Testing the repression hypothesis: Effects of emotional valence on memory suppression in the think – No think task

https://doi.org/10.1016/j.concog.2009.09.004Get rights and content

Abstract

It has been proposed that performance in the think – no think (TNT) task represents a laboratory analogue of the voluntary form of memory repression. The central prediction of this repression hypothesis is that performance in the TNT task will be influenced by emotional characteristics of the material to be remembered. This prediction was tested in two experiments by asking participants to learn paired associates in which the first item was either emotionally positive (e.g. joy) or emotionally negative (e.g. hatred). The second word was always emotionally neutral (e.g. socks). Consistent with the repression hypothesis, significant memory suppression was observed in both experiments following ‘no think’ instructions for memories associated with emotionally negative material. No suppression was observed for memories associated with emotionally positive information. Implications of these findings for the relationship between performance in the TNT task and the controversial notion of memory repression are considered.

Introduction

In this paper we report two experiments exploring effects of emotional valence in the think – no think (TNT) task developed by Anderson and colleagues (Anderson and Green, 2001, Anderson et al., 2004). These authors have claimed that performance in the TNT task represents a laboratory analogue of the voluntary form of repression described in Freud’s early writing (see Erdelyi, 2006). We refer to this proposal below as the repression hypothesis. We manipulated the emotional valence of memory materials in a TNT task in order to test the prediction of this hypothesis that memory impairment in the think – no think task will be more apparent for items associated with emotionally negative than emotionally positive information.

Although being unable to remember is usually regarded negatively, and characterised as cognitive failure, we can all bring to mind events and situations that we might prefer to forget. Indeed, the adaptive value of forgetting has been stressed in a recent theoretical analysis of memory by Anderson and Schooler (2000). Forgetting might be advantageous for at least two reasons. Firstly, forgetting might be useful for optimising cognitive efficiency. The ability to forget irrelevant and unwanted details may be necessary for efficient mental functioning, and may be integral to the development of memory schemas and the capacity for abstract thought. A second, more controversial proposal is that forgetting emotionally painful and traumatic episodes might play a role in emotional regulation. The notion of memory repression has, of course, long been a controversial topic in psychology, and the heat and vigour of this controversy show little sign of abating (e.g. see Erdelyi, 2006).

The adaptive utility of forgetting and the issue of conscious control, whether one can recruit executive processes in order to deliberately forget an episode, are related but logically separable issues. Forgetting might be adaptively useful but operate unconsciously, remaining impermeable to conscious control. Alternatively, forgetting might be accessible to executive control mechanisms. Intriguing evidence in favour of the latter possibility was reported by Anderson and Green (2001) in the influential journal, Nature. These authors devised an ingenious variant of the classic paired associate learning task. Their think – no think (TNT) version of this task proceeded in three stages. In the learning stage participants learned to associate pairs of unrelated words, such as ordeal – roach or street – book, and learning was assessed by the ability to produce the second (response) word when provided with the first (cue) word. After achieving a learning criterion of at least 50% correct, participants proceeded to the think – no think stage, in which participants were presented with cue words and were instructed to either say the response word (think condition) or to attend to the cue word while trying to keep the response word out consciousness (no think condition). In both conditions the number of times each cue word was presented was varied (1, 8 or 16 repetitions). In the final (memory test) stage participants’ memory for all word pairs, including those in the no think condition, was assessed. Uncontroversially, memory improved with repeated recall and rehearsal attempts in the think condition. The interesting and surprising feature of their results was that in the final test, memory for no think items was worse than a baseline condition comprising items that were learned initially, but which did not appear in either the think or no think condition. Furthermore, performance became progressively worse as the number of cue word repetitions in the no think condition increased (1, 8 or 16 repetitions). That is repeated attempts at not thinking about a previously learned response word, while attending to a cue word, led to a progressive inability to recall these words subsequently. Performance at the memory test stage was assessed in two ways. In the same probe condition participants were shown the original cue words and attempted to produce the response words that were learned earlier. In the independent probe test, participants were presented with a category cue together with an initial letter (e.g. insect – R) and attempted to produce an appropriate word from the initial learning list. Their observation that memory impairment in the no think condition was comparable with same and independent probe tests of memory was seen as important, because it suggested that the memory difficulty was not specific to the learned associations between item pairs. Memory impairment in this situation might be due to an interference effect which occurs in the no think condition because participants think of new associations (e.g. ordeal – bungy jump, ordeal – prostate exam) as a way of keeping the original word (roach) out of consciousness. These new association(s) would then interfere with the original memory for the link between ordeal and roach. However, Anderson and Green (2001; see also Levy & Anderson, 2002) argued that interference of this kind would not affect performance when memory is tested by the independent probe technique, because in this case the retrieval process is independent of the specific associative links formed during the initial learning phase. The presence of significant memory impairment in the independent probe condition was, argued Anderson and Green, consistent with the proposal that the memory representation itself is less accessible, due a process of inhibition operating under executive control under ‘no think’ instructional conditions.

The theoretical interpretation offered by Anderson and Green (2001) for these intriguing findings appealed to two broad explanatory frameworks. One aspect of their explanation appealed to the notion of executive control processes (see also Levy & Anderson, 2008). That is, after the initial learning phase, the response word becomes a relatively automatic and prepotent response to the cue word. The instruction to avoid thinking about the response word in the no think condition can therefore be viewed as analogous to response over-ride instructions in other tasks, such as the go/no go task, which have been linked to the executive functions associated with the frontal lobes (Garavan & Ross et al., 2002).

A second feature of their interpretation was to propose that the memory suppression effect observed in the TNT task could plausibly be viewed as a laboratory analogue of the voluntary form of Freudian memory repression (see also Conway, 2001). This latter claim attracted critical commentary from authors such as Kihlstrom (2002) who pointed out that a modest reduction in the accuracy of remembering a set of emotionally neutral words in the laboratory may not be comparable to the wholesale loss of entire autobiographical episodes as a result of emotional trauma, which is proposed to occur in clinical cases of memory repression (see also Hayne, Garry, & Loftus, 2006).

The degree of disagreement that currently exists among memory researchers concerning the scientific status of repression could hardly be greater. For example, according to Erdelyi (2006) “Repression has become an empirical fact that is at once obvious and problematic.” (p. 499). In stark contrast, Kihlstrom (2002) proposes that “The repression (or suppression) of trauma appears to be a clinical myth in search of scientific support” (p. 502). A major source of difficulty here is that different authors have adopted radically different definitions of memory repression. Erdelyi (2006) defines repression broadly, as a “consciousness-lowering process” that “consists of a class of operations that reduces the accessibility to consciousness of some target material” (p. 502). Erdelyi’s conception of repression includes both inhibitory (subtractive) and elaborative (additive) memory processes, and this broad view of repression clearly encompasses paradigms which involve deliberate cognitive avoidance, such as the think – no think and directed forgetting tasks (Bjork, Bjork, & Anderson, 1998). In contrast, authors such as Kihlstrom, 2006a, Kihlstrom, 2006b adopt a narrower definition, in which repression is viewed as a (hypothetical) unconscious defence mechanism that causes dramatic forgetting of traumatic experiences. Repression defined in this way has little or nothing to do with findings from the think – no think paradigm because the latter involves: (i) conscious control, (ii) modest rather than dramatic forgetting, and (iii) does not involve trauma of any kind. In the current research, the conception of repression that we adopted was closer to the broad view of Erdelyi (2006) and also closer to the original description by Freud, who wrote that “the essence of repression lies simply in turning something away, and keeping it at a distance from the conscious” (Freud, 1957). Indeed this simple definition of repression also captures the central feature of the think – no think task, in which participants are asked to keep a recently learned association out of consciousness.

Following their publication, Anderson and Green’s findings attracted widespread attention, and numerous attempts to replicate their findings have now been reported. The success of these replication studies has been decidedly mixed. Bulevich, Roediger, Balota, and Butler (2006) reported three experiments in which the procedures described by Anderson and Green were followed carefully. In none of the three experiments was there any evidence, or even tentative evidence, for impaired memory in the no think condition compared to baseline. In an electrophysiological study, Bergstrom, Velmans, De Fockert, and Richardson-Klavehn (2007) contrasted ERP waveforms elicited by cue words under think and no think conditions. Although ERPs in response to think and no think items displayed different temporal and topographic characteristics, these electrophysiological features were not accompanied by any memory effects: Accuracy in the no think condition was almost identical to baseline. Hertel and Gerstle (2003) presented participants with adjective-noun pairs that varied in emotional valence (e.g. cosy – chair vs. electric – chair). They found that non-dysphoric (control) participants recalled less than dysphoric participants in the no think condition. However, statistically significant memory suppression was not observed in either group. Hertel and Calcaterra (2005) also used related adjective–noun pairs, but in this study the pairs were emotionally neutral (e.g. porcelain – doll). The aim of the study was to discover whether thought substitution (i.e. thinking about something else during no think trials) played an important role in the memory suppression effect. To this end, performance was compared between two groups. The aided group were provided with substitute words and were encouraged to rehearse these substitute associates as a way of avoiding thinking about the original word during no think trials (e.g. porcelain – goblet as a substitute for porcelain – doll). Significant memory suppression was observed in this group, but not in an unaided group, who were simply instructed to avoid thinking about the original response word, as in the original Anderson and Green (2001) study. However, within the unaided condition significant memory suppression was observed for those participants who spontaneously adopted a substitution strategy and reported using it frequently. Because Hertel and Calcaterra (2005) failed to include an independent probe test of memory the most parsimonious interpretation of their findings is perhaps in terms of classic interference theory, rather than inhibition of memory representations. That is, rehearsal of substitutes during ‘no think’ trials produced a classic retroactive interference effect, impairing recall of the original response words.

In a study by Joormann, Hertel, Brozovich, and Gotlib (2005) participants learned pairs in which a neutral cue word was paired with a response word that could be either positive or negative in emotional valence (e.g. lamp – pain, curtain – humour). The performance of participants who met DSM-IV criteria for major depressive disorder was compared with a control group comprising non-depressed individuals. The overall performance of non-depressed participants failed to display a memory suppression effect, although significant suppression did emerge after a somewhat complex procedure was applied, in which one third of the participants were excluded on the basis of apparent non-compliance with instructions, assessed via questionnaire. Although, earlier work by the same group (Hertel & Gerstle, 2003) had shown that dysphoric participants recalled more items than control participants in the suppression condition, in this study it was the clinically depressed group who showed significant memory suppression. However, this effect was modulated by emotional valence. Somewhat counter-intuitively, the depressed participants were able to suppress memory for emotionally negative noun pairs (e.g. lamp – pain), but were unable to suppress memory for the positive pairs (e.g. curtain – humour). Wessel, Wetzels, Jelicic, and Merckelbach (2005) compared TNT task performance of individuals with high and low scores on a questionnaire measure of dissociative experiences. Emotionally neutral noun pairs were used, and significant memory suppression was observed in the same-probe test, but not in the independent probe test. Contrary to the authors’ prediction there was no difference between high and low dissociative individuals with respect to the degree of memory suppression. Salame and Danion (2007) predicted that healthy controls would show memory suppression in the TNT task, but schizophrenic participants would not. Surprisingly, they found the reverse: in the same probe condition significant memory suppression was observed for the schizophrenic group, but not for the healthy controls. In the independent probe condition, there were no memory suppression effects, for either schizophrenic participants or controls.

Set against these failures, or partial failures to replicate Anderson and Green’s (2001) memory suppression effect are several studies reporting positive results. Anderson et al. (2004) reported a behavioural replication of their earlier finding, together with brain imaging data suggesting that memory suppression may involve a network in which hippocampal activity is modulated by executive control via activity in dorsolateral prefrontal cortex. Depue, Banich, and Curran (2006) reported two experiments in which emotionally neutral faces were paired with emotionally neutral or emotionally negative words (Experiment 1), and with emotionally neutral or emotionally negative pictures (Experiment 2). In both experiments memory suppression effects appeared to be greater for emotionally negative than for neutral material. In a recently published study, Depue, Curran, and Banich (2007) replicated their memory suppression effect using emotionally neutral faces paired with emotionally negative pictures. In addition, they reported fMRI data suggesting that memory suppression may entail a two stage process, in which frontal regions interact initially with brain regions associated with sensory processing, and in a second stage with the hippocampus and amygdala, which are known to be involved with memory and emotional processing. In another recently published study Lee, Lee, Lee, and Tsai (2007) reported an interesting and potentially important finding. These authors used (Chinese) word pairs as stimuli, and compared memory performance in a condition where the respond and suppress cues were presented relatively briefly (3 s) or for a slightly longer period (5 s). The original Anderson and Green study had exposed respond – suppress cues for 4 s. Lee, Lee et al. (2007) assessed memory at final test using the same probe technique, and observed a clear memory suppression effect with briefly exposed (3-s) respond – suppress cues, but not with longer (5-s) cues. Without further data it is difficult to arrive at a clear interpretation for this finding, but it suggests that temporal parameters of the TNT task may be a crucial factor mediating memory suppression.

The overall impression gained from this survey is that memory suppression in the think – no think task is a decidedly elusive phenomenon. In two of the studies just described the effect emerged in clinical samples (depression and schizophrenia) but not in the non-clinical control groups (Joormann et al., 2005, Salame and Danion, 2007). Temporal parameters of the task (Lee, Lee et al., 2007), participant strategies (Hertel & Calcaterra, 2005) and the emotional valence of the materials all appear to be important (Depue et al., 2006, Joormann et al., 2005).

In both their original report Anderson and Green (2001) and in Anderson et al., 2004) memory suppression the TNT task was explicitly related to Freud’s notion of memory repression. Perhaps surprisingly, there was no explicit manipulation of emotional valence in either of these studies (see Schacter, 2001). If TNT memory suppression and Freudian memory repression do indeed arise from common mechanisms then one would expect emotional valence to be an important factor, because the latter is associated with emotionally negative and distressing experiences. Depue et al. (2006) compared memory suppression with neutral and emotionally negative material, and found reliable memory suppression with the latter but not the former. However, their design did not include an emotionally positive condition. Therefore, at this stage it is unclear whether the findings of Depue et al. (2006) are dependent on emotionality itself (emotional vs. neutral material) or whether they are dependent on the valence of the emotional materials used, which was negative in tone. If TNT suppression and repression are related then one might expect emotional valence to be critically important, because repression is specifically thought to affect memory for negative rather than positive experiences. Therefore in the two experiments reported below we compared TNT suppression for material associated with emotionally negative (e.g. hatred) and emotionally positive (e.g. joy) words. If TNT suppression and Freudian repression arise from common mechanisms (the repression hypothesis) then one would expect to observe suppression in response to emotionally negative material but not in response to emotionally positive material.

A notable feature of our design is that we used a paired associate learning task in which the emotional valence of the cue word was manipulated. The emotional valence of the response word was always neutral. In several previous studies investigating effects of emotionality on memory suppression, the valence of the response word has been manipulated (e.g. Depue et al., 2006, Joormann et al., 2005 – to our knowledge the current study is the first to examine effects of cue word valence on memory suppression. An important advantage of manipulating emotional valence of the first (cue), rather than the second (response) item of each word pair is methodological. With this design our manipulation of emotional valence was deconfounded from stimulus factors affecting word memorability, because participants attempted to remember the same set of neutral words in both valence conditions.

In several publications Anderson and colleagues have stressed the importance, in terms of theoretical interpretation, of observing memory suppression with both same probe and independent probe tests of memory (see Anderson and Green, 2001, Anderson et al., 2004, Levy and Anderson, 2002). As explained earlier, the presence of memory suppression in the independent probe condition is important for evaluating whether classic retroactive interference, or inhibition of memory representations via executive control is the best explanation for the effect. Although Anderson et al. (2004) reported memory suppression in both a same probe and independent probe condition, all the replications and partial replications by independent investigators, as described above, have employed same probe memory tests. To our knowledge no independent replication of memory suppression with an independent probe memory test has yet been reported. Accordingly, the experiments reported here included both same probe and independent probe conditions.

We report two experiments below. Experiment 1 used a between participants design in which emotional valence was used as a cue to either rehearse or suppress a previously learned association. One group of participants were instructed that during the think – no think phase they should rehearse a previously learned association when presented with an emotionally positive word, but avoid thinking of a previously learned association when presented with an emotionally negative word. A second group of participants were given the reverse instructions. We refer to these groups below as the Suppress Negative and Suppress Positive groups, respectively. Experiment 2 used a within participants design in which participants used word colour as the cue to either rehearse or suppress. Participants were instructed that during the think – no think phase they should rehearse in response to words presented in green and suppress in response to words presented in red. The design of Experiment 2 enabled us to perform within participant comparisons of memory suppression with negative and positive words. It was important to establish the reliability of effects observed in the first experiment using a within participants design, as a prelude to a neuroimaging study involving within participant comparisons of both memory performance and brain activation as a function of cue word valence.

Section snippets

Participants

Twenty eight adults volunteered to take part in the Suppress Negative Condition and 26 adults volunteered to participate in the Suppress Positive Condition. The research was approved by the University of Auckland Human Participants Ethics Committee.

Materials

Twelve emotionally positive and 12 emotionally negative words were selected from the Affective Norms for English Words (ANEW) of Bradley and Lang (1999). Mean word frequencies (Kucera & Francis, 1967) of the positive and negative word sets was 41.08

Suppress negative condition

Results from the Suppress Positive condition are illustrated in Fig. 1. The percentage of correct responses in each condition was entered into a three way analysis of variance (Anova) with two repeated measures factors: think – no think (rehearse vs. suppress) and number of repetitions (0, 4, 16); and one independent groups factor, memory test (same probe vs. independent probe). Both the main effect of think – no think, F(1, 26) = 17.44, p < .001, and the interaction of this factor with repetitions,

Discussion

Results from the experiment were in agreement with the repression hypothesis. When participants attempted to suppress memory for items associated with emotionally negative words they were successful: memory was significantly worse than baseline. Furthermore, the degree of memory suppression observed in the experiment was comparable in the same probe and independent probe conditions, an observation which has important implications for theoretical interpretation of the suppression effect, as

Experiment 2

Before discussing our memory suppression findings in detail, a further experiment designed to test the repression hypothesis will be described. Our motivations for performing Experiment 2 were twofold. Firstly, in view of the patchy record of replicating the memory suppression effect it seemed important to establish the reliability of our finding that participants are able to suppress memory for items associated with negative words, but are unable to suppress memory for items associated with

Results

Results from Experiment 2 are summarised in Fig. 3. In the analyses reported below memory suppression effects were evaluated by comparing baseline performance with the suppression condition, and rehearsal effects were evaluated by comparing baseline performance with the rehearsal condition. Two further factors were included in each analysis: memory test (same probe vs. independent probe) and test order (same probe first vs. independent probe first).

Discussion

In this experiment, as in Experiment 1, participants were able to voluntarily suppress memory for words associated with emotionally negative cues, but were unable to suppress memory for items associated with emotionally positive cues. It is noteworthy that the memory suppression effects observed in Experiment 1 were replicated in Experiment 2, because the two experiment differed in two significant aspects of methodology. In Experiment 1 the emotional valence of words presented during the think

General discussion

In the two experiments reported here participants were able to suppress memory successfully for items that were associated with emotionally negative stimulus words. In contrast, participants failed miserably when trying to suppress memory for items associated with emotionally positive words. In both experiments, memory in the repeated suppression conditions was almost identical to baseline, for items associated with emotionally positive words. These findings are in agreement with the repression

Conclusion

In agreement with the repression hypothesis participants were able to suppress memory for items associated with emotionally negative words, but were unable to suppress memory for items associated with emotionally positive words. The degree of memory impairment was substantial (16% in Experiment 1, 10% in Experiment 2), and the effect was observed regardless of whether memory was probed with the original cue word (same-probe test), or with a category name and first letter cue (independent probe

Acknowledgment

This research was supported by the University of Auckland Research Committee.

References (33)

  • J.R. Anderson et al.

    The adaptive nature of memory

  • E.L. Bjork et al.

    Varieties of goal directed forgetting

  • Bradley, M., & Lang, P. (1999). Affective norms for English words (ANEW): Instruction manual and affective ratings....
  • J.B. Bulevich et al.

    Failures to find suppression of episodic memories in the think/no-think paradigm

    Memory & Cognition

    (2006)
  • M.A. Conway

    Repression revisited

    Nature

    (2001)
  • B.E. Depue et al.

    Suppression of emotional and nonemotional content in memory – Effects of repetition on cognitive control

    Psychological Science

    (2006)
  • Cited by (46)

    • Emotional contexts modulate intentional memory suppression of neutral faces: Insights from ERPs

      2016, International Journal of Psychophysiology
      Citation Excerpt :

      Hence, our third hypothesis is that neutral-faces-in-positive-scenes may be harder to forget due to attentional broadening phenomena that might hamper the intentional memory control process. This prediction agrees with evidence that positive information may impede the inhibitory control processes in think/no think (Lambert et al., 2010), attentional (Rowe et al., 2007) and motor tasks (Albert et al., 2010). If so, we could expect a lack or inversion of behavioral DF effect alongside with a reverse electrophysiological pattern in N2/P3 components (reduction of N2 and enhancement of P3 for TBF faces-in-positive-scenes).

    • Cue-independent forgetting by intentional suppression - Evidence for inhibition as the mechanism of intentional forgetting

      2015, Cognition
      Citation Excerpt :

      Finally, memory for all of the target words is tested (e.g., ordeal-r__). Results have shown that recall for the suppressed targets is worse than recall for the baseline targets (on which neither Think nor No-think training has been given), providing the first evidence that intentional suppression is able to cause memory impairment (e.g., Benoit & Anderson, 2012; Bergstrom, de Fockert, & Richardson-Klavehn, 2009; Depue, Curran, & Banich, 2007; Joormann, Hertel, LeMoult, & Gotlib, 2009; Kim & Yi, 2013; Lambert, Good, & Kirk, 2010; Levy & Anderson, 2008; Racsmany, Conway, Keresztes, & Krajcsi, 2012; van Schie, Geraerts, & Anderson, 2013; Waldhauser, Lindgren, & Johansson, 2012). Anderson and Green (2001) suggested that the underlying mechanism of voluntary suppression was different from that of the traditional interference approach.

    • Unsuccessful suppression is associated with increased neuroticism, intrusive thoughts, and rumination

      2015, Personality and Individual Differences
      Citation Excerpt :

      Suls and Martin (2005) describe a neurotic cascade; individuals high in neuroticism are motivated to engage in thought suppression, even though their attempts at suppression may prove unsuccessful, as described above. Emotionally negative information may be better suppressed than emotionally positive information (e.g., Lambert, Good, & Kirk, 2010; Noreen & MacLeod, 2013). According to the Neurotic Cascade, highly neurotic individuals appraise events as more harmful or threatening, and negative affect from stressful experiences may carry over to adjacent experiences or thoughts which may not be negative per se.

    View all citing articles on Scopus
    View full text