The response-signal speed-accuracy tradeoff (SAT) procedure was used to investigate how proactive interference (PI) affects retrieval from working memory. Participants were presented with 6-item study lists, followed immediately by a recognition probe. A variant of a release from PI design was used: All items in a list were from the same semantic category (e.g., fruits), and the category was changed (e.g., tools) after three consecutive trials with the same category. Analysis of the retrieval functions demonstrated that PI decreased asymptotic accuracy and, crucially, also decreased the growth of accuracy over retrieval time, indicating that PI slowed retrieval speed. Analysis of false alarms to recent negatives (lures drawn from the previous study list) and distant negatives (lures not studied for 168+ trials) suggests that PI slowed retrieval by selectively eliminating fast assessments based on familiarity. There was no evidence indicating that PI affected slow processes involved with the recovery of detailed episodic information.