Faster on easy items, more accurate on difficult ones: Cognitive ability and performance on a task of varying difficulty

Abstract

Using more complex items than those commonly employed within the information-processing approach, but still easier than those used in intelligence tests, this study analyzed how the association between processing speed and accuracy level changes as the difficulty of the items increases. The study involved measuring cognitive ability using Raven's Advanced Progressive Matrices and examining changes in the accuracy–ability and speed–ability correlations due to the increasing difficulty of the items. As expected, high-ability individuals demonstrated a higher accuracy rate and faster performance than low-ability individuals. The accuracy–ability association became stronger as the item difficulty increased. The speed–ability correlations tended to decrease in absolute values as the item difficulty increased, although this trend, when formally examined, was non-significant.For a more precise analysis, the entire variance of each response time and accuracy was decomposed into the components that described constancy in performance over difficulty levels and the rate of change in performance caused by increasing task difficulty. Estimating these components for both response latencies and accuracy scores within the same latent growth model demonstrated that the speed–accuracy correlation was evident only for baseline performance; the rate of slowing with increasing task difficulty was not correlated with the respective rate of change in accuracy level. High-ability participants differed from low-ability participants in the speed (not accuracy) of baseline performance and in the rate of difficulty-related change in accuracy (not in processing speed).