I am in no respect an expert in neuroscience, but naturally I am aware of the main technical developments of the past few decades — especially functional MRI — which now provide neuroscientists with amazing imagery related to brain activity. I am also aware of the broad effort in the field to establish correlations between brain activity and mental states.
I will not deny that I am mildly discomfited by this effort, not because there is anything suspect about such correlations but because they are so often conjoined with a strange presumption that somehow brain scans are particularly probative windows on human behaviour, whereas in fact they are usually just fancy proxies for things we already know by other means (as has been convincingly argued). One also routinely runs into a tacit neurological reductionism according to which minds are “really just” brains, and you and I are, at bottom, “really just” fleshy computers processing stimuli. In this view of things, the notion of persons as bearers of freedom, dignity, and moral responsibility tends to become, at best, occluded.
My discomfort is only mild because I am aware that, whatever the merits of any particular scientific study, the minds-are-brains view is plagued by conceptual problems and, at least within the ambit of the reigning philosophy of nature in which matter is defined to be devoid of mental properties, is doomed to failure.
But, quite apart from the question of how we should interpret findings of correlations between mental states and brain activity, there remains the question of whether we should believe that such correlations exist in the first place. It seems that we should, but with reservations, for the evidence is not as strong or as straightforward as one might think.
For instance, a few years ago an important paper identified problems with common analysis techniques in fMRI studies. The authors showed that using such techniques they could produce nice correlations using data that were pure noise. Studies which avoided such confused methods uniformly showed comparatively low correlations. The authors speculated that a significant number of the findings claimed by the field might be illusory. I do not know what revisions resulted when (or if) the data were analyzed again.
And now, in this month’s Nature Reviews, comes another paper that criticizes the results of a wide swath of neuroscience work. The authors argue that a significant fraction of neuroscience studies suffer from low statistical power, meaning that both the sample sizes and the effects being studied are generally small. The problems with low power studies are many: the probability of missing true effects is fairly high, as is the probability of falsely “discovering” something that isn’t there. Even when a finding is true, low power studies tend to exaggerate it. Here is a popular level summary of the paper and the issues at stake.
Obviously it is up to the specialists to sort these issues out, and I have no doubt that they will. But there does seem to be warrant for wariness the next time you hear a claim that the neural correlate of this-or-that aspect of your mental life has been found. Sometimes things are just not that simple.