Comment on Devlin and Poldrack

The basic tenet developed by Devlin and Poldrack is to thepoint: by definition functional neuroimaging is an attempt to map‘function’ onto ‘anatomy’. Hence it is critical to cast the functionalresults, albeit indirectly measured, in precise anatomical terms.Furthermore, the authors correctly point out that in many instancesinsufficient care is devoted to this crucial ‘matching’ step betweenfunction and anatomy. The authors make an analysis of thedifferent types of anatomical descriptions presently used for theanalysis of human functional imaging data but the conclusionsreached in the paper are inherently limited by the choice of‘reference frame’ or ‘template’ brain upon which functional dataare projected. Clearly, the authors reject the use of a brain from asingle subject as the ‘template’ (e.g., the Tournoux and Talairachatlas). Moreover, the authors acknowledge that ‘Brodmann areas’are only one of the possible parcellations of human cortex, just asthe Tournoux and Talairach atlas is only a particular referencebrain. Indeed, multiple parcellation schemes are possible based oncytoarchitectonics and myeloarchitectonics.Importantly, however, this micro-architecture is in fact only oneof four criteria for identifying cortical areas. Monkey studies haveclearly shown the limitation of the Brodmann areas and theimportance of defining cortical areas by multiple criteria. Forexample area 18 in the monkey contains several visual retinotopicrepresentations, and these have been definitively recognized asseparate cortical areas (V2, V3). It is our position that, just as in themonkey, it is essential in humans to evolve from Brodmann areasto cortical areas defined by multiple criteria: cyto- and myeloarch-itecture, anatomical connections, the topographic organization(where possible), and functional properties. It may take some time,however, before we have the tools to apply all four criteria to thehuman brain.In the interim, we agree with the authors and the ICBM that theMNI atlas should be considered as the standard anatomical referenceframe, which can be complemented by population-based probabil-istic surface maps such as the PALS atlas (Caret software) andautomatically labeled T1-weighted images.Inaddition,wewishtopointout,however,thatBrodmannareasdonot allow one to draw any conclusions about the homology betweenthehumanandmonkeybrains.Weareconvincedthattheonlywaytoaddress the question of homologies is by introducing fMRI in theawake monkey (Vanduffel et al., 2001, 2002 ) as this technique linkshumanfMRIwithsinglecellstudiesandinvasivestudiesperformedinthe monkey (Orban et al., 2004). Functional imaging in the monkeyhas the additional benefit that regions of interest can be definedanatomically. For example, using monkey fMRI, Nelissen et al.(2005)we were able to show that two cytoarchitectonically definedventral premotor areas (F5c and F5a) differ in functional properties.More recently it has been shown that the latter two regions alsodiffer in anatomical connections (Luppino et al., 2006), whichstrongly suggests that they are indeed different cortical areas. Ofcourse much less is known about the architectonics of the humanbrain, but as more atlases become available (Germann et al., 2005;Eickhoff et al., 2006) they can be used to define anatomical ROIs.Connections with other brain regions is another anatomicalcriterion with which to define cortical areas. It is tempting to thinkthat probabilistic fiber tracking, as mentioned by the authors, mightsupply such an anatomical criterion in humans. Yet currentlyavailable DTI-based tractography methods are beset with severalproblems: (1) none of the current DTI/HARDI methods can easilydistinguish between detailed intra-voxels fiber crossings, and (2)all of the DTI-based tractography algorithms are model-driven andhence not free of assumptions. Thus all these tools still need to bethoroughly validated by monkey studies. The lack of appropriatevalidation in monkeys can lead to unsound conclusions (Sawamuraet al., 2006), underscoring the need to validate the manyassumptions made in human brain imaging by studies in nonhumanprimates, especially before using them in clinical setting.References