Influence of repression on autobiographical memories and expectations of the future

Abstract

The present study investigated repressors' self-referential processing of autobiographical memories and future events in terms of specificity and response latency. Repressors (n = 20) and nonrepressor controls (n = 17) selected from a larger pool (N = 166) completed an autobiographical memory task (AMT) and a future event task (FET). As expected, repressors were less specific than controls in reporting unpleasant experiences (anxious and depressive) but were no different to the controls in the specificity of their reports of pleasant experiences. This pattern of group differences applied to both past and future time frames although all participants were more specific in reporting pleasant and anxious experiences than depressive experiences. Unexpectedly, repressors' response latencies were not slower for unpleasant experiences than that of controls. All participants were faster in anticipating future events than in recalling past memories. In keeping with theoretical accounts, anticipation of anxiety-related experiences was faster than recall of anxiety-related memories, whereas latencies for depression-related or pleasant-related experiences did not differ.

Keywords:

• Autobiographical memory and imageability of the future
• repression

People with a defensive cognitive style, as in repression, are thought to avoid negative affective experiences and to create a positive response bias on self-report measures of psychological functioning (e.g., Creswell & Myers, 2002). Cognitive models suggest that people possess stable cognitive structures that create biases in the way they attend to and organise their world in memory, especially as this organisation relates to the self (Conway & Bekerian, 1987). Therefore, if cognitive processing styles are characteristic of individuals' distinctive organisation of knowledge and feeling, then it follows that there should be a degree of coherence in how the self is constructed over time. Just as personal memories influence people's lives, present and future processing also relate powerfully to individuals' self-concepts, emotions, and personal meanings (Hermans, 1999; Nurmi, 1991). In this study we examined the degree of concordance in repressors' biased views of the self within autobiographical memories and images of the future.

In identifying repressive coping, Weinberger, Schwartz, and Davidson (1979) proposed four distinct coping styles characterised by different combinations of anxiety and defensiveness scores. In their typology, repression consisted of low anxiety and high defensiveness and was distinguished from other control groups: low anxious (low anxiety, low defensiveness), high anxious (high anxiety, low defensiveness), and defensive high anxious individuals (high anxiety, high defensiveness). The present investigation aimed to study the defensive processing style of repressors, relative to low anxious controls (i.e., low anxious, low defensive nonrepressors).

Several authors have proposed that repressors show a positive bias through their avoidance of unpleasant experiences and use of self-enhancements to maintain a fragile positive self-concept (e.g., Mendolia, 1999; Myers & Brewin, 1996; Shedler, Mayman, & Manis, 1993). Myers (2000) found that even when placed in potentially stressful situations, repressors reported low levels of distress despite physiological reactions such as increased heart rate and raised blood pressure. Repressors also tend to avoid threatening stimuli while interpreting ambiguous stimuli in a nonthreatening manner (Eysenck, 1997). These biases of selective attention and interpretation may explain in part the desynchrony between repressors' self-reports and observational measures. Studies have shown that repressors have a bias toward an avoidant style of information processing (e.g., Fox, 1993; Myers & McKenna, 1996). Recent research has also indicated that repressors report higher levels of romantic avoidant attachment than nonrepressors (Vetere & Myers, 2002).

A study by Davis (1987) established that repressors are further characterised by their limited access to affective memories, particularly pronounced for fearful and self-conscious experiences, relative to low anxious controls. Similarly, the Davis and Schwartz (1987) study showed that repressors, relative to controls, recalled significantly fewer negative memories, lending further support to the view that repression involves limited access to negative affective memories. In this latter study, repressors were also found to recall fewer positive memories when requested to report autobiographical memories associated with emotional word cues in a short time period. These results have been replicated and shown to apply even when repressors are provided with rich descriptions of events to aid recall (Newman & Hedberg, 1999).

Davis (1987) suggested that repressors' autobiographical memory performance may reflect an overall memory deficit in the retrieval of emotional experiences per se. In contrast, a study by Holtgrave and Hall (1995) identified distinct variations in repressors' memory recall for different types of negative stimuli. When repressors and nonrepressors were presented with positive and negative emotion cue words in a depth of processing task, the experiences recalled by repressors were significantly more positive than those recalled by nonrepressors. On a follow-up surprise memory test for the words used to elicit recall, repressors (relative to nonrepressors) recalled significantly fewer of the negative emotional words that had not elicited the recall of an experience. Although the repressors demonstrated a general memory decrement for negative emotional words, this was not the case for anxiety-related words and the authors argued that repression may not involve a memory deficit for all emotional material.

While previous research has examined repressors' processing styles using past and present experimental time frames, little research has used future time frames in the study of repression. One exception is a study by Myers and Brewin (1996), which provided evidence to suggest that repressors' interpretative biases can extend to their views of the future. That study required participants to endorse the likelihood of negative events happening to them in the future. Repressors rated the likelihood of negative events occurring as significantly less likely than did low anxious controls. This finding invites further research on the nature of views of the future in repression. Moreover, assuming that one's future represents a higher degree of uncertainty than the past, it is likely to evoke higher levels of personal threat for anxious individuals and as such would seem to be particularly relevant in the study of repression.

To date, research on memory and anticipatory biases in repression has focused on the number of memories that can be recalled and likelihood ratings for future events. However, the quality of memories accessed or anticipated for one's own experience may be a more sensitive indicator of biases. According to Williams (1996), emotional disturbance influences the way in which individuals process material related to their concept of self. For example, in the case of depression, Williams showed that cognitive biases can lead not only to a decrease in the numbers of memories related to self but also to a tendency to overgeneralise these memories, particularly positive memories. Thus, the person loses access to specific information about the self and can access only categorical information. Williams (1996) argues that accessing only categorical memories can be a defensive manoeuvre that limits the person's access to specific negative experiences (e.g., in abuse victims). Thus, in contrast to the lack of specificity for positive memories in depression, it would seem likely that the positive bias of repressors would be evidenced by a deficit in their capacity to recall specific negative memories and in framing specific negative future scenarios.

The current study set out to extend previous research on self-referential biases associated with repression by examining the specificity and accessibility of past memories and future personal experiences. It was hypothesised that repressors would be less specific in describing unpleasant experiences than low anxious controls and would be less specific in describing unpleasant experiences relative to pleasant experiences. No difference was expected between repressors and control participants on specificity for pleasant experiences. Further, as an extension of the Holtgrave and Hall (1995) preliminary findings, it was anticipated that repressors would be less specific in describing depressive experiences than anxious experiences. In terms of latencies, it was hypothesised that repressors would take longer to recall unpleasant experiences than would low anxious controls. Finally, both repressors and controls were expected to be faster to retrieve and anticipate personal pleasant experiences than unpleasant experiences.

Method

Participants

A pool of 166 undergraduate psychology students volunteered to participate in the study. This sample comprised 121 women and 45 men, with a modal age of 20 – 24 years. Participants were selected for the repressor and low anxious control groups based on their scores for self-reported anxiety, depression and social desirability (details are discussed in the section on “Selection criteria for experimental groups”).

Materials

Manifest Anxiety Scale

The 23-item Manifest Anxiety Scale (MAS; Butcher, Dahlstrom, Graham, Tellegen, & Kaemmer, 1991) was used to measure respondents' levels of trait anxiety. Items were rated on a 6-point Likert scale ranging from 1, very untrue of me, to 6, very true of me, with five items reverse scored. Butcher et al. (1991) reported high internal consistency for the MAS (α = .82).

Marlowe – Crowne Social Desirability Scale – short form

A 13-item true/false version of the Marlowe – Crowne Social Desirability Scale short form (MCSDS; Reynolds, 1982) was used to measure social desirability and defensiveness. Eight items are reverse scored. Reynolds (1982) reported that this version demonstrates acceptable alpha reliability (.76) and compares favorably with the standard 33-item MCSDS alpha reliability (.79). Myers and Brewin (1994) contest that, in contrast to other social desirability scales that measure response bias, the MCSDS can identify individuals predisposed to repressive coping styles.

Beck Depression Inventory

The Beck Depression Inventory (BDI; Beck & Steer, 1993) comprises 21 items relating to symptoms and attitudes reflecting severity of depressive symptomatology (Beck, Steer, & Garbin, 1988). Items are rated on a 4-point scale ranging from 0 to 3, yielding a score range from 0 to 63. Alpha coefficients have been cited as .87 with psychiatric patients and .91 with nonpsychiatric patients (Beck et al., 1988).

Autobiographical Memory Task

The Autobiographical Memory Task (AMT; Williams & Broadbent, 1986) yields measures of response latency and specificity of respondents' autobiographical memories. The AMT used here required participants to retrieve and write down specific personal memories corresponding to six cue words of pleasant and unpleasant hedonic tone. Two cues referred to past pleasant memories (successful, confident) and four cues referred to unpleasant memories: two, depression-related (lonely, failure) and two, anxiety-related (scared, humiliated).

Future Event Task

The Future Event Task (FET; Williams et al., 1996) measures response latency and specificity in imagining future personal events. Respondents imagine and write down future events associated with six pleasant and unpleasant emotional cue words. The two future pleasant-related cues used were capable and loved and the unpleasant future stimuli comprised two depression-related cues (dejected, isolated) and two anxiety-related cues (afraid, edgy).

Selection of cue words

Cue words were selected for the AMT and FET based on the results of a preliminary study. A pool of 115 stimulus words, considered to be neutral, pleasant-related and unpleasant-related (depression and anxiety) was drawn from words used in past research (Bentall, Kanley, & Bowen-Jones, 1995; Dalgleish, Cameron, Power, & Bond, 1995; MacLeod, 1995). The use of specific anxiety- and depression-relevant words permitted detection of any possible differentiation in participants' responses to depression- and anxiety-related cues. Separate sets of ratings were made by three independent raters and an independent clinical psychologist on a 5-point Likert Scale (1, not at all related to 5, extremely related) according to (a) pleasantness, (b) depression, and (c) anxiety relevance. The final selection was made by the independent clinical psychologist based on cues that met the required criteria among all four independent raters.

Stimulus cues were chosen following a procedure developed by Bradley, Mogg and Williams (1995). The four depression stimulus cues were drawn from those words rated ⩾ 4, for depression relevance, and < 2 for anxiety relevance. Conversely, the four anxiety stimulus cues were selected from those words rated ⩾ 4, for anxiety relevance, and < 2 for depression relevance. In addition, each unpleasant cue word was judged to be < 2 on the pleasantness scale. The four pleasant words were rated < 2 for relevance to anxiety and depression and > 4 on the pleasantness scale.

Selection criteria for experimental groups

Repressor and low anxious control groups were selected from the volunteer sample on the basis of their scores on the MAS, MCSDS and BDI. Repressors were matched with controls on depression scores to remove any differences between the groups due to depression levels. Of the 37 people who met the criteria for repression based on MAS and MCSDS scores, 20 rated ⩽ 5 and > 0 on the BDI and therefore were included in the repressor group. Participants with zero scores were excluded because some authors contend that scores of 0 on the BDI may be characterised by other forms of psychopathology (e.g., Gotlib & Cane, 1989).

Repressor group (n = 20)

Those participants who were below the total sample mean (71) on the MAS, ⩽ 5 and > 0 on the BDI, and in the upper third on the MCSDS (⩾ 10) were included in the repressor sample. This sample comprised 14 women and 6 men.

Control group (n = 17)

For inclusion into the low anxious control sample, participants were required to score below the sample mean (71) on the MAS, ⩽ 5 and > 0 on the BDI, and in the lower third on the MCSDS (⩽ 6). This group comprised 11 women and 6 men.

Procedure

Testing took place in small groups and the researcher was blind to each participant's identity and experimental status. Presentation of the AMT and FET was counterbalanced to control for any effects of order. An independent assistant randomly assigned half of the participants in each group to the FET – AMT and half to the AMT – FET format. There were no order effects for specificity or response latency (ps > .05). At the commencement of the experimental session students were paired with a trained timekeeper.

In responding to AMT and FET items participants were instructed to give enough information in their written responses for a reader to get an idea of what happened or was considered likely to happen. It was emphasised, however, that they should try to report a specific memory or specific future situation that included such aspects as, place, time and event. It was also explained that it did not matter how long ago the event had occurred or when in the future it was likely to happen. Once the participant had read each stimulus sentence he/she indicated this to the timekeeper. The timekeeper then recorded the response time taken between finishing the sentence and commencing a written response. If the person failed to recall a memory or to imagine a future situation after 60 s the timekeeper recorded 60 s on the time record sheet and directed the participant to the next cue. There was no time limit for written responses.

Prior to the experimental cues, participants were given examples of general and specific responses to past and future cued sentences and completed two practice items with relatively neutral cue words. When it was clear to the researcher that their responses met the requirements they proceeded to the experimental trials. These trials were private and confidential and neither the researcher nor the timekeepers were privy to participants' responses.

Coding of the AMT and FET data

The criteria used for coding specificity were based on previous research (Williams et al., 1996). Responses were coded as specific, moderate or general for past and future scenarios. A response was deemed specific if it accounted for a unique event that happened (or was anticipated to happen) at a particular time and place, and lasted no longer than a day. A moderate response consisted of a unique event that happened (or was anticipated to happen) but did not refer to a particular time or place, whereas a general response did not describe a unique event and did not include references to time or place. Based on procedures developed by Baddeley and Wilson (1986), responses were ranked on a scale of specificity. A specific response scored 3 points; a moderate response, 2 points and a general response, 1 point. One participant in the repression group failed to give responses to the depression-related cues in the past and future time conditions. The repression group means were assigned for this participant in the past and future specificity conditions, respectively. Possible specificity scores ranged from 2 to 6 for the two anxious, depressed and pleasant cues presented in the past and future time condition, respectively. Mean specificity across time conditions (past, future) represent averaged scores, thus possible scores also ranged from 2 to 6 for each valence condition. An independent coder rated 40% of the responses to establish reliability. Agreement between the first author and the independent coder was high (κ = .89) using Cohen's Kappa.

Results

Two three-way analyses of variance (ANOVAs) were performed to examine participants' specificity levels and reaction times for personal memories and imageability of future self-scenarios. Each ANOVA included two within-subject factors: cue valence (pleasant, anxious, depressive) and time (past, future), and one between-subject factor of group (low anxious controls, repressors). Simple main effects were examined for significant results. Because sample sizes were relatively small, the effect sizes are included in the relevant results. Clark-Carter (2001) converted the Cohen (1988) recommendations for effect sizes into values for η². Accordingly, an η² of 0.01 represents a small effect size; η² of 0.059, a medium effect size; and η² of 0.138, a large effect size.

Specificity of autobiographical memories and future events

ANOVA revealed significant large main effects of Group, F(1,35) = 22.49, p < .001, η² = .39; Valence, F(2,34) = 52.69, p < .001, η² = .76; and Time, F(1,35) = 32.20, p < .001, η² = .48. As predicted, there was also a large interaction between Group and Valence, F(2,34) = 15.56, p < .001, η² = .48. No other interactions were significant (ps > .05).

Comparisons of means for the Group main effect revealed that, overall, the low anxious controls (M = 5.11) were significantly more specific in their responses than were the repressors (M = 4.12). Comparisons for the main effect of Valence showed that specificity for pleasant cues (M = 5.43) was higher than anxious cues (M = 4.68): F(1,36) = 13.08, p < .01; which in turn showed higher specificity than depressive cues (M = 3.75): F(1,36) = 22.22, p < .001. The main effect of Time showed that, overall, participants' responses were less specific about the future (M = 4.26) than about the past (M = 4.98). The Group × Valence interaction is represented in Figure 1.

Figure 1. Group specificity means for valence experiences.

As shown in Figure 1, results were generally as hypothesised. Repressors were less specific in their accounts of depressive experiences (M = 2.82) than controls (M = 4.67), t(35) = 5.95, p < .001, and also less specific in reporting anxiety-related experiences (M = 4.12) than controls (M = 5.23), t(35) = 3.09, p < .01. As predicted, repressors (M = 5.42) and controls (M = 5.44) did not differ on specificity for pleasant experiences.

As predicted, repressors showed a higher level of specificity in their responses to pleasant cues than both depressive, t(19) = 13.89, p < .001, and anxious cues, t(19) = 4.02, p < .01. As predicted, repressors were more specific in reporting anxious experiences than depressive experiences, t(19) = 3.94, p < .01.

Response latencies for recall of autobiographical memories and anticipation of future events

The ANOVA for response latencies revealed a large-sized main effect of Valence, F(2,34) = 10.72, p < .05, η² = .39, and a medium effect of Time, F(1,35) = 4.57, p < .05, η² = .12. Contrary to expectations, a Group main effect was not shown on response latencies (p > .05). Overall, reaction times were faster for pleasant experiences (M = 19.33) than unpleasant experiences (depressive, M = 29.94; anxious, M = 26.87). As well, reaction times overall were faster for anticipating future events (M = 23.17) than recalling past memories (M = 27.59). These main effects were qualified by a Valence × Time interaction, F(1,34) = 4.50, p < .05, η² = .11. No other interactions were significant. Means for the Valence × Time interaction are presented in Figure 2.

Figure 2. Mean response latencies for past and future valence experiences.

As illustrated in Figure 2, participants' latencies for retrieval of depressive (M = 28.83) and anxious memories (M = 34.05) were slower than for recall of pleasant memories (M = 20.89): t(36) = 2.38, p < .05; t(36) = 3.47, p < .01, respectively. There were no differences in response latencies for anxious and depressive memories (p > .05). For future cues, the mean response latencies for both anxious (M = 20.29) and pleasant experiences (M = 18.16) were faster than latencies for future depressive experiences (M = 31.43): t(36) = 3.77, p < .01; t(36) = 2.83, p < .05, respectively. There were no differences in response latencies for anticipation of anxious and pleasant future events (p > .05).

Comparison between past and future cues showed that reaction time for anticipation of future anxiety-related experiences (M = 20.29) was significantly faster than recall of anxiety-related memories (M = 34.05): t(36) = 2.89, p < .01. There were no significant differences in latencies for pleasant or depressive cues.

Discussion

The current findings partially support the view that repressors show selective accessibility of personal memories and future events. Repressors' defensive denial of unpleasant painful experiences was reflected by their difficulty in being specific about remembering and imagining unpleasant personal events relative to controls. As predicted, repressors' specificity for pleasant experiences was comparable to that of low anxious controls. As expected, there was a consistency on specificity processing styles for groups over time, although all participants were less specific about describing future events than past events. Counter to prediction, repressors were not slower in retrieving unpleasant experiences than controls.

One interpretation of the repressors' results is that their encoding of aversive experiences is less elaborate than that of pleasant experiences. This under-encoding thus results in a less complex representation of unpleasant emotions, hence limiting accessibility of painful memories or future events. The lack of associative links in memory for unpleasant experiences may block retrieval of specific events and hinder imageability of future events.

Notably, repressors showed greater specificity in reporting anxious experiences than depressive experiences, which suggests they may have a stronger investment in repressing unpleasant events related to depression than to anxiety. Importantly, this result is consistent with the Holtgrave and Hall (1995) finding that repressors display a memory decrement for certain types of negative words (shame, guilt, sadness, pain, disgust, anger), but not for anxiety-related words. It may be that depressive experiences (e.g., failure) are more intimidating to a repressor's sense of self-esteem because they relate more directly to the self than anxious experiences that may perhaps be more readily externalised (e.g., scared). Alternatively, repressors' differential processing of anxiety and depression cues may be associated with their characteristically high responsiveness to threat despite their denial of anxiety (e.g., Weinberger et al., 1979; Weinberger & Davidson, 1994). The salience of anxiety-related material may facilitate integration more readily than depressive material and thereby render anxious experiences more accessible (Graf & Mandler, 1984).

Certainly, the current study demonstrates the value of studying repressors' specificity response-variations to different forms of unpleasant stimuli (e.g., depression, anxiety). Future research on specificity could extend this examination of response variations to more distinct subtypes of negative (e.g., guilt, anger, sadness) and positive (happiness, excitement, relaxation) emotional material. Study in this area awaits further research.

Although specificity levels for pleasant experiences were comparable for repressors and low anxious controls, there may be qualitative differences in their remembering and imagining of pleasant events. Myers and Brewin (1996), and Shedler et al. (1993) have contended that repressors display an illusory positive self-image and unrealistic optimism. Consistent with this view, examination of the content of repressors' responses for pleasant events showed an element of unrealistic optimism, whereby the absence of any anticipated problem was a consistent theme. For example, in response to the cue success one repressor responded, “When I go for a job interview to be an accountant, I will have my degree, I will feel capable of doing the job properly, without ever having any problems”. Further, repressors specificity for pleasant valence material, relative to unpleasant valence material, could be interpreted as supporting the view of repressors' apparent motivation to maintain a positive self-image and sense of self as psychologically well-adjusted (Shedler et al., 1993). However, the fact that repressors and controls did not differ significantly on specificity for positive experiences but rather by repressors' difficulty in recalling specific negative memories and in framing specific negative future events, relative to controls, also supports the Creswell and Myers (2002) view that repressors tend to avoid negative information rather than being overly positive.

All participants had greater difficulty imagining future unknown and uncertain events than recalling memories that had actually taken place. Anticipation of specific future experiences is presumably a more difficult cognitive task than recall of actual past experiences. This finding indicates that imageability is a critical cognitive factor required in constructing one's future. Nevertheless, the lack of any specified time frame in which to recall autobiographical memories or to imagine future events may have had an inhibitory effect on participants' reporting of specific future scenarios. For example, if participants were given a future time frame such as next week in which to image valenced experiences, the responses may have been more detailed or specific. The impact of instructions on specificity awaits further examination. The absence of group differences for response latencies may suggest that repressors' motivation to maintain a positive self-image renders them more sensitive to time-demand characteristics in group experimental settings. Further, some authors argue that latencies are a less reliable index of differences in accessibility than specificity (Kuyken & Brewin, 1995; Williams, 1996). However, the present study's failure to identify group differences in response latencies for negative experiences may be the result of using only four distinct types of negative stimuli for each task. Extending the range of negative cue words may produce different findings. The theoretical notion that repressors are slower to recall and anticipate aversive experiences warrants further investigation.

A feature of the latency results was that all participants were faster in accessing future anxiety-related experiences than past memories and tended to be slower in accessing depression-related cues, past or future. This pattern of differences in response is consistent with theoretical notions of anxiety and depression. Within cognitive explanations, anxiety is characterised by an attentional focus on future threatening and harmful experiences (Beck & Emery, 1985). In contrast, depression is marked by generalised themes of loss and sadness that permeate one's past, present and future (Clark & Beck, 1989).

In terms of the experimental measures used, the AMT has been used quite extensively in the study of memory specificity (e.g., Williams, 1996). Although the FET is a new measure, it is based on the AMT. Both measures used do have high face validity and the pattern of results for specificity do provide evidence of convergent and discriminant validity.

In summary, the current study has contributed further evidence that individuals with repressive coping styles preferentially bias self-referential valenced memories and future scenarios in a manner consistent with their emotional style. This research illustrates the complex, yet predictable, interaction between emotional disposition and cognitive processing strategies that is consistent across views of the past and of the future.

Additional information

Notes on contributors

Joanne M Dickson

Accepted for publication by the previous Editor, Professor Bill Noble, University of New England

Notes

Accepted for publication by the previous Editor, Professor Bill Noble, University of New England