Default-Mode and Task-Positive Network Activity in
Major Depressive Disorder: Implications for Adaptive
and Maladaptive Rumination
J. Paul Hamilton, Daniella J. Furman, Catie Chang, Moriah E. Thomason, Emily Dennis, and Ian H. Gotlib
Background: Major depressive disorder (MDD) has been associated reliably with ruminative responding; this kind of responding is
network (TPN), as well as activity in structures that influence DMN and TPN functioning, may represent important neural substrates of
maladaptive and adaptive rumination in MDD.
eyes-closed rest in 17 currently depressed and 17 never-disordered adults. We calculated correlations between this metric of DMN
between these measures both with one another and with severity of depression. Finally, we estimated and compared across groups right
fronto-insular cortex (RFIC) response during initiations of ascent in DMN and in TPN activity.
Results: In the MDD participants, increasing levels of DMN dominance were associated with higher levels of maladaptive, depressive
rumination and lower levels of adaptive, reflective rumination. Moreover, our RFIC state-change analysis showed increased RFIC activation
at the onset of increases in DMN activity.
depression, and the RFIC, when prompted by increased levels of DMN activity, initiates an adaptive engagement of the TPN.
tex, rumination, self-reflection, task-positive network
sequences (1,2). A ruminative response style has been found to
uals (3), perhaps because of disrupted allocation of cognitive re-
sources and increased recall and rehearsal of negative life events
maladaptive in MDD, recent conceptualizations suggest that de-
pressive rumination is a multidimensional construct with both
adaptive and maladaptive components (5,6). Investigators using
Responses Scale (RRS) (7), a frequently used self-report measure of
rumination, have identified three types of items in this measure:
depression-related items (RRS-D, e.g., “How often do you think
about how you don’t seem to feel anything anymore?”); items as-
sociated with brooding (RRS-B, e.g., “How often do you think, ‘Why
do I have problems other people don’t have?’”); and items associ-
what you are thinking and analyze it?”) (8). Treynor et al. (8, p. 256)
note that whereas the cognitions represented in the RRS-D and
RRS-B subscales are “a passive comparison of one’s current situa-
tion with some unachieved standard,” items from the RRS-R sub-
defined as a recurrent, self-reflective, and unintentional fo-
cus on depressive symptomatology and its causes and con-
to engage in cognitive problem solving to alleviate one’s depres-
sive symptoms.” Consistent with these interpretations, while de-
pressed persons generally endorse more items from all three RRS
subscales than do nondepressed control subjects, scores on the
associated with lower levels of depressive symptoms at follow-up
(8), whereas high scores on the RRS-B (but not on the RRS-R) sub-
scale have been found to be associated with a maladaptive atten-
tional bias to negative stimuli in MDD (9).
Although the neural substrates of adaptive versus maladaptive
rumination in depression have not been examined, recent work
gests an intriguing neural substrate of ruminative responding in
MDD. Analyses of resting state and task paradigm blood oxygen-
tional organization in the brain composed of two spatially distinct
the task-positive network (TPN) (10,11). During performance of at-
tention-demanding tasks, prefrontal and parietal structures com-
trast, DMN structures, including posterior cingulate and medial
prefrontal cortices, are characterized by decreased activity. During
wakeful rest, the opposite pattern emerges, with the DMN becom-
ing more active and the TPN less active (12).
adaptive rumination in MDD, the DMN has been proposed to un-
dergird passive, self-relational processing (e.g., autobiographical
recall, prospection (13), whereas the TPN has been postulated to
tion, and working memory) (11). Given the evidence cited above
that ruminative responding in MDD may involve passive and mal-
adaptive as well as active and adaptive processes, examining the
relation of DMN versus TPN functioning with ruminative respond-
(CC), Stanford University, Stanford, California.
Address correspondence to J. Paul Hamilton, Ph.D., Stanford University,
Department of Psychology, 450 Serra Mall, Jordan Hall, Building 420,
Stanford, CA 94305; E-mail: email@example.com.
Received Aug 26, 2010; revised Feb 3, 2011; accepted Feb 3, 2011.
BIOL PSYCHIATRY 2011;70:327–333
© 2011 Society of Biological Psychiatry
ing in MDD may help to advance neural theory of this disorder.
Indeed, a body of research documenting aberrant responding of
components of the DMN (14-16) and of the TPN (17,18) in MDD
underscores the importance of examining the interaction of these
two systems in this disorder.
for several reasons. First, recent work implicates this structure in
switching between states of relative dominance of the DMN and
TPN (19). Moreover, this neural structure has been posited to be
involved in awareness of emotion (20) and, more specifically, in
interoceptive error detection, that is, in signaling a discrepancy
between actual and desired somatic states (21). Further, increased
the pathophysiology of this disorder is not known. To the extent
that states of relative TPN and DMN dominance represent desired
ing during switching between TPN and DMN dominance should
advance our understanding of the role of anomalous insula activa-
tion in MDD.
In the present study, we computed relative levels of DMN and
ined the associations of DMN versus TPN activation (henceforth
and adaptive rumination. Because our metric of DMN dominance,
presented below, indexes levels of passive, self-relational thinking
relative to effortful cognition, we hypothesized that depressed in-
dividuals would show increased DMN dominance and that in-
creased DMN dominance in MDD would be associated with
adaptive rumination. In addition, we measured activation in the
depressed and nondepressed participants. We hypothesized that
depressed persons would recruit the RFIC to a greater extent than
would never-disordered individuals at the initiation of states of
relative TPN dominance over DMN.
Methods and Materials
Seventeen adults diagnosed with MDD and 17 control (CTL)
ticipated in this study. All depressed participants met criteria for a
DSM-IV diagnosis of MDD based on their responses to the Struc-
tured Clinical Interview for DSM-III-R Personality Disorders (24) as
administered by trained diagnostic staff; none of the CTL partici-
pants met diagnostic criteria for any current or past Axis I disorder.
any Axis I disorder, with the exception of social anxiety disorder,
not current, generalized anxiety disorder were included in the
study. Participants completed the Beck Depression Inventory-II
(BDI-II) (25), the Hamilton Depression Rating Scale (HAM-D) (26),
validated self-report measures of the severity of depressive symp-
consequences. After a complete description of the study was pre-
parameters were the same as those from a previous study (27)
over 180 temporal frames (NFRAMES) using a repetition time of
2000 msec/frame. Further, 11 of 17 depressed persons and 2 of 17
current study. We present fMRI data preprocessing procedures in
Section 1 of Supplement 1.
Identifying and Comparing Between Groups the DMN and
procedure adapted from Fox et al. (11). For details, please see Sec-
for each participant: one of the DMN and the other of the TPN. To
verify the effectiveness of the procedure we used to identify the
DMN and TPN, we used these binary masks to create voxel-wise
frequency maps depicting for each group the number of subjects
for which each voxel belonged to the DMN or the TPN. We then
conducted voxel-wise, between-group, chi-square analyses (p ?
.05, corrected) on these masks to examine regions in which the
MDD and CTL groups differed with respect to the spatial extent of
DMN or TPN maps, so that we could exclude these regions from
Operationalizing DMN Dominance over TPN. To compute
the extent to which levels of DMN activity exceeded TPN activity
over the course of the resting scan, we first extracted from each
covariate corrected time-series data (see Section 2, Step 1 in Sup-
plement 1 for details regarding our noise covariate correction pro-
cedure), excluding those regions in which depressed and control
participants differed with respect to the spatial extent of DMN or
TPN.1We then constructed an NFRAMES-long vector that was as-
signed a value of 1 for temporal frames for which DMN BOLD was
greater than TPN BOLD and a value of ?1 for temporal frames for
which TPN BOLD was greater than DMN BOLD. We summed this
vector to produce an index of DMN dominance over TPN. This
dominance measure between groups (p ? .05, one-tailed, given a
priori prediction of increased DMN dominance in MDD relative to
of comparing DMN and TPN activity using brain-blood perfusion
a per-subject basis the parts of the TPN and DMN that were most
strongly anticorrelated with each other (i.e., that were in the great-
est competition with each other). Given the respective roles of the
DMN and TPN in self-reflection and effortful cognition, identifying
and comparing the parts of these networks at greatest apparent
odds would seem to have the greatest bearing on understanding
opposing ruminative processes indexed by the RRS-D and RRS-R.
tions supported by these networks. Supporting this assumption,
other studies have found BOLD signal duration to be associated
1As an additional precaution, we calculated the total number of voxels
comprising, and the center of mass of, the TPN and DMN for each
participant and examined group differences in these indexes. The MDD
and CTL groups did not differ in the x, y, and z extents of the TPN and
DMN (all p ? .10). The two groups also did not differ with respect to the
did the size of the TPN or the DMN correlate with measures of TPN
dominance or of rumination (all p ? .10).
328 BIOL PSYCHIATRY 2011;70:327–333
J.P. Hamilton et al.
guish groups at high risk for depression from groups at low risk for
Correlating DMN Dominance over TPN with Rumination. We
DMN dominance and rumination in depressed and never-disor-
dered persons. Specifically, we determined from the pattern of
correlations among the measure of DMN dominance, the three
to account for to determine the associations between unique as-
pects of the RRS subscales and DMN dominance in each group.
r ? .6; p ? .05) but not in the depressed group (all interscale r ? .4;
p ? .05). Consequently, in the control group, but not in the de-
pressed group, we conducted analyses on an aggregate RRS index
between the RRS-D and BDI-II, a marginally significant correlation
between the RRS-D and RRS-B, and a significant correlation be-
in MDD, we first regressed out associations of the RRS-D with the
the BDI-II. Third, in calculating the correlation between unique fea-
tures of the RRS-B and DMN dominance in MDD, we factored out
BDI-II, as well as the relation between DMN dominance and the
between the RRS aggregate score and BDI-II in the control group
score before correlating the RRS with DMN dominance in control
of outlier effects by subjecting significant correlations between
appropriately residualized variables to a procedure in which indi-
vidual cases were iteratively excluded from the correlation calcula-
tion; a given correlation was considered significant only if it re-
mained significant at a noncorrected threshold when individual
cases were excluded from the calculation. Finally, to keep the pos-
sibility of family-wise type I error at p ? .05, we used the Holm-
Bonferroni correction (30) to adjust the significance threshold for
the four correlation calculations (DMN dominance with RRS-D,
Figure 1. (A) Depiction using actual data of procedure for calculating default-mode network (DMN) dominance over task-positive network (TPN). Examples
of onset vectors (red) for DMN (B) and TPN (C) in the context of TPN (green) and DMN (blue) time-series data.
Table 1. Group-Wise Correlation Matrix of Neural and Behavioral
TNP DomRRS-D RRS-BRRS-R BDI-II
BDI-II, Beck Depression Inventory-II; Dom, dominance; RRS-B, Rumina-
sion; RRS-R, Ruminative Responses Scale-Reflection; TNP, task-positive
aSignificant (p ? .05, one-tailed) effects.
bMarginally significant (.10 ? p ? .05, one-tailed) effects.
cMajor depressive disorder.
J.P. Hamilton et al.
BIOL PSYCHIATRY 2011;70:327–333 329
RRS in the control group).
State-Change Analysis of Right Fronto-Insular Cortex. We
activity and at initiations of ascent in TPN activity. To do this, we
constructed delta-function vectors for each participant corre-
sponding to DMN and to TPN onset and regressed these vectors
against preprocessed time-series data from voxels within the RFIC.
The DMN onset vector was a vector of length NFRAMES that was
assigned a value of 1 for temporal frames at which there was a
descent); we made this correspondence a criterion to ensure that
ascents in the DMN time series were meaningful in terms of their
implications for the DMN-TPN system. The DMN onset vector was
and DMN descent and a value of 0 otherwise. Detection of troughs
and peaks in the DMN and TPN time series was performed with a
nonderivative-based algorithm (http://billauer.co.il/peakdet.html;
National Institutes of Health, Bethesda, Maryland) implemented in
MATLAB (http://www.mathworks.com; Mathworks, Natick, Massa-
chusetts). Examples of DMN and TPN onset vectors are shown in
Figures 1B and 1C, respectively. These onset vectors were con-
response and entered into a voxel-wise regression against prepro-
cessed voxel time-series data from the RFIC. The RFIC region of
? 0 and the part of the right inferior frontal gyrus bounded by the
box described by 27 ? x ? 48, 0 ? y ? 28, and ?19 ? z ? 15. This
regression for identifying the DMN and the TPN. To address in our
regression the possibility that the convolved TPN onset function
simply aliased the TPN-averaged time series—which could be the
case if TPN fluctuations were of the same duration as the hemody-
namic response—we also included in the regression the TPN-aver-
necessary to include the DMN-averaged time series in this regres-
sion because of its high collinearity in all participants with the
TPN-averaged time series. The resulting fit coefficients from this
variance with one between-subjects factor (group: MDD, CTL) and
one within-subject factor (network: DMN onset, TPN onset). We
Demographic and Clinical Variables
Demographic and clinical characteristics of the depressed and
nondepressed participants are presented in Table 2; case-by-case
demographic and clinical data for participants in the MDD group
are presented in Table 3. The two groups of participants did
not differ significantly in age, t(32) ? .84, or gender composition,
?2(32) ? .0, both p ? .10. As expected, the depressed participants
had higher scores on the BDI-II, HAM-D, RRS-D, RRS-B, and RRS-R
Table 2. Participant Demographic and Clinical Characteristics
Hamilton Depression Rating Scalea
Back Depression Inventory-II
RRS, Ruminative Responses Scale.
ap ? .05; Mean and SE (standard error of the mean) reported where
Table 3. Demographic and Clinical Characteristics of Depressed Sample
Duration (Months) of
Note: Neither dividing the depressed sample according to the presence of a concurrent diagnosis of SAD
nor dividing the sample according to the presence of a concurrent or past diagnosis of any anxiety disorder
yielded a significant effect of anxiety or any of the neural variables measured in this study: ps ? .10.
BDI-II, Back Depression Inventory-II; F, female; HAM-D, Hamilton Depression Rating Scale; M, male; SAD,
social anxiety disorder; U, undetermined or too many to correctly recall.
330 BIOL PSYCHIATRY 2011;70:327–333
J.P. Hamilton et al.
7.01, and 4.31, respectively, all p ? .05.
Spatial Extent of DMN and TPN in the Depressed and
Maps summarizing the spatial extent of DMN and TPN in the
participants did not differ with respect to the spatial extent of the
extent of the right fronto-insular cortex (center of mass ? 32, 11,
?5; k ? 54 voxels) (Figure 2B).
DMN Dominance over TPN and Its Association with
Major depressive disorder and CTL participants did not differ
with respect to dominance of DMN over TPN, t(32) ? 1.49, p ? .10.
of the RRS with our measure of DMN dominance indicated that
greater DMN dominance was significantly associated with higher
RRS-D scores, r(15) ? .48, p ? .026 (marginally significant, given
r(15) ? ?.65, p ? .002 (less than Holm-Bonferroni criterion of p ?
.013) (Figure 3). Importantly, both of these correlations remained
The RRS-B scores were not correlated significantly with level of
group, the residualized RRS measure was not significantly corre-
lated with DMN dominance, r(15) ? .03; p ? .10.
State-Change Analysis of Right Fronto-Insular Cortex
conducted on voxels comprising the RFIC identified a region (cen-
ter of mass ? 40, 21, 6; k ? 73 voxels; no overlap was observed
between this region and the region identified in the spatial extent
analysis) that responded differentially at the onset of increases in
4A for a statistical map of this interaction). As shown in Figure 4B,
whereas during the initiation of a rise in TPN activity the RFIC
showed increased activation in MDD but not in CTL participants,
during the initiation of a rise in DMN activity this region showed
increased activation in CTL but not in MDD participants. See Table
S1 in Supplement 1 for results obtained when this same analysis
was conducted at the whole-brain level.
In the present study, we examined the relative dominance of
DMN over TPN and its association with adaptive and maladaptive
rumination in major depression. In addition, we examined RFIC
responding during initiations of ascent in the DMN and the TPN in
depressed and in never-disordered participants. We found that
levels of adaptive, reflective rumination. Further, our RFIC state-
change analysis showed that, relative to healthy control partici-
pants, depressed participants showed increased RFIC activation at
the onset of increases in TPN activity (and decreases in DMN activ-
ity); in contrast, healthy control participants exhibited increased
RFIC response at the onset of increases in DMN activity (and de-
creases in TPN activity).
composed of the DMN and TPN performs similar operations in
depressed and nondepressed persons but does so based on mark-
edly different information. It is important to note that the predic-
tion of maladaptive and adaptive rumination by individual differ-
ences in relative levels of DMN and TPN activity in depression is
TPN derived from research with nondepressed samples. For exam-
that subserves passive, self-relational processes such as recall of
Figure 2. (A) Frequency maps for default-mode network
(cool colors) and task-positive network (warm colors) de-
rived from regression-defined masks for individuals in
major depressive disorder and control groups. (B) Chi-
square statistic map showing increased frequency of in-
clusion of right fronto-insular cortex in the task-positive
trol; MDD, major depressive disorder.
Figure 3. Negative correlation of default-mode network dominance with
Ruminative Responses Scale-Reflection (top) and positive correlation of
default-mode network dominance with Ruminative Responses Scale-De-
pression (bottom) in the major depressive disorder group. DMN, default-
mode network; RRS, Ruminative Responses Scale; TR, repetition time.
J.P. Hamilton et al.
BIOL PSYCHIATRY 2011;70:327–333 331
autobiographical memories (13) and mind wandering (31)—was
associated with higher levels of less effortful, maladaptive, depres-
shortcomings, failings, faults, mistakes?”). Symmetrically, we also
found in MDD that greater dominance of TPN—a network that is
active during performance of cognitively demanding tasks that
associated with higher levels of effortful, reflective processing
(RRS-R; e.g., “How often do you analyze your personality to try to
understand why you are depressed?”).
RFIC response at the onset of increases in TPN (and decreases in
DMN) activity and at the onset of increases in DMN (and decreases
in TPN) activity: whereas depressed participants activated the RFIC
at TPN troughs (DMN peaks) but not at DMN troughs (TPN peaks),
not at TPN troughs (DMN peaks). Given that the RFIC plays a role in
switching between states of relative dominance of DMN and TPN
(19) and that its role in interoceptive awareness (20) enables it to
detect discrepancies between desired and actual somatic states
and TPN are operating on different information in depressed and
nondepressed individuals. If the RFIC monitors for the presence of
should initiate a DMN-TPN state-change call when a peak in DMN
activity occurs, potentially enacting TPN-based affect regulatory
mechanisms. Indeed, this is the pattern of results obtained in this
It is important to consider that, while our interpretation of our
RFIC state-change findings links the literature concerning the role
RFIC initiates DMN-TPN state change. Indeed, the present findings
in interoceptive awareness (20), the RFIC responds to the initiation
of TPN dominance in MDD, perhaps reflecting the salience of this
switch. The fact that we obtained the opposite pattern of RFIC
responding in healthy control subjects—the RFIC was engaged
during TPN peaks (DMN troughs)—is intriguing and may be ex-
plained by recent conceptualizations of the DMN as central to pos-
itive, creative processes in psychologically healthy persons (32).
from more typical analytical processing and engage in more cre-
ative DMN-mediated thought. Of course, it is also plausible that in
healthy control subjects increased RFIC responding serves simply
to mark the onset of DMN dominance.
study, the metric of DMN dominance over TPN, is novel and in-
volves interpreting relative BOLD signal values that can be influ-
enced by factors not related to neural activity. While the strong
positive correlation of our measure of DMN dominance with ante-
rior insula responding during initiations of ascent in DMN activity
serves as preliminary validation of our metric of DMN dominance
(see Section 3: Validation of our metric of DMN dominance, in
Supplement 1), this metric nevertheless requires more direct vali-
dation. Additional research examining the relation of our measure
that provide more direct estimates of brain metabolism (e.g., posi-
tron emission tomography) is required to strengthen the conclu-
We should also note that we used trait measures of rumination
scan about whether they were ruminating or about the content of
possible rumination. We took this approach both because rumina-
tion is a reliable phenomenon in depression and because we did
not want to interfere with either the process of rumination or with
TPN-DMN dynamics by probing participants during the resting
scan. We note further that, while our findings relating DMN domi-
nance to measures of rumination in MDD are consistent with cur-
rent conceptions of DMN and TPN function, these findings are
group interaction. (B) Impulse response functions from region in (A) as a
function of network onset and group. CTL, control; DMN, default-mode
network; MDD, major depressive disorder; TPN, task-positive network; TR,
332 BIOL PSYCHIATRY 2011;70:327–333
J.P. Hamilton et al.
nonetheless correlational and, consequently, may be mediated by Download full-text
one or more unmeasured variables.
The present study provides unique insights about the relation
between the intrinsic functional organization of the brain and
adaptive and maladaptive rumination in depression. The data pre-
sented here support a formulation in which the DMN supports
representation of negative, self-referential information in depres-
sion, and when prompted by increased levels of DMN activity, the
Future work examining the relation between DMN-TPN dynamics
tionnaires or experience sampling to measure the presence and
quality of rumination during scanning.
Preparation of this manuscript was supported by Grant MH59259
from the National Institute of Mental Health awarded to Ian H. Gotlib
All authors report no biomedical financial interests or potential
on the remediation of depressive affect. J Pers Soc Psychol 58:519–527.
the duration of depressive episodes. J Abnorm Psychol 100:569–582.
3. Kuehner C, Weber I (1999): Responses to depression in unipolar de-
theory. Psychol Med 29:1323–1333.
4. Nolan SA, Roberts JE, Gotlib IH (1998): Neuroticism and ruminative
response style as predictors of change in depressive symptomatology.
5. Trapnell PD, Campbell JD (1999): Private self-consciousness and the
6. Watkins E, Teasdale JD (2004): Adaptive and maladaptive self-focus in
7. Nolen-Hoeksema S, Morrow J, Fredrickson BL (1993): Response styles
and the duration of episodes of depressed mood. J Abnorm Psychol
ered: A psychometric analysis. Cognit Ther Res 27:247–259.
biases. Behav Ther 37:269–280.
10. Corbetta M, Kincade JM, Shulman GL (2002): Neural systems for visual
orienting and their relationships to spatial working memory. J Cogn
11. Fox MD, Snyder AZ, Vincent JL, Corbetta M, Van Essen DC, Raichle ME
GL (2001): A default mode of brain function. Proc Natl Acad Sci U S A
13. Spreng RN, Mar RA, Kim ASN (2009): The common neural basis of
autobiographical memory, prospection, navigation, theory of mind,
and the default mode: A quantitative meta-analysis. J Cogn Neurosci
(2009): The default mode network and self-referential processes in de-
15. Drevets WC, Price JL, Furey ML (2008): Brain structural and functional
abnormalities in mood disorders: Implications for neurocircuitry mod-
els of depression. Brain Struct Funct 213:93–118.
(2009): Altered negative BOLD responses in the default-mode network
during emotion processing in depressed subjects. Neuropsychophar-
17. Siegle GJ, Thompson W, Carter CS, Steinhauer SR, Thase ME (2007):
Increased amygdala and decreased dorsolateral prefrontal BOLD re-
18. Mayberg HS, Liotti M, Brannan SK, McGinnis S, Mahurin RK, Jerabek PA,
et al. (1999): Reciprocal limbic-cortical function and negative mood:
Converging PET findings in depression and normal sadness. Am J Psy-
19. Sridharan D, Levitin DJ, Menon V (2008): A critical role for the right
fronto-insular cortex in switching between central-executive and de-
21. Paulus MP, Stein MB (2006): An insular view of anxiety. Biol Psychiatry
23. Fu CHY, Williams SCR, Cleare AJ, Brammer MJ, Walsh ND, Kim J, et al.
sion by antidepressant treatment: A prospective, event-related func-
tional magnetic resonance imaging study. Arch Gen Psychiatry 61:877–
24. First MB, Spitzer RL, Gibbon M, Williams JBW (1995): The Structured
Clinical Interview for DSM-III-R Personality-Disorders (SCID-I).J Pers Dis-
sion. New York: Guilford Publications.
26. Hamilton M (1960): A rating scale for depression. J Neurol Neurosurg,
tigating neural primacy in Major Depressive Disorder: Multivariate
granger causality analysis of resting-state fMRI time-series data [pub-
lished online ahead of print May 18]. Mol Psychiatry.
28. Siegle GJ, Steinhauer SR, Thase ME, Stenger VA, Carter CS (2002): Can’t
shake that feeling: Event-related fMRI assessment of sustained
amygdala activity in response to emotional information in depressed
individuals. Biol Psychiatry 51:693–707.
amygdala activation during sad mood elaboration as a function of
5-HTTLPR [published online ahead of print April 1]. Soc Cogn Affect
30. Holm S (1979): A simple sequentially rejective multiple test procedure.
31. Christoff K, Gordon AM, Smallwood J, Smith R, Schooler JW (2009):
system contributions to mind wandering. Proc Natl Acad Sci U S A 106:
32. Christoff K, Gordon A, Smith R (2011): The role of spontaneous thought
Decision Making. London: Psychology Press.
J.P. Hamilton et al.
BIOL PSYCHIATRY 2011;70:327–333 333