Making remembering more memorable

Transcription

1 Memory ISSN: (Print) (Online) Journal homepage: Making remembering more memorable Marcus L. Leppanen & Keith B. Lyle To cite this article: Marcus L. Leppanen & Keith B. Lyle (2018) Making remembering more memorable, Memory, 26:7, , DOI: / To link to this article: Published online: 01 Jan Submit your article to this journal Article views: 109 View related articles View Crossmark data Full Terms & Conditions of access and use can be found at

2 MEMORY 2018, VOL. 26, NO. 7, Making remembering more memorable* Marcus L. Leppanen and Keith B. Lyle Department of Psychological and Brain Sciences, University of Louisville, Louisville, KY, USA ABSTRACT Memory retrieval is a cognitive operation that itself can be remembered or forgotten, with potentially important consequences. To study memory for prior remembering, we had participants first study target words (e.g., bark) alongside semantically related cue words (e.g., dog). Then, on Test 1, participants retrieved targets in response to either the study cue or a changed cue that was semantically related to a homograph of the target (e.g., birch). Finally, on Test 2, participants retrieved all targets in response to the original study cues, and participants judged whether targets were previously retrieved on Test 1. As in previous research, cue change on Test 1 rendered target retrievals less memorable, suggesting context changes harm memory for prior remembering. We hypothesised that the negative effect of context change could be ameliorated by reminding participants of the original study cues during Test 1. We had participants either retrieve (Experiments 1 and 3, Ns = 46 and 62) or view (Experiment 2, N = 118) the study cue following each target retrieval. Reminding significantly reduced the negative effect of cue change, with self-generation being especially potent. This indicates that reminding can make remembering more memorable in the face of context change. ARTICLE HISTORY Received 27 June 2017 Accepted 21 December 2017 KEYWORDS Memory for prior remembering; reminding; context; cued recall Think about the last memory you retrieved before reading this paper. Was it a memory you had retrieved many times before (e.g., of an important conversation you had with a colleague last week and have been thinking about ever since) or one you were recalling for the first time? Recalling previous occasions of memory retrieval has been called remembering prior remembering (Arnold & Lindsay, 2002). Failure to remember prior remembering can have various consequences. In everyday life, for example, forgetting prior remembering might lead people to repeatedly share the same memory with a conversational partner an experience that can elicit emotional reactions ranging from mere embarrassment to serious doubt about one s cognitive health. A more dramatic example of what can happen when people forget prior remembering is provided by Schooler (2001), who documented cases in which victims of childhood sexual abuse forgot instances of remembering their abuse. This led individuals to falsely believe that they had repressed memories for long periods of time an unsettling distortion of selfknowledge. In addition to personal consequences, remembering prior remembering also has implications for research on studies of involuntary and intrusive memories (Bernsten, 2010), which often rely on participants self-reports of whether and how often events have been remembered. Unless people document their recollections immediately after they happen, self-reports will necessarily have a retrospective aspect and be dependent on people s ability to remember prior remembering (for related commentary, see Hintzman, 2011). Forgetting prior remembering therefore poses a threat to the validity of research on involuntary and intrusive remembering (for related discussion, see Takarangi, Strange, & Lindsay, 2014). The variety of potential consequences associated with forgetting prior remembering motivates us to understand the circumstances under which it is more or less likely to occur. Prior research indicates that one cause of forgetting prior remembering is change in retrieval context over time. Evidence comes from the dominant laboratory paradigm used to study memory for prior remembering (Arnold & Lindsay, 2002). This paradigm comprises a study phase followed by two distinct test phases. First, participants are presented with semantically related cue-target pairs (e.g., hand palm, dog bark). Test 1 comprises cued recall of target words. Cues for some targets are the same as at study (e.g., hand is the cue for palm) while cues for other targets are changed to a word that is semantically related to a homograph of the target (e.g., dog is changed to birch). Other targets are not tested at all in this phase. Test 2 comprises a second round of cued recall. All cues are the same as at study and all targets are tested. The manipulation of cue identity on Test 1 CONTACT Marcus L. Leppanen marcus.leppanen@louisville.edu Department of Psychological and Brain Sciences, University of Louisville, 2301 S. 3rd Street, Louisville, KY 40292, USA *Results of this research were presented, in part, at the fifty-sixth annual meeting of the Psychonomic Society Informa UK Limited, trading as Taylor & Francis Group

3 MEMORY 961 means that, for the targets that were tested, there is either a semantic change in retrieval context between tests or there is not. After attempting retrieval of each target on Test 2, participants are asked whether the target was also retrieved on Test 1; this constitutes the critical test of memory for prior remembering. The key finding is that participants are less likely to remember retrieving targets on Test 1 when the cue changed between tests versus remained the same (sometimes called the forgot-it-allalong effect), suggesting that semantic changes in retrieval context harm memory for prior remembering (Arnold & Lindsay, 2002, 2005; Raymaekers, Peters, Smeets, Abidi, & Merckelbach, 2011). Negative effects of retrieval-context change are not particularly surprising if we assume, as Arnold and Lindsay (2002) did, that memory for prior remembering is governed by the same principles that govern memory for other types of events. One of these principles is that events are more likely to be remembered when retrieval contexts closely resemble encoding contexts (e.g., Godden & Baddeley, 1975; Smith, Glenberg, & Bjork, 1978; Tulving & Thomson, 1973). In Arnold and Lindsay s(2002) procedure, the critical to-be-remembered events are instances of memory retrieval that occur during Test 1. These are encoded in the specific context of Test 1 cues and, presumably, neural activity associated with acts of retrieval is bound, via medial-temporal-lobe processes (Squire & Zola, 1996), to Test 1 cues. The retrieval context for these acts is established by presentation of Test 2 cues. The reason changed cues (versus ones that are the same as on Test 1) are often ineffective at providing access to previously encoded retrieval acts is presumably that changed cues were not present during Test 1 and therefore were not bound to retrieval-related neural activity that occurred on Test 1, essentially violating the encoding-specificity principle (Tulving & Thomson, 1973). The encoding-specificity perspective described above suggests a means of increasing memory for prior remembering, at least under the conditions typically investigated using Arnold and Lindsay s (2002) procedure. In that procedure, as previously stated, Test 2 cues for all targets match study-phase cues. For example, if bark is initially studied with the cue dog, then the Test 1 cue may be either dog or birch, but the Test 2 cue will invariably be dog. Given this setup, suppose individuals were reminded of the study cue (i.e., dog) after retrieving bark during Test 1. Although reminding could happen on both sameand changed-cue trials, it is on changed-cue trials that it would presumably be most impactful because it would provide an opportunity to bind the study cue to the act of memory retrieval. This opportunity otherwise does not exist on changed-cue trials because the study cue never appears (unlike on same-cue trials, wherein the study cue and retrieval act may be bound together during the cued-recall attempt itself). Reminders could allow study cues, when presented on Test 2, to effectively cue memory for Test 1 retrievals, even when retrievals initially occurred in the context of a changed cue. Hence, our proposal is that reminding people of the study cue immediately following the Test 1 act of remembering might improve memory for prior remembering in the face of retrieval-context change (provided the study cue is reinstated on Test 2). A hypothetical real-world analogy might make our proposal easier to understand. Imagine you have an interesting experience with a waiter at Restaurant A (the encoding/study context). You later tell the story of that experience to a friend at Restaurant B, without dwelling on contextual details specific to Restaurant A (analogous to a changed Test 1 retrieval context). Later still, you and the same friend end up together at Restaurant A. Being back in Restaurant A might make you think of the experience you had with the waiter there (analogous to Test 2 retrieval in the original study context). Due to the change in retrieval context between the restaurants, you might not remember previously retrieving the story. Consequently, you might launch into a retelling of a story your friend has already heard. Such an accidental retelling would be less likely, we argue, if you had been specifically reminded of Restaurant A soon after you initially retrieved the story in Restaurant B (analogous to a study-cue reminder during Test 1 retrieval). We chose to examine the role of reminding in memory for prior remembering because it seems likely that, when remembering occurs under conditions of retrievalcontext change in everyday settings, it is sometimes accompanied by a reminder of the original context and sometimes not. For example, events encoded in an upsetting emotional context might be remembered in a calmer emotional context after the crisis has passed. When recalling the events, someone who experienced them with us may or may not be present to remind us of our original emotional state. A thorough understanding of the effects of retrieval-context change on memory for prior remembering should therefore take into account the presence or absence of reminders. The proposal that reminding might influence memory for prior remembering is consonant with recent arguments that reminding has wide-ranging implications for human cognition (Benjamin & Ross, 2010; Hintzman, 2011; Jacoby & Wahlheim, 2013; Putnam, Sungkhasettee, & Roediger, 2017; Wahlheim & Jacoby, 2013). As the cited authors have pointed out, when current events remind people of previous ones, it can have various beneficial effects, such as improving the quality of recency judgments, producing proactive facilitation (versus interference), and protecting people from the misinformation effect. If reminding also improves memory for prior remembering, it would support Arnold and Lindsay s (2002) assumption that remembering prior remembering operates on the same principles as remembering other types of information and it would provide additional evidence for what Jacoby and Wahlheim called the importance of looking back for human cognition.

4 962 M. L. LEPPANEN AND K. B. LYLE Our prediction that a reminder-based intervention will increase memory for prior remembering rests on the assumption that Arnold and Lindsay s (2002) procedure does not ordinarily involve a high rate of involuntary reminding. People are often reminded of past events involuntarily (Hintzman, 2011). One could imagine that, whenever people retrieve a target from the study phase, they are automatically reminded of the study cue, even if the Test 1 cue is different (e.g., being reminded of dog when bark is retrieved in response to birch). This sort of involuntary reminding might be expected based on the associativesymmetry hypothesis (Kahana, 2002), according to which A B word pairs (such as dog-bark) are represented holistically in memory and therefore B can bring to mind A, just as A can bring to mind B. If involuntary reminding normally occurs during Test 1 target retrieval, then explicit reminders of study cues during Test 1 would be redundant and there would be little reason to expect them to improve memory for prior remembering. We reasoned, however, that involuntary reminding is probably uncommon in Arnold and Lindsay s (2002) procedure, at least when Test 1 cues are changed from study. The changed-cue condition was intentionally developed to create a situation in which participants would think about target words differently on Test 1 than during study (i.e., participants were intended to think about the targets homographs on Test 1). We reasoned that activation of a target s homograph with its distinct conceptual and semantic features is unlikely to prompt spontaneous retrieval of the study cue (e.g., when thought of as part of a tree, bark is unlikely to remind people of the sound a dog makes). We therefore assumed that bringing reminding under task control (Jacoby & Wahlheim, 2013) would be necessary to make participants think of study cues during Test 1. The present experiments Three experiments (conducted in the order of report) utilised Arnold and Lindsay s (2002) basic procedure, but with additions to the Test 1 phase (see Figure 1 for a graphical depiction of our methodology). Our central question was whether the negative effect of Test 1 cue change on memory for prior remembering could be ameliorated by reminding participants of the study cue during their first attempt at retrieving a target. Typically, the Test 1 phase in studies utilising Arnold and Lindsay s procedure only requires participants to recall targets, with either the same cue as at study or a changed cue. To bring reminding under task control in the present research, we added a follow-up task to each Test 1 cued-recall attempt. The exact nature of this task differed across experiments. Details are provided in the Introduction and Method section of each experiment. In all experiments, we expected to obtain the typical finding, whereby memory for prior remembering would be poorer when Test 1 and Test 2 retrieval cues differed versus remained the same. However, we furthermore expected memory for prior remembering of changed-cue targets to be improved when participants were reminded of the study cue during Test 1 versus when they were not. The same pattern might be expected for same-cue targets, if the association between study cue and Test 1 retrieval were strengthened by retrieving the study cue, but prior research indicates that memory for prior remembering of same-cue targets is, on average, quite high (e.g., Arnold & Lindsay, 2002), potentially making increases negligible. Experiment 1 In our first experiment, we examined the effect of having participants remind themselves of study cues during Test 1 by prompting participants to recall the study cue following each Test 1 retrieval act. We assumed each participant would be able to recall some study cues but not others, allowing us to observe the impact of self-generated reminding within-participants. We furthermore assumed that when participants were able (versus unable) to bring the study cue to mind, it would create or strengthen an association between the study cue and the memory retrieval. The Test 2 phase was the same as in prior research, with participants recalling targets in response to cues from the study phase and then indicating whether they had retrieved each target on Test 1. We hypothesised that memory for prior remembering of changed-cue targets would be better following successful retrieval of study cues versus unsuccessful retrieval of study cues. Method Participants Participants were 46 undergraduates (33 women; aged years, M = 19.5) who earned course credit for participating. The procedure was approved by the Institutional Review Board at the University of Louisville. Materials A set of 113 homographs were selected as target words (selection criteria are explained in Arnold & Lindsay, 2002). Four of these words served as primacy buffers and 4 as recency buffers in all experiments, leaving 105 critical pairs for analysis. Cue-target pairs in the study phase were formed by pairing one of two possible cues with each target. Critically, each target word was a homograph and each cue corresponded to one of two possible meanings for a given target (e.g., dog and birch were cues for the target bark). Assignment of cue to target was counterbalanced in the study phase such that each cue served equally often as the study cue. During Test 1, 35 targets were assigned to each of 3 conditions: tested with the study cue (same-cue trials), tested with the other possible cue (changed-cue trials), and not tested. Assignment of targets to Test 1 treatment was counterbalanced such that each target received each treatment for an equal

5 MEMORY 963 Figure 1. Visual representation of the experimental paradigm. Modified procedure from Arnold and Lindsay (2002). The sequence of events for a changedcue target are depicted. (A) The study phase was the same in all three experiments. (B) On all Test 1 trials, participants attempted cued recall of studied targets. (C) Cued recall was followed with an instruction to either recall the study cue (Experiments 1 and 3) or type the study cue that was provided on the screen (Experiment 2). In Experiment 3, prior to the instruction to recall the study cue, participants were informed whether the study cue was the same as or different than the just-presented test cue. (D) Test 2 was the same in all three experiments. See text for additional details. number of participants. All targets were tested with the study cue on Test 2, regardless of Test 1 treatment. Procedure Participants were tested in groups of up to five. Each participant completed a version of the Arnold and Lindsay (2002) procedure on an individual computer. The procedure was modified to be run without requiring an experimenter to record participants recall responses. During the study phase participants viewed 105 critical cue-target word pairs (e.g., hand-palm, birch-bark), preceded by 4 primacy buffers and followed by 4 recency buffers. Critical pairs were presented in a new random

6 964 M. L. LEPPANEN AND K. B. LYLE order for each participant. Each homographic target was paired with one of two possible cues (e.g., bark was paired with either dog or birch). On each trial a cue-target pair was presented in the centre of the screen for 2000 ms. The pair was then replaced with a sentence containing the cue word and three asterisks where the target word could logically go (e.g., I heard a dog *** next door). The sentence appeared on screen alone for 3500 ms and then the target word appeared above the sentence for an additional 1000 ms. After offset of the sentence and target, there was a 1000 ms inter-trial interval before the next trial. After the study phase, participants advanced immediately to the Test 1 phase. This phase comprised a cuedrecall test for two-thirds of the studied targets. Participants were informed that some targets would be cued with the original cue (same-cue targets), while others would be cued with a previously unseen word that was semantically related to a target (changed-cue targets). In the latter case, they were to respond with a target from the study phase. If participants could not recall a target, they were instructed to type pass. Participants were instructed that they would also be asked to retrieve the cue they originally studied with the target and were given an example of how to respond in the case of cues that remained the same and those that changed. If participants could not recall the cue from the study phase they were instructed to type pass. During each Test 1 trial participants first performed cued recall of a target. Cue-target pairs consisted of a cue and the first and last letters of the target word separated by dashes (e.g., birch b k). Participants were instructed to type the entire word. Following each cued-recall attempt, participants were given feedback on whether their recall was correct or incorrect. Then, on a separate screen, participants were instructed to Type the cue from the STUDY phase below and press (ENTER). After pressing the ENTER key the next trial began. After completion of Test 1, participants worked on a distractor task for five minutes. Participants then began Test 2, which participants were informed was another cued-recall test. Participants were informed they would be attempting to recall all the targets they had originally studied. Participants were furthermore told that all targets would be cued with words from the study phase. Participants responded to the cued-recall portion of each Test 2 trial identically as during Test 1. As during Test 1, following each cuedrecall attempt, participants were given feedback on whether their recall was correct or incorrect. However, feedback on Test 2 differed, with participants being shown the target when cued recall was incorrect. After retrieving a target participants were also asked: Did you retrieve this target during the first test? Yes or No and were told to respond irrespective of the cue the target had been paired with on Test 1. If a participant had not correctly retrieved a target on Test 1, or if they had responded with pass, they were instructed to respond No. After the Test 2 phase, participants were debriefed and dismissed. Results Subsections labelled Cued Recall of Targets and Effect of Cue Change on Memory for Prior Remembering are common to the Results sections of all three experiments. These subsections report effects typically obtained in Arnold and Lindsay s (2002) procedure. All dependent measures and analyses referenced in those subsections were the same across experiments, unless otherwise noted. Note that, because analyses were conditionalised on the presence of certain response types, and participants sometimes gave no responses of a given type, degrees of freedom sometimes differ between tests. Cued recall of targets We first examined cued recall of targets on Tests 1 and 2.Rememberthat,onTest1,cuesforsometargetswere the same as in the study phase, cues for others were changed, and still other targets were not tested at all. This manipulation had sensible effects on cued-recall performance. On Test 1, the proportion of targets recalled was significantly higher on same-cue trials (M =.80) than changed-cue trials (M =.69),t(45) = 6.16, p <.001, d =.95. We next examined the proportion of targets recalled on Test 2 (on which all cues were the same as in the study phase) as a function of how targets were treated on Test 1 (same-cue, changed-cue, or not tested) using a repeated-measures one-way ANOVA, which yielded a significant result, F(2, 90) = 6.90, p =.002, η 2 =.133. Recall did not differ between targets assigned to the same-cue or changed-cue conditions (Ms =.84 and.82, respectively), t(45) = 1.09, p =.28, d =.18, but these levels of recall were significantly higher than for targets that had not been tested (M =.79), smallest t(45) = 2.26, p =.03, d =.33. Superior recall of previously tested versus nottested targets is consistent with the idea that testing is an active process that increases the accessibility of memories (e.g., Roediger & Karpicke, 2006; van den Broek et al., 2016). Effect of cue change on memory for prior remembering We next analysed judgments of prior remembering as a function of retrieval-context stability or change. Table 1 shows the proportion (and raw number) of targets judged to have been retrieved on Test 1 when memory for prior remembering was queried on Test 2. Following Arnold and Lindsay (2002), we analysed proportion correct for only those targets successfully retrieved on both tests. Proportions correct were submitted, as in Arnold and Lindsay (2002), to a repeated-measures oneway ANOVA with Test 1 cue (same or changed) as the within-participants factor. Participants were significantly more likely to correctly indicate they had retrieved targets assigned to the same-cue condition (M =.93)

7 MEMORY 965 Table 1. Mean number of targets and mean proportion of targets judged as recalled as a function of recall status on Test 1 and Test 2 for Experiment 1. Test 1/Test 2 recall status Number of targets Proportion judged as recalled on Test 1 Same cue* Recalled/Recalled (.01) Recalled/Not recalled (.09) Not recalled/not (.05) recalled Not recalled/recalled (.06) Changed cue* Recalled/Recalled (.03) Recalled/Not recalled (.06) Not recalled/not (.04) recalled Not recalled/recalled (.05) Not tested* NA/Recalled (.06) NA/Not recalled (.03) Notes: NA = Not applicable. * Test 1 context. Numbers in parentheses are standard errors of the mean. Lines in bold are those for which statistical analyses are reported in the text. than the changed-cue condition (M =.71), F(1, 45) = 65.42, p <.001, η 2 =.592, confirming that retrieval-context change harms memory for prior remembering. Study-cue retrieval and its effect on memory for prior remembering We next analysed performance regarding the novel feature of this experiment: The requirement that participants retrieve the study cue after attempting target retrieval on Test 1. We first examined study-cue retrieval, contingent on correct retrieval of the target, as a function of whether the same cue or a changed cue appeared on Test 1. Proportion correct cue retrieval was much higher on same-cue trials (M =.86) than changed-cue trials (M =.22), t(37) = 21.93, p <.001, d = On nearly half of changed-cue trials (M =.49), participants erroneously recalled the Test 1 cue when asked for the study cue, suggesting that participants had difficulty monitoring the source of their memories for Test 1 cues (Johnson, Hashtroudi, & Lindsay, 1993). Finally, we addressed our central question: Was the negative effect of Test 1 cue change on memory for prior remembering ameliorated by self-generated retrieval of study cues? We submitted judgments of prior remembering to a 2 (Test 1 cue: same or changed) 2 (study-cue retrieval: success or failure) within-subjects ANOVA. The main effect of Test 1 cue was significant, F(1, 33) = 47.21, p <.001, η 2 =.589, reflecting the previously documented effect whereby judgments of prior remembering were more often correct when Test 1 cues were kept the same (M =.90) versus changed (M =.77). The main effect of study cue retrieval accuracy was also significant, F(1, 33) = 51.79, p <.001, η 2 =.611. Participants were more likely to remember prior remembering following retrieval success (M =.95) versus failure (M =.72). Critically, the interaction was also significant, F(1, 33) = 61.60, p <.001, η 2 =.651. Decomposing the interaction (see Figure 2), we Figure 2. Proportion correct judgments of prior remembering in Experiment 1. Proportion of correct judgments of prior remembering in Experiment 1 as a function of Test 1 cue type and successful or unsuccessful study-cue retrieval. Error bars represent one standard error of the mean. found that the negative effect of Test 1 cue change was significant when participants failed to retrieve the study cue (Ms =.87 and.59 in the same- and changed-cue conditions, respectively), t(40) = 9.07, p <.001, d = 1.51, but there was no negative effect of cue change when participants successfully retrieved the study cue (Ms =.94 and.97 in the same- and changed-cue conditions, respectively), t(37) = 1.39, p =.174, d =.25. For both trial types, retrieving the study cue during Test 1 improved memory for prior remembering compared to when the cue was not retrieved, same-cue t(40) = 2.68, p =.01, d =.52; changedcue t(37) = 8.99, p <.001, d = Unique contribution of reminding to memory for prior remembering Our finding that memory for prior remembering was greater for targets whose study cue was retrieved (versus not) could conceivably be due to item effects because it could be that targets whose Test 1 retrieval was remembered just happened to be ones with particularly memorable study cues. To argue against this possibility, we sought to show that successful retrieval of study cues in the changed-cue condition was positively related to memory for prior remembering, even after controlling for the ease of remembering targets prior retrieval in the same-cue condition. Our analysis was modelled after analyses reported in other recent examinations of the effects of reminding (Jacoby & Wahlheim, 2013; Putnametal., 2017; Wahlheim&Jacoby,2013). We used hierarchical multiple regression to predict proportion correct judgments of prior remembering for each target in the changed-cue condition, collapsed across participants. In the first step, we entered each target s proportion of correct judgments of prior remembering in the samecue condition. In the second step, we entered the proportion of changed-cue trials on which the study cue was retrieved. Only study-cue retrieval predicted a significant amount of the variance in memory for prior remembering in the changed-cue condition (step 1 ΔR 2 =.004, p =.541, step 2 ΔR 2 =.16, p <.001).

8 966 M. L. LEPPANEN AND K. B. LYLE Discussion Like previous research (e.g., Arnold & Lindsay, 2002), Experiment 1 showed that changes in retrieval context causepeopletoforgetpriorremembering. When retrieval cues on Test 1 and Test 2 differed, versus remained the same, participants were less likely to remember Test 1 retrievals. Our novel prediction was confirmed, in that reminding participants of the study cue immediately after Test 1 retrieval on changed-cue trials increased memory for prior remembering. Analysing memory for prior remembering in the changed-cue condition on a trial-by-trial basis, we found that memory was greater when participants successfully brought the study cue to mind following Test 1 target retrieval versus when they failed to do so. The beneficial effect of reminding completely offset the negative effect of retrieval-context change. Isolating cases in which study-cue retrieval was successful, memory for prior remembering was no worse in the changed-cue condition than in the same-cue condition and was, in fact, numerically better in the former than the latter (see Figure 2). These results were not likely due to item effects and instead suggest that bringing study cues to mind following acts of retrieval on Test 1 made a unique contribution to the subsequent memorability of retrievals. Although reminding was powerful in the changed-cue condition, it was infrequent. Participants successfully recalled the study cue on only a minority of trials. Although we cannot say definitively why this was so, we found that participants often provided the Test 1 cue when asked for the study cue and this suggests that source confusion may have played a role. Participants may have misattributed their just-formed memory of the Test 1 cue to the study phase of the procedure. It is also possible that retroactive interference was involved, since the changed-cue condition involves pairing targets with a new cue immediately before testing memory for an old cue. Whatever its cause, the fact that study-cue recall was infrequent in the changed-cue condition, even though participants were actively attempting to retrieve the cues, bolsters our supposition, stated in the Introduction, that spontaneous or involuntary reminding of study cues in response to changed Test 1 cues is probably rare. Theoretically, we contend that study-cue reminders allow study cues to be bound to Test 1 retrieval acts via medial-temporal-lobe-dependent associative processes (Squire & Zola, 1996). When those same cues reappear on a test of memory for prior remembering, they can effectively cue retrieval for earlier retrieval acts. This is especially important in the changed-cue condition, which otherwise provides little or no opportunity to bind study cues to Test 1 retrieval acts. In the same-cue condition, retrieval acts occur in response to the presentation of study cues, providing an initial opportunity for binding. A reminder of the study-cue immediately subsequent to retrieval can provide a second opportunity. Experiment 2 In Experiment 1, reminding of study cues depended on participants being able to retrieve cues from memory. When participants were able to successfully remind themselves of the study cues, it had a pronounced facilitative effect on memory for prior remembering, but, unfortunately, retrieval attempts often failed. Therefore, in our second experiment, we jettisoned the requirement that participants actively retrieve study cues and instead simply re-presented the study cue following each cued-recall trial on Test 1. Participants were required to type the study cue using the computer keyboard (essentially, a copy response). This was the same motor response as in Experiment 1, but it followed a passive presentation in Experiment 2 versus active retrieval in Experiment 1. We hoped that reminding-via-presentation would increase memory for prior remembering similarly to reminding-via-retrieving, but with the advantage that participants would be able to view all study cues, versus retrieving only a minority. Another methodological difference between Experiment 1 and Experiment 2 is that, in Experiment 2, the reminder condition was compared in between-participants fashion to a condition in which participants were not reminded of study cues in any way during Test 1. This no-reminder control condition was essentially identical to the procedure developed by Arnold and Lindsay (2002, Experiment1). Method Participants Ninety-three individuals were assigned to the reminder condition (62 female; aged years, M = 20.0) and 56 to the control condition (45 female; aged years, M = 21.5). All individuals earned course credit for participating. The procedure was approved by the Institutional Review Board at the University of Louisville. Procedure The only procedural difference between Experiments 1 and 2 lay in what happened following each target-retrieval attempt on Test 1. In the control condition, there was no activity following target-retrieval attempts. In the reminder condition, participants were instructed that after each target-retrieval attempt, they would be shown the cue word that had been paired with the target during the study phase and they should type it on the computer keyboard. On each trial in Test 1, after participants had either retrieved a target and typed it in or typed pass, the study cue and a response box were displayed, with the instructions: Please type the word below in the box and press (ENTER). Data were collected in the reminder and control conditions in separate semesters. Initially, we were interested solely in the reminder condition, anticipating that it

9 MEMORY 967 would yield statistically indistinguishable memory for prior remembering regardless of whether Test 1 trials were same-cue or changed-cue (replicating what we found in Experiment 1 given reminding-via-retrieval). As reported under Results, however, we found that memory for prior remembering in the reminding-viapresentation condition was, contra our expectations, significantly poorer following changed-cue testing versus same-cue testing. We then ran the control condition to assess whether the effect of retrieval-context change was reliably different in the reminder condition than in a standard no-reminder condition. Participants in the two conditions were drawn from the same participant pool and run in the same laboratory by similarly trained research assistants. When collecting data in both conditions, we ran participants throughout the entire semester and a greater availability of participants in one semester than the other led to oversampling in the reminder condition compared to the control condition. Results Cued recall of targets We examined cued recall of targets on Tests 1 and 2 in the same manner as in Experiment 1, but taking into account the between-participants factor of condition. We analysed the proportion of targets recalled on Test 1 via a 2 (cue type: same or changed) 2 (condition: reminder or control) mixed-design ANOVA. The main effect of cue type was significant, F(1, 110) = , p <.001, η 2 =.575, with greater recall on same-cue trials (M =.84) than changed-cue trials (M =.72). The main effect of condition was also significant, F(1, 147) = 14.58, p <.001, η 2 =.09, with greater recall in the control condition (M =.81) than the reminder condition (M =.74). The effect of condition was unanticipated and we consider its implications in the Discussion section below. The two-way interaction was not significant, p =.834. We analysed the proportion of targets recalled on Test 2 via a 3 (Test 1 treatment: same-cue, changed-cue, or not tested) 2 (condition: reminder or control) mixed-design ANOVA. The main effect of Test 1 treatment was significant, F(2, 294) = 25.79, p <.001, η 2 =.149. Recall was, on average, identical for targets assigned to the same- and changedcue conditions (M =.86), and that level of recall was significantly higher than for targets that had not been tested (M =.81), smallest t(148) = 6.32, p <.001, d =.63. The main effect of condition was also significant, F(1, 147) = 10.83, p =.001, η 2 =.069, with greater recall in the control condition (M =.88) than the reminder condition (M =.83). The effect of condition mirrors the one obtained on Test 1 and indicates that, on both tests, participants in the control condition unexpectedly recalled more targets than participants in the reminder condition. The two-way interaction was not significant, F(2, 294) = 2.95, p =.054, η 2 =.02. Effect of cue change on memory for prior remembering Table 2 shows the proportion (and raw number) of targets judged to have been retrieved on Test 1 when memory for prior remembering was queried on Test 2. We analysed proportion correct as a function of cue change, while also considering the between-participants factor of condition. Proportion correct was analysed via a 2 (Test 1 cue: same or changed) 2 (condition: reminder or control) mixed-design ANOVA. Results are depicted in Figure 3. While the main effect of Test 1 cue was significant, F(1, 147) = 88.26, p <.001, η 2 =.375, the main effect of condition was not significant, F(1, 147) = 3.16, p =.078, η 2 =.021. The key result was the significant two-way interaction, F(1, 147) = 8.49, p =.004, η 2 =.055. As shown in Figure 3, correct judgments of prior remembering for same-cue targets did not differ significantly between the reminder and control conditions (Ms =.91 and.92, respectively), p =.811, but correct judgments of prior remembering for changed-cue targets were significantly more likely in the reminder condition than the control condition (Ms =.82 and.73, respectively), t(110) = 2.63, p =.01, d =.37, indicating that reminders of the study cue increased memory for prior remembering in the face of retrievalcontext change. Retrieval-context change (versus stability) nonetheless significantly impaired memory for prior remembering in both conditions, smallest t(92) = 6.16, p <.001, d =.71. The magnitude of this impairment, Table 2. Mean number of targets and mean proportion of targets judged as recalled as a function of recall status on Test 1 and Test 2 for Experiment 2. Proportion judged as Test 1/Test 2 recall status Number of targets recalled on Test 1 Reminder condition Same cue* Recalled/Recalled (.01) Recalled/Not recalled (.05) Not recalled/not recalled (.03) Not recalled/recalled (.04) Changed cue* Recalled/Recalled (.02) Recalled/Not recalled (.04) Not recalled/not recalled (.04) Not recalled/recalled (.03) Not tested* NA/Recalled (.03) NA/Not recalled (.03) Control condition Same cue* Recalled/Recalled (.01) Recalled/Not recalled (.08) Not recalled/not recalled (.05) Not recalled/recalled (.06) Changed cue* Recalled/Recalled (.03) Recalled/Not recalled (.06) Not recalled/not recalled (.05) Not recalled/recalled (.04) Not tested* NA/Recalled (.05) NA/Not recalled (.04) Notes: NA = Not applicable. * Test 1 context. Numbers in parentheses are standard errors of the mean. Lines in bold are those for which statistical analyses are reported in the text.

10 968 M. L. LEPPANEN AND K. B. LYLE remembering was assessed only for items successfully retrieved on both tests. Although participants in the control condition were able to recall a somewhat greater number of targets, what matters for our purposes is that these participants remembered prior remembering in the face of retrieval-context change more poorly than participants who received reminders. Experiment 3 Figure 3. Proportion correct judgments of prior remembering in Experiment 2. Proportion of correct judgments of prior remembering in Experiment 2 as a function of condition and Test 1 cue type. Error bars represent one standard error of the mean. however, was about half as large in the reminder condition (M diff =.10) as in the control condition (M diff =.19). Discussion Under conditions of retrieval-context change, memory for prior remembering was significantly better when participants were reminded (versus not) of each target s study cue immediately after target retrieval on Test 1. Reminding was achieved in Experiment 2 by re-presenting the study cue on the screen and having participants type it on the keyboard. Participants were instructed before the test began that the cue they were being asked to type had been paired with the target they had attempted to retrieve during the study phase. We theorised that this sort of reminding would provide an opportunity to bind study cues to retrieval acts. It appears, though, that reminding participants of a target s study cue after retrieving the target in response to a different test cue is not mnemonically equivalent to retrieving the target in response to the study cue. Even with reminding, changed-cue retrieval events were significantly less memorable than same-cue retrieval events. Hence, reminding-via-presentation reduced the negative effect of retrieval-context change on memory for prior remembering, but did not eliminate it. This contrasts with the finding in Experiment 1 that target retrievals on changed-cue trials were as likely to be remembered as those on same-cue trials, provided participants were able to recall study cues following retrieval. An unexpected finding in this experiment was that cued recall of targets was poorer in the reminder condition than the control condition. This was true on both Tests 1 and 2. A possible explanation is that, for participants in the reminder condition, viewing and typing study cues during Test 1 produced output interference, which has previously been observed in paired-associate recall procedures (Roediger & Schmidt, 1980). It may also be that the result is spurious and unrelated to the experimental manipulation. In any event, it is unclear that differences in recall performance change the conclusions we would draw from this experiment, since memory for prior In Experiment 3, we sought to replicate Experiment 1 in showing that the negative effect of retrieval-context change on memory for prior remembering is completely eliminated when participants recall study cues immediately following an act of retrieval. Two considerations made us feel a replication attempt was warranted, especially given recent intense interest in the replicability of psychological findings (e.g., Open Science Collaboration, 2015). First, the negative effect of retrieval-context change on memory is usually quite robust and, in Experiment 2, we found that it endured, albeit in diminished form, when we externally presented the study cue to participants on every trial. Second, the beneficial effect of study-cue recall was based on a small number of observations because successful recall was rare on changed-cue trials. For either or both of these reasons, one might question the reliability of the finding. Rather than conducting an exact replication, we changed the procedure in the hope of increasing studycue recall on changed-cue trials. We found that participants in Experiment 1 often mistook the Test 1 cue for the study cue. That is, when asked to recall the study cue, participants reported the Test 1 cue. This is understandable from a source-monitoring perspective (Johnson et al., 1993), because, having just seen the Test 1 cue, participants may have experienced source confusion and thought the just-seen cue had also appeared in the study phase, not unlike what happens when eyewitnesses are exposed to post-event misinformation (see Loftus, 2005, for a review). We reasoned that participants might be better able to resolve this source confusion if they were explicitly alerted to instances in which the Test 1 cue was not the study cue and they would therefore need to search their memories for a cue word other than the justpresented one. We therefore presented participants with a statement following each cued-recall attempt on Test 1 that specified whether the cue they had just seen was the same as the study cue or different. We hoped this would encourage participants to thoroughly search for study cues on changed-cue trials, thereby increasing the likelihood of study-cue recall and providing us with more opportunities to observe the effect of study-cue recall on memory for prior remembering. Even if recall remained infrequent, we could test whether this relatively rare event was as potent in Experiment 3 as it was in Experiment 1.

11 MEMORY 969 Method Participants Participants were 62 undergraduates (42 female; aged years, M = 20.1) who earned course credit for participating. The procedure was approved by the Institutional Review Board at the University of Louisville. Procedure Experiment 3 was procedurally identical to Experiment 1 except for the method of reminding during Test 1. After attempting to retrieve each target on Test 1, participants received one of two informational messages that appeared on the computer screen. On same-cue trials, participants saw: This cue remained the SAME. Please type the cue from the STUDY phase below and press (ENTER). On changed-cue trials, participants saw: This cue was DIFFER- ENT than the one you studied. Please type the cue from the STUDY phase below and press (ENTER). Results Cued recall of targets The proportion of targets correctly recalled on Test 1 was significantly higher on same-cue trials (M =.80) than changed-cue trials (M =.68), t(61) = 9.34, p <.001, d = A repeated-measures ANOVA on Test 2 target recall revealed a significant effect of Test 1 treatment, F(2, 122) = 11.71, p <.001, η 2 =.161. Unlike in Experiments 1 and 2, the proportion of targets correctly recalled differed significantly, but only slightly, between targets assigned to samecue (M =.84) and changed-cue trials (M =.82), t(61) = 2.30, p =.025, d =.30. Recall of targets that had not previously been tested (M =.79) was significantly lower than both types of previously tested targets, smallest t(61) = 2.59, p =.01, d =.34. Table 3. Mean number of targets and mean proportion of targets judged as recalled as a function of recall status on Test 1 and Test 2 for Experiment 3. Test 1/Test 2 recall status Number of targets Proportion judged as recalled on Test 1 Same cue* Recalled/Recalled (.01) Recalled/Not recalled (.07) Not recalled/not (.05) recalled Not recalled/recalled (.05) Changed cue* Recalled/Recalled (.03) Recalled/Not recalled (.05) Not recalled/not (.04) recalled Not recalled/recalled (.04) Not tested* NA/Recalled (.04) NA/Not recalled (.03) Notes: NA = Not applicable. * Test 1 context. Numbers in parentheses are standard errors of the mean. Lines in bold are those for which statistical analyses are reported in the text. Effect of cue change on memory for prior remembering Table 3 shows the proportion (and raw number) of targets judged to have been retrieved on Test 1 when memory for prior remembering was queried on Test 2. Participants were significantly more likely to correctly indicate that they had retrieved targets assigned to same-cue trials (M =.92) than changed-cue trials (M =.68), F(1, 61) = , p <.001, η 2 =.629. Study-cue retrieval and its effect on memory for prior remembering We first examined retrieval of study cues as a function of whether the same cue or a changed cue appeared on Test 1. Proportion correct cue retrieval was significantly higher on same-cue trials (M =.90) than changed-cue trials (M =.28), t(57) = 24.61, p <.001, d = Recall on changed-cue trials was numerically higher than in Experiment 1 (by cross-experimental comparison), but only slightly and non-significantly so, t(94) = 1.47, p =.144, d =.24. Finally, we examined whether the negative effect of Test 1 cue change on memory for prior remembering was eliminated by reminding participants of the study cue. We submitted judgments of prior remembering to a 2 (Test 1 cue: same or changed) 2 (study-cue retrieval: success or failure) within-participants ANOVA. There were significant main effects of Test 1 cue, F(1, 45) = 28.64, p <.001, η 2 =.389, and study-cue retrieval, F(1, 45) = 40.73, p <.001, η 2 =.475, but these were superseded by a significant interaction, F(1, 45) = 45.34, p <.001, η 2 =.502. As shown in Figure 4, and mirroring the result from Experiment 1, the interaction arose because the negative effect of Test 1 cue change was significant when participants failed to retrieve study cues (Ms =.89 and.59 for same and changed cues, respectively), t(51) = 7.64, p <.001, d = 1.06, but was absent when participants successfully retrieved study cues (Ms =.92 and.94 for same and changed cues, respectively), t(55) =.216, p =.174, d =.04. Retrieval success (versus failure) significantly increased memory for prior remembering following changed-cue testing, t(54) = 9.82, p <.001, d = 1.35, while the increase was only numerical following same-cue testing, t(52) = 1.02, p =.314, d =.16. Unique contribution of reminding to memory for prior remembering To examine the unique contribution of reminding, above and beyond potential item effects, to memory for prior remembering, we conducted the same hierarchical multiple regression analysis as in Experiment 1. Study-cue retrieval predicted a significant amount of the variance in memory for prior remembering in the changed-cue condition, above and beyond a significant contribution from memory for prior remembering of targets in the same-