Log In | Sign Up | Help

• Home
• Explore
• Community

• Partners

Transmitting false memories in social groups BARBARA H. BASDEN, MATTHEW B. REYSEN, AND DAVID R. BASDEN California State University at Fresno The American Journal of Psychology Summer 2002, Vol. 115, No. 2, pp. 211–231 Content in the AJP database is intended for personal, noncommercial use only. You may not reproduce, publish, distribute, transmit, participate in the transfer or sale of, modify, create derivative works from, display, or in any way exploit the AJP content in whole or in part without the written permission of the copyright holder. To request permission to reprint material from The American Journal of Psychology, please find us online at: http://www.press.uillinois.edu/about/permission.html or email us at: UIP-RIGHTS@uillinois.edu © 2002 by the Board of Trustees of the University of Illinois Transmitting false memories in social groups BARBARA H. BASDEN, MATTHEW B. REYSEN, AND DAVID R. BASDEN California State University at Fresno In 2 experiments, subjects studied word lists drawn from Roediger and McDermott (1995) and then participated in perceived group recall (PGR) tests that were intended to lead each subject to believe that she or he was participating in collaborative recall in a 4-person group. Some of the lists were followed by PGR tests containing the nonpresented critical word, some lists were followed by PGR tests not containing the nonpresented critical word, and some lists were not followed by PGR tests. Subjects then completed individually administered recall tests and subsequent immediate or delayed recognition tests that required remember or know judgments. The major finding was that critical words contained within PGR tests were as likely to be falsely recalled, recognized, and consciously remembered as original list items. These findings show that false memories can be socially transmitted. Although many studies of false memory have been conducted in the past few years, few have examined social influences on false remembering. In a sense this is surprising in that Bartlett (1932), who conducted the earliest investigations of memory distortion, concluded that what others recall may strongly influence personal recollections. The major purpose of these studies was to determine whether false memories occurring during an initial “collective” memory test would be preserved on subsequent tests administered to individual subjects. Roediger and McDermott (1995) adopted and modified a procedure originally devised by Deese (1959) to study false memory in individuals. In Deese’s procedure, word lists of semantic associates were presented (e.g., dream), and subjects were subsequently asked to recall the items. Although a single critical word that was strongly related to the list members was not presented (e.g., sleep), that word was nonetheless quite likely to be recalled. Roediger and McDermott replicated the high levels of false recall for nonpresented critical words reported by Deese. Furthermore, they extended Deese’s original design to include measures of recognition and phenomenological experience. In Roediger and McDermott’s research, subjects often falsely recognized the critical AMERICAN JOURNAL OF PSYCHOLOGY Summer 2002, Vol. 115, No. 2, pp. 211–231 © 2002 by the Board of Trustees of the University of Illinois false memories in groups 213 words; furthermore, they claimed that they consciously recollected the critical words. That is, rather than reporting the falsely recognized items as merely familiar or “known,” subjects reported the items as “remembered” (Tulving, 1985). In the present experiments, the Deese (1959) and Roediger and McDermott (1995) (DRM) lists and procedures were used to examine the social transmission of false memory. Social influences on false memory Social influences on memory have been demonstrated using a variety of testing procedures. For example, Deutsch and Gerard (1955) modified Asch’s (1951) well-known studies showing distortion of perceptual judgments of line length. The modified design had subjects, either individually or in a group situation, make judgments from memory of the length of lines. Furthermore, the degree of prior commitment to responses was manipulated. In one condition, each subject wrote a response on a piece of paper and then discarded it before providing judgments in the presence of the group. In a second condition, each subject wrote a response and signed it before providing judgments in the presence of the group. As Asch had discovered for perceptual judgments just a few years before, subjects tested individually were more accurate than members of groups. Thus, Deutsch and Gerard successfully extended Asch’s findings to a memory task. They also observed that subjects who had made a public commitment before answering were less likely to conform. Despite the fact that subjects apparently knew which responses were correct, as indicated by their private judgments on paper, they answered incorrectly in the group to avoid contradicting responses given aloud by other group members. Social pressure was said to have resulted in conformity in memory for line judgments. Subjects were not tested individually after having been tested in groups. Deutsch and Gerard’s results imply that conformity to the group might occur in the absence of internalization of the false information. Similar results were reported by Betz, Skowronski, and Ostrom (1996). They asked subjects to respond to multiple-choice items. Next they presented bogus responses of other group members that were in high-consensus disagreement or low-consensus disagreement with the true subject’s choice. True subjects were more likely to change their response on a subsequent recall test if there was high-consensus disagreement rather than low, again suggesting that people succumb to group pressure. More recently, Schneider and Watkins (1996) conducted a study that analyzed the effects of social influence on recognition memory. In their experiment, subjects studied lists of related words and then engaged in a recognition task in the presence of a confederate. Both the subject 214 basden et al. and the confederate were asked to decide whether each word presented to them was on one of the original study lists or whether the word was new. The subject and the confederate took turns responding. Schneider and Watkins observed a high rate of conformity to the confederate’s responses on the part of the subject. That is, when the subject responded second, his or her answers tended to mirror those of the confederate. Through this procedure, Schneider and Watkins demonstrated that another person’s responses could play a powerful role in shaping even recognition performance. Again, their results are ambiguous in the same way as Deutsch and Gerard’s (1995) results. Incorrect responses may have resulted from conformity rather than from internalization of the incorrect information. Socially induced memory distortion has been investigated in experiments on collaborative remembering (e.g., Basden, Basden, Bryner, & Thomas, 1997; Basden, Basden, Thomas, & Souphasith, 1998; Weldon & Bellinger, 1997). In these experiments, rather than having another person introduce the false information, the frequency of memory errors was compared when social influence was present or absent. Basden et al. (1998) compared the recall and recognition performance of threeperson collaborative and nominal groups. In a nominal group, people are tested individually, and their nonredundant recall is subsequently combined. Conversely, a collaborative group is one in which people work together to produce a single recall output. An equal number of respondents contribute to recall in both conditions, permitting valid comparisons of collective and noncollective memory. Using a modification of the DRM procedure, Basden et al. (1998, Experiment 2) presented subjects with categorized lists from which the five most typical exemplars were omitted. They found that collaborative groups recalled and recognized more nonpresented critical exemplars than did nominal groups. In agreement with Deutsch and Gerard (1955), Basden et al. concluded that social pressure underlies the phenomenon of greater memory distortion in social groups. When pressured to provide a response during collaborative recall, a group member may provide one of the more typical exemplars. Providing one of those exemplars does not represent conformity in that the subject is producing a response not provided by others. Basden et al.(1998) also asked subjects to provide remember or know judgments for the items they recognized. In Experiment 2, remember judgments were given as frequently to critical words that were falsely recognized as to target words that were correctly recognized. These results confirm that subjects internalize memory errors, even those that occur as a result of group pressure. Roediger, Meade, and Bergman (2001) recently reported an experi- false memories in groups 215 ment demonstrating the phenomenon of social contagion. Pairs of subjects, one of whom was a confederate of the experimenter, studied six different scenes (e.g., kitchen) containing approximately 24 common objects. During a subsequent collaborative recall test, the confederate interjected either a high-expectancy item (e.g., a toaster) or a low-expectancy item (e.g., oven mitts) for three of the six scenes. On a subsequent individual recall test, the true subject recalled the critical items more often when they were introduced by the confederate than when they were not. High-expectancy critical items were more likely to be recalled than low-expectancy critical items. Furthermore, critical items that were recalled were more likely to be designated as known rather than remembered. Overall, their results show that what another person reports remembering yields conformity effects similar to those described in the research of Deutsch and Gerard (1955) and others. Meade and Roediger (in press) extended findings on social contagion reported by Roediger et al. (2001). They showed that warning subjects that the confederate may introduce false information reduced false memory. However, both false recall and false recognition persisted despite the warning. When asked to provide remember or know judgments for recognized items, participants judged critical items introduced by the confederate as known rather than remembered. Meade and Roediger interpreted their overall findings as evidence that false memory occurs not as a function of public conformity but rather as a function of the subject’s personal belief that the false information actually occurred. Plan of the present research In the present experiments, rather than being tested in face-to-face groups, each subject believed that she or he was working together with three other students to recall as many words as possible from word lists. In actuality, each subject worked independently on a computer program designed to introduce nonpresented critical words in varying serial positions predetermined by the experimenter. However, from the subject’s perspective, others within the group had recalled the words occurring during this initial test. Perceived group recall (PGR) tests were followed by individual recall and recognition tests for each member of the group. In both experiments, subjects arrived at the laboratory simultaneously with three others who were perceived to be group members. Each subject studied either 9 (Experiment 1) or 15 (Experiment 2) 12-item word lists involving the following four steps. First, a 12-item word list was presented. Second, each list was followed by an initial task that involved either a PGR test in which the critical word occurred, a PGR test in which 216 basden et al. the critical word did not occur, or no test. As far as the subjects were concerned, a PGR test involved all members of the group. In Experiment 1, nine lists were used, and three lists were tested in each of three conditions (with critical word, without critical word, and no PGR test). Experiment 2 included two new initial recall tests, one with the critical word and one without it. The two new tests were described to the subjects as computer-aided recall (CAR) tests (i.e., the computer’s presentation of the words was purportedly provided to help them to recall more of the remaining words on their own). The true purpose of the CAR tests was to determine whether the influence of the computer’s presentation of the critical word would differ from the influence of the critical word’s purportedly having been recalled by another person. Third, after the initial recall test, each subject completed an individual recall test. The procedure thus involved the following tasks in turn: (1) list presentation, (2) PGR test, CAR test, or no test, and (3) individual recall test. This cycle was repeated for each subject until all lists had been presented. Finally, each subject took a recognition test including items from all of the lists tested. On that test, subjects were asked to render remember or know judgments for recognized items. The recognition test was completed either immediately or after a 1-week delay in Experiment 1 and immediately in Experiment 2. EXPERIMENT 1 Prior research has shown that delay often increases false recall. For example, when Bartlett (1932) and later Bergman and Roediger (1999) had subjects read and recall “The War of the Ghosts,” their memory was increasingly distorted with the passage of time. Schmolck, Buffalo, and Squire (2000) reported that memory distortions increased with delay for events occurring in O. J. Simpson’s criminal trial. Subjects’ memory was first assessed 3 days after the trial. A greater number of memory errors occurred in memory tested 32 months after the trial than in memory tested 15 months after the trial. Spiro (1980) conducted a study that examined the effects of delay on recall of a prose passage. Subjects read the passage once and then attempted to recall it 2 days, 3 weeks, or 6 weeks later. False recall increased as a function of delay. McDermott (1996) analyzed the effect of delay on recall for word lists. Subjects were given five presentations of a 45-item list of words that was created using the Roediger and McDermott (1995) lists. Immediately after each list presentation, a recall test was administered. After completing the five recall tests, subjects returned 48 hours later to complete a delayed recall test. On the de- false memories in groups 217 layed test, McDermott observed a decrease in recall of studied words coupled with an increase in recall of critical words. Thus, delay decreased veridical recall but increased false recall. The influence of delay on false memory has also been studied with recognition tests. Sulin and Dooling (1974) had subjects read brief passages about either a well-known figure (e.g., Helen Keller) or a fictitious person (e.g., Carol Harris). Subjects then completed a recognition test for individual sentences either 5 min or 1 week later. Sulin and Dooling observed more false alarms for sentences involving famous people than nonfamous people. Furthermore, false recognition tended to occur for nonpresented sentences that were related to the famous person (e.g., “She was deaf, dumb, and blind”). Most important for the present purpose, Sulin and Dooling observed a greater level of false alarms to nonpresented sentences after a 1-week delay than after a 5min delay. Thus, as measured by recognition performance, false memory increased with delay. In an experiment that also examined the effect of delay on false recognition, Barclay and Wellman (1986) had six graduate students record autobiographical events for a 4-month period. Over the next 21⁄2 years, the students completed five tests involving recognition memory, temporal ordering, and dating accuracy. Although the students’ confidence in their recognition accuracy remained high throughout all five tests, recognition performance decreased as the retention interval increased. With tests delayed 1 to 3 months, recognition of experienced events decreased, but false recognition of lure items that were similar to experienced events increased. Thus, Barclay and Wellman’s results conform to those reported by Sulin and Dooling (1974). Payne et al. (1996, Experiment 1) examined the effect of delay on both recall and recognition of items with the DRM procedure. Payne et al. had subjects complete a recall test followed by either an immediate or delayed (24-hr) recognition test. They found that nonpresented critical words were recalled and recognized almost as often as studied words, replicating results reported by Roediger and McDermott (1995). On the delayed test, hits to studied words decreased, and false alarms to critical lures remained constant. Thus, although the general conclusion that delay decreases veridical memory was again supported, false recognition did not increase with delay. Their results contradict those reported by Sulin and Dooling (1974) and by Barclay and Wellman (1986), perhaps because the retention interval was shorter. Our hypothesis was that false recognition would increase with the DRM procedures with delay when a retention interval longer than 24 hours was used. The purpose of Experiment 1 was to address the following research questions. First, does the occurrence of the critical word in a PGR test 218 basden et al. increase recall of that word on a subsequent individual recall test? Second, does the PGR test increase subsequent individual recall of the studied associates (true memory)? Third, does a 1-week delay decrease recognition of studied associates (true memory) but increase recognition of critical lures (false memory)? And finally, are phenomenological judgments (remember or know judgments) influenced by others’ recall, by repeated testing, and by delay? METHOD Subjects Forty California State University at Fresno undergraduate psychology students participated to fulfill course requirements. Results obtained from two subjects were excluded from the analyses because they failed to attend the second test session. Materials and design Nine 16-item word lists drawn from Roediger and McDermott’s (1995) study were used. Each list consisted of 15 primary associates to the critical word. For example, for the nonpresented critical word doctor, the first five associates were nurse, sick, lawyer, medicine, and health. From the nine 15-item lists, selecting the first 12 associates to each critical word created nine 12-item sublists. Each 12item sublist was presented in a block during list presentation. The three remaining items in each list (items 13, 14, and 15) were used as lures on the recognition test. The lists were presented via personal computer at the rate of 2 s per word. Immediately after presentation of each of six lists, a PGR test was conducted. During the PGR test, each subject believed she or he was the fourth member of a four-person group in which subjects took turns recalling studied associates. A PGR test began with the occurrence of three list members (words “recalled” by perceived group members) followed by a prompt for the subject to recall a word. This sequence was repeated three times. Thus, during PGR tests, the recall of actual subjects was obtained only in the fourth, eighth, and twelfth output positions. After the remaining three lists were presented, no PGR test was conducted. After each of the nine lists, an individual recall test was administered. This sequence of list presentation, PGR test or no test, and individual recall test was repeated until all nine lists had been presented and individual recall tests conducted. During three of the PGR tests, the critical word corresponding to that list was included randomly as one of the words “recalled” by a perceived group member in one of the first three output positions. On the other three PGR tests, the critical word did not occur. For the remaining three lists, no PGR test was administered. Thus, the three types of initial tests used were PGR test with the critical word, PGR test without the critical word, and no PGR test. After each list presentation and each PGR test, subjects solved simple addition and subtraction problems for 15 s. When no PGR test occurred, the math false memories in groups 219 task after the list presentation was extended to fill the time required for a PGR test. This manipulation controlled the amount of time elapsing between list presentation and the individual recall test in the three initial test conditions. Each subject typed his or her recall at the computer keyboard. After finishing the last of the individual recall tests, subjects completed a 90item recognition test that included remember or know judgments. The recognition test comprised nine 10-item blocks, each corresponding to one of the nine lists. Each 10-item block contained three studied associates, three associated lures (items 13, 14, and 15 from the original 16-item list), three unrelated lures, and one critical lure. Items were randomly ordered within blocks. Procedure All subjects arrived at the test center in groups of four. The initial instructions notified subjects that they were going to be tested as a group on their ability to recall words from a list. They were told that word lists were going to be presented and that they should try to remember as many of the words as they could. Subjects were further informed that after the presentation of some lists, they would be asked to recall the lists as a group and that during these tests, they were to take turns responding. The computers were said to be networked together to permit testing the entire group at once. The subjects were also told that the computer would prompt them when it was their turn to recall a word. After listening to the initial instructions, subjects were each led to a separate test cubicle to begin the testing process. At that time, the nine 12-item lists were presented in blocks, with a PGR test following six of the nine lists. Items were presented at 2-s intervals in descending order of associative strength. The order and condition in which each list was presented were determined randomly for each subject. During each PGR test, subjects were cued for recall by a prompt on the monitor stating, “Recall a word.” To simulate recall typed by others, responses of perceived group members occurred at a random rate of exposure of words and letters within words (i.e., words were exposed one letter at a time). During the test, occasional “typing errors” occurred, and letters were deleted and then replaced to simulate the typing of an actual subject. Presentation rate for words varied from 4 to 7 s. Each PGR test was self-paced. After each of the six PGR tests (and after the initial presentations when no PGR tests occurred), an individual free-recall test was administered before the next list was presented. The individual recall tests were self-paced. Upon completion of the ninth individual recall test, half of the subjects (those in the delayed condition) were dismissed and asked to return 1 week later to complete the experiment. The other half of the subjects heard the directions for the recognition test. These instructions informed subjects that words would appear one at a time on the monitor. For each word, they were to decide whether the word was on one of the original study lists (“old”) or whether the word was not on one of the original study lists (“new”). Subjects were instructed to type “O” for “old” or “N” for “new.” For words that were designated “old,” subjects were then asked to provide a remember or know judgment. Subjects were told to choose remember if they could consciously recollect experiencing the item 220 basden et al. (e.g., they could remember the position the word occupied in the list). Subjects were told to choose know if the word seemed familiar but they could not remember specific information about its occurrence. Several examples were provided. After listening to the directions, subjects completed the recognition test at their own pace and were debriefed. RESULTS Recall Preliminary analyses indicated that during the PGR tests, 100% of the subjects produced a presented associate in the fourth output position, 93% produced a presented associate in the eighth output position, and 45% produced a presented associate in the twelfth output position. The mean proportion of critical words that occurred as intrusions in these output positions during PGR tests was only .03. An anova conducted on the mean proportions of critical words recalled on the subsequent individual recall tests yielded a significant main effect for type of prior test condition, F(2, 72) = 38.88, MSE = .067. Planned comparisons revealed that the differences between the means were statistically significant. The mean proportion of critical words recalled was greatest after the PGR test in which the critical word occurred, M = .79; next greatest after the PGR test in which the critical word did not occur, M = .39; and least when no PGR test was given, M = .30. The difference between the two types of PGR tests shows that the actual presence of the critical word enhanced its subsequent recall more than a PGR test without the critical word. The influence of the critical word introduced on the PGR test reflects social influence on memory. A similar analysis conducted on the mean proportions of studied associates recalled yielded a significant main effect for type of prior group test, F(2, 72) = 15.66, MSE = .003. Subsequent planned comparisons showed that studied associates were recalled better when there was either a prior PGR test in which the critical word occurred, M = .62, or a prior PGR test in which the critical word did not occur, M = .61, than when there was no prior PGR test, M = .30. Recall of studied associates did not differ after PGR tests with and without the critical word. Thus, the PGR test improved recall of studied associates, presumably because true memory can also be enhanced through social influences. Recognition The mean proportions of items recognized in each condition are displayed in Table 1. A significant main effect for type of prior test was observed for critical lures, F(2, 72) = 13.58, MSE = .038. Planned comparisons revealed that the proportion of critical lures recognized dif- false memories in groups 221 Table 1. Mean proportions of items recognized as a function of type of prior PGR test, item type, and time of test in Experiment 1 Type of prior PGR test With critical word Immediate Studied associates Critical lures Associated lures Unrelated lures Delayed Studied associates Critical lures Associated lures Unrelated lures .88 .90 .31 .11 .85 .98 .41 .16 Without critical word .89 .77 .27 .13 .80 .89 .37 .17 No test .81 .58 .27 .06 .66 .83 .35 .21 fered significantly as a function of type of prior recall test (PGR with critical word > PGR without critical word > no PGR test). Thus, social influence increased false recognition of critical words. A similar finding was observed for studied associates. An analysis of variance on the mean proportion of studied associates recognized revealed a significant main effect for type of prior recall test, F(2, 72) = 10.70, MSE = .017. Subsequent planned comparisons revealed that studied associates were recognized more often after PGR tests than when there was no PGR test (PGR with critical word = PGR without critical word > no PGR test). Thus, the PGR test enhanced recognition of studied associates. The effect of delay was examined by comparing the recognition performance of subjects tested immediately with those tested after a 1-week delay. Separate 2 (time of test: immediate or delayed) × 3 (type of test: PGR with critical word, PGR without critical word, no PGR test) anovas were performed on each type of word included on the recognition test: critical lures, studied associates, associated lures, and unrelated lures. The main effect of delay was significant for all item types. The mean proportion of hits to studied associates was greater on the immediate test than on the test after the 1-week delay, F(1, 36) = 8.74, MSE = .029. Conversely, the mean proportion of false alarms to critical lures was less on the immediate test than on the test after the 1-week delay, F(1, 36) = 7.75, MSE = .084. Thus, veridical memory decreased and false memory increased with a 1-week delay. In addition, the mean proportions of false alarms to both associated lures and unrelated lures were greater after a 1-week delay than on the immediate test, F(1, 36) = 4.22, 222 basden et al. MSE = .054 and F(1, 36) = 10.31, MSE = .030, respectively. However, although false memory increased for all lures after a 1-week delay, the mean increase was greater for critical lures, M = .15, than for associated lures, M = .09, and unrelated lures, M = .08. Phenomenological judgments As shown in Table 2, the mean proportions of remember judgments increased for all item types as a function of delay. Remember judgments occurred less often on the immediate test than on the delayed test for studied associates, F(1, 36) = 10.53, MSE = .153; critical lures, F(1, 36) = 7.20, MSE = .032; associated lures, F(1, 36) = 4.30, MSE = .028; and unrelated lures, F(1, 36) = 8.61, MSE = .016. Thus, subjects claimed greater conscious recollection of all item types after a 1-week delay than after no delay. Although remember judgments increased for all item types after a 1-week delay, the difference was more pronounced for critical lures, M = .13, and associated lures, M = .14, than for studied associates, M = .05, and unrelated lures, M = .01. DISCUSSION The result of greatest interest in this study was that critical words thought to have been recalled by others strongly influenced individual recall performance of the members of the group. Thus, the frequency of false recall increased when group members were perceived to have mentioned events that never happened. Furthermore, once those critical words were recalled on the individual tests, they were subsequently Table 2. Mean proportions of remember judgments as a function of type of prior PGR test, item type, and time of test in Experiment 1 Type of prior recall test With critical word Immediate Studied associates Critical lures Associated lures Unrelated lures Delayed Studied associates Critical lures Associated lures Unrelated lure .29 .54 .13 .04 .30 .57 .26 .06 Without critical word .29 .36 .12 .11 .32 .65 .27 .06 No test .20 .36 .16 .10 .30 .54 .22 .10 false memories in groups 223 recognized and judged to be remembered (consciously recollected). The significance of these findings is that it shows that one’s memory of events is strongly influenced by others’ recollections. There was a general deterioration in recognition performance when the length of the retention interval was increased. This finding is similar to results obtained by researchers studying the effects of delay on memory (Bartlett, 1932; Bergman & Roediger, 1999). More importantly, recognition of critical words increased with delay, supporting our hypothesis that false recognition increases if the delay is long enough (cf. Payne et al., 1996). The presence of critical words on PGR tests did not appear to be critical to the increase in false alarms with delay. However, because false alarms increased from .90 to .98 when the critical word had occurred on the PGR tests, ceiling effects might have obscured that effect. Increased errors in recognition were associated with increased remember judgments. These results suggest that delayed recognition and delayed remember judgments may be less likely to draw on memory of specific detail than do their immediate counterparts. In concluding that being in a social group fosters transmission of false memory, we are assuming that subjects believed that they actually were working in groups. The primary purpose of Experiment 2 was to test the validity of our assumption. Subjects may have seen through the deception and thought that the word list was simply being presented an additional time. To determine whether this was the case, we compared the influence of presentation of the critical word by computer with the influence of inclusion of the critical word in a perceived group test. McDermott (1997) demonstrated that the likelihood of recall of the critical word was much greater when it was presented than when it was not presented. Therefore, the increased recall of the critical word after PGR tests that included the critical word in Experiment 1 may simply reflect the influence of that item’s actual presentation.1 If subjects did not believe that they were working in groups, then the critical word might have been perceived as having been presented. Experiment 2 was conducted in an attempt to differentiate between these two possible explanations for our results (social influence vs. subjects’ propensity to recall previously viewed critical words). EXPERIMENT 2 The primary aim in conducting Experiment 2 was to determine whether subjects in Experiment 1 believed that the critical words they had viewed during the PGR tests were actual subject responses and not 224 basden et al. computer-generated responses (i.e., did they “see through” the deception?). This goal was achieved by introducing two new test conditions. These test conditions were a computer-aided recall (CAR) test that included the critical word and a CAR test that did not include the critical word. The rationale for the addition of these tests was the following: If differences were observed between the number of critical words recalled after a PGR test and after a CAR test, then subjects probably believed that the PGR test involved actual subjects. METHOD Subjects Forty-eight California State University at Fresno undergraduate psychology students participated in Experiment 2 as part of a course requirement. The results of five subjects were excluded from the analyses because of their failure to follow instructions. Materials and design Experiment 2 was identical to Experiment 1 with two exceptions. First, two new recall test conditions were added to the design of Experiment 1. These two test conditions were a CAR test that included the critical word and a CAR test that did not include the critical word. The CAR tests were similar to the PGR tests described in Experiment 1, the difference being that subjects were told that the computer would present some of the words they were supposed to remember to help them recall the entire list. Accordingly, fifteen 12-item lists were presented in Experiment 2 (rather than the nine lists in Experiment 1). Furthermore, the addition of the two new test conditions increased the number of items on the recognition test from 90 (nine 10-item blocks, Experiment 1) to 150 (fifteen 10-item blocks, Experiment 2). With the exception of the number of lists used, the lists, recall tests, and recognition tests were presented as they were in Experiment 1. The second difference between the two experiments is that delay was not manipulated in Experiment 2. Procedure The procedure followed in Experiment 2 was identical to that of Experiment 1 with three exceptions. First, instructions explaining the CAR tests were added. These instructions informed subjects that after they viewed some lists of words, the computer would aid them in recalling those lists. Subjects were told that the computer would present three of the words that had been studied and would then prompt them to recall a word and then repeat this cycle twice more. Second, during PGR tests, the letters of the words recalled by bogus group members appeared individually, and typing errors occasionally occurred, as in Experiment 1. During CAR tests, each word appeared as a unit at a fixed rate (5 s) of presentation. Third, all subjects completed the 150-item recognition test immediately after completing the last of the individual recall tests. false memories in groups 225 Thus, in Experiment 2, subjects studied 15 lists. The presentation of each list was immediately followed by a math task. In turn, the math task was immediately followed by one of five types of recall tests. These were a PGR test with the critical word, a PGR test without the critical word, a CAR test with the critical word, a CAR test without the critical word, and no initial recall test. Immediately afterward, each subject took an individual recall test over the last list presented. Lists were tested in this manner until the completion of the fifteenth individual recall test. After that, subjects completed the 150-item recognition test and completed remember or know judgments for recognized words. Subjects were debriefed and received credit for participation after completing the final task. RESULTS Recall Recall performance during the PGR and CAR tests followed a pattern similar to that observed in Experiment 1. Overall, 100% of the subjects produced a presented associate in the fourth output position, 91% produced a presented associate in the eighth output position, and 52% produced a presented associate in the twelfth output position. The mean proportion of critical words recalled as intrusions on the PGR or CAR tests was only .02. The mean proportions of words recalled on the subsequent individual recall tests are shown in Table 3. A one-way anova conducted on the mean proportion of critical words recalled yielded a significant main effect for type of prior test, F(4, 168) = 28.52, MSE = .069. Notably, critical words were recalled more often after their inclusion in CAR tests than after their inclusion in PGR tests, F(1, 42) = 6.03, MSE = .069. This Table 3. Mean proportions of items recalled as a function of prior recall test and item type in Experiment 2 Type of prior recall test With critical word Perceived group Studied associates Critical words Computer-aided Studied associates Critical words .62 .58 .57 .72 Without critical word .61 .26 .62 .22 No test .50 .38 .50 .38 226 basden et al. result supports the conclusion that subjects believed that the PGR tests contained the recall of others in the group. In addition, planned comparisons showed that subjects were more likely to recall critical words after their inclusion in a recall test than after their noninclusion in a recall test. The mean proportion recalled after a CAR test with the critical word present was significantly greater than after a CAR test with the critical word absent, F(1, 42) = 81.32, MSE = .067. Similarly, the mean proportion recalled after a PGR test with the critical word present was significantly greater than after a PGR test with the critical word absent, F(1, 42) = 32.90, MSE = .066. The results with PGR tests replicate those of Experiment 1 and suggest that the actual occurrence of the critical word in a recall test makes subsequent false recall of that word more likely, whether that critical word is perceived as recalled by another group member or is perceived as having been presented again by the computer. Finally, critical words were recalled more often after PGR tests and CAR tests than after no test. Recognition The mean proportions of recognized words for each item type are displayed in Table 4. An anova conducted on the mean proportion of hits to studied associates revealed a significant main effect of type of prior test, F(4, 168) = 5.47, MSE = .022. Studied associates were recognized more often after their inclusion in either PGR or CAR tests than when there was no prior test, F(1, 42) = 14.88, MSE = .023 and F(1, 42) = 6.19, MSE = Table 4. Mean proportions of items recognized as a function of prior recall test and item type in Experiment 2 Type of prior recall test With critical word Perceived group Studied associates Critical lures Associated lures Unrelated lures Computer-aided Studied associates Critical lures Associated lures Unrelated lures .86 .86 .16 .09 .84 .88 .13 .07 Without critical word .86 .64 .21 .12 .79 .55 .06 .07 No test .73 .59 .10 .01 .73 .59 .10 .01 false memories in groups 227 .023, respectively. This finding replicates the effects for studied associates observed in Experiment 1. A significant effect of type of prior test was also observed for critical lures, F(1, 42) = 15.82, MSE = .064. Critical lures were falsely recognized more often after both PGR tests and CAR tests than when there was no prior test, F(1, 42) = 8.97, MSE = .054 and F(1, 42) = 7.80, MSE = .042, respectively. False recognition of nonstudied associates and of unrelated words was also greater after PGR and CAR tests than when there had been no test, as shown in Table 4. Phenomenological judgments Mean proportions of remember judgments are shown in Table 5. An anova conducted on the mean proportions of remember judgments to critical words revealed a significant main effect for type of prior test, F(4, 168) = 8.23, MSE = .076. Subjects claimed to consciously remember critical words more often when they had occurred on the prior PGR or CAR test than when they had not, F(1, 42) = 19.75, MSE = .095. Furthermore, the effect of type of prior test was statistically significant for associated lures and unrelated lures, F(1, 42) = 5.81, MSE = .008 and F(1, 42) = 8.86, MSE = .004, respectively. Subjects claimed to consciously remember associated lures and unrelated lures more often after PGR and CAR tests than when there had been no prior test. Planned comparisons showed that remember judgments given to critical words were greater when those words had occurred on the prior test, either PGR or CAR, than when they had not occurred on the prior test, F(1, 42) = 19.75, MSE = .095. Table 5. Mean proportions of remember judgments as a function of prior recall test and item type in Experiment 2 Type of prior recall test With critical word Perceived group Studied associates Critical lures Associated lures Unrelated lures Computer-aided Studied associates Critical lures Associated lures Unrelated lures .52 .47 .08 .06 .51 .59 .06 .04 Without critical word .54 .34 .12 .08 .50 .30 .05 .02 No test .48 .33 .04 .00 .48 .33 .04 .00 228 DISCUSSION basden et al. In summary, the occurrence or nonoccurrence of the critical word in a PGR or CAR test affected subjects’ later individual recall and recognition performance. Subjects were more likely to falsely recall and recognize a critical word if that word had previously been included on a prior test than if the word had not been included. Furthermore, subjects were more likely to claim to consciously remember a critical word if that word had been interjected than if it had not. We also found that PGR and CAR tests increased the recall and recognition of studied associates. Perhaps most importantly, a difference was observed between the number of critical words recalled after PGR tests and CAR tests. Subjects were less likely to incorporate the critical word in their recall when it was recalled by a perceived group member than if it was presented by the computer. When given PGR tests, subjects apparently believed that they were working in groups and discounted the recall of other group members. However, false recall and recognition levels after the critical word occurred during PGR tests were comparable to those in Experiment 1. Thus, the findings from Experiment 2 are consistent with those of Experiment 1 and support the idea that memory errors are indeed subject to social influence. GENERAL DISCUSSION There were four major findings in the present experiments. First, subjects were more likely to falsely recall a critical word if that word had previously occurred on a PGR or CAR test than if that word had not occurred. Second, prior participation in either a PGR test or CAR test increased subsequent remembering of both studied associates and critical words. Third, delay of the recognition test reduced hits to studied words but increased false alarms to critical lures. And fourth, the proportion of remember judgments to critical lures increased after a 1-week delay. The primary goal of the present experiments was to determine whether the recall of a critical word by a group member would influence each subject’s memory as measured by recall, recognition, and phenomenological judgments. These experiments provide strong evidence that the occurrence of a critical word does affect all these measures. One alternative explanation for the increase in false memory for critical words after their introduction by bogus group members is that the subjects may not have believed that they were working in groups. That being the case, one could propose that the increased level of false memory resulted not false memories in groups 229 from social influence but from subjects’ general tendency to recall previously viewed critical words (e.g., McDermott, 1997). However, if this were the case, one would expect equivalent levels of false memory for critical words presented by the computer and those included by a supposed group member. In fact, this was not what was observed. False memories for critical words were greater after their presentation by the computer than after their occurrence in a PGR test. Our results are consistent with previous studies suggesting that social factors can strongly affect memory (e.g., Basden et al., 1997, 1998; Deutsch & Gerard, 1955; Roediger et al., 2000). Participation in an inaccurate group (albeit a perceived group) subsequently induced higher levels of false recall and false recognition than did participation in an accurate group. Deutsch and Gerard showed that the influence of perceived groups is less than that of face-to-face groups when tests are administered in the group setting. However, the present research shows that the transmission of false memory in perceived groups does persist on subsequent individual recall tests. Furthermore, the amount of false information transmitted appears comparable to that observed by Roediger et al. in face-to-face groups. When Meade and Roediger (in press) compared social contagion effects as a result of exposure to a live confederate with those of an implied one, they found that recall was unaffected by the difference. Presumably the influence of the introduction of the critical item by a live confederate would be less than that of actual presentation of the critical item. Given that the deception was successful, one explanation for the high levels of false memory observed is that offered by fuzzy trace theory (Brainerd & Reyna, 1990; Brainerd, Reyna, & Brandse, 1995; Reyna & Brainerd, 1995). According to fuzzy trace theory, subjects’ gist trace for the presented lists may be enhanced by the actual presentation of a critical word. Assuming that the critical word embodies the central theme of each list, one would expect its actual occurrence on a test to lead to high levels of false memory for critical words. This would prove especially true after a period of delay in which the more precise but less enduring verbatim traces had faded. Thus, fuzzy trace theory does a nice job of predicting high levels of false recognition for critical words after a 1-week delay. Yet fuzzy trace theory would also incorrectly predict greater recall of studied associates and continued recognition of studied associates with delay when the critical word appeared on PGR tests than when it did not appear. Fuzzy trace theory also appears unable to account for the increase in remember judgments to critical words after a 1-week delay observed in Experiment 1. This is the case because fuzzy trace theory predicts that verbatim traces fade more quickly than do gist 230 basden et al. traces. However, in Experiment 1, subjects claimed to remember more verbatim information about critical words after a 1-week delay than after no delay. In the absence of methodological or theoretical problems associated with the remember or know judgment test itself, another explanation for the occurrence of the observed results must be sought. Such an explanation can be found in source-monitoring theory (Johnson, Hashtroudi, & Lindsay, 1993; Johnson & Raye, 1981). According to source-monitoring theory, the high levels of false recognition of critical words after a 1-week delay may be a result of subjects’ inability to determine the source of their memories. That is, subjects may not be able to decide whether they saw the critical word on the original list or during the course of a subsequent recall test. That being the case, one would predict higher levels of false memory for critical words that had been presented than for critical words that had not been presented. That is precisely what was observed in the present experiments. In addition, it is interesting to note that re-experiencing the studied associates had less effect on false memory than did the occurrence of a critical word. This finding further supports the idea that subjects were likely to confuse their memories of critical words included on the recall tests with their memories of originally presented associates. Given such an explanation, how could one account for the increase in false recognition for all lures after a 1-week delay, the decrease in correct recognition of studied associates after a 1-week delay, and the increase in remember judgments to critical words after a 1-week delay? We believe that the answer to this question stems from a general decrement in source memory after a 1-week delay. Such a conclusion is consistent with the observed data and with previous research on the effects of delay on memory (e.g., Bergman & Roediger, 1999; Spiro, 1980). These results extend the findings reported by Basden et al. (1997, 1998) and by Weldon and Bellinger (1997) to show that although collaborative remembering may increase the occurrence of intrusions in recall, collaboration also has the serious consequence that once a false memory occurs within the group, it is likely to be adopted as a true memory by the remaining group members. These results are analogous to those obtained with the eyewitness testimony research conducted by Loftus (1993). Misleading postevent information introduced in the form of questions or a narrative presumably authored by an “expert,” such as the experimenter, may distort remembering. The present research expands on these findings by showing that postevent information introduced by a nonexpert, such as a classmate, can profoundly influence memory as well. However, the expert’s influence, as seen in the CAR test in Experiment 2, may be greater than a classmate’s influence, as seen in the PGR test. false memories in groups Notes 231 Matt Reysen is now at Department of Psychology, Purdue University. This research was presented previously at the 39th Annual Meeting of the Psychonomic Society, Dallas, TX (November 1998). We deeply appreciate the helpful comments provided by Mary Sue Weldon. Correspondence about this article should be addressed to Barbara H. Basden, Department of Psychology M/S PH 11, California State University, Fresno, CA 93740–8019 (e-mail: barbb@csufresno.edu). Received for publication May 4, 2000; revision received November 20, 2000. 1. It must be recognized that the studied associates are presented twice. They are presented a second time during the PGR test. The critical word occurs only once, during the PGR test. References Asch, S. E. (1951). Effects of group pressure upon the modification and distortion of judgments. In H. Guetzkow (Ed.), Groups, leadership and men (pp. 177–190). Pittsburgh: Carnegie Press. Barclay, C. R., & Wellman, H. M. (1986). Accuracies and inaccuracies in autobiographical memories. Journal of Memory and Language, 25, 93–103. Bartlett, F. C. (1932). Remembering: A study in experimental and social psychology. Cambridge, England: Cambridge University Press. Basden, B. H., Basden, D. R., Bryner, S., & Thomas, R. L., III. (1997). A comparison of group remembering: Does collaboration disrupt retrieval strategies? Journal of Experimental Psychology: Learning, Memory, and Cognition, 23, 1176–1189. Basden, B. H., Basden, D. R., Thomas, R. L., III, & Souphasith, S. (1998). Memory distortion in group recall. Current Psychology, 16, 225–246. Bergman, E. T., & Roediger, H. L., III (1999). Can Bartlett’s repeated reproduction experiments be replicated? Memory & Cognition, 27, 937–947. Betz, A. L., Skowronski, J. J., & Ostrom, T. M. (1996). Shared realities: Social influence and stimulus memory. Social Cognition, 14, 113–140. Brainerd, C. J., & Reyna, V. F. (1990). Gist is the grist: Fuzzy trace theory and the new intuitionism. Developmental Review, 10, 3–47. Brainerd, C. J., Reyna, V. F., & Brandse, E. (1995). Are children’s false memories more persistent than their true memories? Psychological Science, 6, 359– 364. Deese, J. (1959). On the prediction of occurrence of particular verbal intrusions in immediate recall. Journal of Experimental Psychology, 58, 17–22. Deutsch, M., & Gerard, H. B. (1955). A study of normative and informational social influences upon individual judgment. Journal of Abnormal Psychology, 51, 629–636. Johnson, M. K., Hashtroudi, S., & Lindsay, D. S. (1993). Source monitoring. Psychological Bulletin, 114, 3–28. Johnson, M. K., & Raye, C. L. (1981). Reality monitoring. Psychological Review, 88, 67–85. 232 basden et al. Loftus, E. F. (1993). Made in memory: Distortions in memory after misleading communications. In D. L. Medin (Ed.), The psychology of learning and motivation: Advances in theory and research. (pp. 187–215). New York: Academic Press. McDermott, K. B. (1996). The persistence of false memories in list recall. Journal of Memory and Language, 35, 212–230. McDermott, K. B. (1997). Priming on perceptual implicit memory tests can be achieved through presentation of associates. Psychonomic Bulletin & Review, 4, 582–586. Meade, M. L., & Roediger, H. L., III. (in press). Explorations in the social contagion of memory. Journal of Experimental Psychology: General. Payne, D. G., Elie, C. J., Blackwell, J. M., & Neuschatz, J. S. (1996). Memory illusions: Recalling recognizing, and recollecting events that never occurred. Journal of Memory and Language, 35, 261–285. Reyna, V. F., & Brainerd, C. J. (1995). Fuzzy trace theory: An interim synthesis. Learning and Individual Differences, 7, 1–75. Roediger, H. L., III, & McDermott, K. B. (1995). Creating false memories: Remembering words not presented in lists. Journal of Experimental Psychology: Learning, Memory, and Cognition, 21, 803–814. Roediger, H. L., III, Meade, M. L., & Bergman, E. T. (2001). Social contagion of memory. Psychonomic Bulletin & Review, 8, 365–371. Schmolck, H., Buffalo, E. A., & Squire, L. R. (2000). Memory distortions develop over time: Recollections of the O. J. Simpson trial verdict after 15 and 32 months. Psychological Science, 11, 39–45. Schneider, D. M., & Watkins, M. J. (1996). Response conformity in recognition testing. Psychonomic Bulletin & Review, 3, 481–485. Spiro, R. J. (1980). Accommodative reconstruction in prose recall. Journal of Verbal Learning and Verbal Behavior, 19, 84–95. Sulin, R. A., & Dooling, D. J. (1974). Intrusion of a thematic idea in retention of prose. Journal of Experimental Psychology, 103, 255–262. Tulving, E. (1985). Memory and consciousness. Canadian Psychologist, 26, 1–12. Weldon, M. S., & Bellinger, K. D. (1997). Collective memory: Collaborative and individual processes in remembering. Journal of Experimental Psychology: Learning, Memory, and Cognition, 23, 1160–1175.