1 Forgetting (and Recovering) the Seemingly Unforgettable Steven M. Smith, David R. Gerkens1, Hyun Choi, & Sarah C. Moynan2 Department of Psychology, Texas A&M University, College Station, Texas, 77843, USA 1 Present address: Department of Psychology, Mississippi State University, Mississippi State, MS, 39762, USA 2 Present address: Department of Psychology, Washington University, St. Louis, MO, 63130-4899, USA Seven experiments demonstrate substantial forgetting and memory recovery effects using a behavioral methodology designed to produce strong retrieval bias. The procedure, which strengthens memories of a large subset of items (fillers) presented in an experiment, reduced recall of the unstrengthened subset (targeted critical items) as much as 63% on a free recall test. On a subsequent cued recall test, appropriate cues produced nearly complete recovery; cued memories for subjects in the Forget group were as accurate as the Control group. Although free recall was strongly influenced, implicit memory was not affected by the forgetting procedure. Distinctive and emotionally-charged materials (e.g., expletives, death-related words) were quite susceptible to the forgetting and recovery effects. Tests using independent probes1 repeatedly failed to find that these forgetting effects involved an inhibitory mechanism. The findings thus demonstrate a forgetting effect that is quite large, is not caused by conscious intentions to forget, operates even with distinctive, emotional materials with sexual and violent content, shows implicit memory of explicitly forgotten material, can be reversed with appropriate cues (memory recovery), and does not appear to involve inhibition.2 Some things may seem unforgettable. Can memorable events be blocked from consciousness, and if so, can they be subsequently recovered? The notion that repressed memories can be blocked, and later recovered has been challenged in the false memory debate2-4; abundant findings of false memories5-7 leave open the possibility that seemingly recovered memories might simply be fabrications of a reconstructive long-term memory system. Laboratory research has provided little evidence of high levels of potentially reversible forgetting, particularly with emotional and distinctive materials that are typically well-remembered. Here we report repeated findings of high levels of blocked and recovered memories, even for materials that have sexually explicit and violent content. A widely-held view of repression8-10 involves forgetting of traumatic experiences that, under normal circumstances, might seem unforgettable. The mechanisms Freud postulated for such forgetting and recovery effects are not under scrutiny here; we focus instead on cognitive mechanisms, inhibition, retrieval bias, and cueing, as causes of these putative phenomena. Anderson & Green1 characterised repression as occurring in stages, the first involving deliberate suppression of unwanted memories, giving rise to inhibition of those memories, and the second involving a habitual pattern of sustained inaccessibility of unwanted memories. Their data support the first stage of their theory, but their claim that inhibitory control of memories could provide a basis for repression has been criticized on several counts11,12, including: (1) Recall following inhibition remains fairly high, (2) The mechanism involved is not unconscious, as might be expected in repression, (3) No evidence shows that inhibited memories are recoverable, (4) There is no evidence that inhibited memories have implicit effects (i.e., what Freud10 called the “return of the repressed”), and (5) There is no evidence of high levels of inhibition with materials with emotional content, such as aggressive or sexually explicit material. The present study addresses these criteria, focusing on the second stage of3 Anderson & Green's theory, and the results show that our approach provides useful empirical tools for studying these aspects of repression. The present experiments used a laboratory procedure called the Retrieval Bias method, which reliably produces habitual patterns of inaccessibility13. In each experiment two groups of subjects view many categorized lists of words (21 in Experiments 1-4, 24 in Experiments 5-7), such as "TOOLS: hammer, screwdriver...," with 10 words per list. The Control group sees no further lists before a free recall test, whereas the Forget group has several additional exposures to all except the critical lists before free recall. Both Control and Forget groups see critical lists only once (Fig. 1). Forget subjects recalled far fewer names of critical lists than did Control subjects (p < .001 for all experiments; see Fig. 2). The size of this forgetting effect (i.e., the difference in free recall between Control and Forget conditions, Fig. 2, panel b) is greater than forgetting effects found in other paradigms, including directed forgetting, retrieval-induced forgetting, and the think/no-think method (Fig. 2, panel a). Forget conditions often exhibited very low levels of recall; for example, the "Death" list was recalled by only 9% of the subjects in Experiment 7, as compared to 72% in the Control condition. Even a list of curse words, judged by an independent group of subjects as highly emotional, distinctive, and memorable, and which was recalled by an average of 77% of the Control group in Experiments 5, 6, and 7, was recalled by only 45% of subjects in the Forget group. Furthermore, our forgetting effect is reversible. When cued with independent probes (e.g., “carpenter-T____”) in Experiments 1, 2, and 3, high levels of critical items were recalled in both Forget and Control conditions. When category names (e.g., "TOOLS") were given as cues in Experiments 5, 6, and 7, equal numbers of list members (e.g., "hammer") were recalled in both conditions (F < 1 in all experiments).4 In spite of the sizeable forgetting effects in free recall, the forgotten items were remembered equally well on a subsequent test when appropriate cues were provided (Fig. 3). Are these forgetting effects due to inhibition of critical items? We tested this possibility using the independent probe method1,18. Experiments 1, 2, and 3 found that when cued by independent probes (e.g., "carpenter-T_____"), subjects in Control and Forget conditions recalled critical list names (e.g., "TOOLS") equally well (all F’s < 1). Performance on independent probe tests (Fig. 3, panel a) was no