Consideration of cognitive variance components potentially solves Beauchamp's paradox.

Beauchamp (1) presents a paradox: There is genetic selection favoring reduced time spent in education, or educational attainment, among the contemporary population of the United States, amounting to an expected decline of 1.5 mo per generation. Despite this finding, the average time spent in education among this population has increased by 24 mo per generation (1). This paradox is analogous to the one that confronted psychologists when it was discovered that the total-scores obtained from intelligence tests (IQ) were rising sharply (this being the Flynn effect), rather than falling over generations, as predicted by negative associations between IQ and fertility (2). Beauchamp treats educational attainment as if it were the target of selection, which is problematic. Educational attainment is better conceptualized as an outcome that arises from facultative calibration over childhood and young adulthood in response to the action of heritable phenotypic characteristics, chief among these being IQ, with which it shares ∼60% of its (linkage-pruned) genetic variance (3). Consistent with this finding, polygenic “educational attainment” scores similar to those used by Beauchamp (1) also directly predict variation in IQ (4). IQ itself is comprised of several variance components, with a general cognitive ability (GCA) factor accounting for ∼50% of the performance variance among individuals, and various specialized abilities accounting for the remainder. GCA is the most heritable variance component of IQ (5). Studies investigating the negative relationship between IQ and fertility in Western populations have revealed that the magnitude of the association is positively moderated by the GCA-loading of IQ measures, indicating that the higher heritability of GCA may in turn be positively moderating the sensitivity of these measures to potential selection (6). Thus, the target of genetic selection in Beauchamp’s (1) data is likely to be GCA, this being the principal source of the “reactive” heritability of educational attainment. GCA may also be decreasing because of selection, consistent with the observation of long-term (80–100 y) performance declines among United States and United Kingdom populations on working memory and processing-speed measures, both of which are considered to be sources of the positive manifold among cognitive abilities, and thus key endophenotypes of GCA (7, 8). The Flynn effect, however, is more pronounced on IQ measures that are less GCA-loaded, and less heritable as well (9). Thus, specialized cognitive skills are being enhanced to a greater degree than GCA is declining, masking the decline at the aggregate IQ level. This finding is consistent with the observation that educational exposure raises IQ at the level of specialized cognitive skills, not at the level of GCA (10), making increasing time spent in education a plausible driver of the Flynn effect. Economic growth could furthermore be placing a premium on cognitive specialization as a source of increased comparative advantage among the workforce, hence the requirement that populations now spend longer periods of time in education. Thus, Beauchamp’s (1) paradox can potentially be resolved by recognizing that cognitive variance components need to be properly defined in future studies, as they can respond in different ways to selection and environmental change.