Does survival processing enhance implicit memory?

Memory & Cognition 2010, 38 (8), 1110-1121 doi:10.3758/MC.38.8.1110 Does survival processing enhance implicit memory? CHI-SHING TSE Chinese University of Hong Kong, Hong Kong, China AND JEANETTE ALTARRIBA University at Albany, State University of New York, Albany, New York Recent research has shown that human memory may have evolved to remember information that has been processed for the purpose of survival, more so than information that has been processed for other purposes, such as home-moving. We investigated this survival-processing advantage using both explicit and implicit memory tests. In Experiment 1, participants rated words in one of three scenarios: survival, pleasantness, and moving, followed by a timed stem-cued recall/stem-cued completion task. Items were completed more quickly in the survival scenario, as compared with the other two for the explicit task, but no differences were found across conditions in the implicit task. In Experiment 2, the implicit task was changed to concreteness judgments to encourage more conceptual processing. Again, the survival-processing advantage occurred in the explicit task (speeded item recognition), but not in the implicit task. These results suggest that a survival-processing advan- tage may benefit participants’ memory performance only during explicit retrieval. Although the evolution of human memory has long Several studies have used the functional approach to been considered (see, e.g., Glenberg, 1997; Sherry & examine memory performance (e.g., Kang, McDermott, Schacter, 1987), it has not been empirically studied until & Cohen, 2008; Nairne & Pandeirada, 2008a, 2010; very recently in cognitive psychology. As was pointed out Nairne, Pandeirada, Gregory, & Van Arsdall, 2009; by Klein, Cosmides, Tooby, and Chance (2002), struc- Nairne, Pandeirada, & Thompson, 2008; Nairne, Thomp- ture follows function. To understand the characteristics son, & Pandeirada, 2007; Otgaar, Smeets, & van Bergen, of a memory system or process, one should specify the 2010; Rudine, Craig, Overbeek, & Green, 2009; Wein- problems that the system or process has evolved to solve. stein, Bugg, & Roediger, 2008). These studies have ex- Nairne and Pandeirada (2008b) proposed three character- amined whether memory might be tuned to remember istics of evolved memory mechanisms. First, since there information that has been processed for survival, perhaps is little adaptive value in reproducing the veridical past, as a result of fitness advantages accrued in the ancestral memory should reconstruct previous episodic experience past (see, e.g., Lu & Chang, 2009, for studies related to the flexibly, rather than reproduce the past like a tape recorder. evolution of memory yet not directly related to survival). It Second, memory should be geared especially to help us has generally been predicted that items that have been pro- perform actions that enhance our reproductive fitness, cessed for the purpose of survival should be better remem- such as remembering the location of food. Third, memory bered than those that have been processed via means that should be tuned to remember certain kinds of domain- are irrelevant for survival (e.g., pleasantness ratings). specific information that is relevant to survival/fitness. To Most of the studies have tested this hypothesis by using explain why memory can be boosted by a specific strat- a design introduced by Nairne et al. (2007). Participants egy, such as self-reference, one should seek to understand first rate, without anticipating a later memory test, lists the proximate cause, such as elaboration, as well as the ul- of words on the basis of their relevance in one of these timate adaptive value of the corresponding memory “tun- two scenarios: survival in the grasslands of a foreign land ing.” To study the characteristics of the evolved memory (survival scenario) or moving to a new home in a foreign system, Nairne and his colleagues have used a functional land (moving scenario), and in an additional condition, approach (Nairne, 2005; Nairne & Pandeirada, 2008a): participants are simply asked to rate the pleasantness of First identify the selection pressures that may have shaped the words ( pleasantness scenario). The latter two condi- the evolution of memory, generate a priori predictions, tions serve as controls—whereas rating words on the basis and then test these hypotheses empirically. of pleasantness induces meaning-driven, item-specific C.-S. Tse, [email protected] © 2010 The Psychonomic Society, Inc. 1110  SURVIVAL PROCESSING 1111 processing of those words, rating words on the basis of typically used in implicit memory studies (e.g., Roediger, their relevance to a moving scenario induces schema acti- Weldon, Stadler, & Riegler, 1992). Apart from testing the vation, which is also likely the case when words are rated survival-processing advantage in implicit memory, we on the basis of their relevance to a survival scenario. tried to generalize the advantage to the explicit stem- Supporting the idea that memory may have evolved, in cued recall test, which, to our knowledge, has never been part, to remember information that has been processed for reported in the literature. survival, in previous studies, participants demonstrated su- perior memory for the words rated in a survival scenario, EXPERIMENT 1 relative to those rated in moving and pleasantness scenar- ios. This finding suggests that the mnemonic advantage Method Participants. Two hundred forty English-speaking undergradu- of survival processing cannot be merely attributed to any ates with normal or corrected-to-normal vision participated for par- item-specific or schematic memory processing and sug- tial course credit. Sixty received an explicit memory test (explicit gests that the advantage reflects an adaptive bias that is ac- group), and 180 received an implicit one (implicit group). More tivated when participants rate the words in a survival sce- participants were tested in the implicit group in order to ensure suf- nario. This survival-processing advantage occurs whether ficient statistical power to detect a potential survival-processing ad- rating scenarios are manipulated within or between sub- vantage in implicit memory (see below). jects (Nairne et al., 2007), whether relevance ratings refer Design and Materials. A 2 (group: explicit or implicit)  4 (sce- nario: pleasantness, moving, survival, or nonstudied) mixed-factor to a character depicted in a video clip or to participants design was used. Group was a between-subjects variable, whereas themselves (Weinstein et al., 2008), whether or not the scenario was manipulated within subjects but between blocks for the memory test is anticipated (Nairne et al., 2007), and when first three rating scenarios. The participants were given three blocks performance in the survival scenario is compared with of 20 words in the study phase, with a different rating scenario in that in other encoding tasks (e.g., imagery, generation, each block. A total of 170 words were chosen as stimuli, 80 of which and self-reference; see Nairne et al., 2008) and scenarios were critical items, 12 were primacy buffer items on the memory test, and the remaining 78 were filler items on the memory test. (e.g., planning a bank heist in Kang et al., 2008; enjoying The lexical characteristics of the stimuli are summarized in Table 1. a vacation at a fancy resort in Nairne et al., 2008; organiz- Because stem cues were used in the memory test, care was taken to ing a charity event with animals at a local zoo in Nairne ensure that the first three letters of each word were unique among & Pandeirada, 2007; and surviving in a city scenario in all 170 words and that the baseline completion rate when the stem Weinstein et al., 2008). cues were given (estimated by Shaw, 1997, when the item has not So far, most of the published studies have manipu- been studied) was below .30 for critical items (see Table 1), so as not to mask the effect due to the rating scenarios in the implicit/explicit lated encoding strategies and scenarios and then com- memory tests. Unlike in prior studies in which all the stimuli were pared the degree to which survival processing enhances concrete words (but see Rudine et al., 2009), half of our critical memory performance as other strategies/scenarios do in items were concrete words, and half were abstract. To ensure that explicit memory tests, including free recall and recogni- items were not preexperimentally associated with the meaning of tion. To our knowledge, no study has examined whether survival more than they were with the meaning of moving, the mean the survival-processing advantage occurs in an implicit semantic similarity between critical items and survival was matched test of memory. In contrast to explicit memory, which with the mean semantic similarity between critical items and mov- ing. Semantic similarity, as reflected by cosines in latent semantic reflects conscious recollection of prior episodes, implicit memory is an unintentional manifestation of the retention of previously acquired information. When taking an im- Table 1 plicit memory test, participants perform a task apparently Mean Statistics for Lexical Characteristics unrelated to a study phase, such as filling out a word stem of Stimuli in Experiment 1 or judging whether a word refers to a concrete concept, Primacy Buffer rather than remembering any items they have seen be- Critical Items and Filler Items fore. Implicit memory is reflected in repetition priming Characteristic M SD M SD that refers to facilitation in processing speed (shorter re- Word length 6.04a 0.97 5.90a 0.92 sponse time [RT]) and/or higher accuracy for studied Word valence 5.36a 1.81 5.55a 2.06 items, relative to nonstudied items. In the present experi- Word arousal 4.92a 0.91 5.06a 0.88 Word dominance 5.13a 0.93 5.13a 0.98 ments, we explored whether or not a survival-processing Word concreteness 4.67a 1.63 4.77a 1.54 advantage would generalize to implicit memory. To satisfy Log HAL word frequency 9.27a 1.32 9.06a 1.28 the retrieval intentionality criterion (Schacter, Bowers, & Word connectivity 1.65a 0.69 1.79a 0.73 Booker, 1989) as closely as possible, we used the analo- Word connection strength 2.87a 1.22 3.11a 1.29 Baseline completion rate 0.17a 0.05 0.26b 0.28 gous explicit memory tests, in which all stimuli, design, Word stem set size 10.74a 3.75 10.00a 4.86 and procedures were the same as those in the implicit “Survival”-LSA cosines 0.08a 0.07 0.08a 0.07 memory tests, except that we instructed participants to re- “Moving”-LSA cosines 0.08a 0.06 0.09a 0.06 member what they had seen in the rating phase. In Experi- Note—The values within each row with different superscripts are sig- ment 1, we used a stem-cued completion test and a stem- nificantly different from each other ( p  .05, two-tailed). Word valence, cued recall test, in which participants were instructed to arousal, and dominance are from Bradley and Lang (1999). Word con- creteness, connectivity, and connection strength are from Nelson, Mc- fill out a word stem with the first word that came to mind Evoy, and Schreiber (2004). Baseline completion rates and word stem set (i.e., implicit memory) or with the studied items in the sizes are from Shaw (1997). The survival-LSA cosines and moving-LSA rating phase (i.e., explicit memory), respectively—tasks cosines were estimated via http://lsa.colorado.edu. 1112 TSE AND ALTARRIBA analyses (LSA; Landauer & Dumais, 1997) was estimated using In the explicit/implicit memory test, participants were presented the “General Reading up to 1st year in college” database, since our with 170 word stems (e.g., TOA-), one at a time, at the center of the participants were mostly first-year college students. screen. These word stems consisted of 12 primacy buffer items (to The 80 critical items were divided into four groups, three of which familiarize the participants with the test procedure), which were pre- were assigned to be study items in the three rating scenarios, and sented at the beginning of the test, 60 studied items from the three the remaining was assigned to be nonstudied items, which were the rating tasks, 20 nonstudied items, which served as a baseline, and baseline test items in the memory test. The four groups were rotated 78 filler items. The latter three groups of stimuli were randomly across participants, such that each served in one of four scenario intermixed for each participant. Although including 12 nonstudied conditions equally often. The order of the three rating scenarios was primacy buffer items might bias participants not to respond or to re- counterbalanced across participants. Equal numbers of participants ject test items, this would affect memory performance equally across in the explicit and implicit groups received each of the 12 counter- the three rating scenarios. None of the primacy buffer or filler items balancing lists. was presented in the rating phase, so the proportion of studied items Procedure. PC-compatible computers were used to display the in the memory test was .35 (i.e., 60/170). Upon presentation of the stimuli and to collect data. The participants were tested in a quiet stem cue, the participants were instructed to type the first word that computer lab in groups of 2–8 and were seated 60 cm away from came to mind (for the implicit group) or an item that they had seen in the screen. All stimuli were presented in Courier New Bold with the rating tasks (for the explicit group). They were asked to respond a font size of 18 in white on a black background. There were three within 12 sec on each trial, and their RT and accuracy were recorded. phases: rating phase, filler task phase, and surprise final memory At the beginning of the test, the participants in the explicit group were test phase. In the rating phase, prior to each of the three blocks, reminded that not all of the stem cues referred to the items they had the participants received one of the three scenarios (pleasantness, seen in the rating phases. Both implicit and explicit groups could skip moving, or survival) and then rated the words on the basis of the sce- trials by typing XXX if they failed to come up with any answers. nario. On each trial of the rating task, a word appeared at the center Following Bowers and Schacter (1990), after the experiment, the of the screen, and participants were asked to rate it using a 5-point participants in the implicit group were asked the following ques- scale, where 1  totally irrelevant/unpleasant and 5  extremely tions concerning their test awareness and intentionality to retrieve: relevant/pleasant. The rating scale appeared on the screen below (1) “What did you think was the purpose of the stem completion each presented item, and the participants responded by pressing the task that you just finished?” (2) “What was your general strategy appropriate key on the number pad. Neither the participants in the in completing the word stems?” (3) “Did you notice any relation implicit group nor those in the explicit group anticipated that they between the words I showed you earlier and the words produced on would be given a memory test after the rating task. The participants the stem-cued completion test?” and (4) “While doing the stem-cued were asked to respond within 5 sec on each trial, and response times completion test, did you notice whether you completed some of the (RTs) were recorded. If a response was not made within 5 sec, the stems with the words studied on the earlier list?” After completing next trial began automatically. Adapted from Nairne et al. (2007), all these questions, the participants were thanked and debriefed.1 the descriptions of the three scenarios were the following. Survival scenario. “In this task we would like you to imagine Results that you are stranded in the grasslands of a foreign land, without any basic survival materials. Over the next few months, you’ll need Unless otherwise specified, the significance level was to find steady supplies of food and water and protect yourself from set at .05, two-tailed. All analyses were planned in ad- predators. We are going to show you a list of words, and we would vance to test for a survival-processing advantage. Effect like you to rate how relevant each of these words would be for you in sizes h 2p and Cohen’s d were reported for F and t statistics, this survival situation. Some of the words may be relevant and others respectively. In general, participants had little difficulty may not—it’s up to you to decide.” providing the relevant ratings within 5 sec. The mean Moving scenario. “In this task we would like you to imagine that you are planning to move to a new home in a foreign land. Over the proportions of unrated words were 2.2% (moving, 1.5%; next few months, you’ll need to locate and purchase a new home and pleasantness, 2.2%; survival, 2.8%) for the explicit group transport your belongings. We are going to show you a list of words, and 1.8% (moving, 1.6%; pleasantness, 1.8%; survival, and we would like you to rate how relevant each of these words 1.9%) for the implicit group. These proportions did not would be for you in accomplishing this task. Some of the words may differ across the three rating scenarios (all ps  .11, h 2p s  be relevant and others may not—it’s up to you to decide.” .01). Due to the small number of unrated words and to Pleasantness scenario. “In this task, we are going to show you a avoid item selection problems, we used the full set of data list of words, and we would like you to rate the pleasantness of each word. Some of the words may be pleasant and others may not—it’s in the following analyses. A response on the memory test up to you to decide.” was scored as correct when it was correctly spelled or The participants were given self-paced breaks in between the blocks matched the grammatically derived forms of the answer of the rating tasks. Following completion of all three rating blocks, (e.g., recalling games for game). they performed a series of filler tasks to avoid the possibility of ceiling Rating phase. The mean ratings/RTs for moving, performance on the explicit memory test, as well as to mask the nature pleasantness, and survival scenarios were 2.85/1,742 msec, of the implicit memory test. In these tasks, they were instructed to type in (1) as many U.S. states as possible within 2.5 min, (2) as many 3.04/1,802 msec, and 2.91/1,786 msec, respectively, for U.S. presidents’ last names as possible within 4 min, and (3) as many the explicit group, and 2.68/1,780 msec, 3.05/1,811 msec, male and female names as possible within 2.5 min. Immediately after and 2.88/1,805 msec, respectively, for the implicit group. these filler tasks, the participants in the implicit group received a There was no interaction associated with group for RTs stem-cued completion test, and those in the explicit group received or ratings [both Fs(2,476)  2.13, h 2p s  .01]. Collapsed a stem-cued recall test. These procedures are in line with Schacter across groups, all of the comparisons were significant in et al.’s (1989) retrieval intentionality criterion for implicit memory studies, since both explicit and implicit groups received the same set ratings (all ps  .001, ds  0.40), but not in RTs (all ps  of procedures for the rating phase, filler task phase, and this surprise .24, ds  0.11). The ratings were higher in the pleasantness memory test phase, with the only difference being the memory test scenario than in the survival scenario and were higher in instructions. the survival scenario than in the moving scenario.  SURVIVAL PROCESSING 1113 Proportion of Correct Recall/Completion .70 Moving Pleasantness .60 Survival Nonstudied .50 .40 .30 .20 .10 0 Implicit Explicit 3,500 Moving RT for Correct Trials (msec) Pleasantness Survival 3,000 Nonstudied 2,500 2,000 1,500 Implicit Explicit Figure 1. Mean proportions and median response times (RTs) for correct recall/completion for explicit and implicit groups as a function of scenario and group in Experiment 1. Error bars indicate standard errors of the means. Memory phase. Figure 1 presents the mean proportion pleasantness scenario [t(59)  2.01, d  0.37], but the of correct recall/completion and median RT for correct RT difference between the latter two conditions was not trials as a function of scenario and group. The proportion significant [t(59)  0.10, d  0.02]. None of the RT of correct recall for nonstudied items in the explicit group comparisons was significant for the implicit group [all was based on their mere guessing. ts(179)  0.89, ds  0.09]. For the proportions of correct recall/completion, On the basis of Cohen’s d (see Cohen, 1988, p. 48, for items rated in the moving, pleasantness, and survival how it is computed) for the explicit group’s within-subjects scenarios were recalled or completed better than were differences in RT between survival and pleasantness sce- nonstudied items in both explicit and implicit groups [all narios and between survival and moving scenarios (.37 ts(59) or ts(179)  9.90, ds  1.80]. However, none of and .48, respectively), the power to detect similar differ- the comparisons among studied items rated in different ences (with p  .05, two-tailed) for the implicit group was scenarios approached significance in explicit or implicit .92 and .99, respectively, with a sample size of 180. With a groups [all ts(59) or ts(179)  1.48, ds  0.16]. For RT, cautionary note that the effect size estimated from sample only items rated in the survival scenario yielded shorter data might have overestimated the effect size observed in RTs than did nonstudied items for the explicit group the population, the absence of the effects in the implicit [t(59)  2.12, d  0.38]. None of the other comparisons group is unlikely to have been due to insufficient statisti- with nonstudied items in the explicit or implicit groups cal power. was significant [all ts(59) or ts(179)  1.09, ds  0.11]. Test awareness/intentionality to retrieve in implicit The explicit group responded faster to items rated in memory. Using their responses to the four questions at the survival scenario than to those rated in the moving the end of the experiment, the participants in the implicit scenario [t(59)  2.59, d  0.47] and those rated in the group were divided into three groups on the basis of their 1114 TSE AND ALTARRIBA test awareness (test aware vs. test unaware) and intention- rated in the moving or pleasantness scenarios. This pattern ality to retrieve (intend vs. not intend to retrieve studied remained the same when test awareness and intentionality items during the test): aware–intend (n  34), aware– to retrieve were taken into account. However, this result unintend (n  128), and unaware–unintend (n  18). could be complicated by the following reasons. Since individuals who were unaware of the nature of the First, because in our explicit stem-cued recall test, memory test did not claim that they retrieved the studied the survival-processing advantage was weak, its analo- items on the memory test, there was no unaware–intend gous implicit stem-cued completion test might not have participant. The test-aware participants were those who been sensitive enough to detect a survival-processing responded “yes” to Questions 3 and/or 4. The participants advantage. who were classified as intending to retrieve were those Second, rating concepts (i.e., words) for their relevance who answered on Questions 1 and/or 2 that they studied in a survival scenario is a meaning-based encoding strat- the items for subsequent tests, thought the purpose of the egy, whereas completing the word stem with the first word experiment was to test whether they could remember words that comes to mind, a perceptual implicit memory test, presented before, and/or admitted that they did retrieve the might not necessarily tap word meaning (see Roediger & studied items during the test. Only the main effect and McDermott, 1993, for more discussions regarding type the interaction associated with group are discussed below. of processing in the stem-cued completion test). Since The group (aware–intend, aware–unintend, or unaware– previous studies showed that perceptual implicit memory unintend)  scenario interaction was not significant for tests were not sensitive to a levels-of-processing manipula- accuracy [F(6,531)  1.85, MSe  0.01, h 2p  .02] or tion (e.g., Roediger et al., 1992), it may not be too surpris- for RT [F(6,531)  0.33, MSe  154,687, h 2p  .004]. ing that relative to other meaning-based encoding strat- The main effect of group was significant for accuracy egies (e.g., pleasantness ratings), a survival-processing [F(2,177)  5.53, MSe  0.01, h 2p  .06] but not for RT advantage could not be revealed in a perceptual implicit [F(2,177)  0.28, MSe  543,162, h 2p  .003]. Follow-up memory test. analyses on accuracy showed that the unaware–unintend Third, a few participants claimed that they did intention- group (.33) yielded lower accuracy than did the aware– ally retrieve the studied items during the implicit memory unintend group (.39) and the aware–intend group (.40) test. These claims could be attributed to the long (12-sec) (both ps  .01, ds  0.36), but there was no difference response deadline, as well as the involvement of word pro- between the latter two ( p  .45, d  0.08). Thus, although duction in the stem-cued completion task. Even though test awareness boosted overall accuracy, intentionality to the overall pattern of results remained the same after retrieve (with the presence of test awareness) did not. Of eliminating the 34 test-aware participants who intended course, one should interpret these results with caution, to retrieve during the implicit memory test, it is important due to the small sample sizes in the aware–intend and to use an implicit memory test that may further minimize unaware–unintend groups. opportunities to use explicit retrieval strategies. One could argue that the overall accuracy for studied Fourth, the dissociation between implicit and explicit items was quite low for the implicit group (~43%), such memory could be attributed to the differential reliability of that any survival-processing advantage might have been implicit and explicit memory measures (Buchner & Wip- masked. However, this range of accuracy was not unusual pich, 2000). Since the participants could come up with any in previous studies (e.g., Roediger et al., 1992). Also, words, as long as they fit in the stems in the implicit stem- an analysis of individuals with above-median accuracy cued completion, the variability of their responses should yielded results similar to those for the full set of partici- be higher than that in the explicit stem-cued recall test, pants, whether or not test awareness and intentionality in which participants needed to fill out specific words, were taken into account. thus reducing the reliability of the measures. Buchner and In summary, we replicated the survival-processing ad- Wippich (p. 248) noted that implicit memory tests that vantage in an explicit stem-cued recall test, although the require participants to make rapid and restricted responses effect was weak and occurred only in RTs. The reason why (e.g., word identification) were more reliable, since this this effect occurred only in RTs was not clear but was procedure could limit the variety of processes involved in perhaps due to participants’ strategies of trading off ac- task performance. Hence, a speeded test with only binary curacy at the expense of RT. The weak effect could also be responses may provide a more reliable implicit memory due to the perceptual nature of the explicit memory test; measure. that is, participants might rely on the stem–target ortho- To address all these concerns, in Experiment 2, we used graphic association, rather than the meaning of the targets, a speeded conceptual implicit memory test where partici- in producing their recall responses. This could be tested pants made binary responses to judge whether or not the in future studies by presenting a related cue stem (e.g., test item referred to a concrete concept. We chose this task happy–ple), rather than only the stem (e.g., ple) with the because (1) we wanted to use the same set of study items study/test items, which would encourage more meaning- as in Experiment 1, in which half of our study items were based encoding. concrete words and half were abstract words, (2) this task Regarding the implicit memory test, although we ob- taps the meaning of test items, (3) the speeded nature of tained robust overall priming effects (i.e., relative to non- this task could minimize opportunities for using explicit studied items), we did not find stronger priming effects retrieval strategies (see, e.g., Hourihan & MacLeod, 2007, for words rated in the survival scenario, relative to those for a discussion), and (4) the binary responses (as well as  SURVIVAL PROCESSING 1115 its speeded nature) could boost the reliability of implicit across groups, all of the comparisons were significant memory measures. To follow as closely as possible the re- for ratings (all ps  .05, ds  0.15), but not for RTs (all trieval intentionality criterion (Schacter et al., 1989), we ps  .10, ds  0.10). The ratings were higher for the used another explicit memory test, an item recognition test, pleasantness scenario than for the survival scenario and, which was reported to demonstrate the survival-processing in turn, the moving scenario, yielding a pattern similar to advantage in previous research (e.g., Nairne et al., 2007). that in Experiment 1. To render the task demands of the implicit and explicit Memory phase. Figure 2 presents the mean propor- memory tests more comparable, we used a speeded item tion of correct judgment/old responses and median RT recognition test in which participants were instructed to for corrected trials/old responses as a function of scenario respond as quickly and as accurately as possible. and group. The old responses for nonstudied items in the explicit group, referred to as a false alarm rate, could not EXPERIMENT 2 be compared with the old responses for studied items (i.e., hit rates) in the three rating scenarios. However, given Method that the participants’ responses to the test items in the Participants. Three hundred twelve English-speaking under- graduates with normal or corrected-to-normal vision participated concreteness judgment task should be the same whether in exchange for partial course credit. Ninety-six received an explicit the items were studied or nonstudied, the proportion of memory test (explicit group), and 216 received an implicit memory correct judgment and correct trial RT of nonstudied items test (implicit group). An equal number of participants in the explicit could be compared with those of studied items in the three and implicit groups received each of the 24 counterbalancing lists rating scenarios and serve as a baseline to measure the (see the key assignment counterbalancing procedure, below). priming effect. Design, Materials, and Procedure. The design, materials, and procedure were identical to those used in Experiment 1, except that For accuracy, items rated in moving, pleasantness, an item recognition task and a concreteness judgment task were used and survival scenarios were judged better than were for explicit and implicit memory tests, respectively. The participants nonstudied items in the implicit group, showing robust in the explicit and implicit groups received the same set of proce- priming effects [all ts(215)  2.20, ds  0.21]. Regarding dures for the rating phase, filler task phase, and the surprise memory the comparisons among studied items rated in different test phase, with the only difference being the memory test instruc- scenarios, whereas for the implicit group none of the tion. In both the explicit and implicit memory tests, the participants were presented with 170 items intact, one at a time, at the center of comparisons approached significance [all ts(215)  the screen. Upon the presentation of the item, half of the participants 1.04, ds  0.10], for the explicit group, items rated in the in the implicit group and in the explicit group were instructed to re- survival scenario yielded higher hit rates than did those spond by pressing the “l” key to indicate that the word was concrete rated in the pleasantness [t(95)  2.73, d  0.39] and (for the implicit group) or studied before (for the explicit group) or moving [t(95)  3.04, d  0.44] scenarios, and the hit rate the “s” key to indicate that the word was abstract (for the implicit difference in the latter two conditions was not significant group) or not studied before (for the explicit group). For the remain- [t(95)  0.49, d  0.07]. (The signal detection measures ing half, the “s” and “l” key assignment was reversed. The partici- pants were asked to respond as quickly and as accurately as they such as dŒ that take into account the false alarm rates for could within 5 sec on each trial, and their RTs and accuracy were re- nonstudied items could be computed. See Figure 2 for corded. For the implicit group, the test-awareness/intentionality-to- the proportion of old responses for nonstudied items. retrieve questions became (1) “What did you think was the purpose However, because the rating scenarios were manipu- of the concreteness judgment task that you just finished?” (2) “What lated within subjects, the analyses of dŒs are functionally was your general strategy in judging word concreteness?” (3) “Did identical to the analyses of hit rates, given that the same you notice any relation between the words I showed you earlier and the words presented on the concreteness judgment test?” and z-transformed false alarm rate was subtracted from the (4) “While doing the concreteness judgment test, did you notice z-transformed hit rates for various rating scenarios.) whether you saw some of the words studied in the earlier list?” For RT, items rated in moving, pleasantness, and survival scenarios were judged faster than nonstudied items in the Results implicit group, again showing robust priming effects [all The analytic procedure was the same as that in Ex- ts(215)  2.56, ds  0.25]. Regarding the comparisons periment 1. The mean proportions of unrated words among studied items rated in the different scenarios, were 1.2% (moving, 0.9%; pleasantness, 1.2%; survival, whereas for the implicit group none of the comparisons 1.6%) for the explicit group and 1.2% (moving, 1.1%; approached significance [all ts(215)  1.06, ds  0.10], pleasantness, 1.2%; survival, 1.4%) for the implicit group. for the explicit group, items rated in the survival scenario The proportion of unrated words was very low and did not yielded faster hit responses than did those rated in the differ across the three rating scenarios (all ps  .08, ds  pleasantness scenario [t(95)  4.71, d  0.68] and those 0.24). Hence, we used the full set of data in the following rated in the moving scenario [t(95)  5.74, d  0.83], and analyses. the hit response RTs in the latter two conditions did not Rating phase. The mean ratings/RTs for the moving, differ [t(95)  1.25, d  0.18]. Thus, the overall pattern pleasantness, and survival scenarios were 2.76/1,939 msec, duplicated that of the accuracy data. 3.08/1,908 msec, and 2.80/1,983 msec, respectively, for On the basis of smallest Cohen’s d among the explicit the explicit group, and 2.89/1,799 msec, 3.08/1,779 msec, group’s within-subjects differences in accuracy and in RT and 2.96/1,852 msec, respectively, for the implicit group. between the survival and pleasantness scenarios and be- There was no interaction associated with group for RT tween the survival and moving scenarios (i.e., .39), the or ratings [both Fs(2,620)  2.75, h 2p s  .01]. Collapsed power to detect similar differences (with p  .05, two- 1116 TSE AND ALTARRIBA 1.00 1.00 Moving Pleasantness .90 .90 Survival Proportion of Correct Judgment Proportion of “Old” Responses Nonstudied .80 .80 .70 .70 .60 .60 .50 .50 .40 .40 .30 .30 .20 .20 .10 .10 0 0 Implicit Explicit 1,400 Moving 1,400 Pleasantness RT for “Old” Responses (msec) 1,300 Survival 1,300 RT for Correct Trials (msec) Nonstudied 1,200 1,200 1,100 1,100 1,000 1,000 900 900 800 800 700 700 600 600 Implicit Explicit Figure 2. Mean proportions and median response times (RTs) for correct judgments for the implicit group and old responses for the explicit group as a function of scenario and group in Experiment 2. Error bars indicate standard errors of the means. Some partici- pants did not produce any false alarms, so RTs for the old responses for nonstudied test items were based on 84 participants only. tailed) for the implicit group was .98, with a sample size the concreteness judgment task. The group (aware–intend, of 216. Hence, the absence of the effects in the implicit aware–unintend, or unaware–unintend)  scenario inter- group is unlikely to have been due to insufficient statisti- action was not significant for accuracy [F(6,639)  1.00, cal power. MSe  0.01, h 2p  .10] or for RT [F(6,639)  0.48, MSe  Test awareness/intentionality to retrieve in implicit 22,806, h 2p  .004], nor was the main effect of group for memory. As in Experiment 1, we classified participants accuracy [F(2,213)  0.52, MSe  0.03, h 2p  .005] or RT in the implicit group into three groups on the basis of [F(6,639)  0.85, MSe  173,967, h 2p  .008]. Thus, nei- their test awareness and intentionality to retrieve: aware– ther test awareness nor intentionality to retrieve enhanced intend (n  3), aware–unintend (n  201), and unaware– participants’ RT or accuracy in the concreteness judgment unintend (n  12). The speeded nature of the concrete- task, although these results should be interpreted with ness judgment task did discourage the participants from caution, due to the very small sample size in the aware– intentionally retrieving the studied items, relative to the intend and the unaware–unintend groups. (The pattern stem-cued completion task, where the participants were of results remained the same in the overall analyses after given a 12-sec response deadline. In contrast to 34 out taking out the 3 test-aware participants who intended to of 180 participants (18.9%) in Experiment 1’s stem-cued retrieve during the implicit memory test.) completion task, only 3 out of 216 participants (1.4%) re- In summary, using an explicit item recognition test, ported that they intentionally retrieved the studied items in we now obtained a typical survival-processing advan-  SURVIVAL PROCESSING 1117 tage in both accuracy and RT. However, we again did not better when the answer to the initial encoding question obtain any survival-processing advantages (relative to is yes than when it is no. However, one could still ask other conditions—e.g., rating words in the pleasantness whether our explicit memory data could replicate Butler or moving scenario) in the implicit memory test, despite et al.’s Experiment 1 effect of initial relevance ratings the fact that we did find a robust priming effect (relative within each scenario. They showed that participants better to the nonstudied items). Because the conceptual implicit recalled words that were judged to be more relevant (i.e., memory test tapped the meaning of the test items, even higher initial ratings), whether the scenario was survival, in this optimal task condition survival processing still did moving, or pleasantness. More important, in our experi- not boost implicit memory performance, as opposed to its ments, implicit memory might have been enhanced by robust effect on explicit memory performance.2 survival processing when the words were rated very high in relevance in the survival scenario, and thus, a genuine GENERAL DISCUSSION survival-processing advantage on implicit memory might have been masked. To address all these issues, To our knowledge, the present study is the first to exam- we followed Butler et al.’s Experiment 1 procedure and ine whether a survival-processing advantage, which has recoded participants’ accuracy and RT in the explicit and been reported in free recall and item recognition, would implicit memory tests as a function of their relevance generalize to another explicit memory task (i.e., stem-cued ratings in each of the three scenarios. About 30%– 40% recall), a stem-cued completion task that reflects percep- of the participants in the explicit and implicit groups did tual implicit memory, and a concreteness judgment task not use the entire rating scale and might have had missing that reflects conceptual implicit memory. We examined cells in at least one of the rating levels. To avoid excessive both RT and (typical) accuracy measures to test whether removal of data that could distort the pattern of our results, a survival-processing advantage could also occur in RT we randomly combined the raw data of 2 participants, data. The findings are straightforward. Across two experi- within each counterbalancing list, into 1 superparticipant ments, we did not find any survival-processing advantage (i.e., n  30 and 90 for implicit and explicit groups in in perceptual (stem-cued completion) or conceptual (con- Experiment 1 and n  48 and 108 for implicit and explicit creteness judgment) implicit memory tests. This was so groups in Experiment 2) before we computed their even when we found robust priming effects (relative to conditionalized data. nonstudied items) in these two tests and replicated the The explicit and implicit memory data in Experiments 1 survival-processing advantage in their analogous explicit and 2 were separately submitted to 5 (initial rating: 1–5)  memory tests (despite being quite weak in the stem-cued 3 (scenario: moving, pleasantness, or survival) repeated recall task) that were designed to follow Schacter et al.’s measures ANOVAs. Because the items in the three (1989) retrieval intentionality criterion.3 scenarios were different at each level of rating scale, this potential item selection problem complicated the The Effect of Initial Ratings on the comparison among the data in the three scenarios (see also Survival-Processing Advantage Butler et al., 2009). Thus, we focus on the linear trend Butler, Kang, and Roediger (2009, Experiments 2 effects of initial ratings on memory performance overall and 3) reported that the survival-processing advantage and within each scenario to test whether we replicated disappeared in an explicit free recall test when participants previous results and then briefly discuss the comparison were asked to rate survival-related items in a robbery of cell means for the three scenarios (see Table 2). scenario and robbery-related items in a survival scenario. For explicit memory RT data, the omnibus test of This result echoes the findings in memory research that linear trends for initial rating was not significant in Ex- participants remember items better if those items are periment 1 [F(1,29)  3.09, MSe  603,881, h 2p  .10] congruent with the way in which they are processed (cf. or Experiment 2 [F(1,47)  0.56, MSe  51,884, h 2p  Craik & Tulving, 1975) and suggests that the survival- .01]. Despite nonsignificance in the omnibus test, a closer processing advantage reported in prior studies could, in look at Experiment 1’s data showed that only the effect of part, be mediated by the congruency of the study items initial ratings on items rated in a pleasantness scenario was with the survival-related concept. In the present experi- significant [F(1,29)  7.75, MSe  473,029, h 2p  .21], ments, to rule out the possibility that a survival-processing but not the effect of those rated in the survival or moving advantage could be explained solely by the items’ being scenario [both Fs(1,29)  2.50, h 2p s  .08]. This result more relevant to the survival scenario, we controlled indicates that the participants were faster to remember the semantic similarity, as quantified by LSA cosines, between words as a function of initial ratings in the pleasantness the study items and the concepts of survival/moving (see scenario, but not in the moving or survival scenarios. Table 1). The items were also rated as being more relevant The trends look clearer in stem-cued recall accuracy and in the pleasantness scenario than in the survival scenario, recognition hit rates. The overall effect of initial ratings suggesting that the survival-processing advantage, if any, was significant in Experiments 1 [F(1,29)  15.64, MSe  in explicit and implicit memory could not be attributed 0.07, h 2p  .35] and 2 [F(1,47)  25.69, MSe  0.02, h 2p  entirely to a congruity effect (Schulman, 1974) that would .35]. In Experiment 1, the effect of initial ratings was predict that memory performance would be highest for significant for items rated in the pleasantness [F(1,29)  items rated in the pleasantness scenario. The congruity 5.67, MSe  0.06, h 2p  .16] and moving [F(1,29)  8.95, effect refers to the finding that words are remembered MSe  0.07, h 2p  .24] scenarios, but not in the survival 1118 TSE AND ALTARRIBA Table 2 Mean Proportions of Correct Recall/Completion/Recognition Hit Rates/ Concreteness Judgments and Median Response Times (RTs, in Milliseconds) for Corrected Trials As a Function of Initial Ratings, Scenario, and Group in Experiments 1 and 2 Initial Rating 1 2 3 4 5 Experiment 1’s Explicit Group Stem-Cued Recall Rate Pleasantness scenario .37 .43 .45 .48 .50 Moving scenario .40 .42 .52 .53 .57 Survival scenario .41 .46 .47 .48 .52 Stem-Cued Recall RT Pleasantness scenario 3,045 3,087 2,753 2,701 2,685 Moving scenario 2,855 3,083 3,064 2,507 2,779 Survival scenario 2,429 2,691 2,586 2,596 2,710 Experiment 1’s Implicit Group Stem-Cued Completion Rate Pleasantness scenario .51 .53 .40 .43 .41 Moving scenario .44 .46 .39 .39 .41 Survival scenario .43 .50 .41 .40 .41 Stem-Cued Completion RT Pleasantness scenario 2,453 2,481 2,659 2,416 2,522 Moving scenario 2,528 2,497 2,563 2,540 2,433 Survival scenario 2,451 2,278 2,463 2,390 2,524 Experiment 2’s Explicit Group Recognition Hit Rate Pleasantness scenario .79 .85 .85 .87 .88 Moving scenario .83 .84 .85 .85 .91 Survival scenario .83 .85 .89 .89 .91 Recognition Hit RT Pleasantness scenario 837 797 841 807 766 Moving scenario 843 802 841 790 834 Survival scenario 737 814 737 787 763 Experiment 2’s Implicit Group Concreteness Judgment Accuracy Pleasantness scenario .74 .73 .74 .72 .66 Moving scenario .73 .70 .75 .69 .76 Survival scenario .75 .74 .70 .72 .73 Concreteness Judgment RT Pleasantness scenario 917 928 934 872 890 Moving scenario 863 942 868 898 846 Survival scenario 851 905 845 864 856 scenario [F(1,29)  1.72, MSe  0.10, h 2p  .06]. Stem- in the survival scenario [F(1,89)  3.42, MSe  0.04, h p2  cued recall accuracy increased as a function of initial .04] or the moving scenario [F(1,89)  1.48, MSe  0.09, ratings in the pleasantness and moving scenarios. Despite h 2p  .02]. However, the word-stem-cued completion rates being nonsignificant, the trend did occur in the survival in the pleasantness scenario actually decreased as a function scenario (see Table 2). In Experiment 2, despite the near- of initial ratings, and a similar, albeit nonsignificant, trend ceiling performance, the effect of initial ratings was sig- was found for the word-stem-cued completion rates in the nificant in all scenarios [pleasantness, F(1,47)  9.14, survival and moving scenarios. This was contrary to the MSe  0.02, h 2p  .16; moving, F(1,47)  4.68, MSe  explicit memory data, although the decrease was mostly 0.03, h 2p  .09; survival, F(1,47)  12.48, MSe  0.01, between the ratings of 2 and 3. Nonetheless, this pattern h p2  .21]. The recognition hit rates increased as a function was not replicated in Experiment 2, when a conceptual of initial ratings in all three scenarios. Hence, despite using implicit memory test was used. The overall effect of ini- different tasks (stem-cued recall and item recognition vs. tial ratings was not significant for accuracy [F(1,107)  free recall), we replicated the effect of initial ratings on 2.57, MSe  0.04, h 2p  .02] or for RT [F(1,107)  3.44, explicit memory in Butler et al.’s (2009) Experiment 1. MSe  50,173, h 2p  .03]. None of the effects within the For the implicit memory data, in Experiment 1, none individual scenarios was significant for accuracy or RT of the effects of initial rating was significant for RT [all [all Fs(1,107)  1.79, h 2p s  .02], except the accuracy Fs(1,89)  1.30, h 2p s  .01], whereas the overall effect for items rated in the pleasantness scenario [F(1,107)  of initial ratings was significant for word completion 8.19, MSe  0.04, h 2p  .07]. However, closer inspection rates [F(1,89)  14.83, MSe  0.06, h 2p  .14]. A closer of the data showed that the decrease in accuracy in the inspection showed a significant effect in the pleasantness pleasantness scenario was mostly between the ratings of 4 scenario [F(1,89)  18.07, MSe  0.04, h 2p  .17], but not and 5. Contrary to the clear trend in accuracy in explicit  SURVIVAL PROCESSING 1119 memory tests, the effects of initial ratings on implicit scenarios were all higher than .90. Hence, the lack of a memory were unsystematic. We also analyzed the implicit survival-processing advantage in implicit memory tests memory data by combining participants on the basis of should not likely be due to the insensitivity of our tasks their test awareness and intentionality to retrieve, but the or insufficient statistical power. It could be that some overall pattern did not interact with these variables. selection pressure in our ancestral past may have led to the Keeping in mind that there could be an item selection specificity of a survival-processing advantage in explicit confound, in implicit memory tests we did not find a memory, but obviously this idea should be tested further systematic superiority of items rated in the survival in future studies. scenario, relative to those rated in the pleasantness and Although an absence of a survival-processing advan- moving scenarios across five levels of initial ratings tage in implicit memory may not be straightforwardly (see Table 2). Of the 40 pairwise comparisons for sur- explained by the strong view of memory evolution, the vival versus pleasantness/moving scenarios on all levels null effect, particularly in the concreteness judgment task, of initial ratings in all implicit memory measures in two might suggest that the survival-processing advantage that experiments, we found 34 null, 5 positive, and 1 negative occurred in explicit memory could not be attributed merely significant survival-processing advantages. The five to the deep processing that is induced via rating the study positive effects did not lie on any particular level of initial items in a survival scenario. However, this idea should be rating: two at the rating of 3 and one each at the ratings further validated in conceptual implicit memory tests that of 2, 4, and 5. Thus, it is safe to conclude that it is un- have been reported to yield the levels-of-processing effect likely that the absence of a survival-processing advantage (e.g., a modified free association task in Hourihan & Mac- in implicit memory might be due to the masking of higher Leod, 2007). If a survival-processing advantage still did initial relevance ratings. Future research needs to adapt not occur for those tests, one would be more confident that Butler et al.’s (2009) Experiments 2 and 3 manipulations the survival-processing advantage in explicit memory was (i.e., rating survival-related items in a robbery scenario not just another levels-of-processing effect. and robbery-related items in a survival scenario) in an Before concluding the present study, it is important implicit memory test to examine even further whether to consider two more alternative explanations for our a survival-processing advantage would emerge when findings. First, the lack of a survival-processing advan- survival-related items were rated in a survival scenario. tage in a conceptual implicit memory test could also be attributed to the nature of the task. Because participants Theoretical Implications of the Present Findings only need to identify whether or not the test item re- One of the characteristics of evolved memory fers to the concrete objects, the notion that their perfor- mechanisms (see Nairne, 2005; Nairne & Pandeirada, mance may not depend on the “availability” of the item 2008b) was that memory should be tuned to remem- in memory might make this task insensitive to detect any ber certain kinds of domain-specific information that survival-processing advantage, despite its being concep- is relevant to survival/fitness. It could be argued that a tual in nature. However, given that item availability, or the flexible and adaptive memory system should not depend productive nature of the memory test, is critical in detect- on retrieval intentionality, because many advantages that a ing a survival-processing advantage, it is not clear why memory system confers to an organism are more primitive the advantage can be found in explicit item recognition, in nature and less dependent on higher order processing. which also requires the participants to identify only the An organism can be benefited by the retrieval of previous study status of test items. (Of course, some could postu- survival-relevant episodes, independently of whether late that recall-like processes are probably involved in the retrieval is explicit or not. For example, one may refuse to explicit item recognition test [e.g., Yonelinas, 2002].) Fu- eat certain foods without intentionally remembering prior ture researchers should use a conceptual implicit memory experiences of nausea related to them (e.g., stomachache). test that is productive in nature (e.g., category production Hence, the strong view about memory evolution could task in Lee, 2008, and Srinivas & Roediger, 1990) and its be that a truly adaptive memory system should rely on analogous explicit memory test, in order to test whether a prior episodes even in the absence of explicit retrieval. survival-processing advantage in implicit memory could The finding that a survival-processing advantage occurred be revealed. only in explicit memory, but not in implicit memory, Second, one could argue that the absence of a survival- seems to contradict, or at least call for modification of, processing advantage in implicit memory was due to a this strong view. Although previous studies used, for in- response mismatch between the study phase (rating the stance, category verification (Mulligan & Peterson, 2008) items in various scenarios) and the test phase (completing and animacy decision (Zeelenberg & Pecher, 2003), but the word stem or judging word concreteness). On the basis not concreteness judgment, as their conceptual implicit of transfer-appropriate processing (Roediger, 1990), the memory tests, the robust overall priming effect (i.e., reinstatement of prior cognitive operations should boost relative to nonstudied items) observed in the concreteness the priming effect, and the match between encoding and judgment task clearly indicates its sensitivity. A similar retrieval operations can be quite domain specific. For ex- point can also be made for the stem-cued completion task, ample, Pilotti, Gallo, and Roediger (2000) showed that due to the significant overall priming effects. In addition, change in voice in auditory presentation between study and the statistical powers for detecting potential priming test reduced priming. Future studies should test whether differences between survival and moving/pleasantness a survival-processing advantage could occur in implicit 1120 TSE AND ALTARRIBA memory when the task demand is entirely identical at The latent semantic analysis theory of the acquisition, induction, and study and at test (e.g., decide whether items are relevant representation of knowledge. Psychological Review, 104, 211-240. Lee, Y. S. (2008). Levels-of-processing effects on conceptual automatic in a survival scenario both at study and at test; see Hughes memory. European Journal of Cognitive Psychology, 20, 936-954. & Whittlesea, 2003, for a similar design). Lu, H. J., & Chang, L. (2009). Kinship effect on subjective temporal distance of autobiographical memory. Personality & Individual Dif- Conclusion ferences, 47, 595-598. In two experiments, we found that a survival-processing Mulligan, N. W., & Peterson, D. (2008). Attention and implicit mem- ory in the category verification and lexical decision tasks. Journal advantage occurs when the tasks require the intentionality of Experimental Psychology: Learning, Memory, & Cognition, 34, to retrieve (e.g., retrieving a studied item to complete a 662-679. word stem or judging whether the item was studied before), Nairne, J. S. (2005). The functionalist agenda in memory research. In but not when they did not require the intentionality to A. F. Healy (Ed.), Experimental cognitive psychology and its applica- tions: Festschrift in honor of Lyle Bourne, Walter Kintsch, and Thomas retrieve (e.g., filling out a stem cue with the first word Landauer (pp. 115-126). Washington, DC: American Psychological that comes to mind or judging whether the word refers to Association. a concrete object). These findings demonstrate that under Nairne, J. S., & Pandeirada, J. N. S. (2007, November). Adaptive some situations, survival processing may not necessarily memory: Is survival processing special? Paper presented at the 48th boost memory performance, relative to other encoding Annual Meeting of the Psychonomic Society, Long Beach, CA. Nairne, J. S., & Pandeirada, J. N. S. (2008a). Adaptive memory: Is strategies (e.g., pleasantness rating). The survival- survival processing special? Journal of Memory & Language, 59, 377- processing advantage should be further tested using other 385. perceptual and conceptual implicit memory tests in future Nairne, J. S., & Pandeirada, J. N. S. (2008b). Adaptive memory: Re- studies in order to generalize the present findings and to membering with a stone-age brain. Current Directions in Psychologi- cal Science, 17, 239-243. examine whether the advantage in explicit memory could Nairne, J. S., & Pandeirada, J. N. S. (2010). Adaptive memory: An- be attributed solely to the deep processing triggered by the cestral priorities and the mnemonic value of survival processing. Cog- relevance ratings in the survival scenario. nitive Psychology, 61, 1-22. Nairne, J. S., Pandeirada, J. N. S., Gregory, K. J., & Van Arsdall, AUTHOR NOTE J. E. (2009). Adaptive memory: Fitness-relevance and the hunter- gatherer mind. Psychological Science, 20, 740-746. We thank Colin MacLeod for his constructive comments on an earlier Nairne, J. S., Pandeirada, J. N. S., & Thompson, S. R. (2008). Adap- version of this article and Ines Martinovic, Genevieve O’Brian, and Renee tive memory: The comparative value of survival processing. Psycho- Pangburn for their help with data collection. Correspondence concerning logical Science, 19, 176-180. this article should be addressed to C.-S. Tse, Department of Educational Nairne, J. S., Thompson, S. R., & Pandeirada, J. N. S. (2007). Adap- Psychology, Chinese University of Hong Kong, New Territories, Hong tive memory: Survival processing enhances retention. Journal of Ex- Kong, China (e-mail: [email protected]). perimental Psychology: Learning, Memory, & Cognition, 33, 263- 273. REFERENCES Nelson, D. L., McEvoy, C. L., & Schreiber, T. A. (2004). The Uni- versity of South Florida word association, rhyme, and word fragment Bowers, J. S., & Schacter, D. L. (1990). Implicit memory and test norms. Behavior Research Methods, Instruments, & Computers, 36, awareness. Journal of Experimental Psychology: Learning, Memory, 402-407. & Cognition, 15, 763-778. Otgaar, H., Smeets, T., & van Bergen, S. (2010). Picturing survival Bradley, M. M., & Lang, P. J. (1999). Affective norms for English memories: Enhanced memory after fitness-relevant processing occurs words (ANEW): Instruction manual and affective rating (Tech. Rep. for verbal and visual stimuli. Memory & Cognition, 38, 23-28. C-1). Gainesville: University of Florida, NIMH Center for Emotion Pilotti, M., Gallo, D. A., & Roediger, H. L., III (2000). Effects of & Attention. hearing words, imagining hearing words, and reading on auditory im- Buchner, A., & Wippich, W. (2000). On the reliability of implicit and plicit memory tests. Memory & Cognition, 28, 1406-1418. explicit memory measures. Cognitive Psychology, 40, 227-259. Roediger, H. L., III (1990). Implicit memory: Retention without re- Butler, A. C., Kang, S. H. K., & Roediger, H. L., III (2009). Con- membering. American Psychologist, 45, 1043-1056. gruity effects between materials and processing tasks in the survival Roediger, H. L., III, & McDermott, K. B. (1993). Implicit memory in processing paradigm. Journal of Experimental Psychology: Learning, normal human subjects. In F. Boller & J. Grafman (Eds.), Handbook of Memory, & Cognition, 35, 1477-1486. neuropsychology (Vol. 8, pp. 63-131). Amsterdam: Elsevier. Cohen, J. (1988). Statistical power analysis for the behavioral sciences Roediger, H. L., III, Weldon, M. S., Stadler, M. L., & Riegler, (2nd ed.). Hillsdale, NJ: Erlbaum. G. L. (1992). Direct comparison of two implicit memory tests: Word Craik, F. I. M., & Tulving, E. (1975). Depth of processing and the fragment and word stem-cued completion. Journal of Experimental retention of words in episodic memory. Journal of Experimental Psy- Psychology: Learning, Memory, & Cognition, 18, 1251-1269. chology: General, 104, 268-294. Rudine, J., Craig, C., Overbeek, R. W., & Green, K. (2009, Novem- Glenberg, A. M. (1997). What memory is for. Behavioral & Brain Sci- ber). It’s not who you are, but what you have: Survival processing for ences, 20, 1-55. nouns versus traits. Poster presented at the 50th Annual Meeting of the Hourihan, K. L., & MacLeod, C. M. (2007). Capturing conceptual Psychonomic Society, Boston. implicit memory: The item it takes to produce an association. Memory Schacter, D. L., Bowers, J. S., & Booker, J. (1989). Intention, aware- & Cognition, 35, 1187-1196. ness, and implicit memory: The retrieval intentionality criterion. In Hughes, A. D., & Whittlesea, B. W. A. (2003). Long-term semantic S. Lewandowsky, J. Dunn, & K. Kirsner (Eds.), Implicit memory: transfer: An overlapping-operations account. Memory & Cognition, Theoretical issues (pp. 47-65). Hillsdale, NJ: Erlbaum. 31, 401-411. Schulman, A. I. (1974). Memory for words recently classified. Memory Kang, S. H. K., McDermott, K. B., & Cohen, S. M. (2008). The mne- & Cognition, 2, 47-52. monic advantage of processing fitness-relevant information. Memory Shaw, R. J. (1997). Unprimed stem completion is only moderately pre- & Cognition, 36, 1151-1156. dicted by word frequency and length. Behavior Research Methods, Klein, S., Cosmides, L., Tooby, J., & Chance, S. (2002). Decisions Instruments, & Computers, 29, 401-424. and the evolution of memory: Multiple systems, multiple functions. Sherry, D. F., & Schacter, D. L. (1987). The evolution of multiple Psychological Review, 109, 306-329. memory systems. Psychological Review, 94, 439-454. Landauer, T. K., & Dumais, S. T. (1997). A solution to Plato’s problem: Srinivas, K., & Roediger, H. L., III (1990). Classifying implicit  SURVIVAL PROCESSING 1121 memory tests: Category association and anagram solution. Journal of 2. Rudine et al. (2009) found that relative to self-referential and pleas- Memory & Language, 29, 389-412. antness ratings, survival ratings enhanced memory for concrete nouns, Weinstein, Y., Bugg, J. M., & Roediger, H. L., III (2008). Can the but not for items that describe personality traits, and argued that survival survival recall advantage be explained by basic memory processes? processing may be effective only for concrete words, but not for abstract Memory & Cognition, 36, 913-919. words. However, postġhoc analyses for the present data showed that word Yonelinas, A. P. (2002). The nature of recollection and familiarity: type (concrete vs. abstract) did not interact with any variables, including A review of 30 years of research. Journal of Memory & Language, the type of memory test (explicit vs. implicit) in both experiments (all 46, 441-517. Fs  1.04), although it should be noted that in our abstract word pool, Zeelenberg, R., & Pecher, D. (2003). Evidence for long-term cross- very few items referred to personality traits (see the Appendix). language repetition priming in conceptual implicit memory tests. 3. The present study is not intended to tease apart or test any one of the Journal of Memory & Language, 49, 80-94. extant implicit memory theories (e.g., process vs. system view). According to the perspective of transfer-appropriate processing (Roediger, 1990), NOTES priming effects should be greater when the overlap in cognitive operations at study and at test becomes larger. Given that the rating tasks involved 1. We collected data regarding participants’ experiences watching the processing of word meaning, one would expect a larger priming effect survival-related TV programs by asking them, “How often do you watch in a conceptual implicit memory test (Experiment 2) than in a perceptual survival-related television programs (e.g., Survivor, Lost)?” at the end of implicit memory test (Experiment 1). However, these two implicit mem- the experiment. Their responses were categorized into four levels (never, ory tests differed not only in the type of processing (perceptual vs. con- sometimes, frequently, and always). However, this variable did not inter- ceptual), but also in the response mode: Stem-cued completion was a act with scenario in RT or accuracy for explicit and implicit groups in production test, and concreteness judgment was an identification test. either experiment. Therefore, the data from these two tests may not be directly comparable. APPENDIX Stimulus Lists in Experiments 1 and 2 and Their Stem Cues in Experiment 1 Critical Items ambition–amb circus–cir garbage–gar messy–mes scream–scr angry–ang color–col greed–gre modest–mod serious–ser basket–bas cottage–cot guilty–gui moment–mom spouse–spo beach–bea cruel–cru health–hea moral–mor stupid–stu bench–ben death–dea holiday–hol mosquito–mos swamp–swa black–bla devil–dev hospital–hos mother–mot talent–tal bored–bor diamond–dia hotel–hot movie–mov tennis–ten bouquet–bou dirty–dir hungry–hun music–mus trophy–tro brave–bra dress–dre infant–inf offend–off trust–tru breeze–bre engine–eng journal–jou panic–pan twilight–twi bride–bri excuse–exc legend–leg queen–que vacation–vac building–bui fabric–fab lonely–lon quick–qui vanity–van cabinet–cab failure–fai machine–mac reptile–rep virtue–vir candy–can finger–fin manner–man ridicule–rid weapon–wea cellar–cel flower–flo material–mat rigid–rig whistle–whi church–chu friend–fri medicine–med rough–rou window–win Primacy Buffer Items and Filler Items in the Memory Test abuse–abu divorce–div justice–jus poverty–pov sugar–sug adult–adu dollar–dol kindness–kin prairie–pra sunset–sun agony–ago elbow–elb knife–kni prison–pri swift–swi alive–ali fantasy–fan letter–let puppy–pup taste–tas alone–alo fault–fau lottery–lot rabbit–rab terrific–ter answer–ans field–fie luxury–lux rainbow–rai thief–thi autumn–aut filth–fil melody–mel razor–raz tobacco–tob blossom–blo fungus–fun mountain–mou reward–rew travel–tra bottle–bot glory–glo nasty–nas river–riv triumph–tri bullet–bul grief–gri noisy–noi robber–rob unhappy–unh butter–but happy–hap ocean–oce sapphire–sap vehicle–veh cliff–cli highway–hig option–opt secure–sec violent–vio coast–coa honey–hon patient–pat skull–sku warmth–war corpse–cor house–hou pencil–pen snake–sna water–wat crime–cri illness–ill perfume–per sphere–sph woman–wom dagger–dag immoral–imm pillow–pil spring–spr world–wor delight–del injury–inj pizza–piz statue–sta yacht–yac dinner–din jelly–jel pleasure–ple stink–sti young–you (Manuscript received December 22, 2009; revision accepted for publication May 10, 2010.)