In Experiment 1, young and old adults both generated and read the same words either two, five, or eight times (with frequencies combined orthogonally) and then judged the frequency of either the generated or read words. In Experiment 2, young and old adults generated and read different words either one, two, or three times and judged the frequency of both the generated and read words. In both experiments, generation, compared to reading, increased the slope of frequency judgments comparably for the two age groups. In Experiment 1, neither age group could discriminate read from generated frequency. In Experiment 2, a 24-h retention interval reduced the slope of frequency judgments, but did not interact with either age or generation efforts. These results seem most consistent with a semantic activation explanation of the generation effect and a familiarity-based judgment about frequency of occurrence.