Cognitive Therapy and Research

, Volume 38, Issue 3, pp 334–342

Coherence Between Attentional and Memory Biases in Sad and Formerly Depressed Individuals

Authors

    • Department of PsychiatryRadboud University Medical Centre
  • Iris van Oostrom
    • Department of PsychiatryRadboud University Medical Centre
  • Linda Isaac
    • Department of Psychiatry and Behavioral SciencesStanford University School of Medicine
    • Veterans Affairs Palo Alto HealthCare System
  • Eni S. Becker
    • Behavioural Science InstituteRadboud University Nijmegen
  • Anne Speckens
    • Department of PsychiatryRadboud University Medical Centre
Original Article

DOI: 10.1007/s10608-013-9590-8

Cite this article as:
Vrijsen, J.N., van Oostrom, I., Isaac, L. et al. Cogn Ther Res (2014) 38: 334. doi:10.1007/s10608-013-9590-8

Abstract

Cognitive theories assume a uniform processing bias across different samples, but the empirical support for this claim is rather weak and inconsistent. Therefore, coherence between biases across different cognitive domains in a sample of 133 non-depressed (Study 1) and a sample of 266 formerly depressed individuals (Study 2) was examined. In both studies, individuals were selected after a successful sad mood induction procedure. A Dot Probe task, an Emotional Stroop task and a self-referential Incidental Learning and Free Recall task were administered to all participants. Principle component analyses indicated coherence between attentional and memory bias in non-depressed, while in formerly depressed individuals distinct components for attentional biases and for memory bias were uncovered. The data suggest that in formerly depressed individuals, self-referent processing during encoding may be related to memory bias, whereas in non-depressed individuals memory bias may be related to both attentional bias and self-referent processing.

Keywords

DepressionAttentional biasMemory biasImageryCoherence

Introduction

The cognitive theories of depression (Beck 1976, 2008; Bower 1981; Gotlib and Joormann 2010) posit that depressed individuals show preferential processing of negative over positive information (Mathews and MacLeod 1994, 2005). Processing biases endure beyond a depressive episode and are activated by stress or negative mood (Joormann and Gotlib 2007; McCabe et al. 2000). After recovery from depression, exposure to a negative mood induction can trigger biased processing (Scher et al. 2005), suggesting that biases may play an important role in the vulnerability to depression (Gotlib and Joormann 2010). In fact, preferential processing of negative information is regarded as a determinant of the development and recurrence of depression (Beevers and Carver 2003; Gotlib and Joormann 2010). In non-depressed individuals, induced sad mood is also associated with biased processing of emotional information (Bradley et al. 1997; Matt et al. 1992). It has generally been found that sad and (formerly) depressed individuals recall negative information more easily than positive information (Matt et al. 1992; Ridout et al. 2009). Biased memory has most frequently been found for self-referential information (Matt et al. 1992). Sad and depressed individuals also show preferential attentional processing of negative information (Peckham et al. 2010).

The cognitive theories assume uniformity among biases in multiple cognitive domains and the ‘combined cognitive bias hypothesis’ has elaborated on this notion by arguing that biases influence each other (Hirsch et al. 2006). However, only a handful of studies have looked at the coherence between attentional and memory biases in the same sample and these findings are inconsistent (Everaert et al. 2012). In their review paper, Everaert and colleagues discuss studies examining the association between biases which provide information about uniformity among biased cognitive processes. Attentional biases for verbal and facial stimuli and memory bias for verbal stimuli were not interrelated in a sample of depressed individuals, as the different bias indices did not correlate significantly (Gotlib et al. 2004). Another study found coherence between attentional and memory biases for identical verbal stimuli in a sample of dysphoric students, but not in non-dysphoric students (Koster et al. 2010). Similarly, evidence was reported for a meditational role of attentional bias in the association between dysphoria status and memory bias in undergraduate students (Ellis et al. 2011; Wells et al. 2010). In these studies, attention for respectively verbal and facial stimuli measured with eye-tracking free viewing paradigms was associated with dysphoria status as well as with recognition of the previously presented stimuli. In summary, the studies report coherence between biases in dysphoria and no coherence between biases in currently depressed and healthy individuals. This may reflect different processes depending on depressive status. To date, there is little evidence for coherence between biases in clinical samples and coherence has not yet been studied in sad transient mood or formerly depressed individuals.

Gaining more insight into the interrelatedness between biases seems especially important now studies have started examining the causal relationship between biases and their effect on emotional symptoms, using Cognitive Bias Modification (CBM) techniques. In CBM, cognitive biases are altered using computerized training procedures, and the consequences for clinically relevant symptoms are assessed (Mathews and MacLeod 2002). Studies have found that manipulation of a cognitive bias can transfer to another cognitive domain (Hertel and Mathews 2011), indicating coherence between biases. Most CBM studies have, however, focused on healthy individuals and the few studies in depression have yielded mixed results (Beard et al. 2012; Hallion and Ruscio 2011). This also indicates that perhaps biased processing operates under different conditions in clinically depressed than in healthy (sad) individuals. Therefore, examining interrelations between biases in depressive mood state as well as depression vulnerability will advance our understanding about the relationship between biases.

We explored the coherence between attentional and memory biases in sad mood state (Study 1) and remitted depression (Study 2) using three well-established tasks similar to those in the study of Gotlib et al. (2004; Dot Probe task, Emotional Stroop task, self-referential explicit memory task). If, as predicted by the cognitive theories, biased processing is a prominent characteristic of depressive mood state as well as depression vulnerability, uniformity on different bias measures should be uncovered in both studies. This is based on the prediction that individuals displaying a pronounced negative bias on one task should also be more sensitive to negative stimuli on the other tasks.

Study 1

Method

Participants

This study is part of a larger study, which aims to link cognitive biases to genetic susceptibility to depression and includes several computer tasks. A sample of 180 undergraduate female students of the Radboud University Nijmegen participated in this study. All non-depressed participants (NDs) gave informed consent and received course credit in return for their participation. Demographics and sample descriptives are presented in Table 1.
Table 1

Means, percentages and standard deviations (SD) of demographics and BDI-II scores per sample (NDs in Study 1, FDs in Study 2)

Dependent variable

Sample

Dot probe

Emotional Stroop

Incidental learning and free recall

Mean

SD

N

Mean

SD

N

Mean

SD

N

Age (years)

NDs

21.2

3.9

133

20.5

2.5

91

21.1

3.9

116

FDs

47.4

11.9

266

47.1

12.0

226

47.4

11.7

256

Sex (%female)

NDs

100

 

133

100

 

91

100

 

116

FDs

66

 

266

66

 

226

65

 

256

BDI-II (total)

NDs

5.7

5.4

133

6.3

5.9

91

5.6

5.6

116

FDs

14.7

10.0

266

14.2

10.0

226

14.7

9.9

256

NDs non-depressed individuals, FDs formerly depressed individuals

Materials and Apparatus

Mood Induction

Before each task, participants saw a highly emotional sad scene (respectively 12, 6 and 4 minutes in length) from the movie “Sophie’s choice” (see Fitzgerald et al. 2011). Participants were instructed to let the emotionality of the film influence their mood and to maintain the sad mood state. Participants were asked to rate their current mood state using a computerized visual analogue scale, that ranged from −10 (indicating saddest mood) to 10 (indicating happiest mood), at baseline and after each mood induction. Only participants for whom the mood induction was successful (i.e. who rated their mood state after the mood induction as more sad than happy) were selected. This resulted in a sample of 135 NDs, which represents 75 % of the total sample of NDs. Using this selection criterion for all three mood inductions engendered some differences in sample size per task.

Dot Probe Task

The design of the Dot Probe task was identical to the version used by Joormann et al. (2006). Sixty pairs of images were selected from the Karolinska Directed Emotional Faces database (Lundqvist et al. 1998). Each pair depicted an individual with two different facial expressions: 20 pairs had angry and neutral facial expressions, 20 pairs sad and neutral and 20 pairs happy and neutral expressions. The 1,000 ms presentation of a face pair was preceded by a black fixation cross in the middle of the screen (for 500 ms). Following the offset of the pictures, a small gray dot appeared on the screen location where one of the pictures had been and remained on the screen until the participant indicated whether the gray dot was on the left or right side of the computer screen by pressing one of two keys on the keyboard. Participants were instructed to determine the location of the probe as quickly and accurately as possible. The inter trial interval was 1,000 ms. Participants first practiced with only responding to the dot. Then, 12 neutral–neutral face pairs were used for practice trials. Each of the picture pairs was presented in each of four blocks, for a total of 240 trials, with a brief rest period between each block. The four presentations of each picture pair, the side on which the emotion face was presented and the side on which the dot was presented was counterbalanced for the four presentations each of picture pair, such that one of each of the four possible iterations was presented in each block for each pair.

Emotional Stroop Task

A computerized Emotional Stroop color-naming task with word-by-word presentation was used to measure attentional interference. A word-by-word version instead of a card version was chosen because the experimenter could not be present during testing. The order of the three valenced blocks (negative, neutral and positive) was randomized for each participant. Words were presented in red, yellow, green or blue against a white background. The testing phase was preceded by 18 practice trials of mixed valence including color words. Participants were instructed to indicate the ink color by pressing the corresponding key on the key board. Corresponding color dots were presented at the bottom of the screen during practice. During testing, the dots appeared for 1,000 ms in case of an error, simultaneously presented with the Dutch for ‘ERROR’ in red. The valenced blocks each consisted of 18 different words, presented 3 times, resulting in 54 trials per block. The words were presented in fixed order, with the restriction that a certain word would not be presented more than twice in a row. The words were selected from a database (Phaf et al. 2006) and matched for length across the valences.

Incidental Learning Task and Free Recall

Participants were presented with 12 negative and 12 positive words in fixed order, with the restriction that no more than two words of the same valence would be presented consecutively. Words were selected from two databases (Dutch translation of the Affective Norms for English Words database, Bradley and Lang 1999; Phaf et al. 2006). Each word was presented for 10 s in capital black letters against a white background. To make encoding self-referential, participants were instructed to vividly imagine themselves in a scene with the presented word (Rogers et al. 1977). Participants were asked to rate how well they were able to imagine themselves in the scene on a 5-point Likert scale. The next word would appear after the number corresponding with their answer was selected and confirmed. This task was followed by a short paper-and-pencil distraction task (Raven matrices; Raven 1958). Upon completion, participants were instructed to return to the computer for an unannounced free recall of the words from the previous computer task. Participants were instructed to type in all the words they could remember within 3 minutes.

Procedure

Participants first filled out the Beck Depression Inventory (BDI-II; Beck et al. 1996) as well as some questions about sociodemographic information. Then the first mood induction started. Subsequently, participants went through the Emotional Stroop task, the second mood induction, the Incidental Learning Task and Free Recall, a third mood induction and finally the Dot Probe task.

Statistical Analysis

Reaction time (RT) trials associated with incorrect responses and mechanical errors were eliminated from the data (5 % of all Stroop data and 1 % of all Dot Probe data) as well as extreme RT trials in the top and bottom 2 % of the variance. We corrected for the effects of potential outliers by computing the median RT of each participant, on each task, group and trial type. On the Dot Probe task, bias-scores were computed for each participant and for each emotional expression type, by subtracting the compatible trials (dot appeared on the side of the emotional face) from the incompatible trials (dot appeared on the side of the neutral face). The effect of the mood induction and the bias effects were examined using analysis of variance (ANOVA).

To examine the underlying structure of the biases, we created one bias difference-score per task that would best represent the relevant bias. On Dot Probe task: bias-score happy − bias-score sad faces, on Emotional Stroop task: (RT positive − RT neutral) − (RT negative − RT neutral), and on the Incidental Learning task and Free Recall: positive − negative scores. A higher/more positive score would represent stronger bias towards positive information compared to negative information. Correlations between the bias difference-scores were calculated and the coherence among the three different cognitive processing tasks was examined using principle component analyses (PCA) with oblique rotation (oblimin) to assess the covariance between factors. PCA allows for determining linear combinations of the variables while retaining as much information from the original measured variables as possible (Fabrigar et al. 1999). It allows for studying underlying structures above and beyond associations. The dot probe angry condition was left out of the PCA because the aim was to study the coherence between interference, selective attention and memory for similar (i.e. sad) stimuli. Also, we wanted to include comparable variables in Study 1 and 2. The association of the components scores derived from the PCA with depressive symptomatology (BDI-II total score) was analyzed using linear regression.

Results

Mood Induction

NDs showed a significant decline in mood rating between baseline (M = 4.42 SD = 3.69) and after the mood inductions (M = −2.87 SD = 1.96; M = −4.00 SD = 2.62; M = −4.96 SD = 1.99 after the first, second and third mood induction, respectively), F(3,21) = 52.79, p < .001, f = 2.75. Due to a programming error, baseline mood state ratings of only 39 NDs were available.

Bias

The means (these are in fact means of median RTs) and standard deviations are presented in Table 2. On the dot probe task, processing of the three facial expression differed from each other, F(2,131) = 4.33, p < .05, f = .26. Specifically, preferential processing of sad faces was found with F(1,132) = 6.64, p < .05, f = .22 comparing sad to happy faces and F(1,132) = 6.48, p < .05, f = .22 comparing sad to angry faces. No significant difference was found between the angry and happy faces, F(1,132) = .18, p = .67, f = .03. On the Emotional Stroop task, RTs for positive, neutral and negative words differed from each other, F(2,89) = 6.98, p < .005, f = .40. NDs demonstrated shorter RTs for positive words compared to neutral words F(1,90) = 14.11, p < .001, f = .40 and compared to negative words F(1,90) = 5.14, p < .05, f = .24. Neutral and negative words resulted in comparable color naming RTs, F(1,90) = 2.73, p = .10, f = .28. Results of the Incidental Learning task showed that NDs were better able to imagine themselves in a scene with positive than with negative words, F(1,115) = 34.66, p < .001, f = .55. NDs recalled more positive than negative words, F(1,115) = 29.78, p < .001, f = .51.
Table 2

Means and SDs of the reaction times, bias-scores and recall rates on the cognitive tasks per sample (NDs in Study 1, FDs in Study 2)

Task and stimulus type

Non-depressed (Study 1)

Formerly depressed (Study 2)

Mean

SD

Mean

SD

Dot probe (bias score in ms)

 Angry faces

−4.43

35.83

  

 Sad faces

5.17

30.37

−4.19

40.22

 Happy faces

−2.80

27.76

−5.24

47.15

Emotional Stroop (in ms)

 Negative/depr. words

590

90

35,839

7,419

 Neutral words

598

93

36,982

7,618

 Positive words

578

77

33,951

6,642

Incidental learning and free recall (median imagery, #words)

 Imagery neg. words

3.68

.84

3.69

.96

 Imagery pos. words

4.20

.70

3.93

.99

 Memory neg. words

4.36

1.96

3.15

1.56

 Memory pos. words

5.36

1.68

4.38

2.00

NDs non-depressed individuals; FDs formerly depressed individuals

Coherence

Table 3 presents the simple zero order correlations between the bias difference-scores. None of the correlations were significant. Only PCA factor loadings higher than .4 are presented (see Table 4). Two distinct components were uncovered. The Dot Probe task, the Emotional Stroop task and the Free Recall rates on the memory task cluster on one component. The second component consists of the imagery ratings and the recall rates within the Incidental Learning task and Free Recall phase. No co-variation between the factors was uncovered with a correlation of r = −.02.
Table 3

Correlations among the attentional and memory bias difference-scores per sample (NDs in Study 1, FDs in Study 2)

 

Dot probe

Stroop

Imagery

Memory

Non-depressed (Study 1)

 Dot probe

   

 Stroop

.16

  

 Imagery

−.08

−.07

 

 Memory

.04

.13

.04

Formerly depressed (Study 2)

 Dot probe

   

 Stroop

.07

  

 Imagery

−.03

−.00

 

 Memory

−.06

.05

.11

None of the correlations reached p < .05 level of significance

NDs non-depressed individuals, FDs formerly depressed individuals

Table 4

Factor loadings from pattern matrix per bias difference-score and sample (NDs in Study 1, FDs in Study 2)

 

Component

1

2

Non-depressed (Study 1)

 Dot probe

.61

 

 Stroop

.71

 

 Imagery

 

.83

 Memory

.58

.55

Formerly depressed (Study 2)

 Dot probe

.70

 

 Stroop

.76

 

 Imagery

 

.70

 Memory

 

.73

NDs non-depressed individuals, FDs formerly depressed individuals

Association Components with Depressive Symptomatology

The linear regression analyses did not yield a significant association between the attentional and memory bias component score (Dot Probe, Emotional Stroop and Free Recall) and BDI-II total score, β = .10, t = .92, p = .36, nor between the memory bias component score (imagery ratings and Free Recall) and BDI-II total score, β = −.18, t = −1.64, p = .11.

Discussion of Study 1

Upon inspection of the factor loadings of the PCA analysis, the results revealed coherence between bias measures on the Dot Probe task, the Emotional Stroop task and the Free Recall task. This suggests that attentional biases and memory bias may share the same underlying process in non-depressed individuals in a sad mood state. Imagery and memory of the same valenced verbal stimuli clustered together on a second component. However, these components may not be associated with depressive symptomatology in non-depressed individuals. As expected, NDs selectively attended to sad facial expressions and exhibited mood-congruent attention interference of negative words compared to positive words. Contrary to our expectations, sad mood state did not result in preferential imagery or recall of negative words. Fostered by the self-referential instructions, the positive self-image non-depressed individuals generally have might not be overwritten by an induced negative mood state and sad non-depressed individuals might promote strategies to restore their original mood state during distraction (Parrott and Sabini 1990). The finding that angry facial expressions attracted the least attention in this group may indicate that this facial expression is not relevant for sad individuals. Is it important to note that undergraduate students represent a highly homogeneous and non-clinical group. To increase clinical relevance, coherence between attention and memory bias was studied in formerly depressed individuals in Study 2.

Study 2

Method

Participants

A total of 337 formerly depressed individuals (FDs) participated in the study. This sample was recruited at the department of psychiatry of the Radboud University Medical Centre, as well as via HSK (multi-centred clinique). Participants were included if they met the criteria of the DSM-IV (American Psychiatric Association 2000) for a previous depressive episode. Exclusion criteria were: current depressive episode, current or lifetime bipolar disorder, current psychotic symptoms, alcohol or substance abuse within the past 6 months, deafness, blindness, neurological disorder, sensorimotor handicaps and IQ estimate less than 70. Trained professionals interviewed eligible participants with the Structured Clinical Interview for the DSM–IV Axis-I disorders (SCID-I; First et al. 1995). The SCID-I has been demonstrated to have a good reliability (Skre et al. 1991; Williams et al. 1992). All interviewers had extensive training in the use of the SCID-I, as well as previous experience in administering structured clinical interviews to psychiatric patients. Participants received a gift certificate for their participation. Sample descriptives are presented in Table 1.

Materials and Apparatus

Mood Induction

FDs underwent a similar negative mood induction procedure as the sad mood group to activate biased processing (Segal and Ingram 1994). Similar to study 1, only participants for whom the mood induction was successful were selected (N = 291, representing 86 % of the total sample of FDs).

Dot Probe Task

In order to keep demands to a minimum, the trials with angry faces were not part of the Dot Probe task. Besides this difference, the Dot Probe task was identical to Study 1.

Emotional Stroop Task

In this study, a card version of the Emotional Stroop task was used with neutral, positive and negative words. After a practice card containing 15 words of mixed categories, the experimental cards containing 48 words each were presented in random order. Presentation of each card started when the researcher (blind to the words’ valence) pressed the space bar on the computer keyboard. It ended as soon as the participant named the print color of the last word on the card and the researcher pressed the space bar again. Participants were instructed to correct errors, which was represented by a slightly longer response time. Total response time per card was used as the dependent variable. The words were selected from two databases (Bradley and Lang 1999; Phaf et al. 2006) and were matched in length.

Incidental Learning Task and Free Recall

The procedure was the same in Study 2 as it was in Study 1, the only difference was that the negative words for the formerly depressed individuals were selected to be depression-specific.

Procedure and Statistical Analysis

The procedure was similar for Study 1 and Study 2, with the only difference being that the formerly depressed individuals in Study 2 did not receive a third mood induction because a pilot study had learned that a third mood induction after the brief memory task put too much of a burden on the participants. Data cleaning procedures and analyses were similar for Study 1 and Study 2, except that age and sex were included as covariates in the ANCOVAs in Study 2. Furthermore, the association of the PCA component scores with number of past episodes was also analyzed.

Results

Mood Induction

FDs showed a significant decrease in mood during both the first (pre M = 3.49 SD = 3.96, post M = .04 SD = 4.19) and second mood induction (pre M = 3.30 SD = 3.78, post M = −.60 SD = 4.32), respectively F(1,253) = 187.54, p < .001, f = .86 for the first mood induction and F(1,258) = 341.90, p < .001, f = 1.15 for the second.

Bias

Mean bias scores and standard deviations are presented in Table 2. On the Dot Probe task, FDs showed preferential processing of sad versus happy faces, F(1,263) = 4.01, p < .05, f = .12. On the Emotional Stroop task, a trend significant effect of Valence was found, F(2,220) = 2.70, p = .07, f = .16. FDs showed longer RTs for depression-specific negative than for positive words, F(1,242) = 48.29, p < .001, f = .45. They demonstrated most interference from neutral words, F(1,242) = 19.72, p < .001, f = .28 comparing neutral to depression-specific negative and F(1,243) = 112.40, p < .001, f = .68 comparing neutral to positive words. On the Incidental Learning task and Free Recall, FDs showed no difference on imagery of positive and depression-specific negative words, F(1,264) = 1.47, p = .23, f = .08. FDs recalled more positive than depression-specific negative words, F(1,253) = 13.30, p < .001, f = .23.

Although all participants were in remission at the time of testing, there was variability in depressive symptomatology (range BDI-II total scores 0–46). To control for the effect of depressive symptoms on biased processing, the bias analyses were repeated including BDI-II total score as covariate. The effect of Valence remained significant on the Dot Probe task, F(1,262) = 4.70, p < .05, f = .14, and on recall, F(1,252) = 14.61, p < .001, f = .24. The results on the Emotional Stroop task went from trend significant to nonsignificant, F(2,219) = 1.73, p = .18, f = .13. When correcting for variation in depressive symptomatology, FDs are more successful in positive than negative imagery, F(1,263) = 7.74, p < .01, f = .17.

Coherence Among the Tasks

Table 3 presents the simple zero order correlations between the bias difference-scores. None of the correlations reached significance, although the correlation between imagery and recall had a trend significant p-value of .09. The variables of the Dot Probe task and the Emotional Stroop loaded on one factor. Imagery ratings and the recall rates within the Incidental Learning task and Free Recall phase loaded on a second component (see Table 4). The two factors were not related with r = −.03.

Association Components with Depressive Symptomatology

The linear regression analysis yielded a significant association between the memory bias component score and the BDI-II total score, β = −.26, t = −3.84, p < .001, as well as with the number of past episodes, β = −.16, t = −2.37, p < .05. The attentional bias component score (Dot Probe and Emotional Stroop) was not significantly associated with BDI-II total score, β = −.06, t = −.92, p = .36, nor with the number of past depressive episodes, β = −.11, t = −1.63, p = .10.

Discussion of Study 2

With regard to coherence, two clear clusters of biases were present: attentional processing biases on the one hand and imagery and memory biases on the other hand. Only the memory bias component was associated with depression severity measures. Furthermore, FDs showed a mood-congruent bias pattern on attention interference as well as selective attention, but not on memory. Looking at selective attention, experiencing past depressed episodes might facilitate general avoidance of emotional facial expressions, albeit stronger for happy than for sad faces (conform behavioral avoidance; Vrijsen et al. 2013). Current depressive symptoms had little influence on biased processing. However, the pronounced variability in depressive symptomatology in FDs indicates that it represents a heterogeneous group and current symptoms are important to take into account.

General Discussion

The results indicate that in NDs, memory bias cohered with both self-referent imagery during encoding as well as with attentional biases. The coherence pattern in NDs is in accordance with studies that have reported dependence between attentional and memory biases as well as between biased processing of identical information in dysphoric students (Ellis et al. 2011; Koster et al. 2010; Wells et al. 2010) and indicates that in NDs attentional and memory biases may be related to the same underlying process. In FDs, no evidence for one underlying bias process was uncovered, which is in accordance with earlier findings in depressed individuals (Gotlib et al. 2004). Memory bias was associated with self-referent imagery and the attentional bias measures were associated with each other, indicating that in formerly depressed individuals memory bias may be specifically dependent of self-referent processing during encoding. Although speculative, in ND’s as well as FDs, sad mood may activate a negative “associative network” as proposed by Bower (1981), resulting in negatively biased mood-driven information processing. In FDs, self-referential processing may additionally activate latent dysfunctional mental representations about the self, also called “schemas”. Consistent with Beck’s theory (1976, 2008), schema-driven information processing might have affected memory bias, resulting in a distinction between attention and memory in FDs.

The results of the two studies underscore that without more knowledge of the underlying process, we cannot simply regard the concept ‘biased processing’ as a global vulnerability factor for emotional disorders. This is corroborated by the finding that in FDs, the memory bias component and not the attention bias component was associated with depressive symptomatology and number of past episodes. Biased processing in memory may represent an especially strong characteristic of depression vulnerability. In NDs, no such association was found, probably due to a low variation of depressive symptomatology in non-depressed undergraduates. The current findings are especially relevant now studies have started manipulating biases and studying the effects on emotional symptoms using CBM. In affected individuals, biased processing may operate under different conditions than in healthy individuals. This may partly explain why the results of CBM in depressed have been mixed (Beard et al. 2012; Hallion and Ruscio 2011). In non-clinical samples (dysphoric students or NDs), coherence between attentional and memory biases was found with identical as well as different stimuli sets. This indicates that the association between biases may be rather robust in non-affected individuals. Both studies with clinical samples (Gotlib et al. 2004 and Study 2) found no evidence for one underlying bias process using different stimuli across the tasks. In order to establish lack of coherence in affected individuals, efforts should be directed at examining coherence between biased processing of identical stimuli in (formerly) depressed samples.

The limitations of the present studies warrant mention. First, different versions of Emotional Stroop task were used in Study 1 and Study 2. Given that Stroop effects were examined within samples using highly comparable stimuli (see Salo et al. 2001), we are confident that both Emotional Stroop tasks measured the same attentional interference process. Secondly, participants in Study 1 were selected on sad mood state. However, because baseline mood ratings were missing, we were not able to select participants on decrease in mood, which might have been a more appropriate mood measure. Another limitation is the relatively low reliability of our bias measures (Eide et al. 2002; Schmukle 2005), which might have contributed to the lack of unitary coherence in FDs as well as in depressed individuals (Gotlib et al. 2004). Increasing reliability of bias measures is therefore of importance, possibly by adapting current paradigms or through multiple sampling of bias measures within individuals. Furthermore, processing of emotional information on one task may have influenced processing on a subsequent task. It was however expected that any order effects would be eliminated by the fairly strong effect of the mood induction. Although the current setup allowed for elegant examination of coherence within groups, direct comparison between groups was not possible because of the large demographic differences.

The present findings suggest an avenue for intervention, though such possibilities obviously go beyond the data. Cognitive therapy generally focuses on adapting current dysfunctional assumptions and attentional processes (Beck 1983). Our findings indicate that it could be effective to address both attentional as well as memory processes in cognitive therapy; focusing on the here and now as well as challenging dysfunctional beliefs from the past. Moreover, considering the emotionality of memory in trainings such as the MEmory Specificity Training (Raes et al. 2009) may offer a structural methodology for targeting memory bias and improving patients’ wellbeing. In conclusion, biased processing in different cognitive domains may operate according to the same underlying principle in non-depressed individuals, while processing of self-referent information in memory may be a distinct process in affected individuals. Experiencing a depression alters the schema’s about oneself (Beck 1976, 2008), which possibly influences memory processes.

Acknowledgments

We would like to thank the participants for their time and effort and the Behavioural Science Institute for its support.

Copyright information

© Springer Science+Business Media New York 2013