Reply to Dakin et al.

Steven Dakin and colleagues present a detailed analysis of the stimuli used in our study [1] revealing that our method of interpolation is associated with certain changes in the statistical properties of interpolated images. They argue that these changes might represent a possible source of the BOLD signal non-monotonicity that we observed in the primary visual cortex (V1). This finding is interesting, but we’d like to clarify the following point: the main finding of our study, namely the dissociation between non-monotonic V1 BOLD signal [1] and monotonic psychophysical performance [2], is in no way affected by this. In addition, our model showed that Gabor responses are sparser for natural images than for pure noise, suggesting that cortical ‘trade-off’ may actually play a role in the experimentally observed effects. However, we agree with Dakin and colleagues that statistical properties associated with interpolation may have indeed contributed to the BOLD non-monotonicity. We note with interest that phase-only kurtosis appears to show characteristics broadly consistent with V1 BOLD. If this turns out to be true in general, it would provide even stronger evidence for a dissociation between perception and V1 BOLD levels. Future experimental work is needed to address these issues, testing both the generality of the correlation between V1 BOLD levels and phase-only kurtosis, as well as BOLD monotonicity for other interpolation methods such as WMP interpolation suggested by Dakin and colleagues. Testing such hypotheses generated by quantitative techniques is likely to lead to further advances in our understanding of the BOLD signal and its relation to behavior.