No voodoo, just can- and can't-do of brain scans

The brain scanner - has it provided a window into the mind as so many claimed it would? Or is it little more than an expensive toy?

Brain-imaging labs attract generous grants and their output is published in high-impact journals. But in 2009 a paper surfaced prior to publication (tinyurl.com/34mp83u; see News, February 2009) in which Ed Vul and his colleagues claimed that numerous neuroimaging studies in social neuroscience had deployed iffy statistical methods, leading them to identify 'voodoo correlations' between psychological states and regions of brain activity. The charge caused a storm of controversy and the paper, when it finally came out in print, had been renamed with a less provocative, voodoo-free title. Now the dust has settled, the same journal, Perspectives in Psychological Science, has published a special issue on what brain imaging can and can't tell us.

In the lead article, Gregory Miller of the University of Illinois, himself a user of brain-imaging techniques, is highly critical of 'naive' reductionists who claim that psychological phenomena are somehow based in brain processes or located in particular neural structures. 'Functions do not have a location,' he writes, adding later, with reference to a particular study on voters: 'Trust decisions and political attitudes do not occur in the brain. Decisions, feelings, perceptions, delusions, memories do not have a spatial location. We image brain events... We do not image, and cannot localise in space, psychological constructs.'

Miller reminds readers that brain-imaging studies are only able to provide correlations between brain activity and psychological processes and that how each affects the other, if at all, remains largely mysterious. He also reminds us that although many researchers act as though the neural is somehow more fundamental than the psychological, there are many examples of the causal direction apparently flowing the other way - for instance, psychotherapy has been shown to trigger various brain changes.

In her contribution, Diane Beck of the University of Illinois discusses the allure and popularity of brain-imaging data among the media and general public. Part of the story is likely to be the appeal of colourful blobs-on-brain images and the fact that neural correlates are somehow seen as lending biological credibility to behavioural results. Ultimately Beck says that responsibility lies with individual researchers not to allow the media to make unsubstantiated claims based on mere neural correlates. 'We should be careful not to encourage portrayals of our research as explaining a behaviour or condition when it does not,' she writes. 'In short, every time we allow the press to mischaracterise our results or overstate our conclusions, we run the risk of damaging the reputation of our entire field.'

Other contributors defended the value of neuroimaging research to psychological inquiry. Brian Gonsalves and Neal Cohen at the University of Illinois gave the example of brain-imaging memory studies, which consistently reveal activity in the posterior parietal cortex (PPC), an area usually associated with attentional processes. This finding has prompted searches for the specific memory-based factors associated with PPC activity, with retrieval success and the perception of memory 'oldness' emerging as relevant. 'These ideas have led to a great deal more theorising about the interactions of memory and attention,' Gonsalves and Cohen write, 'tested not only with neuroimaging methods but in neuropsychological studies as well.'

For neuroimaging to become more productive, Russell Poldrack at the University of Texas argues that psychologists need to develop a comprehensive and agreed 'cognitive ontology' - that is, 'the component operations that comprise mental function'. To this end, Poldrack and his colleagues have established a collaborative, online tool, the Cognitive Atlas (www.cognitiveatlas.org). Once established, Poldrack believes the Cognitive Atlas will help neuroimagers identify examples of 'selective associations', which is when activity in a given neural structure is associated 'with only one putative cognitive process', thus allowing the inference that 'the reality of this process has been established'. If researchers can predict, based on neuroimaging data, whether a particular mental process was engaged, then, Poldrack says, that would be evidence for a selective association. Poldrack concludes that using 'detailed ontologies along with large-scale data-mining approaches, it may finally be possible to determine the joints at which the brain carves the mind.'

--Christian Jarrett