Responding to a stimulus presumably creates a transient episodic link between individual stimulus- and response-related features, a so-called event-file (e.g., Hommel, 1998). Any feature repetition retrieves the whole event file, leading to impaired performance whenever the retrieved information does not match current processing requirements, so-called partial repetition costs (PRCs). In the widely-used S1R1-S2R2 task (e.g., Hommel, 1998), participants first perform a cued response (R1) after the onset of a stimulus (S1), followed by a binary-choice response (R2) determined by a feature of a subsequent stimulus (S2). In its typical implementation, the binary choice task for the S2 could be applied to stimuli presented as S1. This potentially leads to a conflict at S1 presentation if the to-be-executed response (signaled by the cue) violates the instructed response rule (for the S2). Importantly, this is confounded with the occurrence of partial repetitions between S1-R1 and S2-R2. The present experiment (N = 31) resolved the confound by eliminating the possibility to employ the S2 task at the time of S1. To this end, different stimuli for S1 (rectangle or diamond) and S2 (triangle or circle) were employed in a typical S1R2-S2R2 task with shape as response-relevant dimension for S2. Location, Color, and Response relation were systematically varied between S1-R1 and S2-R2. S2 latency and accuracy revealed PRCs for location, but not for color, suggesting that (1) event files linking spatial stimulus and response features are created in an unconfounded task setting and (2) not all stimulus features enter an event file format.