Journal of Affective Disorders 198 (2016) 127–134 Contents lists available at ScienceDirect Journal of Affective Disorders journal homepage: www.elsevier.com/locate/jad Research paper Where the depressed mind wanders: Self-generated thought patterns as assessed through experience sampling as a state marker of depression Ferdinand Hoffmann a,n, Christian Banzhaf b, Philipp Kanske a, Felix Bermpohl b, Tania Singer a a b Department of Social Neuroscience, Max Planck Institute for Human Cognitive and Brain Sciences, Leipzig, Germany Department of Psychiatry and Psychotherapy, Charité Campus Mitte, Universitätsmedizin Berlin, Germany art ic l e i nf o a b s t r a c t Article history: Received 9 December 2015 Received in revised form 28 February 2016 Accepted 5 March 2016 Available online 8 March 2016 Background: Self-generated thoughts (SGTs), such as during mind wandering, occupy much of our waking life. Individuals with Major Depressive Disorder (MDD) are less in the “here and now” and prone to rumination. Few studies have looked at SGTs in depression using experience sampling methods and no study has so far investigated the specific contents of depressive SGTs and how they vary from one time point to another. Methods: MDD patients (n ¼25) and matched healthy controls (n¼26) performed an established mind wandering task, involving non-demanding number discriminations. Intermittent probe questions ask for participants' current SGTs, that is, how off-task the thoughts are, how positive or negative, self- or otherrelated, and past- or future-oriented. Results: Multi-level modelling revealed that MDD patients engaged in more mind wandering than healthy controls. Their SGTs were predominantly negative and less positive, more self-related and pastoriented. Strongest predictor of depressive SGT was the decreased positive valence of thoughts. MDD patients' future and past-oriented thoughts were particularly more negative compared to healthy controls. Within MDD patients, the less positively valenced thoughts they had and the less variable these thoughts were, the more depressive symptoms they showed. Limitation: No other measures of rumination and worry were used. Conclusion: MDD patients show a very specific SGT pattern, possibly reflecting ruminative and anxious thoughts. This SGT pattern in depression might represent a useful state marker and even constitute an etiological factor of this debilitating disease, considering the importance of current SGT on and individual's cognitive processes and affective states. & 2016 Elsevier B.V. All rights reserved. Keywords: Depression Mind wandering Experience sampling Self-generated thought State marker 1. Introduction Self-generated thoughts (SGTs) arise independently of external stimulation through the environment, and comprise experiences such as mind wandering, day-dreaming, rumination and planning (Smallwood, 2013). It is known that SGT forms a crucial part of human mental activity, occupying up to 50% of our waking mind (Kane et al., 2007; Killingsworth and Gilbert, 2010). Mind wandering occurs particularly when attentional and cognitive demands in relation to the external environment are low (e.g. Smallwood et al., 2004). Some studies have linked SGT such as n Correspondence to: Department of Social Neuroscience Max-Planck Institute for Human Cognitive and Brain Sciences, Stephanstrasse 1a, 04105 Leipzig, Germany. E-mail address: hoffmann@cbs.mpg.de (F. Hoffmann). http://dx.doi.org/10.1016/j.jad.2016.03.005 0165-0327/& 2016 Elsevier B.V. All rights reserved. mind wandering to negative mood and unhappiness (Killingsworth and Gilbert, 2010; Stawarczyk et al., 2013), recent findings, however, suggest that it is crucial to consider ‘where the mind wanders’ and to look more specifically at the content of SGTs. For example, in healthy individuals, past-focused thoughts seem to be related to a reduction in positive mood, more depressive symptoms and increased cortisol levels after stress, while future-focused thoughts lead to an increase in positive mood and an attenuated stress response (Baird et al., 2011; Engert et al., 2014; Ruby et al., 2013a; Smallwood and O’Connor, 2011; Smallwood et al., 2007). Thus, SGT seems to represent a heterogeneous mental phenomenon with variable effects on human cognition, affect and behaviour (Andrews-Hanna et al., 2014). Patients with depression are known to show considerable difficulties with staying in the “here and now”. Investigations have demonstrated that MDD patients tend to engage in maladaptive 128 F. Hoffmann et al. / Journal of Affective Disorders 198 (2016) 127–134 SGTs in the form of excessive rumination and worry (NolenHoeksema, 1991, 2000; Watkins, 2008). Depressive rumination has been characterized as a mode of responding to distress which involves repetitively and passively focusing on symptoms of distress and on the possible causes and consequences of these symptoms (Nolen-Hoeksema, 2000). Worry, in contrast, has been described as a chain of negative uncontrollable thoughts and images, constituting an attempt to engage in problem-solving on issues with an uncertain outcome (Borkovec et al., 1983). While rumination and worry tend to correlate and share some similar features such as their repetitive nature, they have also been found to be statistically distinguishable (Nolen-Hoeksema et al., 2008). It has thus been proposed that rumination is more past and present oriented, focused on issues of self-worth, meaning and loss, whereas worry seems to be future-oriented, focused on anticipating threats (Nolen-Hoeksema et al., 2008; Watkins, 2008). While both rumination and worry have been associated with depressive symptoms, rumination seems to be more strongly related to depression (Hendriks et al., 2014; Hong, 2007). In MDD, rumination seems to represent a vulnerability marker for developing the disorder, and also shows a relation to the duration of MDD episodes and relapse probability (Nolen-Hoeksema et al., 2008). According to the cognitive model of depression, adverse early life events lead to internalization of negative self-referential schemas that subsequently bias information processing, in how stimuli are encoded, organized and retrieved (Disner et al., 2011). Therefore, these negative self-referential schemas might bias thoughts towards rumination of negative past experiences in depressive patients. Previous research on rumination and worry in depression has been largely based on questionnaire and induction methods, which both have considerable limitations and lack ecological validity. Questionnaire methods rely heavily on retrospection which is often impaired in psychiatric populations. Most induction methods prompt participants towards rumination about certain subjects (thinking e.g. about the long-term goals you have set), which then is compared to distraction inductions, asking the participants to focus on non-self-relevant images (Nolen-Hoeksema and Morrow, 1993). However without no-intervention control conditions, the effects between the two different conditions are difficult to disentangle. More critically these induction methods do not capture momentary online thoughts. In contrast growing research on SGTs in healthy individuals suggests that if one wants to make more objective claims about the wandering mind in health or in depression one ought to investigate SGTs using online experience sampling measures. Such online-measures of SGTs would allow picking up disease specific SGT patterns in the moment when they actually occur, and are particularly useful in depression research as ecological momentary assessment avoids the recall bias toward negative memories and also allows for better generalization to real life. To our knowledge only one study looked at online SGTs in clinically depressed participants (Ottaviani et al., 2014). Using an experience sampling method this study categorized SGTs in depression as either normal mind wandering, on task thoughts or perseverative cognition, linking the latter to autonomic inflexibility in lower heart-rate variability. No study however has so far tried to objectively decompose the specific SGT contents in clinical depression, regarding the previously established temporal (past or future oriented), social (self or other related), and emotional (negative or positive) dimensions of SGT (Engert et al., 2014; Ruby et al., 2013a, 2013b). From a clinical perspective identifying the specific SGT pattern in depression seems of great interest, as such SGT pattern could function as a state marker of depression, which could potentially then be utilized as a measure of disease progression and therapeutic change, as well as helping better differentiation of mental disorders. In addition it is unknown how these SGT contents in MDD patients relative to healthy controls vary over time, which could function as a good indicator of the repetitiveness of thoughts, inherent in rumination and also worry. This study thus aimed to comprehensively investigate the space of SGTs in depression in terms of amount, content and variability. Based of clinical observation and description disordered thoughts in depression have been commonly conceptualized as rumination and worry. Such disordered thoughts in depression can however, considering the recent findings in mind wandering research possibly more accurately be described by pathological SGTs in depression. In this study we thus aimed to specifically investigate the online SGT contents and their variability in a sample of clinically depressed patients. We used an established non-demanding choice reaction time task (CRT) that allows spontaneous SGTs in participants. During this task participants are asked at random time points, first, how much they were on task, and, secondly, about the specific content of their thoughts (Engert et al., 2014; Ruby et al., 2013a, 2013b), such as questions asking if their thoughts were focused on certain temporal epochs (future or past), involved different referents (self or other) and varied in valence (negative or positive). This task is particularly useful as an objective online measure of the amount and specific content of SGTs, but also of their variability over time, as participants are asked about the SGTs repeatedly throughout the task. One index of thought variability is the fluctuation of thought ratings from one thought probe to the next, computed as the squared successive difference in ratings, which has been an established method in experience sampling studies (EbnerPriemer et al., 2007; Jahng et al., 2008; Skirrow et al., 2014; Trull et al., 2008). Another index of variability represents the extremity of the thought ratings, calculated as the squared difference of each rating from the mean for that variable. We first hypothesized that MDD patients relative to healthy controls would generally show more mind wandering, that is engage in more SGTs, being less in the ‘here and now’, and that these stimulus independent thoughts would be more negative, self, and past related in accordance with the cognitive model of depression and findings on increased rumination in depression. Secondly, we also hypothesized that MDD patients would think more about negative future events, possibly pointing to their tendency to engage in stronger worrying compared to healthy controls. Thirdly we expected that MDD patients would also show less variability and more rigid thought patterns compared to healthy controls, capturing the repetitive nature of their thoughts. Lastly we expected a relation between symptom severity and specific patterns in SGTs and their variability in depression. 2. Methods 2.1. Participants 25 patients with depression were recruited through the inpatient clinic of the Charité-Universitätsmedizin Berlin, or were referred to us by specialized clinicians (see Table 1). 26 healthy control (HC) participants matched to the patients in terms of years of education, age, and gender with no history of psychiatric or neurological disorders were recruited by public notices and from project databases of the Charité-Universitätsmedizin Berlin. Participants were assessed for mental disorders using the Structured Clinical Interview for the DSM-IV (American Psychiatric Association, 2000) and a diagnosis of acute state of depression was confirmed with no other primary diagnoses. All participants completed the Beck Depression Inventory (BDI, Hautzinger et al., 1995), and were also assessed with the Hamilton Depression Rating Sale (HAMD-17, Hamilton, 1960). Additionally, participants completed a measure of F. Hoffmann et al. / Journal of Affective Disorders 198 (2016) 127–134 Table 1 Demographic and clinical characteristics of the participants. Sample size Gender Age Education (years) Verbal IQ (WST) BDI HAMD-21 Healthy controls MDD patients Significant effects (p o 0.05) 26 10 males 41.9 (13.6) 16.9 (2.9) 105.4 (7.5) 2.5 (3.8) 0.77 (1.6) 25 8 males 41.1 (12.6) 15.2 (3.2) 100.7 (11.4) 32.0 (9.6) 20.4 (3.6) * * 129 5) negatively valenced and 6) positively valenced and 7) how much off task their thoughts were at that point in time. Additionally, they rated their current mood (i.e. how positive and how negative they felt). Importantly healthy controls and MDD patients did not differ in terms of accuracy (healthy controls: 87.2%, MDD patients: 86.4%) and reaction time (healthy controls: 882.29 ms, MDD patients: 923.62 ms) on the CRT task (more detail reported in Section 3). The entire task lasted approximately 14 min. Stimuli were presented using E-prime 2.0 (Psychology Software Tools, Inc., Sharpsburg, PA, USA). 2.3. Data analysis Table 2 Medication of MDD patients. Medication Number of participants SSRI NaSSA SNRI NDRI TCA AAP SSRIþLithium TCA þLithium NaSSA þ Lithium SSRIþAAP SNRIþ AAP SSRIþTCA SSRIþAgomelatine AAPþ Lithium SSRIþNaSSA 6 2 3 2 1 1 1 1 1 1 1 1 1 1 1 crystallized intelligence (Wortschatztest, WST, a vocabulary test part of the HAWIE-R, the German adaptation of the Wechsler Adult Intelligence Scale, Schmidt and Metzler, 1992). All MDD patients except one were on medication (see Table 2). No patient was medicated with benzodiazepines for at least 48 h. Exclusion criteria included any comorbid axis I disorder, current neurological disorder, substance abuse within 6 months before study participation, diagnosis of antisocial personality disorder or borderline personality disorder. None of the patients had a history of electroconvulsive therapy (ECT). The study was approved by the local research ethics committee (Charité-Universitätsmedizin Berlin) and written informed consent was obtained from all participants. 2.2. CRT task We used an established mind wandering paradigm that probes off-task thoughts during a choice reaction time task (Baird et al., 2012; Ruby et al., 2013a; Smallwood et al., 2013) and assessed the content of the participants' thoughts on six different dimensions 1) past, 2) future, 3) self, 4) other, 5) negative valence and 6) positive valence (e.g. Engert et al., 2014; Ruby et al., 2013a). A series of black digits between 1 and 8 was presented. One sixth of the digits was presented in red color signaling participants that they should indicate via button press if this number was odd or even. Black digits were presented for 1000 ms and red digits for 2000 ms. Responses had to be made while the colored digits were still present on the screen. Stimuli were separated by a fixation cross of variable duration (2200–4400 ms). The number of thought probes and their presentation were randomly determined (Ruby et al., 2013a), to avoid any expectancy biases, and thus sampling SGTs in the most unconstrained way (number of probes between four and nine). Participants were asked to rate their current thoughts using a nine-point Likert scale on several dimensions including how much their thoughts were 1) past-oriented, 2) future-oriented, 3) self-related, 4) other-related, For the main analyses, we used multi-level models because they take correlated observations within individuals into account and perform well with missing data or unequal numbers of data points within individuals (Jahng et al., 2008). Linear mixed models were calculated in SPSS version 22 with the number of the particular sampling point within the session (e.g., sample 5) as covariate and with a random intercept. Sampling point was used as a covariate within the models to account for different numbers of sampling points between the participants. In addition to the rating level (e.g. to what extent was a certain thought negative or selfrelated), we investigated two indices of variability, i.e., fluctuations and extremity in ratings. To obtain a measure of how much individuals fluctuate in their single SGT ratings from one thought probe to the next, we calculated squared successive differences, which has been established in experience sampling studies (Ebner-Priemer et al., 2007; Jahng et al., 2008; Skirrow et al., 2014; Trull et al., 2008). Fluctuation scores were calculated for each SGT dimension separately. To obtain a measure of how extreme the individual ratings were, we calculated the squared difference of each rating from the total sample mean (including healthy controls and MDD patients) for that variable. In contrast to the fluctuations, this does not take into account how big successive changes are, but rather indicates how much a certain rating differs from the “norm”. As a measure of effect size for the central main effects and interaction effects Omega-squared (ω2) was calculated by taking the difference from 1 and the variance of the residuals of the full model divided by the variance of the residuals of the model without the respective fixed factor of interest (Xu, 2003). A value of ω2 ¼.010 represents a small effect size, a value of ω2 ¼.059 indicates a medium effect size and a value of ω2 ¼.138 represents a large effect size (Kirk, 1996). In a first step, differences between MDD patients and HCs in ratings levels, fluctuations in ratings, and extremity in ratings (see below for details) were subjected to multi-level models as described above. In a second step we tested for group differences and in particular interrelations of the different content dimensions of SGT, on which healthy controls and MDD patients were found to differ. This allowed us to investigate whether certain SGTs were more strongly correlated with one another in MDD patients compared to healthy controls. Finally in a third step, we investigated how symptom severity in healthy controls and MDD patients related to the different SGT measures (rating, fluctuation, extremity). Symptom severity in healthy controls and MDD patients might relate to very specific SGTs. In the case of healthy controls such SGTs might represent a vulnerability factor towards the development of depression. 3. Results 3.1. Performance Independent samples t-tests showed no significant differences between healthy controls and MDD patients in accuracy (t(49) ¼ 130 F. Hoffmann et al. / Journal of Affective Disorders 198 (2016) 127–134 Table 3 Differences between MDD patients and HCs in rating levels, fluctuations in ratings and extremity in ratings as estimated with multi-level modelling. Model parameters Predictors Rating levels b Off task Fluctuations in ratings S.E. p-value ω2 b S.E. Extremity of ratings p-value ω2 b S.E. p-value ω2 Intercept Group Sample 6.634 24.736 2.571 4.363 5.432 0.734 0.140 0.001 0.001 0.108 0.050 359.591 277.831 43.210 203.236 180.207 45.631 0.078 0.124 0.345 0.009 0.004 846.379 75.813 60.207 147.244 184.016 17.665 o 0.001 0.682 0.001 0.002 0.033 Intercept Group Sample 25.859 2.930 1.268 4.221 5.403 0.622 o 0.001 0.593 0.042 0.004 0.012 933.098 43.052 2.681 206.792 268.797 49.229 o 0.001 0.874 0.957 o 0.001 0.004 680.979 120.726 11.640 104.074 113.300 21.552 o 0.001 0.287 0.590 0.006 0.003 Intercept Group Sample 45.729 21.686 2.879 3.634 4.446 0.869 o 0.001 o 0.001 0.007 0.070 0.052 857.550 300.419 21.361 246.294 218.004 55.763 0.001 0.169 0.702 0.008 0.001 990.713 108.317 10.471 125.197 143.679 19.568 o 0.001 0.451 0.593 0.004 o 0.001 Intercept Group Sample 45.036 26.220 3.109 5.374 7.024 0.660 o 0.001 0.001 o 0.001 0.191 0.053 268.231 122.903 6.681 117.744 104.220 26.658 0.024 0.239 0.802 0.006 o 0.001 1032.565 115.103 7.895 133.213 170.661 17.227 o 0.001 0.503 0.647 0.011 o 0.001 Intercept Group Sample 40.583 28.135 1.449 4.665 5.724 0.492 o 0.001 o 0.001 0.006 0.226 0.013 524.571 2.515 45.744 108.168 105.093 21.686 o 0.001 0.981 0.036 o 0.001 0.018 845.119 62.209 5.986 132.122 166.982 16.267 o 0.001 0.711 0.713 0.003 o 0.001 Intercept Group Sample 28.7787 16.268 1.661 3.417 4.344 0.672 o 0.001 0.001 0.014 0.129 0.050 352.832 461.218 54.674 219.309 244.840 43.935 0.108 0.062 0.214 0.020 0.018 852.037 304.787 45.124 215.769 215.674 20.314 o 0.001 0.166 0.029 0.016 0.017 Intercept Group Sample 24.160 10.964 3.680 5.001 6.068 0.725 o 0.001 0.077 o 0.001 0.021 0.087 954.850 40.411 48.768 235.007 208.597 52.983 o 0.001 0.847 0.358 o 0.001 0.003 787.485 105.544 47.501 125.673 145.361 29.654 o 0.001 0.470 0.112 0.014 0.020 Negative mood Intercept Group Sample 49.204 19.361 1.820 3.943 5.877 0.594 o 0.001 0.002 0.003 0.134 0.030 688.076 199.076 55.060 194.558 181.977 32.040 0.001 0.279 0.087 0.009 0.009 662.681 22.053 44.862 124.650 162.188 13.689 o 0.001 0.892 0.001 0.001 0.003 Positive mood Intercept Group Sample 37.993 25.131 1.357 3.667 5.048 0.491 o 0.001 o 0.001 0.006 0.240 0.010 257.217 23.414 2.991 73.350 65.278 16.133 0.001 0.720 0.853 o 0.001 0.032 607.443 32.923 20.523 118.0167 152.701 12.877 o 0.001 0.830 0.112 0.001 0.002 Other Self Negative Positive Past Future Note: sample represents the covariate of the number of sampling points of the thought probes. 0.13, p ¼.894; healthy controls: 87.2% 717.3, MDD patients: 86.4% 724.7) or reaction times (t(49) ¼1.26, p ¼.212; healthy controls: 882.29 ms 7106.24, MDD patients: 923.62 ms 7126.74) in the CRT. 3.2. Mood probes Multilevel models revealed that MDD patients showed elevated levels of negative and decreased levels of positive mood (see Table 3 and Fig. 1 Supplemental material). The fluctuations and extremity of the mood ratings were not significantly different between groups. 3.3. Thought probes 3.3.1. Rating levels Multilevel models revealed that MDD patients show significantly increased off-task thoughts, relative to the healthy controls (see Fig. 1 and Table 3). Regarding the particular SGT contents, MDD patients reported more negative and less positive, as well as more past- and self-related thoughts compared to healthy controls. MDD patients also engaged in marginally more future-oriented SGTs. Number of sampling points (“Sample”), was included as a covariate in all analyses. Non-significant results of the number of sampling points particularly for fluctuation and extremity suggest that variability measures were not as strongly influenced by the number of sampling points, thus did not change as much over the course of the experiment, as the ratings themselves. Including current mood as a covariate showed significant relations between mood and thought valence, but did not change the group differences in positive (b ¼7.800, S.E. ¼2.940, p ¼ .010) and negative thoughts (b ¼ 15.4435, S.E. ¼4.124, p ¼.001). There were no group differences in the amount of other-related thoughts. While healthy controls had more positive thoughts than negative thoughts (t(25) ¼ 4.022, p o .001), MDD patients had marginally more negative thoughts than positive thoughts (t(24)¼ 1.981, p¼ .059). 3.3.2. Variability There were no significant differences between the groups in extremity of SGTs (see Table 3). There was however a marginally significant group difference in fluctuations of past-related thoughts, with MDD patients showing less fluctuating past-related SGTs (see Table 3). F. Hoffmann et al. / Journal of Affective Disorders 198 (2016) 127–134 131 Table 4 Significant differences in interrelations between SGT contents in MDD patients and healthy controls as estimated with multi-level modelling. Model parameters Predictors Rating levels b Negative Negative Fig. 1. Amount of SGT and SGT contents for healthy controls and MDD patients. Depressed patients showed increased negative and decreased positive as well as increased self- and past-related thoughts, relative to healthy controls (error bars represent standard errors). 3.3.3. SGT contents as predictors of MDD Looking at what SGT content was the most accurate in distinguishing between MDD patients and healthy controls a stepwise logistic regression was performed on the two groups, with the six SGT contents as predictors (past, future, other, self, negative, positive). The results showed that the amount of positive thoughts was most discriminative between the two groups (B ¼ 0.45, Wald ¼ 59. 05, p o.001, Nagelkerke R square¼ .323). 3.3.4. Interrelations between SGT contents in MDD patients and healthy controls Looking specifically at interrelations of the different content dimensions of SGT, multi-level models were used on the SGT contents in which MDD patients and healthy controls were found to differ (self, past, negative, positive, see Section 3.3.1). These analyses were run to investigate whether certain SGTs tended to interrelate more strongly in MDD patients relative to healthy controls. Each SGT content was used as a covariate to investigate interrelation with another SGT (as dependent variable), while group was used as an independent factor (healthy controls vs MDD patients). This resulted in twelve separate models, with only the significant models reported in Table 4. In particular, it was found that MDD patients' temporal thoughts of past and future were significantly more negatively valenced in comparison to the healthy controls (see Fig. 2 and Table 4), suggesting a stronger preoccupation with negative past and negative future events. In contrast positive thoughts in healthy controls were more pastrelated compared to MDD patients. In addition past-oriented thoughts were more self-related in MDD patients compared to healthy controls. 3.4. Relation of SGT variables to symptom severity 3.4.1. MDD patients To test for a relation of the observed effects with symptom severity in the MDD group, we ran two separate liner-mixed models with BDI and Hamilton scores as covariates. Both symptom severity measures, BDI and Hamilton scores showed a significant positive correlation (healthy controls: r ¼0.68, p o0.01; MDD patients: r ¼0.43, p¼ 0.03). BDI was a marginal predictor for the amount of mind wandering in general (b ¼ 0.942, S.E.¼ 0.500, p ¼0.071). Looking specifically at SGT contents, BDI was a predictor of increased levels of past-related thoughts (b ¼1.109, S.E. ¼0.497, Positive Past S.E. p-value ω2 Intercept Group Sample Past*Group 38.418 20.527 2.803 0.217 5.543 6.406 0.760 0.092 o 0.001 0.002 0.001 0.019 0.074 0.017 0.032 Intercept Group Sample Future*Group 41.157 22.254 3.025 0.173 5.746 6.901 0.794 0.083 o 0.001 0.002 o 0.001 0.038 0.106 0.043 0.014 Intercept Group Sample Past*Group 46.546 20.628 1.419 0.245 4.896 6.156 0.516 0.085 o 0.001 0.001 0.009 0.004 0.107 o 0.001 0.041 Intercept Group Sample Self*Group 13.961 4.647 1.299 0.179 5.927 6.337 0.543 0.085 0.022 0.466 0.017 0.036 o 0.001 0.032 0.019 Note: sample represents the covariate of the number of sampling points of the thought probes. p ¼0.045), and a marginally significant predictor of decreased positive valence (b ¼ 0.863, S.E. ¼0.420, p¼ 0.050). Hamilton scores were a significant predictor of future-related thoughts (b ¼4.271, S.E. ¼0.623, p o0.001) and negative valence (b ¼2.369, S.E.¼1.293, p¼ 0.080). These findings suggest that more severe symptoms are associated with an increase in specific SGT contents. Relating symptom severity to SGT fluctuations, BDI scores were found to be a predictor of the fluctuations in self-related thoughts (b ¼ 39,527, S.E.¼14.031, p ¼0.006) and fluctuations in negative (b ¼ 18,634, S.E.¼ 8.369, p ¼0.027) and positive valence (b ¼ 20,029, S.E. ¼7.460, p ¼0.008). Hamilton scores were found to be a predictor of the fluctuations in self-related thoughts (b ¼ 74,202, S.E.¼ 34.163, p ¼0.032) and fluctuations positive valence (b ¼ 54,834, S.E.¼18.460, p ¼ 0.004). These findings suggest that, the more severe the depressive symptoms in MDD patients were the less variable the reported experience of specific SGTs in the course of the experiment was. Relating symptom severity to SGT extremity, BDI was found to be a predictor of extremity in past-related thoughts (b ¼34,396, S. E. ¼713.425, p ¼0.015). Hamilton was a predictor of extremity in future-related thoughts (b ¼ 91,014, S.E. ¼27.059, p ¼0.001). 3.4.2. Healthy controls Another interesting question was how depressive symptomatology in healthy controls related to SGTs, possibly indicating that certain SGTs can indicate a vulnerability to developing depression. BDI was found to be a marginal predictor of the amount of pastrelated thoughts (b ¼1.492, S.E. ¼0.775, p ¼ 0.067). Hamilton scores were also found to be a significant predictor of the amount of past-related thoughts (b ¼ 4.025, S.E.¼ 1.718, p¼ 0.023. Symptom severity was found to be unrelated to fluctuations in SGTs in healthy controls. BDI (b ¼ 54.687, S.E.¼ 16.886, p¼ 0.001), as well as Hamilton (b ¼ 70.058, S.E.¼ 41.067, p ¼0.090) were found to be a predictor of extremity in the amount of general mind wandering. 132 F. Hoffmann et al. / Journal of Affective Disorders 198 (2016) 127–134 Fig. 2. (A) In contrast to healthy controls, MDD patients show a significant positive relation between the negative valence and past-relatedness of SGTs. (B) In contrast to healthy controls, MDD patients show a significant positive relation between the negative valence and future-relatedeness of SGTs. 4. Discussion In this study we aimed to investigate thinking patterns related to mind wandering in clinical depression relative to healthy controls using an objective online-measure, a mind wandering laboratory task that probes the amount and specific content of taskunrelated thoughts in the moment in which they occur. More specifically the content of thoughts was assessed on following six dimensions: 1) past, 2) future, 3) self, 4) other, 5) negative valence and 6) positive valence (Engert et al., 2014; Ruby et al., 2013a, 2013b). In addition to typically used average scores of STG frequencies, we also looked here at fluctuation and extremity of SGTs as a measure of variability over time. We hypothesized that MDD patients relative to healthy controls would engage in more mind wandering, showing a specific SGT pattern possibly related to rumination and worry. In addition we expected that the amount and also the variability of depression specific SGTs would relate to symptom severity. As hypothesized MDD patients engaged in a greater amount of SGTs compared to healthy controls, showing more off-task mind wandering, being less in the “here and now”. The SGT pattern in MDD patients relative to healthy controls was characterized by significantly more self- and past-related thoughts that were more negative and less positive in valence. In particular, MDD patients' past-related thoughts were strongly coupled to negatively valenced SGTs. These findings suggest that distorted thought processes in depression can be very well captured and described in online depressive mind wandering. This specific depressive SGT pattern might in part relate to pathological rumination about negative past events, which is commonly found to be increased in depression (Nolen-Hoeksema, 1991, 2000). Supporting this suggestion, the “feelings of guilt” item of the Hamilton scale including questions about rumination was particularly associated with pastrelated thoughts in MDD patients (see Supplemental Information). This SGT pattern is also in accordance with the three proposed elements of rumination (Disner et al., 2011): altered emotion and memory processing, increased self-referential processing, and decreased top-down inhibition of these processes. Interestingly, the strongest predictor of depressive SGT was the positive valence of the thoughts. This is in accordance with some findings showing that the absence of positive affect might be actually more indicative of the depressive state, as increased negative affect is shared by most mood and anxiety disorders (Watson et al., 1988). Further, symptom severity also related with the amount of positive thoughts within the group of the MDD patients, as well as their variance. This suggests that the low and flat positive valence of thoughts plays a crucial functional role in depression and its severity. Findings of decreased positive affect in response to rewards and abnormal functioning of reward sensitive brain areas such as the nucleus accumbens in depression (Heller et al., 2009), might be seen as further support. In the case of the healthy controls depressive symptoms were related to increased amount of past-related thoughts as previously reported (Ruby et al., 2013a; Smallwood and O’Connor, 2011; Smallwood et al., 2007). Past-related thoughts might therefore represent a vulnerability factor for depression, possibly related to a ruminative response style, which has been shown to be associated with onset, maintenance and reoccurrence of depressive episodes (NolenHoeksema, 2000; Nolen-Hoeksema et al., 2008). While MDD patients only marginally engaged in more futurerelated thoughts, their future-oriented thoughts were significantly more negative than in healthy controls. This could be interpreted as a SGT pattern of worry, and indeed the somatic anxiety scale of the Hamilton scale was positively related to future thoughts in MDD patients, lending support for this suggestion (see Supplemental information). Interestingly future-related thoughts in healthy controls have mostly been associated with adaptive and beneficial effects such as increasing subsequent positive mood and attenuated cortisol levels at baseline and after a stressor (Engert et al., 2014; Ruby et al., 2013a). It has also been suggested that future-oriented SGT is instrumental in autobiographical planning and consolidation (Baird et al., 2011; Smallwood et al., 2011). Our results do show however that in the case of clinical depression future thoughts are negative and potentially maladaptive being related to worry that exacerbates the depressive state. In this study we also looked at SGT variability over time given the multiple sampling points of the thought probes. We used two indices of SGT variability, namely the fluctuation of the SGT F. Hoffmann et al. / Journal of Affective Disorders 198 (2016) 127–134 ratings, and the extremity of the SGT ratings, which both can be seen as a good indicator of the repetitiveness or perseverance of thoughts, such as present in depressive rumination and worry. Significant group differences were only found for the fluctuation of past-related thoughts, with MDD patients showing less variability in past-related thoughts compared to healthy controls. In addition, symptom severity in MDD patients related to fluctuations in SGTs. Particularly, the stronger the symptoms that the MDD patients portrayed, the less self-related, positive and negative thoughts fluctuated. Similarly symptom severity in MDD patients was associated with the extremity of future-related SGTs. In general these findings suggest that the stronger the depressive symptoms portrayed by MDD patients, the more rigid certain SGTs are over time. In particular the less variable the positive valence of such SGTs is, the more repetitive and perseverative certain thoughts and the corresponding affect are. Increasing inflexibility in SGTs in depression might be associated with reported cognitive inflexibility in depression (Altamirano et al., 2010; Joormann et al., 2011; Whitmer and Gotlib, 2013), due to serotonergic dysregulation within prefrontal cortex (Clarke et al., 2004). From a clinical point of view, identifying SGT patterns of MDD and other mental disorders seems to be of relevance. Disease specific SGT patterns can function as state markers, which can be utilized to monitor disease progression and more importantly therapeutic changes and shifts towards more adaptive SGTs. Mapping disease specific SGT patterns, could be also further helpful for diagnostic purposes. As decreased positive valence of SGT was most representative of the depressive mind, one could envision how therapies fostering self-compassion and positive affect might be particularly useful for clinical interventions (Van Dam et al., 2011). In addition mindfulness-based therapies have been shown to be very effective in treating depression, by increasing momentary positive emotions and reward experience (e.g. Geschwind et al., 2011) and decreasing negative mind wandering in terms of ruminative thinking (e.g. Ramel et al., 2004). For future studies it would be interesting to compare SGT patterns between different mental disorders. Naturally, investigating SGT patterns in anxiety disorders relative to depression would be very interesting, particularly as there exists a high comorbidity between these two disorders. It could be assumed that the SGT pattern of anxiety disorders relative to MDD shifts more strongly in the temporal domain towards future, representing pathological worry, while being possibly equally negative and self focused. Additionally it would be interesting to explore how SGT patterns change in remission from depression. Investigating SGT “online” with tasks such as the CRT combined with assessment of the different thought content dimensions, as used in this study could represent a more objective measure than clinical interviews and self-reports. Particularly self-reports demand a certain amount of retrospection and introspection, which are often deficient in many mental disorders. 4.1. Limitations One limitation of the present study is that no other measures of rumination and worry were available to relate these to the SGT pattern in depression, and strongly validate the interpretation of rumination and worry. In addition the study was based on a relatively small sample size. 5. Conclusion In conclusion this study aimed to investigate the specific contents of online SGTs and their variability in depression. The findings show that MDD patients engaged in more SGTs than healthy 133 controls that were seemingly ruminative in nature, being more negative and less positive in valence as well as more self- and past related. Particularly in the temporal domain, MDD patients' thoughts about the past and the future were more negatively valenced. The decreased positive valence of SGTs was the best predictor for depression. Additionally the variability in SGTs related negatively to symptom severity in depression. Investigating different SGT patterns of mental disorders such as depression using event-sampling online measures can be useful for deriving state markers that may help diagnosis and intervention. Appendix A. Supporting information Supplementary data associated with this article can be found in the online version at http://dx.doi.org/10.1016/j.jad.2016.03.005. 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. Psychol. Sci. 21, 1–6. American Psychiatric Association, 2000. Diagnostic and Statistical Manual of Mental Disorders. American Psychiatric Association, Washington DC. Andrews-Hanna, J.R., Smallwood, J., Spreng, R.N., 2014. The default network and self-generated thought: component processes, dynamic control, and clinical relevance. Ann. N. Y. Acad. Sci. 1316, 29–52. Baird, B., Smallwood, J., Mrazek, M.D., Kam, J.W., Franklin, M.S., Schooler, J.W., 2012. Inspired by distraction: mind wandering facilitates creative incubation. Psychol. Sci. 23, 1117–1122. Baird, B., Smallwood, J., Schooler, J.W., 2011. Back to the future: autobiographical planning and the functionality of mind-wandering. Conscious. Cogn. 20, 1604–1611. Borkovec, T.D., Robinson, E., Pruzinsky, T., DePree, J.A., 1983. Preliminary exploration of worry: some characteristics and processes. Behav. Res. Ther. 21, 9–16. Clarke, H., Dalley, J., Crofts, H., Robbins, T., Roberts, A., 2004. Cognitive inflexibility after prefrontal serotonin depletion. Science 304, 878–880. Disner, S.G., Beevers, C.G., Haigh, E.A., Beck, A.T., 2011. Neural mechanisms of the cognitive model of depression. Nat. Rev. Neurosci. 12, 467–477. Ebner-Priemer, U.W., Kuo, J., Kleindienst, N., Welch, S.S., Reisch, T., Reinhard, I., Lieb, K., Linehan, M.M., Bohus, M., 2007. State affective instability in borderline personality disorder assessed by ambulatory monitoring. Psychol. Med. 37, 961–970. Engert, V., Smallwood, J., Singer, T., 2014. Mind your thoughts: associations between self-generated thoughts and stress-induced and baseline levels of cortisol and alpha-amylase. Biol. Psychol. 103, 283–291. Geschwind, N., Peeters, F., Drukker, M., van Os, J., Wichers, M., 2011. Mindfulness training increases momentary positive emotions and reward experience in adults vulnerable to depression: a randomized controlled trial. J. Consult. Clin. Psychol. 79, 618. Hamilton, M., 1960. A rating scale for depression. J. Neurol. Neurosurg. Psychiatry 23, 56–62. Hautzinger, M., Bailer, M., Worall, H., Keller, F., 1995. BDI Beck-Depressions-Inventar. Verlag Hans Huber. Bern. Heller, A.S., Johnstone, T., Shackman, A.J., Light, S.N., Peterson, M.J., Kolden, G.G., Kalin, N.H., Davidson, R.J., 2009. Reduced capacity to sustain positive emotion in major depression reflects diminished maintenance of fronto-striatal brain activation. Proc. Natl. Acad. Sci. 106, 22445–22450. Hendriks, S.M., Licht, C.M., Spijker, J., Beekman, A.T., Hardeveld, F., de Graaf, R., Penninx, B.W., 2014. Disorder-specific cognitive profiles in major depressive disorder and generalized anxiety disorder. BMC Psychiatry 14, 1–9. Hong, R.Y., 2007. Worry and rumination: differential associations with anxious and depressive symptoms and coping behavior. Behav. Res. Ther. 45, 277–290. Jahng, S., Wood, P.K., Trull, T.J., 2008. Analysis of affective instability in ecological momentary assessment: indices using successive difference and group comparison via multilevel modeling. Psychol. Methods 13, 354–375. Joormann, J., Levens, S.M., Gotlib, I.H., 2011. Sticky thoughts depression and rumination are associated with difficulties manipulating emotional material in working memory. Psychol. Sci. 22, 979–983. Kane, M.J., Brown, L.H., McVay, J.C., Silvia, P.J., Myin-Germeys, I., Kwapil, T.R., 2007. For whom the mind wanders, and when an experience-sampling study of working memory and executive control in daily life. Psychol. Sci. 18, 614–621. Killingsworth, M.A., Gilbert, D.T., 2010. A wandering mind is an unhappy mind. Science 330, 932. Kirk, R.E., 1996. Practical significance: a concept whose time has come. Educ. Psychol. Meas. 56, 746–759. Nolen-Hoeksema, S., 1991. Responses to depression and their effects on the duration of depressive episodes. J. Abnorm. Psychol. 100, 569–582. 134 F. Hoffmann et al. / Journal of Affective Disorders 198 (2016) 127–134 Nolen-Hoeksema, S., 2000. The role of rumination in depressive disorders and mixed anxiety/depressive symptoms. J. Abnorm. Psychol. 109, 504–511. Nolen-Hoeksema, S., Morrow, J., 1993. Effects of rumination and distraction on naturally occurring depressed mood. Cogn. Emot. 7, 561–570. Nolen-Hoeksema, S., Wisco, B.E., Lyubomirsky, S., 2008. Rethinking rumination. Perspect. Psychol. Sci. 3, 400–424. Ottaviani, C., Shahabi, L., Tarvainen, M., Cook, I., Abrams, M., Shapiro, D., 2014. Cognitive, behavioral, and autonomic correlates of mind wandering and Perseverative cognition in major depression. Front. Neurosci. 8, 1–9. Ramel, W., Goldin, P.R., Carmona, P.E., McQuaid, J.R., 2004. The effects of mindfulness meditation on cognitive processes and affect in patients with past depression. Cognit. Ther. Res. 28, 433–455. Ruby, F.J., Smallwood, J., Engen, H., Singer, T., 2013a. How self-generated thought shapes mood—the relation between mind-wandering and mood depends on the socio-temporal content of thoughts. PLoS One 8, e77554. Ruby, F.J., Smallwood, J., Sackur, J., Singer, T., 2013b. Is self-generated thought a means of social problem solving? Front. Psychol. 4, 1–10. Schmidt, K.-H., Metzler, P., 1992. Wortschatztest. WST. Beltz. Skirrow, C., Ebner-Priemer, U., Reinhard, I., Malliaris, Y., Kuntsi, J., Asherson, P., 2014. Everyday emotional experience of adults with attention deficit hyperactivity disorder: evidence for reactive and endogenous emotional lability. Psychol. Med. 44, 3571–3583. Smallwood, J., 2013. Distinguishing how from why the mind wanders: a processoccurrence framework for self-generated mental activity. Psychol. Bull. 139, 519. Smallwood, J., Davies, J.B., Heim, D., Finnigan, F., Sudberry, M., O’Connor, R., Obonsawin, M., 2004. Subjective experience and the attentional lapse: task engagement and disengagement during sustained attention. Conscious. Cogn. 13, 657–690. Smallwood, J., O’Connor, R.C., 2011. Imprisoned by the past: unhappy moods lead to a retrospective bias to mind wandering. Cogn. Emot. 25, 1481–1490. Smallwood, J., O’Connor, R.C., Sudbery, M.V., Obonsawin, M., 2007. Mind-wandering and dysphoria. Cogn. Emot. 21, 816–842. Smallwood, J., Ruby, F.J., Singer, T., 2013. Letting go of the present: mind-wandering is associated with reduced delay discounting. Conscious. Cogn. 22, 1–7. Smallwood, J., Schooler, J.W., Turk, D.J., Cunningham, S.J., Burns, P., Macrae, C.N., 2011. Self-reflection and the temporal focus of the wandering mind. Conscious. Cogn. 20, 1120–1126. Stawarczyk, D., Majerus, S., D’Argembeau, A., 2013. Concern-induced negative affect is associated with the occurrence and content of mind-wandering. Conscious. Cogn. 22, 442–448. Trull, T.J., Solhan, M.B., Tragesser, S.L., Jahng, S., Wood, P.K., Piasecki, T.M., Watson, D., 2008. Affective instability: measuring a core feature of borderline personality disorder with ecological momentary assessment. J. Abnorm Psychol. 117, 647–661. Van Dam, N.T., Sheppard, S.C., Forsyth, J.P., Earleywine, M., 2011. Self-compassion is a better predictor than mindfulness of symptom severity and quality of life in mixed anxiety and depression. J. Anxiety Disord. 25, 123–130. Watkins, E.R., 2008. Constructive and unconstructive repetitive thought. Psychol. Bull. 134, 163–206. Watson, D., Clark, L.A., Carey, G., 1988. Positive and negative affectivity and their relation to anxiety and depressive disorders. J. Abnorm. Psychol. 97, 346–353. Whitmer, A.J., Gotlib, I.H., 2013. An attentional scope model of rumination. Psychol. Bull. 139, 1036–1061. Xu, R., 2003. Measuring explained variation in linear mixed effects models. Stat. Med. 22, 3527–3541.