Recent work has highlighted that the generation of thoughts unrelated to the current environment may be both a cause and a consequence of unhappiness. The current study used lag analysis to examine whether the relationship between self-generated thought and negative affect depends on the content of the thoughts themselves. We found that the emotional content could strongly predict subsequent mood (e.g. negative thoughts were associated with subsequent negative mood). However, this direct relationship was modulated by the socio-temporal content of the thoughts: thoughts that were past- and other-related were associated with subsequent negative mood, even if current thought content was positive. By contrast, future- and self-related thoughts preceded improvements of mood, even when current thought content was negative. These results highlight the important link between self-generated thought and mood and suggest that the socio-temporal content plays an important role in determining whether an individual's future affective state will be happy or sad.
Citation: Ruby FJM, Smallwood J, Engen H, Singer T (2013) How Self-Generated Thought Shapes Mood—The Relation between Mind-Wandering and Mood Depends on the Socio-Temporal Content of Thoughts. PLoS ONE 8(10): e77554. doi:10.1371/journal.pone.0077554
Editor: Luis M. Martinez, CSIC-Univ Miguel Hernandez, Spain
Received: March 26, 2013; Accepted: September 3, 2013; Published: October 23, 2013
Copyright: © 2013 Ruby et al. This is an open-access article distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited.
Funding: The study was partly supported by Grant R305A110277 from the US Office of Education awarded to J.W. Schooler and J. Smallwood. The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript. No additional external funding was received for this study.
Competing interests: The authors have declared that no competing interests exist.
Thoughts and feelings do not always arise from events in the here and now. Self-generated thoughts (SGT), as reflected by experiences such as mind-wandering and day-dreaming, illustrate that our mind can produce thoughts in a stimulus-independent fashion , using previously stored information. Whether their initiation is spontaneous or voluntary, SGT are generated based on intrinsic changes that take place within the individual rather than immediate perceptual input. Studies suggest that these SGT are a core form of human cognition and occupy as much as half of waking mentation –.
It is relatively common for SGT to be focused on events that may occur in the future. A prospective bias to SGT is prominent in Europe , the USA ,  as well as in China  and Japan  and content analysis has documented that these future thoughts often involve autobiographical planning . Presumably people use SGT to take advantage of the benefits that prospection affords: they use previously-acquired knowledge to prepare for events that have not yet happened, so that their actions can be more effective if the opportunity to act ever arises –.
Consistent with the notion that SGT conveys a long-term benefit, individuals who mind-wander under non-demanding circumstances tend to delay gratification  and generate more creative solutions to problems . SGT, however, is not always beneficial and when it occurs during complex tasks such as reading, it is often associated with reduced performance (e.g. , ). Moreover, in daily life, mind-wandering has been linked to automobile accidents . Evidence of both costs and benefits therefore suggests that SGT is not a homogenous experience .
An important negative consequence of SGT emerges through its association with mood. Using experience sampling in more than 2000 participants, Killingsworth and Gilbert  observed that episodes of SGT were followed at the next sampling point (hours later or the following day) by lowered mood. Based on the temporal precedence of mind-wandering episodes, they suggested that “mind wandering […] was generally the cause […] of unhappiness”. Similarly, inducing negative mood in participants increases mind-wandering  and shifts its temporal focus from the future to the past , . In addition, the association between negative affect and past-related thoughts has been documented in individuals with depressive disorders, who excessively ruminate about past failures (e.g. , ). Together, these results suggest that SGT, especially when focused on the past, may be both the cause and the consequence of negative mood.
Based on their data, Killingsworth & Gilbert (2010) suggest that “a wandering mind is an unhappy mind”, an assumption that would be correct if all types of mind-wandering impacted on mood in a homogeneous manner. However, given that SGT can have heterogeneous consequences in other domains (i.e. both costs and benefits), we explored whether its influence on mood might also be heterogeneous. For example, past-related thought may be especially likely to be associated with low mood  while other types of thought (e.g. future-focused) may not.
To test these competing hypotheses, we measured mood and SGT in a set of participants while they performed a simple choice reaction time task (CRT). To capture potentially heterogeneous types of SGT, participants answered a series of questions regarding the content of their thoughts i.e. whether they were task-related, focused on different temporal epochs (past or future), involved different referents (self or other) and varied on their emotional tone (positive or negative). Using Principal Component Analysis (PCA), we decomposed these reports based on the patterns of co-variance across different questions, which allowed different types of thoughts to be defined. We then implemented lag analyses using linear mixed models in order to explore the relation between different types of SGT and subsequent mood.
We recruited 85 German-native speakers from the Max Planck Institute for Human Cognitive and Brain Sciences database. Three participants were excluded as they had an extremely low accuracy on the CRT task. The average age of the remaining participants was 25.5 years (range: 21–31 years) and all had normal or corrected-to-normal vision. Two individuals were left-handed, 35 were females.
The study was approved by the Ethics Commission of the Medical Faculty of the University of Leipzig under the code 360-10-13122010. All the participants gave written consent before the beginning of the experiment and were remunerated 21 Euro for their participation.
Similar versions of the CRT task have been routinely used in studies on mind-wandering (e.g. , ).The task lasted 14 min. Stimuli were presented using E-prime 2.0 , . Participants observed a sequence of black and colored digits on a computer screen. Only when a colored digit was presented, participants had to indicate whether the digit was odd or even with a button push. Black digits were presented for 1000 ms and colored ones for 2000 ms. Responses had to be made while colored digits were still presented on the screen, or else the trial was considered as missed. Stimuli were separated by a fixation cross of variable duration (2200–4400 ms). Colored and non-colored digits were presented with a ratio of approximately 1/6.
Intermittently throughout the task, participants were interrupted and instead of being presented with a digit, they were probed about the content of their thoughts. Participants were asked to answer 9 questions using a 9-point Likert scale (see ,  for previous uses of this method). Answers were made using the keyboard and question presentation was self-paced. Participants reported whether their thoughts were related or unrelated to the task, the temporal, social and emotional aspects of the thoughts and their current mood (see Text S1 for a list of questions). The number of probes and their occurrence were randomly determined (Mean number of probes: 7.10, SE = .18, range: 3–12; Mean duration between two probes: 2 min, SE = .05 min, range: 0.5–8.5 min).
Prior to the experiment, participants completed a battery of online questionnaires including the Beck Depression Inventory (BDI ) using the LimeSurvey Tool . The BDI was administered to acquire an established measure of sustained negative affect independent from the mood reports obtained during the laboratory session.
Behavioral and Subjective Data
Participants had normal accuracy and response time (RT) during the CRT task (Mean accuracy = .94, SE = .01; Mean RT = 799.5 ms, SE = 14.35). Overall, 590 probes were recorded across 83 participants. For each probe, participants took on average 41 s to answer the 9 questions (mean RT to answer one question: 4531 ms, SE = 71.5 ms). Replicating previous findings (e.g. , , ), SGT were more frequently directed towards the future than the past (paired t-test, t(82) = 5.95, p<.001; Fig. 1A). Thoughts were also more frequently self-related than other-related (t(82) = 4.32, p<.001) and rated as more positive than negative (t(82) = 9.39, p<.001), similar to Bernsten and Bohn's findings . The large number of questions that composed the experience sampling procedure may have affected participants' SGT experience. However, the standard performance levels and the presence of the perspective bias in our sample suggest that our experience sampling method was similar to previously reported methods (e.g. , , ).
A) Mean probe ratings. *** indicates t-tests with p values<.001. B) Scree Plot from the PCA, showing the Eigen values of the 7 components obtained. The first three components explained almost 70% of the overall variance in our data. C) Heatmap of the 3 Principal Components, showing the component loadings for each question. The Socio-temporal Past Other Component (ST-PO) weighted positively on Off-task, Other and Past ratings, and marginally on Future ratings. The Affect Component positively weighted on Negative ratings and negatively on Positive ratings (i.e. high scores on this factor reflect more negative SGT). Finally, the ST Future and Self Component (ST-FS) weighted positively on Self, Future and Off-task ratings.
We used lag-analyses in order to investigate the links between SGT and mood reports. For example, to investigate the link between Off-task ratings and mood, we used Off-task ratings at a given time (e.g. t0) to predict the mood ratings of the probe immediately following (e.g. t1). This analysis therefore requires that our predicted variable has to be lagged by one period (in this example, the mood ratings). This lagging technique requires that the last probe from every subject has to be discarded (e.g. if the last Off-task rating is used as a t0 probe, there is no more mood ratings to use as a t1 probe). The lag-analyses reported hereafter are therefore based on data from 507 and not 590 probes.
Linear Mixed Models.
Linear mixed models (LMMs, ) were used to perform lag analyses as they allow to estimate both fixed effects (effects that one is interested in e.g. the effect of t0 Off-task ratings on t1 Mood) and random effects (effects that arise due to groups within the data e.g. multiple sampling within a subject). All predictor variables (but not dependent variables) were z-transformed prior to performing lag analyses. All LMMs included one random effect (the intercept for each Subject) to control for the dependency arising due to repeated sampling of data within subjects. Time of t1 probe onset was included as a fixed effect to account for the fact that mood is likely to decrease as the task continues. Finally, for LMMs predicting t1 Mood, t0 Mood was also included in the analysis as a fixed effect in order to control for the auto-correlation between t0 and t1 Mood. Data were analyzed and plotted using R.15.2, lme4, languageR and ggplot2 packages –. Because of the large size of our data set and because we only estimate a single random effect the significance of the fixed effects can be estimated using Markov chain Monte Carlo methods, via the pvals.fnc function provided in languageR package . Before plotting a dependent variable as a function of a single predictor (e.g. Fig. 2 and Fig. 3), it was adjusted for all the other predictors specified in the model. Median splits were used to visualize results when significant interactions were obtained.
A ) Predicting t1 Mood from t0 Off-task ratings. When t0 Mood was high (right panel), t0 off-task focus was linked to a decrease in t1 Mood. Data was plotted following a median split on t0 Mood. B) Predicting t1 Off-task ratings from t0 Mood. Negative mood at t0 was linked to an increase in t1 Off-task, but only when t0 Off-task ratings were low (left panel). Data was plotted following a median split on t0 Off-task. Thick black lines represent best-fitting linear regressions and gray ribbons represent 95% confidence intervals.
A ) Main effects. All three thought components at t0 significantly predicted t1 Mood. Increase in t0 ST-PO was linked to a decrease in t1 Mood (left). Increase in t0 positive thoughts was linked to an increase in t1 Mood (middle). Please note that for purpose of display, the Affect Component was reversed so that higher values indicate more positive thoughts. Finally, an increase in t0 ST-FS was linked to an increase in t1 Mood (right). B), Interaction between ST-PO and Affect Content. Especially when thoughts were positive, t0 ST-PO was associated with a decrease in t1 Mood. Data was plotted following a median split on t0 Affect Component. C). Interaction between ST-FS and Affect Content. Especially when thought content was negative, t0 ST-FS was associated with an increase in t1 Mood. Thick black lines represent best-fitting linear regressions and gray ribbons represent 95% confidence intervals.
Relation between off-task thinking and mood
First, we aimed to replicate the findings that SGT can both precede and follow unhappy mood , . We conducted a lag-analysis predicting t1 Mood from t0 Off-task ratings, including t0 Mood in the analysis. This revealed a significant interaction between t0 Off-task and t0 Mood (See Table 1 for a full description of the results). As seen in Fig. 2A, there was a negative relation between t0 Off-task and t1 Mood especially when t0 Mood was positive (right panel). This suggested that off-task thinking preceded decreases in t1 Mood, especially when t0 Mood was positive. A second lag-analysis (predicting t1 Off-task from t0 Mood) also revealed a negative relation between t0 Mood and t1 Off-task ratings, especially when t0 Off-task ratings were low (Table 2 and Fig. 2B, left panel). This suggests that low t0 Mood was associated with high t1 Off-task ratings, especially when t0 Off-task ratings were low.
These results broadly replicate previous studies: i) SGT followed negative mood – and ii) SGT preceded negative mood, although only when previous mood was positive . We suspect that t0 Off-task did not significantly predict t1 Mood when t0 Mood was negative (Fig. 2A, left panel) because of the auto-correlation of mood between t0 and t1 (r = 0.70, p<.001). In other words, only in the case where t0 Mood is positive can the negative effect of t0 Off-task be dissociated from the autoregressive properties of mood.
Role of thought content in the relation between mind-wandering and mood
Next, we explored the role of content in the relationship between negative mood and SGT. PCA with Varimax rotation was used to decompose the 9 probe questions based on the co-variance within our data. Three Principal components were obtained, explaining nearly 70% of the total variance (Fig. 1B and 1C). We obtained two socio-temporal (ST) factors and a single affective factor: 1) A Past and Other Component (ST-PO), weighting strongly on past, other and Off-task ratings. 2) An Affect Component, weighting strongly on the negative and positive dimensions (because the Affect component positively weights on negative ratings, high scores imply negative SGT). 3) A Future and Self Component (ST-FS), weighting on future, self and Off-task ratings.
Lag analysis was used to examine whether the socio-temporal or the emotional content of SGT was linked to subsequent mood. In this analysis, t1 Mood was our dependent variable and the three t0 PCA factors were independent variables. The analysis revealed five significant effects (Table 3). i) An effect of the Affect Component, suggesting that positive thought was associated with positive mood at the next probe (Fig. 3A, middle); ii) An effect of ST-PO, indicating that this type of socio-temporal content was associated with a subsequent reduction in mood (Fig. 3A, left); iii) An interaction between ST-PO and Affect Component (Fig. 3B), indicating a negative relation between ST-PO and mood even if the current content of thought was positive; iv) A main effect of ST-FS, indicating that this type of socio-temporal thought was linked to an improvement of subsequent mood (Fig. 3A, right); v), An interaction between the ST-FS and Affect Components, indicating that the positive relation between mood and future and self thinking was more pronounced when the current content of thought was negative (Fig. 3C). As in the analysis with Off-task ratings, the strong relation between t0 Affect component and subsequent mood is likely to partially mask the effects of ST components. Overall, these results suggest that the relation between SGT and mood depends on its content.
Controlling for co-linearity between Emotional Content and Mood.
Mood and Affect Component were highly correlated, both at t0 and between t0 and t1 lags (Table 4). To understand whether this co-linearity might have been responsible for the heterogeneous effects of SGT types, we calculated the difference in Mood between t0 and t1 lags (MoodDiff). Using a similar lag analysis as before (see Table S1) we found comparable results: ST-PO accompanied by positive content was associated with an increase in negative mood (Fig. S1A), while ST-FS was linked to increases in mood, especially when participants experienced negative thoughts (Fig. S1B). In this analysis, the correlation between t0 Affect component and MoodDiff was substantially reduced, suggesting that co-linearity in our initial model cannot be responsible for the heterogeneous effects of thought types on mood.
Link to BDI Score.
Finally, we investigated whether the content of SGT could predict the BDI. Replicating previous studies indicating that past-related thought is a characteristic of dysphoria , we found that past- other-related thoughts as well as negative thoughts were associated with higher BDI scores (Univariate ANOVA predicting logged BDI score from PCA Components; main effects, ST-PO: F = 4.30, p = .04; Affect Component: F = 3.62, p = .06). By contrast, there was no effect of ST-FS Component (F = 0.27, p>.6). This demonstrates that an established measure of chronic low mood is accompanied by SGT directed towards the past rather than the future.
Our study examined whether the relation between SGT and negative mood depends on the content of thought. Using experience sampling during a short undemanding task, we replicated the findings that SGT can both precede and follow negative mood , –. However, we found that this relationship depends on the content of thought. Past- other-related thoughts were linked to decreases in mood, even if SGT content was positive. By contrast, SGT focused on the future and self was linked to increases in positive mood, even if the content of thought was negative. As the most pronounced impact of a future focus on mood was found when thoughts were negative, our data underline that unhappiness and SGT are inextricably linked. However, the opposing effects of ST-PO and ST-FS components demonstrate that the occurrence of certain kinds of SGT may constrain rather than prolong negative mood.
However, the correlational design of our study does not allow us to conclude that certain types of SGT caused subsequent decrease or increase in mood. In order to establish a causal link between SGT and mood, future research will have to provide specific models that explain the possible mechanisms that may allow SGT to directly influence mood. For example, it has recently been proposed that it is necessary to distinguish between influences that determine the occurrence of SGT and those that control the continuity of the experience once initiated . In this context, because our results indicate a link between past-other-related SGT and negative mood, it raises the possibility that the reprocessing of past events may initiate a cyclical process that underlies the link between SGT and unhappiness. In this regard our study remains informative as it reveals that only certain but not all types of SGT have a negative relationship to mood, an observation which provides a boundary condition on more specified models in the future.
Although we replicated the findings that SGT can precede negative mood, this was only the case when t0 Mood was positive. As mentioned earlier, this may be caused by a ceiling effect (the negative effect of SGT may only be observed when t0 Mood is not too low) but it may also be caused by a regression to the mean i.e. high t0 Mood may be linked with high t0 positive Off-task thoughts. However at t1, the Off-task thoughts may no longer be as positive, leading to a decrease in mood back to average levels. Even though this effect may explain the link between Off-task and Mood, regression to the mean cannot account for the opposing effect of ST-PO and ST-FS as the emotional content was included in the corresponding lag analyses. For example, positive emotional content at t0, in the absence of ST-PO, did not lead to less positive mood.
The conditions under which SGT were recorded may also influence our conclusions. Following studies showing that SGT are most common during non-demanding situations (e.g. , ), we measured SGT during an easy CRT task which requires limited cognitive resources to be performed properly. The task therefore puts individuals into a non-demanding situation and allows them to frequently generate SGT if desired. In addition, based on studies showing similarities between SGT under laboratory conditions and in everyday life (e.g. ), we suppose that our results are likely to generalize to daily life especially under circumstances where current perceptual input is not especially salient (e.g. while being stuck in a traffic jam).
In addition to demonstrating the heterogeneity of the relation between SGT and mood, our study also demonstrates that SGT can be informatively characterized according to the covariation between distinct constituents of the content. Our application of PCA revealed a statistical overlap between self-related and future-related thoughts that corroborates prior work. For example, a brief period of self-reflection caused an increase in future-related thoughts . In another study, ratings of open-ended reports of SGT revealed that future-related thought were also highly self-related . The PCA also grouped together the past and other constituents into one component which was associated with negative mood. This statistical characterization of SGT is consistent with prior work  linking past-related thought to negative affect both at a transient level (i.e. following mood induction) and more sustained level (i.e. BDI). The fact that PCA revealed statistical categories of SGT whose psychological properties mimic the results of experimental manipulations provides independent support for the existence of psychologically distinct types of SGT.
Finally, the current concerns hypothesis ,  provides a valuable perspective on the observation that past and future-related SGT have opposing links to mood. According to this framework, mind-wandering often arises because unfulfilled goals or ambitions have greater salience than current environmental inputs (see also ). If individuals simply simulated a current concern, without attempting to generate possible solutions, this could prolong the influence that the unfulfilled goal has on mood. By contrast, future-related thoughts allow individuals to create plans ,  and so could provide the individual with mental strategies or heuristics that could then be used to resolve these issues in the future (e.g. ). In this way self and future-related thoughts may reduce the negative influence that current concerns have on mood. Although the capacity to limit mind-wandering may be the best way to improve happiness in the long run, the strong tendency of the mind to spontaneously generate thoughts may hinder this possibility. Our data suggests that, if mentally setting aside a problem is not an option, moving forward by adopting a future focus may be the next best strategy.
Effect of thought content on mood change. A) t0 ST-PO was associated with a negative change of mood, especially when t0 thought content was positive. B) t0 ST-FS was linked to a positive change of mood, especially when t0 thought content was negative.
LMM predicting mood change from previous thought content. Fixed effects estimates for the linear mixed model predicting MoodDiff (i.e. the difference between t1 Mood and t0 Mood) from t0 content. We included the Subject as a random effect.
Thought sampling questions. The following 9 questions were presented during the thought sampling procedure. Participants responded using a 9-point Likert scale (1: not at all, 9: completely). For the off-task question, 1 indicated “I was thinking exclusively about the task” and 9 indicated “I wasn’t thinking at all about the task”.
We thank Daniel Schad for valuable advice regarding R and LMMs, and Johannes Golchert and Felix Weirich for help with data collection.
Conceived and designed the experiments: FJMR HE JS. Performed the experiments: FJMR HE. Analyzed the data: FJMR JS. Wrote the paper: FJMR HE JS TS.
- 1. Smallwood J (2013) Distinguishing how from why the mind wanders: A process-occurrence framework for self-generated mental activity. Psychol Bull 139: 519–535. doi: 10.1037/a0030010
- 2. Smallwood J, Nind L, O′Connor RC (2009) When is your head at? An exploration of the factors associated with the temporal focus of the wandering mind. Conscious Cogn 18: 118–125. doi: 10.1016/j.concog.2008.11.004
- 3. Baird B, Smallwood J (2011) Schooler JW (2011) Back to the future: autobiographical planning and the functionality of mind-wandering. Conscious Cogn 20: 1604–1611. doi: 10.1016/j.concog.2011.08.007
- 4. Killingsworth MA, Gilbert DT (2010) A wandering mind is an unhappy mind. Science 330: 932. doi: 10.1126/science.1192439
- 5. Andrews-Hanna JR, Reidler JS, Huang C, Buckner RL (2010) Evidence for the default network's role in spontaneous cognition. J Neurophysiol 104: 322–335. doi: 10.1152/jn.00830.2009
- 6. Song X, Wang X (2012) Mind wandering in Chinese daily lives--an experience sampling study. PLoS One 7: e44423. doi: 10.1371/journal.pone.0044423
- 7. Iijima Y, Tanno Y (2012) [The effect of cognitive load on the temporal focus of mind wandering]. Shinrigaku Kenkyu 83: 232–236. doi: 10.4992/jjpsy.83.232
- 8. Tulving E (2002) Episodic memory: From mind to brain. Annual Review of Psychology 53: 1–25. doi: 10.1146/annurev.psych.53.100901.135114
- 9. Schacter DL, Addis DR, Hassabis D, Martin VC, Spreng RN, et al. (2012) The future of memory: remembering, imagining, and the brain. Neuron 76: 677–694. doi: 10.1016/j.neuron.2012.11.001
- 10. Gollwitzer PM (1999) Implementation intentions: strong effects of simple plans. American Psychologist 54: 493. doi: 10.1037//0003-066x.54.7.493
- 11. Smallwood J, Ruby FJM, Singer T (2013) Letting go of the present: Mind-wandering is associated with reduced delay discounting. Consciousness and Cognition 22: 1–7. doi: 10.1016/j.concog.2012.10.007
- 12. Baird B, Smallwood J, Mrazek MD, Kam JW, Franklin MS, et al. (2012) Inspired by distraction: mind wandering facilitates creative incubation. Psychol Sci 23: 1117–1122. doi: 10.1177/0956797612446024
- 13. McVay JC, Kane MJ (2012) Why Does Working Memory Capacity Predict Variation in Reading Comprehension? On the Influence of Mind Wandering and Executive Attention. Journal of Experimental Psychology-General 141: 302–320.
- 14. Smallwood J, McSpadden M (2008) Schooler JW (2008) When attention matters: The curious incident of the wandering mind. Memory & Cognition 36: 1144–1150. doi: 10.3758/mc.36.6.1144
- 15. Lagarde E, Gabaude C, Maury B, Lemercier C, Salmi L-R, et al. (2012) Mind wandering and driving. Injury Prevention 18: A200. doi: 10.1136/injuryprev-2012-040590t.6
- 16. Smallwood J, Fitzgerald A, Miles LK, Phillips LH (2009) Shifting moods, wandering minds: negative moods lead the mind to wander. Emotion 9: 271–276. doi: 10.1037/a0014855
- 17. Smallwood J, O′Connor RC (2011) Imprisoned by the past: unhappy moods lead to a retrospective bias to mind wandering. Cogn Emot 25: 1481–1490. doi: 10.1080/02699931.2010.545263
- 18. StawarczykD, MajerusS, D′Argembeau A (in press) Concern-induced negative affect is associated with the occurrence and content of mind-wandering. Conscious Cogn.
- 19. Watkins E, Teasdale JD (2001) Rumination and overgeneral memory in depression: Effects of self-focus and analytic thinking. Journal of Abnormal Psychology 110: 353–357. doi: 10.1037/0021-843x.110.2.333
- 20. Nolen-Hoeksema S, Wisco BE, Lyubomirsky S (2008) Rethinking Rumination. Perspectives on Psychological Science 3: 400–424. doi: 10.1111/j.1745-6924.2008.00088.x
- 21. Schneider W, Eschman A, Zuccolotto A (2002) E-Prime User's Guide. Pittsburgh: Psychology Software Tools Inc
- 22. Schneider W, Eschman A, Zuccolotto A (2002) E-Prime Reference Guide. Pittsburgh: Psychology Software Tools Inc.
- 23. Christoff K, Gordon AM, Smallwood J, Smith R (2009) Schooler JW (2009) Experience sampling during fMRI reveals default network and executive system contributions to mind wandering. Proceedings of the National Academy of Sciences of the United States of America 106: 8719–8724. doi: 10.1073/pnas.0900234106
- 24. Mrazek MD, Smallwood J, Franklin MS, Chin JM, Baird B, et al. (2012) The role of mind-wandering in measurements of general aptitude. J Exp Psychol Gen 141: 788–798.
- 25. Beck AT, Steer RA, Brown GK (1996) BDI-II, Beck depression inventory: manual. San Antonio, Tex. Boston: Psychological Corp.; Harcourt Brace. vi, 38 p. p.
- 26. LimeSurvey Project Team, Schmitz C (2012) LimeSurvey: An Open Source survey tool,. Hamburg, Germany: LimeSurvey Project
- 27. Berntsen D, Bohn A (2010) Remembering and forecasting: The relation between autobiographical memory and episodic future thinking. Mem Cognit 38: 265–278. doi: 10.3758/mc.38.3.265
- 28. Pinheiro JC, Bates DM (2000) Mixed-effects models in S and S-PLUS: Springer Verlag.
- 29. Baayen RH (2011) languageR: Data sets and functions with "Analyzing Linguistic Data: A practical introduction to statistics".
- 30. Bates D, Maechler M, Bolker B (2012) lme4: Linear mixed-effects models using S4 classes.
- 31. R Core Team (2012) R: A Language and Environment for Statistical Computing. Vienne, Autria: R Foundation for Statistical Computing.
- 32. Wickham H (2009) Ggplot2: elegant graphics for data analysis. New York: Springer. viii, 212 p. p.
- 33. Smallwood J (2006) Schooler JW (2006) The restless mind. Psychol Bull 132: 946–958. doi: 10.1037/0033-2909.132.6.946
- 34. McVay JC, Kane MJ, Kwapil TR (2009) Tracking the train of thought from the laboratory into everyday life: an experience-sampling study of mind wandering across controlled and ecological contexts. Psychon Bull Rev 16: 857–863. doi: 10.3758/pbr.16.5.857
- 35. Smallwood J, Schooler JW, Turk DJ, Cunningham SJ, Burns P, et al. (2011) Self-reflection and the temporal focus of the wandering mind. Conscious Cogn 20: 1120–1126. doi: 10.1016/j.concog.2010.12.017
- 36. Klinger E (1978) Dimensions of Thought and Imagery in Normal Waking States. Journal of Altered States of Consciousness 4: 97–113.
- 37. Klinger E (1999) Thought flow: Properties and mechanisms underlying shifts in content. In: Salovey JASP, editor. At play in the fields of consciousness: Essays in honor of Jerome L Singer. Mahwah, NJ, US: Lawrence Erlbaum Associates Publishers. pp. 29–50.