A meta-analysis of the worst performance rule

Abstract

The worst performance rule (WPR) describes the phenomenon that individuals' slowest responses in a task are more predictive of their intelligence than their fastest or average responses. Because the WPR supposedly amplifies in heavily g-loaded tasks and in samples whose cognitive abilities factor structure is dominated by a strong g-factor, it has been suggested that whatever mechanism is giving rise to the positive manifold may not promote peak performance, but may rather limit performance in a wide range of cognitive tasks. The aim of the present meta-analysis was to provide a meta-analytically determined estimate of the strength, consistency, and generalizability of the WPR. Across 19 studies containing 23 datasets with a total of 3767 participants, there was robust evidence for the WPR. However, the increase in correlations across quantiles of the RT distribution did not follow a linear, but a logarithmic trend, suggesting that those cognitive processes contributing to fast responses in reaction time tasks are less strongly related to cognitive abilities (r = −0.18) than other cognitive processes contributing to average (r = −0.28) and slow responses (r = −0.33). There was no evidence that the strength of the worst performance rule increased with greater mean reaction times, in tests of general intelligence, or in samples with lower or average cognitive abilities. Instead, it was attenuated in less intelligent samples and greater when correlated with speed instead of intelligence or memory tests. Hence, the WPR may not be as characteristic for g and may play a smaller role for theoretical accounts of the positive manifold than previously thought.