The verbal fluency task requires generation of category exemplars and appears to be an example of what M. Moscovitch (1995) calls a strategic test of memory retrieval. Four experiments explored the role of individual differences in working memory (WM) capacity on verbal fluency under various secondary load conditions. High WM participants consistently recalled more exemplars. However, load conditions caused a decline in recall only for high WM participants. Low WM participants showed no effect of secondary workload on exemplar generation. WM group differences and load effects were observed even in the 1st min of retrieval, which suggests that differences were not due to differences in knowledge. A model of retrieval is supported that relies on cue-based-automatic activation, monitoring of output for errors, controlled suppression of previously recalled items, and controlled strategic search.