Age-related performance variance on substitution coding tests has been found to account for much of the age-related variance in tests of fluid and other abilities, leading to the conclusion that cognitive decline is due to slowing. Although it is an easy task, which could easily be performed accurately given adequate time, the substitution coding task is not a pure measure of cognitive speed. Evidence from growth curve analyses involving 3,708 volunteers (49-95 years of age) from the Manchester and Newcastle Studies of Cognitive Aging (P. Rabbitt, C. Donlan, N. Bent, L. McInnes, & V. Abson, 1993) indicates that, with practice on this task, improvement is related more to memory than to age, reasoning, vocabulary, or perceptual speed. In other words, faster performances are related primarily to memory. Operational similarities between speeded measures and measures of higher order abilities, which weaken the argument for causal relationships, are discussed.