ABSTRACT
As demonstrated in chapter 2, any correlation between cognitive ability and performance on a reasoning or decision-making task can be interpreted as indicating some degree of algorithmic-level limitation—at least for some individuals. In theory, correlations close to unity represent the idealized version of the Apologist’s position—every individual performing at the limits of their computational capacity. The descriptive model would exactly equal the prescriptive model of performance in such a case, and there would be no reason to impute irrationality to any subject. However, the actual results fell considerably short of this extreme. Even when corrected for attenuation due to the modest reliabilities of many of the reasoning tasks (Schmidt & Hunter, 1996), the associations observed in several of the studies were considerably less than perfect. Thus, the computational limitations that have been implicated in performance on some of the tasks do not represent absolute limitations. The moderate correlation between cognitive ability and performance on these tasks thus leaves considerable residual variance. Given that computational limitations can explain only a portion of the normative/descriptive gap, this chapter examines whether there are clues in the residual variance that might help to explain why a normative/descriptive gap remains.
