Superstition is not restricted to primitive tribes, but afflicts academia. The publication of a preliminary notification of results in conference proceedings should not adversely affect the likelihood of a paper being accepted for publication, but why take a chance? A tentative conclusion in even the most restrained abstract may inflame the ire of a potential Malevolent Reviewer, who waits for their moment, accepts the invitation to be a dispassionate evaluator of truth and then savages the paper when it is proffered for publication. The Risk-Averse Editor flees from the dilemma, and through inertia or moral turpitude spins out the review process until our noble researcher dies penniless in a garret, their insights lost to humanity.
So, the following paper is a work still in progress, and this is just advance notice.
Spearman’s hypothesis strongly confirmed
Many studies have been conducted on differences in mean intelligence test scores between ethnic groups. Often cultural factors are used to explain these differences. However, Spearman's hypothesis states that the group differences on the subtests of IQ batteries can best be explained in terms of differences in the complexity of the tasks, that is, the demands they make on the general factor of mental ability, the g factor. In a review of many studies Jensen (1998) reports a correlation of .63 between subtests’ g loadings and mean group differences on those subtests. Jensen hypothesizes that a psychometric meta-analysis will show that this value is a strong underestimate of the true effect.
Psychometric meta-analysis (Hunter & Schmidt, 2004) estimates what the results of studies would have been if all studies had been conducted without methodological limitations or flaws. The results of perfectly conducted studies would allow a clearer view of underlying construct-level relationships (Schmidt & Hunter, 1999). Psychometric meta-analysis is based on the principle that there are statistical artifacts in every dataset and that most of these artifacts can be corrected. In the present meta-analyses we corrected for five statistical artifacts that alter the value of outcome measures: (1) sampling error, (2) reliability of the vector of g loadings, (3) reliability of the vector of group differences in intelligence, (4) restriction of range of g loadings, and (5) deviation from perfect construct validity.
A psychometric meta-analysis of Spearman's hypothesis was carried out using the software package developed by Schmidt and Le (2004) showing Jensen was correct. Our psychometric meta-analysis of IQ batteries showed a strong relation between ethnic group differences and the subtests’ g loading: a true correlation of .91, based on a very large total N. Group identity was tested as a moderator, but no evidence was found: all different ethnic groups showed similar or in some cases virtually identical results.
Spearman’s hypothesis is now supported for most major test batteries and most major racial groups and can be considered to be an empirical fact. Mean differences in intelligence between groups can be almost perfectly explained by the complexity of the subtests in an IQ battery. There is simply no support for cultural bias as an explanation of group differences.