Abstract): "determined" is too strong a word, especially for a phylogenetic analysis that relies on assessing correlation. There is an implication of causality here for which you do not present evidence to assert

(No need for re-review of course. I'm sure the alternative word you come up with will suffice)

Other than that, this was a pleasure to (re-)read. Thank you for the opportunity to review this work, and best wishes to the authors and anyone else involved.