Experiment 1 examined the relationship between item retrievability and item discriminability using a generate-rate task. All the items of an input list came from a single category first names. The generate-rate task required subjects to generate, over a 10-min period, as many first names as possible and to rate each name for list membership. Retrievability was assessed either by how early names were generated or by the number generated in a particular subclass. Retrievability was higher for high frequency names and discriminability was higher for low frequency names. High frequency list names generated early were better discriminated than high frequency list names generated late: in contrast, the discriminability of low frequency names was high whenever they were generated. Experiment 2 replicated the main results of Experiment 1, examined the effect of personal significance on retrievability and discriminability, and, for comparison, included a group given a recall instead of a generate-rate instruction. Personally significant names were both generated earlier and discriminated better than other names were. However, the overall conclusion from the two experiments is that the intrinsic relationship between item retrievability and item discriminability is surprisingly weak. The output of list names by the generate-rate and recall groups was sufficiently similar to warrant the conclusion that results obtained by the use of the generate-rate task have implications for the analysis of recall, including the effect of the recall criterion on recall.