The authors present an interesting manuscript that conducts surface-based morphometric analyses of dogs and other mammals, finding that domestic dogs have a similar, but more variable relationship between brain size and gyrification index. I have several comments that, if addressed, would improve the manuscript.

Comments:

1) I would strongly encourage the authors to consider their peer reviewers when formatting manuscripts for peer review. Many if not most reviewers in 2020 now review PDFs electronically and so placing text, figures, and figure legends in separate locations results in an enormous amount of scrolling (which is also worsened by double spacing). It is much easier to review manuscripts that are single spaced with figures and legends at the appropriate location in the main text.
2) Lines 142-143 multiple typos.
3) Please list the in-plane resolution for convenience of all scans.
4) Please list basic parameters like the spatial resolution for all scans, even if previously published. This is necessary to assess the voxel size relative to the cortical thickness of this study. It would probably be a good idea to have a data table listing important parameters for each data source.
5) What resolution was everything “standardized” to? Such steps cannot create information—only destroy it. Also, some resampling algorithms (e.g. trilinear) will induce blurring above and beyond that which would occur with a lower resolution acquisition.
6) The panels (e.g. Figure 1) are so small I cannot evaluate the quality of the surface reconstructions. Given the wide variety of scans used here, it would be helpful to have examples for each unique MR acquisition.
7) A variety of methods are used to compute GI in this study. Are there any significant differences in the measure across methods within species?
8) The approach to parcellations seems very rudimentary. Why not at least use the gyral and sulcal landmarks from the Datta et al 2012 PLOS paper? Having created surface meshes for all datasets, it should be reasonably simple to use a folding-based surface registration to that template as is commonly done in humans.
9) Again the figures are super small and even when zoomed in text is difficult to read.
10) I didn’t follow how the numbers in table 2 were means of cortical thickness that were higher than the regional values for cortical thickness in table 3.
11) I wondered if human selection of dogs for size leads to the differences in brain size (scaling with body weight), but does not exert as much of an effect on GI, which might be more closely related to the specific brain functions of dogs? Perhaps it would be worth controlling for body weight in the domesticated dog sample and seeing if brain size becomes more predictive of GI?

The authors have largely addressed my concerns. I only note the following mostly minor issues (except #1, which I have more serious concerns about):

1) I don’t understand the difference in image quality between Figure 1 and Figure 2a. The authors deleted a statement about image resampling, but it does appear that Figure 2a is downsampled relative to Figure 1. The authors need to be more transparent about what happens to the images between Figure 1 and Figure 2. Are the surfaces and segmentations generated at the acquisition data resolutions or some other resolution? Are the images resampled and if so, what algorithm was used for resampling. These issues can have important impacts on segmentation and surface quality, which will affect the cortical thickness and GI measures.
2) I would talk about 3D isotropic resolutions for those scans with equal resolution in all directions and slice thickness and in plane resolutions for those scans with unequal slice thickness and in plane resolution, rather than “scan resolution.”
3) Line 189: The Siemens Trio is a 3T scanner.
4) Most of the data are of very high quality; however, one dataset has a slice thickness of 1.5mm, which is on the order of cortical thickness. Because the cerebral cortex has convolutions in all 3 dimensions, thick slices will likely cause biases in the surface reconstruction, which will affect cortical thickness and GI. In general, one wants at least two voxels spanning the thinnest cortex in a brain to get accurate white and pial surfaces. It seems that only a couple of specimens are affected by this issue; however, and that there are many other higher quality datasets for that species. This could be briefly noted as a limitation of these datasets in the discussion.