Two experiments investigated the impact of memory set size on effects of irrelevant spatial S-R correspondence. One vs. four stimuli were assigned to either hand, with random variation of the location of the target stimulus (Experiment 1) or of an accessory tone (Experiment 2). Correspondence effects decreased with increasing set size, thus disproving the buffer model of response selection suggested by Mewaldt, Connelly and Simon (1980). Instead, results are interpreted in terms of automatic stimulus-induced response activation that is subject to rapid decay.