September 18, 2006

More massaged data from Richard Lynn

I had previously pointed out that Richard Lynn has massaged his data based on one study that I had access to and could thus cross-check his against Lynn's reported values. In a review (http://dx.doi.org/10.1016/j.intell.2006.08.001) of Race differences in intelligence: An evolutionary hypothesis in Intelligence, N.J. Mackintosh brings up more falsifications of data by Lynn:
Even if we wanted the data, can we rely on Lynn to have given an accurate account of them? I do not pretend to have read the originals of more than a handful of the papers and books cited by Lynn, but it just so happens that I wrote two of them myself, and Lynn has simply got their data wrong. Table 6.2 gives the results of 13 estimates of the IQ scores of south Asians in Britain and Australia. The median score is said to be 89, and similar for Indians and Pakistanis (as is to be expected since they are the same racial group). In fact, three British studies have given the same IQ tests to Indian and Pakistani children, and in all three, Indian children have outscored the Pakistanis by 4–6 IQ points. Mackintosh and Mascie-Taylor (1986) reported IQ scores of 97 and 93 for 10-year-old children of Indian and Pakistani origin respectively, but Lynn incorrectly attributes both scores to Pakistanis, one said to be for children resident in Britain more than 4 years, the other for more recent immigrants. West, Mascie-Taylor, and Mackintosh (1992) reported IQ scores of 91 and 85 for Indian and Pakistani children, but in Lynn's table these have mysteriously turned into scores of 87 and 88. The errors may not be particularly important, and I do not know how typical they are. But they do not increase my confidence in Lynn's scholarship.

A more egregious example is provided by his treatment of the Eyferth (1961) study of two groups of illegitimate children fathered by (mostly) American black and white servicemen and brought up by their (carefully matched) German mothers. Eyferth reported an average IQ of 96.5 for the mixed race children and of 97.2 for the whites. Lynn reduces the former number to 94 to compensate for use of an old test, and compares it, not with the score of the white sample, but with an average IQ of 100 for German children. He is thus able to conclude that the IQ of these mixed race children is half way between that of Americans and Africans. He derives the same conclusion from the Weinberg, Scarr, and Waldman (1992) transracial adoption study since, at the 10-year follow-up, the mixed race children had an average IQ of 94, mid-way between the 102 of the white children and the 89 of the black children. He omits to mention one of the more salient features of this follow-up, namely, that there had been substantial attrition in the white sample—with a loss of those children with lower IQ scores, resulting in an overestimate of the white group's IQ by some 6 points.

More from the review:
Group differences in IQ, as a general rule indeed, are said to be largely genetic. Although he argues that the results of adoption studies point to this conclusion, Lynn's main focus is on brain size and his evolutionary hypothesis, which is that genetic differences between different races must have produced differences in such a genetically determined characteristic as intelligence, since they have, after all, produced differences in brain size, with groups remaining in or migrating to benign tropical climates undergoing less selection pressure for increases in intelligence and brain size than those exposed to harsh northern climates and ice ages. It would have been news to the first Europeans who ventured there that the Australian outback was an undemanding environment, and, unfortunately for Lynn's argument that increases in brain size were a consequence of migration to the northern hemisphere and the need to survive ice ages, the cranial capacity of early Homo sapiens, 100,000 or more years ago, was rather greater than that of modern Europeans (Aiello & Dean, 1990). Moreover, even by his own calculations (in Table 16.3), differences in brain size explain only a quarter of the observed IQ differences between Europeans and Africans, Native Americans and south Asians. In fact, this is a serious overestimate, because he assumes a correlation between brain size and IQ of .40, which is true for MRI studies; but the racial data he uses are based on cranial capacity, which correlates only .18 with IQ. Differences in cranial capacity between different groups, therefore, could explain no more than 12% of the IQ differences between them.

...

What can one say? Much labour has gone into this book. But I fear it is the sort of book that gives IQ testing a bad name. As a source of references, it will be useful to some. As a source of information, it should be treated with some suspicion. On the other hand, Lynn's preconceptions are so plain, and so pungently expressed, that many readers will be suspicious from the outset.

No comments: