Commentary on Devlin and Poldrack

Functional brain imaging is fundamentally an anatomical technique. This means that if a submitted manuscript locates an activation in the wrong place, the paper should be rejected out of hand. This may sound harsh, but the situation would be analogous to someone claiming that they had recorded from single units in the SMAwhen in fact they were recording from premotor cortex. So I strongly endorse the paper by Devlin and Poldrack (2007) who argue that we must all agree on a valid method for the localization of activations. But a cognitive psychologist might say that what matters is that there is a difference between two conditions, and that it is of less interest exactly where the difference in activity lies. That would be fine if the psychologist was simply constructing box diagrams or connectionist models. But the whole point of imaging is that it does tell you where the difference lies. The cognitive psychologist might argue that what is important is the fact that patient DF is poor at identifying objects but is able to reach for and grasp them normally. But what imaging adds is that she has a lesion in LOC and that parietal area AIP is active when she grasps (James et al., 2003). Why is that a valuable addition? The answer lies in understanding the anatomical basis of localization of function. Passingham et al. (2002) have shown that each cortical area has a unique pattern of inputs and outputs (a ‘connectional fingerprint’), and they argue that this is a major determinant of the function of that area. The reason is that the inputs provide the information on which the area operates and the outputs determine the areas that it can influence. We have a very detailed knowledge of the corticocortical connections in the macaque monkey brain (Young, 1993), and the connections have been summarized on the Web site CoCoMac (Stephan et al., 2001). The reason why it is totally misleading if a paper claims that the activation is in area A, whereas in fact it is in area B, is that areas A and B have a different pattern of inputs and outputs. But can we assume that the connections in the macaque brain also hold for the human brain? For the first time, diffusion tensor imaging allows us to find out, at least for the major axonal bundles. For example, Croxson et al. (2005) compared the parietal and temporal lobe projections to the prefrontal cortex in the macaque and human brain. The over-riding impression is that they are very