JOURNAL OF VERBAL LEARNING AND VERBAL BEHAVIOR 8, 2 0 6 - 2 1 4 (1969) Response Perseveration in Binary-Choice Recognition1 ARTHUR S. REBER University of British Columbia, Vancouver 8, British Columbia, Canada A S's memory for a previously given response was investigated in three experiments using a binary-choice recognition paradigm. In Exp. I Ss were shown original lists of 15 consonant trigrams followed by test lists of 30 consonant trigrams. They responded by categorizing each item as "old" or "new." Items on which Ss made an error were retested. Subjects who were informed that only error items would be used on the retest were not able to use such information efficiently and were correct no better than chance. Subjects who were not so informed about the nature of the retested items showed a strong response- perseveration effect and were correct significantly less than chance. In Exp. II the design was expanded to two regular recognition tests prior to the re-presentation of error items. Again Ss were unable to use the information about their previous responses although the response-perseveration effect was even more pronounced. In Exp. III a total of four test trials was run without any information feedback. Response latencies were recorded as well as correct and error responses. The data revealed a strong response-perseveration effect that increased with successive responses, with the most recent responses having the greatest effect. The results were discussed in terms of two possible hypotheses: (a) one based upon S's storage and retrieval of information about his previous responses, and (b) the other based upon predispositional aspects of stimuli. In experiments on human learning we often require the S to learn to make some response to a particular stimulus. When studies of memory are carried out it is implicit that they refer to stimulus memory, with the Sgenerally being required to recall or recognize something about the stimulus. In the studies reported here we tried to analyze the S's memory for the other end of the S-R chain--the response. A binary-choice, recognition memory para- digm was adopted for these studies. The S's memory for his response was investigated by the use of a second "guessing" trial. That is, all items incorrectly responded to on the first recognition-test trial were re-presented for a second guess. The distribution of correct 1 This research was supported by a grant from the University of British Columbia Committee on Research. Portions of this paper were read at the WPA Meetings, San Diego, 1968. John Anderson, Frank Krammer, and Kathleen Reber helped in running subjects and analyzing the d~tta. Thanks are extended to Michael Humphreys and Delos D. Wickens for some interesting insights into the problem. responses made on this guess-trial will reflect what the S can recall about the response made on Trial 1. This design seemed to have several advantages over other possible ones. First, by using a recognition procedure we guarantee that S make some well-specified response on each trial. Second, by using the binary-choice paradigm we reduce the effects of the stimulus display (see Murdock, !963). Subjects can still search the stimulus display of the previously presented items that is held in memory, but the exhaustive explicit examination of all items is eliminated. Third, the probability that the S makes a correct response on the second, guess-trial is essentially determined by the probability that he recalls the error response made to that item on Test 1. Using variations on this basic design, we ran three experiments. EXPERIMENT I Method The stimulus items were 30 consonant trigrams of low association value. Fifteen of them were selected at 206