The experiments presented here demonstrate that normal aging amplifies differences in time production occurring in delayed free-recall testing. Experiment 1 compared the time production ability of two healthy aged groups as well as college-aged participants. During the test session, which followed a 24-h delay and omitted all feedback and examples of the two target intervals, the two groups of aged participants' over-produced a 6s interval. This effect is similar in form to errors shown by young participants, but twice the magnitude. Productions of a 17s interval were generally accurate overall. However, further analysis indicated that the majority of aged participants over-produced the 17s interval while a minority greatly under-produced the 17s interval. Furthermore, aged participants showed violations of the scalar property of timing variability in the training session, while in the test session, only those who under-produced the 17s interval showed this tendency. In contrast, training session performance was good for all participants. Experiments 2 and 3 investigated the ability of the participants in Experiment 1 to reproduce the length of a line from memory, under conditions analogous to those of the time production experiments. These experiments provided tests of the specificity of the errors observed in Experiment 1. Performance in the older participants was accurate, if more variable, compared to the young participants, in contrast to the time production results, indicating that production inaccuracy in free-recall is specific to interval timing in the current context.