A previous investigation reported that cross-sectional age differences in Digit Symbol Substitution (DSS) test performance reflect declines in perceptual processing speed. Support for the tenability of the processing speed hypothesis requires examining whether longitudinal age-related change in DSS performance is largely mediated by changes in speed. The present study used data from the Victoria Longitudinal Study to examine patterns and predictors of longitudinal change in DSS for 512 older adults (M(age) = 68.37 years, SD = 7.43). On the basis of multilevel modeling, baseline DSS performance was poorer for older participants and men, with longitudinal declines more pronounced with increasing age and decreasing speed. In contrast to the present cross-sectional findings, statistical control of change trajectories in perceptual speed using the same data did not substantially attenuate age changes. These discrepancies suggest different sources of variance may underlie cross-sectional age differences and longitudinal age changes for DSS.