Age and method variance in measures of speed and working memory

Ten measures of speed of processing were administered to 157 individuals, aged 18 to 89 years. The 10 measures comprised five pairs, each of which had a paper-and-pencil and a computer, reaction time (RT) based version of the same measure. Three measures of working memory span were also administered. Two structural equation models were fit to the speed data, one with a single latent variable, speed, and another, nested-factor model in which there were also latent variables for the two methods of measurement. The model with the method latent variables provided a better fit. Age was more strongly related to the method latent variables than to the general speed latent variable. Adding the working memory measures showed that there was also shared variance in those measures beyond the general latent variable, also related to age. The results show that any single measure of speed includes variance due to speed but also to the method of measurement. Use of a latent variable approach to speed is recommended.