What do you think?
Rate this book
256 pages, Hardcover
First published May 16, 2013
The problem with such mindless neuroscience is not neuroscience itself. The field is one of the great intellectual achievements of modern science. Its instruments are remarkable. The goal of brain imaging is enormously important and fascinating: to bridge the explanatory gap between the intangible mind and the corporeal brain. But that relationship is extremely complex and incompletely understood. Therefore, it is vulnerable to being oversold by the media, some overzealous scientists, and neuroentrepreneurs who tout facile conclusions that reach far beyond what the current evidence warrants — fits of “premature extrapolation,” as British neuroskeptic Steven Poole calls them. When it comes to brain scans, seeing may be believing, but it isn’t necessarily understanding.
...one cannot use the physical rules from the cellular level to completely predict activity at the psychological level. By way of analogy, if you wanted to understand the text on this page, you could analyze the words by submitting their contents to an inorganic chemist, who could ascertain the precise molecular composition of the ink. Yet no amount of chemical analysis could help you understand what these words mean, let alone what they mean in the context of other words on the page.
The difficulty with reverse inference is that specific brain structures rarely perform single tasks, so one-to-one mapping between a given region and a particular mental state is nearly impossible. In short, we can’t glibly reason backward from brain activations to mental functions.
To be fair, there is nothing wrong with the reverse-inference approach as long as the investigative buck doesn’t stop there. Indeed, the approach frequently offers a valuable starting point for generating fruitful hypotheses that can later be tested in systematic experiments.
Many aspects of Vul’s critique are technical, but his basic point is easy to grasp: If you search a huge set of data— in this case, tens of thousands of voxels— for associations that are statistically significant and then do more analyses on only those associations, you are almost guaranteed to find something “good.” (To avoid this mistake, the second analysis must be truly independent of the first one.) This error is known variously as the “circular analysis problem,” the “nonindependence problem,” or, more colloquially, “double-dipping.”
As Insel observed in a sobering 2009 article, there is no evidence that the past two decades of advances in neuroscience have born witness to decreases in mental disorders’ prevalence or to any impact on patient life span. The failure of brain-imaging techniques to have yet made major inroads into the causes and treatment of mental illness offers a necessary reminder for modesty in our expectations.