Further arguments that ability tilt correlations are spurious: A reply to Coyle (2022)

Ability tilt refers to a within-individual difference between two abilities, e.g. a difference between math and verbal ability. Coyle and colleagues have demonstrated correlations between ability tilts and measures of the constituent abilities. We have previously pointed out that such measures may be spurious as the tilt variable is dependent on the constituent abilities. We have further shown that reported tilt associations are inconsistent with simulations including non-spurious tilt-effects, and concluded that tilt-correlations demonstrated by Coyle and colleagues are spurious. In a recent paper, Coyle responded with a series of arguments, including that the validity of tilt correlations is supported by their agreement with theoretical predictions, and that the analyses we used in our previous critique (regression effects) differ from tilt-correlations. Here, we advance the discussion by responding to the arguments put forward by Coyle. We show that the difference between regression effects and correlations is not material to the validity of our argument. Furthermore, we discuss the relation of tilt correlations to theory, and show that many empirical tilt-correlations, e.g. between the birth rate - death rate difference and fertility in US states, can be observed although such correlations can hardly be explained by differential investment theories. Therefore, we maintain that tilt correlations are spurious and that they offer little support for theories concerning the development of intelligence.

Automated summary

Ability tilt refers to a difference between two abilities within an individual. The authors argue that previous research showing correlations between ability tilts and constituent abilities is false because the tilt variable is dependent on the constituent abilities. They further demonstrate that reported tilt associations are inconsistent with simulations, suggesting that the correlations are spurious. The authors respond to arguments by Coyle, showing that the difference between regression effects and correlations is not relevant and discussing how empirical tilt-correlations can be observed even if they cannot be explained by current theories.