Asymmetry of G

Is the strength of the general factor of intelligence uniform? That is, do the various indicators of crystallized intelligence correlate similarly at all levels of intelligence? This question was first explored in 1927, by Spearman, who, on the basis of his data, suggested a law of diminishing returns (also known as the divergence hypothesis, or in my terms, the asymmetry of g). This observation states that at low levels of intelligence, the various facets of intelligence are very highly correlated, but at higher levels of intelligence, the correlations between the various facets are less strong.

Between 1927 and 1990, this work disappeared from sight, and as late as 1989, Detterman and Daniel could argue that “it was thought that the positive manifold [the loading of ability tests onto a single factor] was uniformly distributed over the whole range of ability.” This finding has been supported in a number of other studies.

The best estimate is for a linear decline in average correlations between IQ measures from r = 0.46 for the least gifted group to average correlations of about r = 0.30 for the more highly gifted. If this picture is sustained through additional research, then perhaps much of the conflict between those arguing for a pervasive g factor and those arguing for specific abilities can be reconciled by considering that both may be true, with g dominating at the lower levels of intelligence and specific factors dominating at higher levels of intelligence.

In practice, this means that we will have to reconsider our use of intelligence as a selection tool in job performance. It is fine to use g at low levels of intelligence, but at high levels, the specific abilities needed for a task will come into play, so predictive studies will have to explore the interaction between g and specific abilities to understand who the high performers will be.