Fluent speech production requires interaction between the language processing system and many other cognitive systems, including working memory (WM). However, there are still unanswered questions of how exactly language and WM relate. Previous evidence suggests that verbal WM consists of separable semantic and phonological components which have different relations to language processing (Martin, Rapp, & Purcell, 2020). Recent neuropsychological work replicated a previously reported relation between semantic WM and the elaboration of content in utterances and also reported a relationship between the rate of spontaneous narrative production and phonological WM (Martin & Schnur, 2019). The latter finding was unexpected because the default scope of phonological planning has been considered to be quite small, making few demands on phonological WM. In the current work, we tested the hypothesis that the relation between phonological WM and the rate of narrative production is due not to phonological WM capacity per se, but rather a shared phonological retrieval process that affects both the efficiency of retrieval of words in narrative speech and the rate of rehearsal in supporting performance in the phonological WM task. In Experiment 1, we examined the relations between single word production latencies, phonological WM, and rehearsal rate using an individual differences approach. If the phonological retrieval hypothesis is correct, we should find a relationship between single word production latencies and phonological WM performance. Additionally, we should observe a relation between phonological WM and rehearsal rate. The first prediction was confirmed in that phonological WM related naming to picture naming latencies for lower but not higher frequency items. However, rehearsal rate was unrelated to phonological WM performance when rehearsal rate was measured with the speed of repeating a single word, thus arguing against a role for rehearsal in phonological WM performance. To further investigate whether phonological retrieval supports rehearsal in the phonological WM tasks, Experiment 2 tested participants’ single word production and phonological WM using tasks that either do (i.e., standard word recall or matching span) or do not (i.e., speeded WM tasks) allow for rehearsal. We predicted that phonological WM would be related to single word production when the phonological WM task allowed for rehearsal, but not when the task discouraged it. In contrast to these predictions, performance on the standard WM tasks did not relate to picture naming latencies, whereas for two of the speeded tasks, such relations were found, with the relations being stronger for low frequency items. The failure to find a relation of WM performance to speech rate in Experiment 1 and the relation of picture naming latencies to WM in speeded WM tasks in Experiment 2 call into question the role of rehearsal in WM performance. Consequently, these findings suggest that other explanations for the relationship between narrative language production and phonological WM—including general processing speed and output phonological buffer capacity—should be explored.