Speed mediation of adult age differences in cognition.

Previous research has established that a large proportion of the adult age-related variance in various measures offluid or process cognition is reduced when statistical control procedures such as hierarchical regression are used to eliminate variation in measures ofperceptual speed. This finding was confirmed in the present study and was extended to include paired-associate and free-recall measures of memory in addition to measures of reasoning and spatial abilities. Most of the speed mediation was associated with speed measures requiring cognitive operations such as comparison or substitution instead of merely motor responses such as copying symbols or drawing lines. These results suggest that the rate of performing elementary cognitive operations is an important proximal mediator ofthe adult age differences in several types ofcognitive tasks. Inspired largely by ideas of Birren (e.g., 1955,1965,1974), Salthouse 0985a) proposed that the adult age differences in certain measures of cognitive functioning might be partially mediated by age-related reductions in the speed ofexecuting relevant mental operations. The evidence available at that time was rather weak, because no published studies had been explicitly designed to investigate this hypothesis. Most of the data relevant to the hypothesis were obtained by locating published articles reporting correlations among age, a measure of processing speed, and a measure of cognition (or studies by Salthouse containing the relevant data), and then computing partial correlations between age and the cognitive measure after partialing out the measure of speed. Only a few data sets with the necessary information could be identified, and the samples were often small and usually involved two extreme age groups rather than a continuous range of ages. Furthermore, in all cases the relevant constructs were assessed with a single measure rather than with potentially more reliable and valid multiple measures. Despite these limitations, the hypothesis received some support because statistical control of the measure ofspeed frequently resulted in an appreciable reduction in the magnitude of the relation between age and the measure of cognition.