Research Article - (2018) Volume 8, Issue 4

The Relationship between Rumination and Autobiographical Memory Specificity in People with Depression

Corresponding Author:
Tatia MC Lee
Professor in Neuropsychology,Laboratory of Neuropsychology, Rm 656, 6th Floor, Building Centennial Campus, The University of Hong Kong, Pokfulam Road, Hong Kong
Tel: +(852) 2857-8394
Fax: +(852) 2819-0978
E-mail: tmclee@hku.hk

Abstract

Rumination and autobiographical memory specificity are neurocognitive factors associated with the course of a major depressive disorder (MDD). Yet, their relationships have not been fully understood. In this study, we explored whether a rumination of different valences had varying effects on autobiographical memory specificity in participants with major depressive disorder. We used a 2 (group: MDD, control) x 2 (rumination: positive, negative) x 2 (time: pre, post) mixed design. Fifty-two currently depressed people and 52 non-psychiatric controls completed this experiment. They completed the Autobiographical Memory Test and the mood ratings before and after completing either the positive or the negative rumination task. In the rumination task, they were requested to focus their attention on some specific thoughts about themselves. Results showed a significant group (depressed, control) x time (pre, post) interaction effect for the number of specific memories. This was a result of a significant decrease in specific memories retrieved after negative rumination in the depressed group but not after positive rumination. No significant differences among the non-psychiatric controls were observed. Our findings suggest that possible inhibitory deficit in people with depression may have compromised the ability to shunt task-irrelevant negative materials from working memory. The overloading of the working memory may be associated with reduced retrieval of specific memories.

Keywords

Rumination, Autobiographical Memory, Valence, Depression, Working memory

Introduction

Reduced autobiographical memory specificity in people with depression was first reported by Williams and Broadbent [1]. who observed that suicide attempters tended to report overgeneral memories when requested to recall specific memories using the cue-word method. Since then this observation has been consistently observed in people with major depressive disorders (MDD) [2]. Findings of a recent meta-analysis study [3] have also indicated that over-general autobiographical memory was a significant predictor of the course of depression. These findings has led to the speculation that autobiographical memory specificity could be a possible neurocognitive marker of MDD [4,5].

Rumination and depression are significantly associated with each other [6] reported that rumination can predict depressive symptoms. Watkins and Teasdale [7] demonstrated that the analytical and abstract nature of rumination might contribute to the over-general memory presentation. A neurobiological basis of rumination has been reported. Kühn found that rumination was associated with reduced structural volume and resting state activity in the anterior cingulate cortex and dorsolateral prefrontal region [7], suggesting that deficits in inhibitory control could represent premorbid vulnerabilities to rumination. Previous findings have suggested that neural activity at the ACC might have been related to a pattern of aggravated self-referential processes typically observed in ruminators. Morphometric changes of the prefrontal region appear to be associated with the frequency of relapse of depression [8]. Piguet have reported that the tendency to ruminate was positively correlated with neural activation in the entorhinal cortex during rest and during cognitive tasks with low attentional load. This demonstrates that individuals with a higher propensity to ruminate likely engage in internally driven trains of thoughts possibly involving the retrieval of self-related memory episodes and information.

Research on rumination of different valences has been scarce. A very first study by Sutherland and Bryant has successfully elucidated the valence specific effect of rumination autobiographical memory specificity such that more overgeneral memories were retrieved after negative rumination than after positive rumination in people in the high-depressed than low-depressed groups [9]. Joormann and Gotlib [10] proposed that depressed people have valence-specific inhibitory deficits and that this inhibition problem is limited to negative materials [9]. Joormann et al. [10] verified that the negative priming effect was only observed for positive but not negative affective stimuli, suggesting that there was minimum negative priming effect during the processing of negative words [10]. For depressed people, a higher level of self-reported rumination is associated with less inhibition of negative materials [11]. A similar result of reduced inhibition of negative materials was also found in individuals who were dysphoric but not clinically depressed [12] and when sad and happy faces were used as stimuli instead of negative and positive words [13]. Williams [15] proposed a comprehensive CaR-FA-X model that explains the mechanism underlying reduced specificity in autobiographical memory. In the CaR-FA-X model, “FA” stands for functional avoidance, “CaR” is capture and rumination, and “X” refers to executive capacity and control [15,16]. Williams [15] further postulated that people with depression tended to ruminate which rendered them susceptible to truncated search of encoded information leading to fewer specific memories being retrieved [15,16].

Knowledge of relationships between depression, rumination, and autobiographical memory specificity is important for understanding the underlying mechanisms of autobiographical memory over-generality in depression. In this study, we compared the autobiographical memory specificity in people having an active episode of depression. We hypothesized that our clinical participants, relative to the healthy controls, would have less specific autobiographical memories. We further examined the effect of ruminations with different valences on autobiographical memory specificity in these clinical participants. We hypothesized that the clinical participants, relative to the healthy controls, would retrieve fewer specific memories after negative than positive rumination.

Materials

▪ Participants

Fifty-eight female Chinese psychiatric participants were recruited from both the in-patient and outpatient units of a regional psychiatric hospital in Hong Kong. All psychiatric participants met the diagnostic criteria for a diagnosis of major depressive disorder (MDD) without psychotic features according to the DSM-IV [17]. Three of them did not complete the experiment because they could not pass the manipulation check. The diagnosis was confirmed with the Chinese version of the Mini International Neuropsychiatric Interview [18,19]. The severity of depression was measured using the Chinese version of Beck Depression Inventory – II [20,21]. The BDI-II is a commonly used assessment of the severity of depression. It is a 21-item self-report inventory measuring the affective, cognitive, and physical symptoms of depression. Research has shown that the BDI-II is a valid and reliable assessment tool for Chinese populations [22-24]. Three participants in the MDD group with scores on the BDI-II of less than 14 were excluded from the study because they were not currently depressed. Patients with organic brain disorder, brain damage, substance abuse or dependence disorder, psychotic disorder, bipolar disorder, mental retardation, or posttraumatic stress disorder, or those who had received electroconvulsive therapy within the previous six months were excluded. Among the participants in the MDD group, only five of them were not currently taking antidepressants. Ten of them had no comorbidity. Thirty of them had dysthymia, 32 of them had generalized anxiety disorder, four had social phobia, two had obsessive compulsive disorder, one had agoraphobia, and two had panic disorder. The average number of depressive episodes was 1.40 (SD = 0.98).

Fifty-two nonpsychiatric Chinese female community controls were recruited from the same hospital. Most of them were supporting staff of the hospital and their friends. They reported neither a history of psychiatric disorders nor brain damage. The Chinese version of BDI-II was administered. All participants in the control group had scores of less than 14on the BDI-II.

There were no significant differences in age and education among those in the MDD group and the control group(p>.05). We also used the Mini- Mental State Examination [25,26] to screen out participants with cognitive impairment. The MMSE is an instrument widely used for this purpose. It is a 30-point test including questions tapping into orientation, immediate recall, delayed recall, working memory, language comprehension, language expression, and visual spatial skills. Both the original version and the Chinese version have been reported to have good psychometric properties [26]. We administered the Chinese version of the Ruminative Responses Scale of the Response-Style Questionnaire [27,28] to measure the participants’ tendency to ruminate when experiencing a depressed mood. The RRS is a 22-item self-report questionnaire. It has been reported to have robust psychometric properties [28].

▪ Design

We used a 2 (group: MDD, nonpsychiatric control) x 2 (rumination: positive rumination, negative rumination) x 2 (time: premanipulation, postmanipulation) mixed design. The dependent variables were the score on the mood scales and the number of different kinds of memories recalled before and after the rumination task.

We categorized the reported memories as specific, categoric, extended, semantic association, and omission. Specific memories were defined as recollections of a particular event that happened on a particular day at a particular time and place in the past and that lasted for no longer than one day. Categoric memories referred to remembrances of some repeated events. Extended memories were those concerned with an event lasting for longer than one day. Semantic associations referred to some words associated with the semantic meanings of the cue words. An omission was recorded when the participants failed to retrieve a memory within the time limit (i.e., one minute).

Measures

▪ Autobiographical Memory Test (AMT)

We chose 42 positive cue words and 42 negative cue words from those used by researchers in previous studies on autobiographical memory specificity [14,9,1]. As the cue words used in this study were all in Chinese, we asked seven Chinese clinical psychologists to rate the translated positive cue words and negative cue words in terms of the degree of familiarity, imageability, and emotionality on an 11-point Likert-type scale, with 0 referring to the least and 10 referring to the most. Based on their ratings, we chose 18 positive cue words and 18 negative cue words. The positive cue words had a mean rating of 8.35 on familiarity (SD = 1.23), 7.79 on imageability (SD = 2.26), and 8.09 on emotionality (SD = 0.53). The negative cue words had a mean rating of 8.13 on familiarity (SD = 1.47), 7.67 on imageability (SD = 2.13), and 7.90 on emotionality (SD = 1.06). We performed three separate t tests to compare the ratings of positive and negative cue words and found no significant differences (p>.05). We matched the positive and negative cue words according to the degree of familiarity, imageability, and emotionality. The words were presented in a fixed order, alternating the positive and negative words.

In the AMT, we asked the participants to retrieve a specific event that had happened to them in response to the cue words. A specific event means an event that happened at a particular time and a particular place and was concluded in one day. At least two practice trials were given to the participants to make sure that they understood the requirement. The participants had one minute to retrieve a memory in response to the cue word. When a participant could not give a memory within one minute, the experimenter proceeded to the next cue word. The memories retrieved by the participants were audiotaped. Twenty percent of the audio recordings were randomly selected and categorized by a second independent rater who was blind to the group membership of the participants. The agreement of the two raters was excellent, K = .95. There were two word lists for the AMT (List A and List B). The order of the word lists (List A first or List B first) and the different valences of ruminations were counterbalanced across participants in the MDD group and in the control group.

▪ Mood scales

We used mood scales to check whether there were any changes in the participants’ moods before and after the rumination task. Participants were asked to rate how happy and how sad they were at that moment on two 10-point Likert-type scales ranging from 0 not at all to 9 extremely.

▪ Procedure

As all the participants were Chinese, all the materials were written in Chinese and the instructions were orally presented in Cantonese. Informed consent was obtained from the participants. They were first interviewed with the Chinese version of the M.I.N.I. [29]. They then completed the Chinese version of the BDI-II [21], the Chinese version of the RRS and the Chinese version of the MMSE The participants were randomly allocated to either the positive or the negative rumination group. Then we administered the pre-experiment assessment, including the mood scales and the AMT. Then they did the rumination task. Following Nolen-Hoeksema and Morrow [30], the rumination task was used to change the direction of the participants’ thoughts as they were requested to focus their attention when we asked them to slowly repeat some sentences on a list and be aware of what they were feeling and imagining. In this study, we used positive and negative statements of about the same length from Sutherland and Bryant [9]. All the rumination statements were translated into Chinese. A back translation was done as a quality check on the Chinese translation. We asked five Chinese clinical psychologists to rate the difficulty people would experience in ruminating on these positive and negative statements using a 10-point Likert-type scale ranging from 1 extremely easy to 10 extremely difficult. On the basis of their ratings, 28 positive and 28 corresponding negative statements were chosen in order to minimize the level of difficulty and the standard deviation. The mean levels of difficulty for the positive and the negative statements were 4.73 (SD = 2.28) and 3.26 (SD = 1.92), respectively. Such levels of difficulty were matched according to the results of a paired sample t test (p>.05). The statements were printed on a piece of paper, and their order was fixed. The participants were asked to focus on the statements for a total of eight minutes.

After the rumination task, the participants received the post-experiment assessment, which included the manipulation check, the mood scales, and the AMT. In the manipulation check, the participants were asked what percentages of their thoughts in the rumination task were positive, negative, or neutral. Three participants who reported less than 60% of thoughts of the right valence were excluded from the study. The manipulation check showed that the MDD group reported a mean of 76.54% positive thought content (SD = 14.13) during the positive rumination. In contrast, the control group reported a mean of 84.81% of positive thought content (SD = 12.37). Regarding negative rumination, the MDD group reported a mean of 76.73% negative thought content (SD = 14.35), and the control group reported a mean of 74.81% (SD = 14.32). We then conducted a 2 (group: MDD, nonpsychiatric control) x 2 (rumination: positive rumination, negative rumination) ANOVA, which showed that neither the main effects nor the interaction effects were significant (p>.05). Therefore, members of the MDD and the control groups had a similar percentage of negative thoughts in the negative rumination task and a similar percentage of positive thoughts in the positive rumination task.

We assessed the demand effects. We asked the participants whether they knew what the researchers expected to find in this study. None of them gave a correct answer. We provided the participants in the negative rumination group with an optional brief happiness induction, allowing them to view comical video clips for about 15 minutes. At the end of the experiment, the participants were debriefed.

Results

▪ Participants’ background variables

Table 1 shows the participants’ average ages, years of education; scores on the BDI, the RRS, and the MMSE; and number of depressive episodes.

  Positive rumination Negative rumination
  MDD Control MDD Control
Age 48.50 (9.34) 42.15 (10.20) 41.73 (10.10) 46.15 (9.49)
Education 7.19 (4.46) 9.04 (2.74) 8.81 (3.59) 8.31 (3.40)
BDI-II 30.08 (9.88) 4.19 (2.71) 33.15 (12.60) 5.50 (3.34)
RRS 50.04 (10.27) 32.54 (8.34) 49.69 (11.59) 37.46 (9.03)
MMSE 27.65 (2.17) 28.58 (1.68) 28.46 (1.39) 28.35 (1.67)
Number of depressive episodes 1.58 (1.21) N/A 1.23 (0.65) N/A

Table 1: Means (and Standard Deviations) of Age; Education; BDI-II, RRS, and MMSE Scores; and Number of Depressive Episodes.

We conducted a one-way ANOVA to check whether the two MDD groups showed a difference in the number of depressive episodes. The result was not significant (p > 0.05). The MDD positive rumination group had the same number of depressive episodes as the MDD negative rumination group.

We conducted separate 2 (rumination: positive rumination, negative rumination) x 2 (group: MDD, nonpsychiatric control) ANOVAs to examine whether these four groups were comparable in terms of the other characteristics listed in Table 1. The main effect of group was found in the BDI-II and RRS scores, with F (1,100) = 271.13, p < 0.001 and F (1,100) = 58.90, p < 0.001, respectively. The interaction effect was not significant (p > 0.05). Irrespective of whether the condition was positive or negative rumination, the MDD groups had higher BDI-II scores and higher RRS scores than the control groups. However, the two MDD groups were comparable in terms of BDI-II and RRS scores. Similarly, the two control groups exhibited similar BDI-II and RRS scores.

As mentioned above, we matched the ages of the participants in the MDD group and the nonpsychiatric control group. However, after randomly assigning the participants to two different rumination groups, there was a significant rumination x group interaction effect on the age of participants (F (1,100) = 7.87, p < 0.01). There was also a significant difference in ages of members of the two MDD groups of different ruminations (F (1, 50) = 6.29, p < 0.05). The MDD participants under the negative rumination condition were significantly younger than those under the positive rumination condition. The two positive rumination groups also differed significantly in age (F (1, 50) = 5.47, p < 0.05). The nonpsychiatric control participants under the positive rumination condition were significantly younger than the MDD participants under the positive rumination condition. Therefore, age was used as a covariate in analyzing the effect of manipulations on the memories reported in the AMT and the mood ratings in order to remove the confounding effect caused by the age differences among participants.

The results of other two-way ANOVAs showed that there was neither a main effect nor an interaction effect related to education and MMSE (p > 0.05). The four groups were comparable in terms of years of education and estimated cognitive abilities.

▪ Baseline Memory Specificity

We compared the baseline memory specificity of the MDD group and the nonpsychiatric control group. Because there were too few extended memories and semantic associates reported by the participants for meaningful analysis, the statistical analyses were done only on the number of specific memories, categoric memories, and omissions. Before the rumination task, extended memories comprised only 1.78% of total possible responses, and semantic associates comprised only 0.89%.

We performed three separate t tests to compare the number of specific memories, categoric memories, and omissions in the MDD group and the nonpsychiatric control group before the rumination task. The results showed that the control group had significantly more specific memories than the MDD group at baseline (t (102) = -3.02, p < .01). In contrast, more categoric memories were retrieved by the MDD group than by the control group, with t (102) = 3.40, p< 0.005. However, there was no difference in the number of omissions between the MDD group and the control group at the baseline (p > 0.05).

▪ The effect of manipulations on memory specificity

The descriptive data related to the number of specific memories for different participant groups, the valence of rumination, and time are shown in Table 2. As mentioned above, there was a significant rumination x group interaction effect on the age of participants. Therefore, age was used as a covariate in the data analysis. A 2 (rumination: positive rumination, negative rumination) x 2 (group: MDD, nonpsychiatric control) x 2 (time: premanipulation, postmanipulation) ANCOVA with age as the covariate was performed to examine the effect of manipulation on the change in the number of specific memories for different participant groups. There was no significant three-way interaction effect (p > 0.05). However, there was a marginally significant time x group interaction effect, with F (1,99) = 3.85, p = .05. Other interaction effects and the main effect were not significant (p >0.05). A post-hoc comparison revealed that the number of specific memories decreased significantly after negative rumination in the MDD group, t (25) = 2.96, p < 0.01. For the control group, the change after negative rumination was not significant (p > 0.05). Similarly, both the control group and the MDD group did not change significantly in the number of specific memories after positive rumination (p > 0.05).

Group Type of memory Positive rumination Negative rumination
    Pre Post Pre Post
MDD Specific 12.92 (2.94) 12.15 (3.66) 12.69 (2.75) 11.27 (3.13)
Extended .38 (.57) .35 (.56) .31 (.55) .58 (.76)
Categoric 2.15 (2.29) 2.58 (2.30) 2.00 (1.98) 2.15 (1.74)
Semantic associate .27 (.60) .42 (.90) .31 (.62) .15 (.37)
Omission 2.27 (2.44) 2.50 (3.00) 2.69 (1.91) 3.85 (3.03)
Control Specific 15.23 (2.85) 14.58 (3.26) 13.92 (3.36) 14.35 (3.52)
Extended .12 (.43) .15 (.46) .46 (.95 ) .27 (.53 )
Categoric .81 (1.36) .77 (.99) 1.00 (1.27) 1.08 (1.62)
Semantic associate .04 (.20) 0 (0.00 ) .04 (.20) .08 (.27)
Omission 1.81 (2.04) 2.50 (2.70) 2.58 (2.67) 2.23 (3.63)

Table 2: Means (and Standard Deviations) of Different Types of Memories as a Function of Participant Groups, Valence of Rumination, and Time.

▪ The effect of manipulations on other types of memories

The descriptive data related to the other types of memories for different participant groups, the valence of rumination, and time are shown in Table 2. Too few extended memory and semantic associates were reported by the participants. Extended memories comprised only 1.83% of total possible responses, and semantic associates comprised only 0.89%. Therefore, no meaningful statistical analyses could be done on the number of extended memories and semantic associates. Two 2 (rumination: positive rumination, negative rumination) x 2 (group: MDD, nonpsychiatric control) x 2 (time: premanipulation, postmanipulation) ANCOVAs with age as the covariate were performed to examine the effect of manipulation on the number of categoric memories and omissions in different participant groups. The ANCOVA revealed that neither the main effect nor the interaction effects were significant for categoric memories (p > 0.05). For the omissions, there was a significant three-way interaction effect, with F (1,99) = 4.85, p < 0.05. No other main effect or interaction effect was significant (p > 0.05). A post-hoc comparison indicated that the MDD group had significantly more omissions after negative rumination, with t (25) = -2.65, p< 0.05. No such significant change was found in the control group after negative rumination (p >.05). Likewise, neither the MDD group nor the control group showed a significant change in the number of omissions after positive rumination (p > 0.05).

▪ The effect of manipulations on mood

The descriptive data related to mood ratings for different participant groups, valence of rumination, and time are shown in Table 3. For the positive and negative mood ratings, we performed two separate 2 (rumination: positive rumination, negative rumination) x 2 (group: MDD, nonpsychiatric control) x 2 (time: premanipulation, postmanipulation) ANCOVAs with age as the covariate to examine the effect of manipulation on the change in mood ratings for different participant groups. For positive mood, the ANCOVA revealed a group x time interaction effect (F (1,99)=9.87, p < 0.01) and a rumination x time interaction effect (F(1,99) = 62.88, p < 0.001), qualified by a group x rumination x time interaction effect (F (1,99) = 8.87, p < 0.01). A post-hoc comparison showed that the positive mood rating after positive rumination increased significantly in the MDD group, with t (25) = -6.14, p < 0.001. In contrast, the results indicated that the positive mood rating increased in the control group after positive rumination, but the change was not significant (p > 0.05). However, the positive mood rating decreased significantly for both the MDD and the control groups after negative rumination, with t (25) = 3.05, p < 0.01 and t (25) = 4.32, p < 0.001, respectively.

Group Rumination Positive mood Negative mood
    Pre Post Pre Post
MDD Positive 3.87 (2.30) 5.65 (2.12) 3.81 (2.53) 2.88 (2.44)
Negative 4.04 (1.80) 3.35 (1.98) 3.54 (2.52) 4.54 (2.76)
Control Positive 6.54 (1.50) 6.96 (1.73) 0.38 (0.94) 0.31 (0.79)
Negative 5.92 (1.57) 5.15 (2.01) 1.65 (1.88) 2.85 (2.09)

Table 3: Means (and Standard Deviations) of Mood Ratings as a Function of Participant Groups, Valence of Rumination, and Time.

For negative mood, the ANCOVA revealed a rumination x time interaction effect, with F (1,99) = 16.09, p < 0.001. There was no other significant result (p >.05). A post-hoc comparison also showed that the negative mood change was not significant for the control group after positive rumination (p > 0.05). However, the negative mood rating significantly decreased for the MDD group after positive rumination, with t (25) = 2.13, p < 0.05. In contrast, the negative mood rating increased significantly for both the MDD group and the control group after negative rumination, with t (25) = -2.13, p < 0.05 and t (25) = -2.82, p < 0.01, respectively.

We computed the change scores for the mood ratings and number of different types of memories by subtracting the post-manipulation scores from the pre-manipulation scores. The correlations between the change score of positive mood and all the change scores of different types of memories were not significant (p > 0.05), with r ranging from -0.07 to 0.07, N = 104. Similarly, the correlations between the change score of negative mood and all the change scores of different types of memories were not significant (p > 0.05), with r ranging from -0.20 to 0.11, N = 104.

Discussion

Findings of our study indicate a robust phenomenon of reduced memory specificity in depression. Our clinical participants in the MDD group recalled fewer specific memories and more categoric memories than the controls. This result is consistent with findings of previous studies [31-34]. Furthermore, our clinical participants recalled fewer specific memories after negative rumination. This observation could not be repeated with positive rumination. This finding is consistent with that reported by Sutherland and Bryant [9] who demonstrated a greater reduction in specific memories after negative than positive rumination. Findings of this study cannot be understood from the perspective of mood differences because memory specificity is independent of moodstate [35,36]. In this study, neither the change in positive mood nor that in negative mood was significantly correlated with the change in types of autobiographical memories.

The observation that people with depression retrieved fewer specific autobiographical memories after negative rumination may relate to cognitive bias commonly seen in people with depression. According to Gotlib and Joormann [37], people with depression might have difficulty with disengaging attention from negative stimuli. According to the working memory model proposed by Baddeley [38,39], the central executive regulates and controls cognitive processing by directing cognitive resources to task-relevant representation and by inhibiting distraction from task-irrelevant information. But the central executive has limited capacity. If there is much interference from the task-irrelevant representation, the capacity of the working memory could be compromised [40]. Previous findings have shown that autobiographical memory specificity was affected by working memory capacity [41,42]. Following these lines of thoughts, people with depression may find disengaging their attention from negative materials challenging. Hence, the capacity of working memory capacity can be compromised, which in turn contributes to reduced memory specificity in these people. In addition, Joormann and Gotlib [37] showed that depressed people have difficulty inhibiting the processing of previously relevant but currently irrelevant negative information [43]. Such interference might affect the retrieval of specific autobiographical memory, which explains why our clinical participants have reduced memory specificity.

According to findings of this study, the fewer specific memories retrieved by the MDD group after negative rumination can be explained by the valence-specific inhibitory deficit in people with depression. They are unable to update the content of working memory effectively by inhibiting the processing of irrelevant negative materials. As the working memory has only limited capacity, this interferes with the performance of the task at hand, so fewer specific memories are retrieved. The reduced inhibition when processing negative materials is associated with rumination [43,11].

There are a few limitations of this study that could affect the generalizability of our findings. In order to control for the sex-related effect, only women with depression were tested in this study. Previous researchers have reported significant differences in the clinical course and symptom profiles of depressed men and women [44]. Also, there can be sex-related effects on memory specificity [45]. Furthermore, women are more prone to rumination than men [46]. Future studies might consider verifying the actual nature of the sex-related effect on cognitive bias presented by people with depression. Another limitation of this study was the use of a convenient sample, which limited the generalizability of the results. In addition, the nature of rumination (concrete or abstract) was not controlled in this study. However, the effects of concrete rumination and abstract rumination have been investigated by other researchers [14,7,47]. It would be interesting for future researchers to study [48-50] the interaction effects of the nature of rumination (concrete versus abstract) and the valence of rumination (positive or negative) .

Acknowledgement

This work was supported by The University of Hong Kong May Endowed Professorship in Neuropsychology.

We would like to thank Prof. J. Mark G. Williams, Dr. Cola S. L. Lo, Dr. K. Sutherland and Prof. R. A. Bryant for sharing their study materials and questionnaires.

References