psy

Psychological Science 22(8) 979 –983 © The Author(s) 2011 Reprints and permission: sagepub.com/journalsPermissions.nav DOI: 10.1177/0956797611415539 http://pss.sagepub.com

Depression is associated with a tendency to respond to nega- tive mood states and life events with ruminative thinking (Nolen-Hoeksema, Wisco, & Lyubomirsky, 2008). Recurrent, unintentional, and uncontrollable ruminative thoughts are not only a symptom of depression, but they have also been associ- ated with vulnerability to the onset and recurrence of depres- sive episodes and with the maintenance of negative affect (Nolen-Hoeksema et al., 2008). Given these findings, it is crit- ical that researchers gain a better understanding of the cogni- tive processes that may increase rumination. Ruminators tend to perseverate on recurring thoughts that revolve around a par- ticular theme, most frequently related to the causes and impli- cations of their depressive symptoms, and have trouble switching to a new train of thought. Indeed, the results of recent studies suggest that deficits in working memory (WM), a system that provides temporary access to a select set of rep- resentations in the service of current cognitive processes (Cowan, 1999; Miyake & Shah, 1999), underlie ruminative responses in depression (Davis & Nolen-Hoeksema, 2000; Joormann, 2010; Joormann & Gotlib, 2008).

WM reflects an individual’s focus of attention because it holds the representations of which a person is aware at any given moment. Given the capacity limitation of this system, it is important that the contents of WM be updated efficiently

and that information important to an ongoing task is kept accessible (e.g., Friedman & Miyake, 2004). It is also impor- tant, however, that people can flexibly manipulate the infor- mation that is accessible in WM to respond to changes in the environment and personal goals. Cognitive inflexibility may lead people to become stuck in a particular mind-set (Davis & Nolen-Hoeksema, 2000), which may lead to rumination and depression. Investigators have documented that depression and likely rumination involve difficulties both keeping irrele- vant emotional information from entering WM (Goeleven, De Raedt, Baert, & Koster, 2006; Joormann, 2004, 2006) and removing previously relevant negative material from WM (Joormann & Gotlib, 2008). It is imperative, however, that research also examines individuals’ ability to manipulate rep- resentations that are currently held in WM. When memories and external information enter WM, their position in WM is represented by associations that are formed between these items (Morris & Jones, 1990). To manipulate information in WM, the representations and the associations among them

Corresponding Author: Jutta Joormann, Department of Psychology, University of Miami, 454 Flipse Bldg., Coral Gables, FL 33124 E-mail: [email protected]

Sticky Thoughts: Depression and Rumination Are Associated With Difficulties Manipulating Emotional Material in Working Memory

Jutta Joormann1, Sara M. Levens2, and Ian H. Gotlib2 1University of Miami and 2Stanford University

Abstract

Cognitive inflexibility may play an important role in rumination, a risk factor for the onset and maintenance of depressive episodes. In the study reported here, we assessed participants’ ability to either reverse or maintain in working memory the order of three emotion or three neutral words. Differences (or sorting costs) between response latencies in backward trials, on which participants were asked to reverse the order of the words, and forward trials, on which participants were asked to remember the words in the order in which they were presented, were calculated. Compared with control participants, depressed participants had higher sorting costs, particularly when presented with negative words. It is important to note that rumination predicted sorting costs for negative words but not for positive or neutral words in the depressed group. These findings indicate that depression and rumination are associated with deficits in cognitive control.

Keywords

attention, depression, memory, cognitive processes

Received 11/1/10; Revision accepted 5/8/11

Research Report

at STANFORD UNIV on October 14, 2011pss.sagepub.comDownloaded from

980 Joormann et al.

must be modified (Morris & Jones, 1990; Shimamura, 2000). In fact, the ability to not only maintain but also manipulate and update information in WM has been associated with individual differences in a range of cognitive processes (Barrouillet, Lépine, & Camos, 2008), including inhibitory control (Minamoto, Osaka, & Osaka, 2010), self-regulatory behavior (Hofmann, Gschwendner, Friese, Wiers, & Schmitt, 2008), and problem solving (De Smedt et al., 2009).

The study reported here was designed to test the hypothesis that depression and rumination are associated with a deficit in the manipulation of negatively valenced material in WM. We modified a WM manipulation task used in previous research (Crone, Wendelken, Donohue, & Bunge, 2006) to investigate participants’ ability to manipulate emotional and neutral mate- rial in WM. This WM manipulation task distinguishes the abil- ity to manipulate information in WM from the ability to maintain information in WM (Crone et al., 2006). We pre- dicted, first, that compared with their nondepressed counter- parts, depressed participants would exhibit deficits in their ability to manipulate negative material in WM, and, second, that difficulties manipulating negative material in WM would be related to individuals’ tendency to ruminate.

Method Participants

Fifty-three individuals participated in this study. Participants were solicited through advertisements posted within the commu- nity. Trained interviewers administered the Structured Clinical Interview for DSM-IV Axis I Disorders (First, Spitzer, Gibbon, & Williams, 1996); individuals were included in the depressed group (n = 26) if they met the criteria for major depressive disor- der (MDD) in the fourth edition of the Diagnostic and Statistical Manual of Mental Disorders (DSM–IV; American Psychiatric Association, 1994), and in the control group (n = 27) if they did not meet diagnostic criteria for any current or past Axis I disor- der. Participants also completed the Beck Depression Inventory- II (BDI; Beck, Steer, & Brown, 1996), a 21-item self-report measure of the severity of depressive symptoms, and the Rumi- native Responses Scale (RRS; Treynor, Gonzalez, & Nolen- Hoeksema, 2003), which assesses how participants tend to respond to sad feelings and symptoms of dysphoria.

Demographic and clinical characteristics of the two groups of participants are presented in Table 1. The two groups dif- fered significantly in age, t(50) = 2.87, p < .05, and education, χ2(1, N = 53) = 5.63, p < .05. As expected, the MDD partici- pants obtained higher scores on the BDI than did control par- ticipants, t(51) = 15.06, p < .01, as well as higher RRS scores, t(51) = 9.23, p < .01.

Stimuli We selected 180 positive, 180 neutral, and 180 negative nouns from the Affective Norms of English Words list (Bradley & Lang, 1999), which provides arousal and valence ratings on a

9-point scale. We took care to ensure that the words in the three groups did not differ in length and that the positive and negative words did not differ in arousal ratings. The mean valence ratings were 7.39 (SD = 0.62) for the positive nouns, 2.66 (SD = 0.62) for the negative nouns, and 5.48 (SD = 0.55) for the neutral nouns.

WM manipulation task On each trial, participants saw three words presented on a computer monitor. The words were presented one at a time for 1,000 ms each (a 750-ms fixation display preceded each word). After the presentation of the three words, a fixation dis- play was presented for 750 ms. This was followed by a cue (either the word “Backward” or the word “Forward” presented for 750 ms) instructing participants either to remember the words in the order in which they were presented (forward) or to reverse the order and re-sort them in WM (backward). The cue was followed by a 3,000-ms delay period to allow partici- pants to rehearse or re-sort the words. Finally, a probe word consisting of one of the three words was presented until the subject responded. Participants were instructed to press a key (“1,” “2,” or “3”) to indicate as quickly and as accurately as possible whether the probe was the first, second, or third word (counting forward or backward, as appropriate) in the set they had been instructed to remember. Responses and response latencies were recorded.

The recognition probe was used to index sorting costs (i.e., differences between response latencies on the forward and the backward trials). Sorting costs reflect difficulties in manipu- lating material in WM independently of individual differences in pure maintenance (Crone et al., 2006). Six conditions were compared (three word valences in each of the three presenta- tion orders; see Table 2). All three words within each trial were positive, negative, or neutral. Whether the first, second, or third word was the correct response was counterbalanced so that the probability of responding “1,” “2,” or “3” was equal across all six conditions. The experiment consisted of two blocks, with each condition presented 15 times in each block. The sequence of trials within blocks and the order of the blocks

Table 1. Demographic and Clinical Characteristics of Participants

Group

Characteristic Depressed Control

Sex 8 males, 18 females 14 males, 13 females Age (years) 46.73 (10.02) 38.42 (10.81) College educated 46% 77% Number of participants with a comorbid diagnosis

12 0

Mean depression score 29.92 (9.60) 1.30 (2.32)

Mean rumination score 2.69 (0.49) 1.47 (0.47)

Note: Standard deviations are presented in parentheses. Depression was measured with the Beck Depression Inventory-II (Beck, Steer, & Brown, 1996), and rumination was measured with the Ruminative Responses Scale (Treynor, Gonzalez, & Nolen-Hoeksema, 2003).

at STANFORD UNIV on October 14, 2011pss.sagepub.comDownloaded from

Working Memory in Depression 981

were randomized. For each participant, a random sample of words was selected from the word lists without replacement.

Results Correct responses

The mean percentage of correct responses in the different con- ditions are presented in Table 2. As expected, overall error rates were low. We conducted a repeated measures analysis of variance (ANOVA) to examine differences in the number of correct responses as a function of group, experimental condi- tion, and valence.1 This ANOVA did not yield significant main effects or interactions.

Decision latencies following probes We restricted our analyses of decision latencies to trials on which participants made correct responses. To eliminate outli- ers, we treated decision latencies that exceeded 3 s as missing values (less than 5% of all reaction times). There was no group difference in the number of outlying latencies. Mean response latencies for MDD and control participants in the different experimental conditions are presented in Table 2. We con- ducted an ANOVA on response latencies, with group as the between-subjects factor and condition (backward, forward) and word valence (neutral, positive, negative) as within- subjects factors. This analysis yielded a significant main effect of valence, F(2, 102) = 16.78, p < .01, η

p 2 = .25, a significant

main effect of condition, F(1, 51) = 56.74, p < .01, η p 2 = .53,

and a significant main effect of group, F(1, 51) = 8.40, p < .01, η

p 2 = .14, as well as significant interactions of group and con-

dition, F(1, 51) = 17.04, p < .01, η p 2 = .25, and condition and

valence, F(2, 102) = 5.28, p < .01, η p 2 = .09. These effects

were qualified, however, by the predicted three-way interac- tion of group, condition, and valence, F(2, 102) = 4.45, p < .02, η

p 2 = .08.2 To examine this interaction further, we com-

puted sorting costs for each participant by subtracting response latencies in the forward trials from response latencies in the backward trials. Figure 1 shows the mean sorting costs for the three word valences in each group.

Analyses of sorting costs

Our main hypotheses involved group differences in sorting costs. Specifically, we predicted that MDD and control partici- pants would differ in sorting costs for positive, negative, and neutral stimuli. We tested this prediction by conducting a repeated measures ANOVA on sorting costs. This analysis yielded significant main effects for both group, F(1, 51) = 17.04, η

p 2 = .25, and valence, F(2, 102) = 5.28, p < .01, η

p 2 = .09.

These main effects were qualified, however, by the predicted significant interaction of group and valence, F(2, 102) = 4.45, p < .05, η

p 2 = .08.3 Follow-up tests indicated that MDD and

control participants differed in sorting costs for positive words, t(51) = 3.79, p < .01, d = 1.04, negative words, t(51) = 4.61, p < .01, d = 1.27, and neutral words, t(51) = 3.03, p < .01, d = 0.83. Follow-up tests within the control group yielded no significant differences between sorting costs for positive versus neutral words, t(26) = 1.42, p = .17, positive versus negative words,

Table 2. Response Latencies and Percentage of Correct Responses in the Two Groups

Depressed group Control group

Sorting order and word valence

Response latency (ms)

Correct responses (%)

Response latency (ms)

Correct responses (%)

Forward Negative 1,314 (553) 95 1,086 (401) 94 Positive 1,280 (629) 95 1,023 (373) 95 Neutral 1,321 (643) 94 999 (351) 94 Backward Negative 1,871 (855) 87 1,227 (430) 89 Positive 1,702 (784) 91 1,130 (347) 87 Neutral 1,726 (861) 89 1,154 (361) 89

Note: Standard deviations are presented in parentheses. Participants were instructed to sort words either in the order in which they were presented (forward) or in the reverse order (backward).

0

100

200

300

400

500

600

700

Positive Neutral Negative

S or

tin g

C os

t ( m

s)

Word Valence

Control Participants

Depressed Participants

Fig. 1. Mean sorting cost as a function of word valence for control and depressed participants. Sorting costs were calculated by subtracting response latencies on trials in which participants were instructed to remember words in the order in which they were presented from response latencies on trials in which participants were instructed to remember words in the reverse order. Error bars show 1 SEM.

at STANFORD UNIV on October 14, 2011pss.sagepub.comDownloaded from

982 Joormann et al.

t(26) < 1, or neutral versus negative words, t(26) < 1. In con- trast, within the MDD group, participants had significantly higher sorting costs for negative words than for positive words, t(25) = 2.95, p < .01, and neutral words, t(25) = 3.72, p < .01; there was no difference in sorting costs for positive versus neu- tral words, t(25) < 1.

Sorting costs and rumination Our second hypothesis was that sorting costs for negative words would be significantly correlated with rumination. Given the high correlation between BDI and RRS scores (r = .52), we conducted a hierarchical linear regression analy- sis, in which we predicted sorting costs for negative words from RRS scores, controlling for BDI scores and sorting costs for positive and neutral words. In the full sample, we also included group and the interaction of group and RRS scores, as well as the interaction of group and BDI scores. The predic- tors explained 81% of the variance in sorting costs for nega- tive material, but neither BDI nor RRS scores was a significant predictor. The interaction of group and RRS scores, however, was a significant predictor of sorting costs for negative words (β = 1.28, p < .02). When examining only the control group, this regression was not significant. In contrast, in the MDD group, the regression explained 83% of the variance in sorting costs for negative words: Both RRS scores (β = 0.27, p < .05) and sorting costs for neutral words (β = 0.55, p < .01) were significant and unique predictors of sorting costs for negative words in the MDD group.4

Discussion A growing body of research has linked depression with a tendency to ruminate in response to negative affect (Nolen- Hoeksema et al., 2008). Moreover, numerous studies have dem- onstrated that rumination is associated with a heightened vulnerability to experiencing depressive episodes (Lyubomir- sky & Nolen-Hoeksema, 1993; Nolen-Hoeksema, Parker, & Larson, 1994). Despite this research, however, it is still unclear why some people are prone to ruminate while others find it rela- tively easy to reorient and recover. One reason may be that ruminators become stuck on recurrent thoughts that revolve around a specific theme and have difficulty flexibly switching to a new train of thought; such perseveration may reflect diffi- culties manipulating information in WM.

As we predicted, MDD participants in our study found it difficult to manipulate information in WM. Moreover, although MDD compared with control participants were char- acterized by higher sorting costs in general, these costs were greatest when the material was negatively valenced. Indeed, whereas control participants did not demonstrate differential sorting costs for positive, neutral, and negative words, MDD participants exhibited significantly greater sorting costs for negative than for positive or neutral words. These findings extend previous research on WM function in depression by

demonstrating that MDD is associated not only with difficul- ties keeping irrelevant negative material from entering WM and discarding previously relevant material from WM, but also with difficulty manipulating negative material in WM.

It is important to note that the backward conditions were more difficult than the forward conditions. Although this may make it difficult to attribute our findings to WM manipulation, other studies that have examined the effect of WM load and task difficulty in depression have not reported that increasing load or difficulty results in a valence-specific deficit in MDD subjects (e.g., Levens & Gotlib, 2010). Therefore, we are con- fident that our manipulation effects, which are restricted to negative valence, reflect differences in WM manipulation that are not fully explained by task difficulty.

We also found that difficulties in the manipulation of nega- tive words, but not of neutral or positive words, were associ- ated with increases in self-reported rumination in the MDD participants. These results add to findings of previous studies documenting an association between rumination and cognitive inflexibility (Altamirano, Miyake, & Whitmer, 2010; Whitmer & Banich, 2007), and between rumination and both the intru- sion of irrelevant material into WM and the ability to discard irrelevant material from WM (Joormann & Gotlib, 2008). In the study reported here, self-reported rumination was related to cognitive inflexibility specifically in the manipulation of negative material, but only in individuals diagnosed with MDD. Most studies investigating rumination and cognitive flexibility have used neutral material (e.g., Altamirano et al., 2010; Whitmer & Banich, 2007) with unselected or dysphoric samples, not people diagnosed with MDD. This is the first study to assess these constructs with valenced stimuli in clini- cally depressed individuals, for whom the salience of negative material is likely greater than it is for unselected participants. Therefore, it is likely that discrepancies between the current findings and the results of other studies are due to differences in both participant samples and stimuli.

Given the cross-sectional design of our study, formulations concerning underlying mechanisms of rumination are neces- sarily speculative. Investigating individual differences in executive functioning in WM has the potential to yield impor- tant insights concerning the maintenance of negative affect and vulnerability to experiencing depressive episodes. Diffi- culties manipulating material in WM may affect people’s abil- ity to regulate negative affect by impairing the formation of new associations within WM. For example, down-regulating an emotional response requires that existing associations be reorganized to alter the meaning of a stimulus or event, whereas reappraisal requires that new associations be formed between accessible negative content and novel positive con- tent. Examining people’s ability to manipulate the contents of WM promises to be critical in identifying individuals who recover easily from negative affect and differentiating them from individuals who tend to get stuck in a vicious cycle of increasingly negative ruminative thinking and deepening sad mood.

at STANFORD UNIV on October 14, 2011pss.sagepub.comDownloaded from

Working Memory in Depression 983

Acknowledgments

Sara M. Levens is now at the University of Pittsburgh.

Declaration of Conflicting Interests The authors declared that they had no conflicts of interest with respect to their authorship or the publication of this article.

Funding This research was supported by National Institute of Mental Health Grant MH59259, awarded to Ian H. Gotlib.

Notes

1. Including age and level of education as covariates did not alter the findings. 2. Including age and level of education as covariates did not alter the three-way interaction, F(2, 94) = 3.95, p < .02, η

p 2 = .08.

3. Including age and level of education as covariates did not alter the two-way interaction, F(2, 94) = 3.95, p < .02, η

p 2 = .08.

4. Using the Brooding and Reflection subscales of the RRS did not yield significant findings in any analysis.

References

Altamirano, L. J., Miyake, A., & Whitmer, A. J. (2010). When mental inflexibility facilitates executive control: Beneficial side effects of ruminative tendencies on goal maintenance. Psychological Science, 21, 1377–1382.

American Psychiatric Association. (1994). Diagnostic and statistical manual of mental disorders (4th ed.). Washington, DC: Author.

Barrouillet, P., Lépine, R., & Camos, V. (2008). Is the influence of working memory capacity on high-level cognition mediated by complexity or resource-dependent elementary processes? Psy- chonomic Bulletin & Review, 15, 528–534.

Beck, A. T., Steer, R. A., & Brown, G. K. (1996). Manual for the Beck Depression Inventory-II. San Antonio, TX: Psychological Corp.

Bradley, M. M., & Lang, P. J. (1999). Affective norms for English words (ANEW): Technical manual and affective ratings. Gainesville: University of Florida, Center for Research in Psychophysiology.

Cowan, N. (1999). An embedded-processes model of working mem- ory. In A. Miyake & P. Shah (Eds.), Models of working mem- ory: Mechanisms of active maintenance and executive control (pp. 62–101). New York, NY: Cambridge University Press.

Crone, E. A., Wendelken, C., Donohue, S. E., & Bunge, S. A. (2006). Neural evidence for dissociable components of task-switching. Cerebral Cortex, 16, 475–486.

Davis, R. N., & Nolen-Hoeksema, S. (2000). Cognitive inflexibility among ruminators and nonruminators. Cognitive Therapy and Research, 24, 699–711.

De Smedt, B., Janssen, R., Bouwens, K., Verschaffel, L., Boets, B., & Ghesquière, P. (2009). Working memory and individual dif- ferences in mathematics achievement: A longitudinal study from first grade to second grade. Journal of Experimental Child Psy- chology, 103, 186–201.

First, M. B., Spitzer, R. L., Gibbon, M., & Williams, J. B. W. (1996). Structured Clinical Interview for DSM-IV Axis I Disorders—

Clinician Version (SCID-CV). Washington, DC: American Psy- chiatric Press.

Friedman, N. P., & Miyake, A. (2004). The relations among inhi- bition and interference control functions: A latent-variable analysis. Journal of Experimental Psychology: General, 133, 101–135.

Goeleven, E., De Raedt, R., Baert, S., & Koster, E. H. W. (2006). Deficient inhibition of emotional information in depression. Journal of Affective Disorders, 93, 149–157.

Hofmann, W., Gschwendner, T., Friese, M., Wiers, R., & Schmitt, M. (2008). Working memory capacity and self-regulatory behavior: Toward an individual differences perspective on behavior deter- mination by automatic versus controlled processes. Journal of Personality and Social Psychology, 95, 962–977.

Joormann, J. (2004). Attentional bias in dysphoria: The role of inhibi- tory processes. Cognition & Emotion, 18, 125–147.

Joormann, J. (2006). The relation of rumination and inhibition: Evidence from a negative priming task. Cognitive Therapy and Research, 30, 149–160.

Joormann, J. (2010). Cognitive inhibition and emotion regulation in depression. Current Directions in Psychological Science, 19, 161–166.

Joormann, J., & Gotlib, I. H. (2008). Updating the contents of work- ing memory in depression: Interference from irrelevant negative material. Journal of Abnormal Psychology, 117, 182–192.

Levens, S. M., & Gotlib, I. H. (2010). Updating positive and negative stimuli in working memory in depression. Journal of Experimen- tal Psychology: General, 139, 654–664.

Lyubomirsky, S., & Nolen-Hoeksema, S. (1993). Self-perpetuating properties of dysphoric rumination. Journal of Personality and Social Psychology, 65, 339–349.

Minamoto, T., Osaka, M., & Osaka, N. (2010). Individual differences in working memory capacity and distractor processing: Possible contribution of top-down inhibitory control. Brain Research, 1335, 63–73.

Miyake, A., & Shah, P. (1999). Models of working memory: Mecha- nisms of active maintenance and executive control. New York, NY: Cambridge University Press.

Morris, N., & Jones, D. M. (1990). Memory updating in working memory: The role of the central executive. British Journal of Psychology, 81, 111–121.

Nolen-Hoeksema, S., Parker, L. E., & Larson, J. (1994). Ruminative coping with depressed mood following loss. Journal of Personal- ity and Social Psychology, 67, 92–104.

Nolen-Hoeksema, S., Wisco, B., & Lyubomirsky, S. (2008). Rethinking rumination. Perspectives on Psychological Science, 3, 400–424.

Shimamura, A. P. (2000). The role of the prefrontal cortex in dynamic filtering. Psychobiology, 28, 207–218.

Treynor, W., Gonzalez, R., & Nolen-Hoeksema, S. (2003). Rumina- tion reconsidered: A psychometric analysis. Cognitive Therapy and Research, 27, 247–259.

Whitmer, A., & Banich, M. (2007). Inhibition versus switching defi- cits in different forms of rumination. Psychological Science, 18, 546–553.

at STANFORD UNIV on October 14, 2011pss.sagepub.comDownloaded from