The serial reaction time data from an experiment producing a negative recency effect at different interstimulus intervals were subjected to an analysis which included an investigation of up to and including fifth-order sequential effects. Third-, fourth-, and fifth-order effects were found to occur, indicating that the negative recency could not be explained by considering sequences of two events only. The implications of these findings are discussed with respect to models seeking to explain sequential effects in serial reaction time tasks.Sequential effects have been found in serial reaction time experiments such that under some conditions, responses to a signal which is the same as the preceding signal are faster than if it is different (e.g., Bertelson, 1961). Under other conditions the opposite has been found, the response to a signal which is different from the preceding signal being faster than if it is the same (e.g., Williams, 1966). The former has been called a positive recency or "repetition" effect (Bertelson, 1961), and the latter a negative recency or "alternation" effect.Some experiments have reported that the repetition effect tends to decrease, and even to change to an alternation effect, with increasing interstimulus interval (e.g., Bertelson, 1961;Hale, 1967), but others have failed to find such a change (e.g., Keele, 1969;Schvaneveldt & Chase, 1969). Generally, however, with two-choice compatible S-R arrangements, repetition effects have been found with interstimulus intervals of less than 1 sec. (e.g., Bertelson, 1961;Hale, 1967), while alternation effects have been found with interstimulus intervals of greater than 1 sec. (e.g., Hale, 1967 ;Williams, 1966).Attempts to explain the repetition effect by Bertelson (1963) and the alternation effect by Williams (1966) have both assumed that they primarily are due to the 1 The author wishes to thank A. T. Welford for his valuable advice, encouragement, and criticism, and R. Willson for writing the computer programs used in this experiment.2 Requests for reprints should be sent to N.