Referee 2's Review:

Review of first revised manuscript:

This revision addresses many of my issues of concern, but I am still not convinced that it provides convincing evidence for a functional distinction between the aFMC and RCZ regions. It is absolutely crucial that a Region x Contrast (or, specifically, ROI x Context x Agency) interaction is reported. Without this, no strong claims about functional specialization can be made. The authors now report the results of such an analysis in their cover letter, but I did not spot it in the manuscript (my apologies if I've missed this; if not, it needs to be reported).

The validity of this analysis turns on the way that ROIs were selected. The authors state that "all voxels of the relevant contrast exceeding the critical threshold in the mean z-map were determined. We then extracted the time course of the signal underlying these activated voxels for each participant from the preprocessed data" (p. 14). Is this correct? This would imply that the ROI analyses averaged over every voxel in the whole brain activated by a particular contrast. I.e., the aFMC ROI would actually represent every voxel in the brain activated by the Context x Agency interaction. This is clearly inappropriate, because the aFMC is not the only region activated by this contrast. It also seems inconsistent with the claim "we also obtained a significant cross-over interaction in the mediodorsal nucleus of the thalamus" (Figure 2 caption), implying that separate ROIs were generated for each brain region. So the authors still need to clearly state how they have defined the ROIs. For instance, does the aFMC ROI refer to every activated voxel in a contiguous cluster of aFMC voxels activated in this contrast?

There is a second issue with the ROI x Context x Agency interaction. If I understand correctly, the aFMC ROI was established by searching for voxels activated by the Context x Agency interaction whereas the RCZ ROI was established by searching for voxels activated by the main effect of Agency. In this case the ROIs are biased, and a ROI x Context x Agency interaction might be expected simply because of the way that the ROIs were defined. It would be far more convincing if a ROI x Context x Agency interaction could be established using ROIs defined from a previous study, or a contrast orthogonal to the ones in this analysis.

Minor points:

Supplementary material, page 1: "The overall accuracy was also computed for the catch trials revealing 89.3% of errors". Presumably this should be accuracy of 89.3% rather than errors.

Page 11: "In total, 100 experimental trials (without catch trials and null-events) were presented, resulting in approximately 25 trials for each combination of context and agency." What were the actual means for the number of trials in each cell of the 2 (Context) x 2 (Agency) design?

Review of the second revised manuscript:

This revision provides more convincing evidence for a functional distinction between aMFC and RCZ regions and I think it makes a useful contribution to the literature.

However, the way in which ROIs were defined could be clarified further still. E.g., p.15: "The ROI for the aMFC was generated from the neutral contrast ... and defined as a sphere with 3mm radius centered on the peak voxel of the activation". Presumably this means the peak voxel of the activation within aMFC. So this then depends on how the aMFC is defined (and likewise for the RCZ region). Was there some anatomical criterion, or was this just the visual impression of the authors? I ask because the aMFC co-ordinate of -14,36,3 (fig. S1) seems rather posterior for BA 10.

Second, I had some concerns about the Monte Carlo simulations (p. 14). It is not clear to me how these were conducted. More specifically, it would be useful to know exactly how many contiguous voxels are required to survive these corrections for multiple comparisons at each initial z-threshold, and how many voxels there actually were for each cluster in the empirical data.

***********
N.B. These are the comments made by the referee when reviewing an earlier version of this paper. Prior to publication the manuscript has been revised in light of these comments and to address other editorial requirements.