Encountering stimuli and responses is thought to create transient episodic links between stimulus- and response-related features, referred to as event files. Reencountering any feature of an event file is assumed to retrieve the whole event file, leading to impaired performance whenever the retrieved information does not match current processing requirements as compared with full or no repetitions (partial repetition costs, PRCs). The current study (N = 34) further investigated this notion by additionally analyzing error types in a variant of the S1R1-S2R2 task. In a typical task, participants first perform a cued response (R1) after the onset of a stimulus (S1), followed by a binary-choice response (R2) determined by a specific feature of a subsequent stimulus (S2). However, with two response alternatives, errors in R2 cannot be unambiguously attributed to the retrieval of the event file created of the S1-R1 episode. In the present task, shape (circle, triangle, or square) and location (above or below) were systematically repeated or changed between S1 and S2. Responses (i.e., “left”, “center”, or “right”) either repeated or changed between R1 and R2. Results revealed significant PRCs in latency and overall accuracy for shape only. Importantly, repeating the shape between S1 and S2, but changing the required response, increased the likelihood of repeating R1 at the time of S2. Moreover, the results suggest that repeating the required response between R1 and R2 retrieves the S1 shape. Together, the results support the notion that stimulus-response information contained in an event file is retrieved upon feature repetition.