According to the worst performance rule (WPR), the correlations between intelligence and sorted performances, for example on reaction time tasks, should strengthen from the best to the worst performance. A commonly proposed explanation for the WPR is that poor performances reflect lapses of attention that are particularly strongly related to intelligence. The correlation of sorted scores rule (CSSR), on the other hand, claims that the WPR arises due to certain statistical properties of the data. Specifically, the magnitude of intelligence-performance correlations will change with the rank order of the test when intelligence is correlated with the within-individual standard deviation (WISD) of the tests. If the latter correlation is negative, a WPR is seen, i.e. intelligence-performance correlations will be lower for tests with higher rank order. If the intelligence-WISD correlation were positive, however, intelligence-performance correlations would instead increase with test rank order. In the present study, through strategic slicing of two samples (N = 1485, and N = 43,987, respectively), we created subsamples with a large range of intelligence-WISD correlations. In accordance with the CSSR, but not the WPR, the association between intelligence-performance correlations and test rank order was found to reflect the intelligence-WISD correlation of the subsample. This supports that the WPR might be a special case of the more general CSSR and that the WPR is crucially dependent on intelligence-WISD correlations. The findings also indicate that the predictions made by the CSSR generalize to other predictors besides intelligence and to other outcomes besides reaction time.
Read full abstract