Higher-order g versus blended variable models of mental ability: Comment on Hampshire, Highfield, Parkin, and Owen (2012)

Abstract

Hampshire et al. (2012) suggested that the positive correlations among all mental ability tasks do not necessarily imply the existence of a general intelligence (g) factor (e.g., Spearman, 1904), but could instead be explained by “task mixing”, that is, the influence of multiple independent ability factors on the various tasks. We note here that the task mixing model is conceptually very similar to the blended variables model proposed in the personality domain (Ashton et al., 2009) as an alternative to higher-order personality factors. Here we use CFA to compare a higher-order g model with a task mixing or blended variable model in relation to the data of Hampshire et al., and we find that the higher-order g model provides a much closer fit to the data. Following Thurstone (1938), we suggest that it is conceptually implausible that every task is influenced by every factor of mental ability. We also suggest that the non-existence of g would be demonstrated by finding mutually orthogonal markers of those factors; however, the data of Hampshire et al. and other mental ability datasets suggest that this cannot be achieved.