Congratulations on getting such rare data! Given the small sample size and the large error bars in [figure 3], it would be useful if you could report individual differences. Because of the paired design, instead of the bar graphs, it would be more informative to report pairwise differences for the different categorical contrasts. To account for individual differences in BOLD signal, an index might be more appropriate, for instance: (face-object) / (face+object). This could be reported for every contrast and every participant in a small table as a reply to this question.

Actually, your supplementary information provides all the information needed. It is interesting to consider the results from individual dogs. From that we can make three observations:

#There seems to be a difference between human and dog faces

The average response to faces is 0.05. In this average, the responses to human and dog faces were themselves averaged, based on a non-significant t-test. However, the t-test is not significant because one dog shows a strong effect in the opposite direction than the 5 other dogs. The average difference between human and dog faces in these 5 dogs is 0.0624, which is larger than the average response to faces.

#There are large individual differences in categorical differences

For the different contrasts, here is the number of dogs showing a difference in the same direction as the group effect:

• face - object: all dogs
• face - place: 4/6 dogs
• face - scrambled: 4/6 dogs

#There are large differences for other contrasts than the face - object one:

Several dogs show categorical differences for other contrasts that are as large as those observed for the face-object contrast:

• object - place: 3/6 dogs
• place - scrambled: 4/6 dogs