Confusing cortical columns.

T he late developmental neurobiologist, and a member of the National Academy of Sciences, Victor Hamburger told me during one of our discussions about the distinction between boring data and exciting concepts, ‘‘one can spend an entire lifetime correcting a flawed paper published in reputable journal and still loose the battle if people like the basic idea’’ (V. Hamburger, personal communication). An example of the longevity of basically incorrect information is the phenomenon of ‘‘The basic uniformity in structure of the neocortex,’’ published in 1980 by Rockel, Hiorns, and Powell (1). This highly influential paper had obvious problems at almost every level: The authors selected an arbitrary 30m-wide, 25m-deep vertical cortical ‘‘column’’ between the pia and the bottom of the cortex, because the ruler in the graticule of the oil-immersion eyepiece on their microscope had a 30m marker and their histological sections were 25 m thick; then, they estimated that the number of neurons within this ‘‘minicolumn’’ is 110 in all cytoarchitectonic areas examined, without any correction for the cell size; and finally, based on this dubious finding, they made a broad generalization that the magic number of 110 is constant in all mammalian species (rodents, carnivore, and primates, including human) in all cytoarchitectonic areas (except the primary visual cortex in primates). This finding led them to conclude that, ‘‘the intrinsic structure of the neocortex is basically more uniform than has been thought and that differences in cytoarchitecture and function reflects differences in connections.’’ Most neuroscientists recognized the problems with both the method used and the data obtained, but many found the simple concept of the uniformity of the cortex across various modalities as well as during evolution of neocortical expansion highly attractive. Although at least six research articles have directly refuted the accuracy of the data of Rockel et al. as well as the validity of their generalization (e.g., ref. 2), according to the Institute of Scientific Information (ISI), their article has been cited 500 times. It is discussed often in the major reference books of neuroscience and brain evolution, and is used widely in computational models of cortical operations. The concept of uniformity across all species appealed to philosophers who liked the idea that the difference between animals and human are just quantitative. It also provided the scientific basis of the so-called ‘‘tabula rasa hypothesis’’ of cortical development, which assumes that all cytoarchitectonic areas are specified from the initially homogeneous and equipotential cortex by input from the periphery. This is why the article by HerculanoHousel et al. (3) in this issue of PNAS serves a useful purpose, even though it does not report any unexpected results. Using a state-of-the-art unbiased stereology method, the authors show convincingly that the density of neurons in the neocortex varies as much as three times even among the highly related primate species. The results from the two