You are on page 1of 9

Distinct electrophysiological signatures of

task-unrelated and dynamic thoughts


Julia W. Y. Kama,b,c,1, Zachary C. Irvingd, Caitlin Millse, Shawn Patelc, Alison Gopnikf, and Robert T. Knightc,f
a
Department of Psychology, University of Calgary, Calgary, AB T2N 1N4, Canada; bHotchkiss Brain Institute, University of Calgary, Calgary, AB T2N 4N1,
Canada; cHelen Wills Neuroscience Institute, University of California, Berkeley, CA 94720; dCorcoran Department of Philosophy, University of Virginia,
Charlottesville, VA 22904; eDepartment of Psychology, University of New Hampshire, Durham, NH 03824; and fDepartment of Psychology, University of
California, Berkeley, CA 94704

Edited by Robert Desimone, Massachusetts Institute of Technology, Cambridge, MA, and approved November 18, 2020 (received for review June 11, 2020)

Humans spend much of their lives engaging with their internal unrelated to an ongoing, typically externally oriented, task (8–10).
train of thoughts. Traditionally, research focused on whether or In the laboratory, subjects’ thoughts are often unrelated to the
not these thoughts are related to ongoing tasks, and has identified experimental task (9). Task-unrelated thought is also frequent in
reliable and distinct behavioral and neural correlates of task- everyday life (11), as when a student becomes distracted during a
unrelated and task-related thought. A recent theoretical frame- lecture or while driving.
work highlighted a different aspect of thinking—how it dynami- Recent theories are less task-centric and instead focus on the
cally moves between topics. However, the neural correlates of dynamics of mind wandering—that is, how internal trains of
such thought dynamics are unknown. The current study aimed thoughts unfold over time (12–17). In particular, the “dynamic
to determine the electrophysiological signatures of these dynam- framework” of spontaneous thought distinguishes three subtypes
ics by recording electroencephalogram (EEG) while participants within the train of thoughts: 1) deliberately constrained, 2) au-

PSYCHOLOGICAL AND
COGNITIVE SCIENCES
performed an attention task and periodically answered thought- tomatically constrained, and 3) freely moving thoughts (12, 14).
sampling questions about whether their thoughts were 1) task-
Constraints on the train of thoughts serve to focus internal at-
unrelated, 2) freely moving, 3) deliberately constrained, and 4)
tention on a topic for extended periods of time. For example,
automatically constrained. We examined three EEG measures across
deliberately constrained thoughts occur when a person actively
different time windows as a function of each thought type: stimulus-
directs her thoughts to goal-relevant information (e.g., when you
evoked P3 event-related potentials and non–stimulus-evoked alpha
power and variability. Parietal P3 was larger for task-related relative
contemplate your latest experiment). This type of constraint is
to task-unrelated thoughts, whereas frontal P3 was increased for implemented through cognitive control. Automatically constrained
deliberately constrained compared with unconstrained thoughts. thoughts focus on affectively or personally salient information that
Frontal electrodes showed enhanced alpha power for freely moving is difficult to disengage from (e.g., when you worry about your niece
thoughts relative to non-freely moving thoughts. Alpha-power var- who will have surgery). This type of constraint is automatic in
iability was increased for task-unrelated, freely moving, and uncon- nature and thought to operate largely outside of cognitive control.
strained thoughts. Our findings indicate distinct electrophysiological In contrast, freely moving thoughts occur when both of these
patterns associated with task-unrelated and dynamic thoughts, sug- constraints are weak, allowing the mind to wander with no over-
gesting these neural measures capture the heterogeneity of our arching purpose and direction (e.g., when your thoughts drift from a
ongoing thoughts. movie, to gardening, to dinner). Notably, the dynamic framework

task-unrelated thoughts | freely moving thoughts | constrained thoughts | Significance


|
mind wandering EEG
Our minds rarely stay still when left alone. Such trains of

C ognitive neuroscience has reached a consensus that the brain


is not idle at rest (1). This rings intuitively true: When left
alone, our minds rarely stay still. Moreover, recent evidence
thought, however, may unfold in vastly different ways. Here,
we combined electrophysiological recording with thought sam-
pling to assess four types of thoughts: task-unrelated, freely
suggests that rest is not a homogeneous state (2), nor is the brain moving, deliberately constrained, and automatically constrained.
truly “at rest” as it fluctuates across time and contexts (3–6). This Parietal P3 was larger for task-related relative to task-unrelated
too is intuitive: What is striking is not just that we move from thoughts, whereas frontal P3 was increased for deliberately
thought to thought unprompted but also the diverse ways that constrained compared with unconstrained thoughts. Enhanced
trains of thought unfold over time. Sometimes, our thoughts frontal alpha power was observed during freely moving
freely wander between topics. You might remember this morn- thoughts compared with non-freely moving thoughts. Alpha-
ing’s run, then imagine gardening, then think about the dinner power variability was increased for task-unrelated, freely mov-
you’ll cook tonight. Other times, we deliberately constrain our ing, and unconstrained thoughts. Our findings indicate these
thought types have distinct electrophysiological signatures,
thoughts and work diligently toward a goal. In a quiet moment,
suggesting that they capture the heterogeneity of our ongoing
you might methodically contemplate the results of your latest
thoughts.
experiment. Still other times, our thoughts get stuck on an affec-
tively salient topic, from which it is difficult to break free. You Author contributions: J.W.Y.K. and Z.C.I. designed research; J.W.Y.K. and S.P. performed
might worry, over and over, about your niece who is going through research; J.W.Y.K., C.M., and S.P. analyzed data; and J.W.Y.K., Z.C.I., C.M., S.P., A.G., and
major surgery next week. R.T.K. wrote the paper.
Research has traditionally examined the internal train of The authors declare no competing interest.
Downloaded at Bolivia: PNAS Sponsored on January 21, 2021

thought in the context of mind wandering. This field has ex- This article is a PNAS Direct Submission.
panded at such a rapid pace that some have dubbed this the “era Published under the PNAS license.
of the wandering mind” (7). To date, the vast majority of mind- 1
To whom correspondence may be addressed. Email: julia.kam@ucalgary.ca.
wandering research has focused on the static content of individual This article contains supporting information online at https://www.pnas.org/lookup/suppl/
thoughts. In particular, mind-wandering studies have primarily doi:10.1073/pnas.2011796118/-/DCSupplemental.
focused on task-unrelated thought (8)—that is, thoughts that are Published January 18, 2021.

PNAS 2021 Vol. 118 No. 4 e2011796118 https://doi.org/10.1073/pnas.2011796118 | 1 of 9


purports that these three subtypes of thoughts are independent of Alpha-power increases recorded over posterior sites have been
task relatedness. In other words, task-related and task-unrelated associated with internal attention (23, 25) as well as spontaneous
thoughts can both be deliberately constrained, automatically brain activity recorded at rest that is not elicited by external
constrained, or freely moving. See SI Appendix for further details stimuli (26, 27). In contrast, frontal alpha has been linked to
about the relationship between dynamic categories. creative, divergent thinking (28, 29). Accordingly, we hypothe-
Empirical research on the dynamics of thought is in its infancy. sized that task-unrelated thoughts would be associated with en-
Behavioral research has focused on contrasting task-unrelated hanced posterior alpha power, whereas freely moving thoughts
thoughts with the other three subtypes of thoughts, with a par- would be associated with increased frontal alpha power. Given
ticular emphasis on freely moving thoughts. These findings sug- that our dynamic thought-sampling questions address variability
gest that self-reported freely moving and task-unrelated thoughts within the train of thought over time, we examined the neural
have distinct behavioral markers. For example, studies of mind correlates of these thought dynamics by capturing momentary
wandering in everyday life found that self-reports of freely moving changes in alpha power (i.e., alpha-power variability) over the
and task-unrelated thought fluctuate at different rates throughout same ERP-free time window. We predicted that freely moving
the day (11). Consistent with predictions of the dynamic frame- thought would be associated with increased alpha-power vari-
work, these studies reported that task-unrelated and freely moving ability, whereas constrained thought would show reduced alpha-
thoughts are only modestly correlated (r < 0.3) and they occurred power variability.
independent of each other (8, 11). Specifically, although task-
related thoughts are often deliberately constrained and task- Results
unrelated thoughts sometimes move freely, the dynamic frame- In a simple attention task, participants were presented with left
work predicts that this is not always the case. In fact, task-related and right arrows and asked to respond by pressing the left and
thoughts can move freely (e.g., when a graphic designer freely right arrow keys on the keyboard, respectively. Throughout the
associate ideas for her new website design) or be automatically task, thought probes were presented at the end of each block, in
constrained (e.g., when someone obsesses over a problem at which participants answered several questions about their thoughts
work). Task-unrelated thoughts can be deliberately or automati- occurring within the last 10 to 15 s preceding the probe. Partici-
cally constrained (e.g., when you construct a grocery list or worry pants were asked to categorize their thoughts by answering the
about your niece’s surgery during a lecture). The only two em- following questions: 1) Were your thoughts related to the task?; 2)
pirical studies to date have focused explicitly on freely moving were your thoughts freely moving?; 3) were your thoughts delib-
thought (8, 11). Although this provides some initial evidence that erately constrained?; and 4) were your thoughts automatically
task-unrelated thought is different from dynamic thoughts, no constrained? Participants were provided with detailed descriptions
studies have assessed the neural correlates of dynamic thought and examples of each type of thought and a training session, as
types. The identification of different electrophysiological signa- described in Materials and Methods. Based on their responses, we
tures of these thought types would thus provide important vali- examined how frequently each thought type occurred. We also
dation that these categories reflect distinct entities. used their responses to each thought probe to categorize the six
In the current study, we examined the electrophysiological arrow trials preceding the probe into one of the four thought
signatures of the four types of thought by recording an electro- types. Some trials occurred during a block characterized by task-
encephalogram (EEG) while participants performed an atten- related thoughts, for example, whereas other trials occurred dur-
tion task. Participants occasionally answered thought-sampling ing a freely moving period. All behavioral (i.e., mean accuracy,
questions about the nature of their thoughts throughout the task. reaction time, and reaction-time variability) and ERP measures
Thought sampling is the standard method in mind-wandering (i.e., frontal and parietal P3) were time-locked to the onset of the
research: Participants are randomly interrupted as they per- arrows. Alpha power and alpha-power variability were also time-
form a laboratory task and answer questions about their imme- locked to the same stimulus onset but examined in a later time
diately preceding thoughts (9, 10). In line with previous studies, window after the offset of the ERP response. To identify the be-
we ask the standard question about whether participants’ thoughts havioral and electrophysiological signatures of each thought type,
were task-unrelated (see ref. 8 for a review). In addition, we asked all measures were then compared within each of the four thought
whether subjects’ trains of thought were freely moving, deliber- types.
ately constrained, and automatically constrained.
We used electroencephalography because this method has the Attentional Results. First, we examined the percentage of atten-
temporal resolution necessary to capture the transient changes in tion reports by thought types, as reported in Table 1. Subjects
neural activity corresponding to our trains of thoughts, including reported having task-unrelated thoughts more often than task-
stimulus-evoked, task-dependent activity and stimulus-independent, related thoughts; however, they reported freely moving thoughts
intrinsic activity. We first examined event-related potentials (ERPs), about as often as thoughts that did not move freely. Further,
which index the electrophysiological response evoked by task- subjects also reported more thoughts that were not deliberately
relevant stimuli. Previous research suggests that task-unrelated or automatically constrained than constrained thoughts.
thought attenuates the magnitude of ERP components associ- Second, in addition to reporting the occurrence of each thought
ated with sensory (18–20) and cognitive (21–24) processing of type, we also implemented post hoc analyses to examine the extent
task-relevant stimuli. ERPs therefore provide an electrophysio-
logical signature of when a participant has disengaged from task-
relevant stimuli. We predicted that task-unrelated thoughts would
be associated with a reduced P3 ERP component. ERP alpha power & variability
To index stimulus-independent, intrinsic neural activity, we
examined spectral power in the alpha band (8 to 14 Hz) during a 0 0.6 1.8
time window after the offset of ERPs, which is unlikely to be Time (s)
Downloaded at Bolivia: PNAS Sponsored on January 21, 2021

impacted by stimulus-evoked responses. This segregation of the


Fig. 1. EEG measures across the poststimulus time window. We examined
poststimulus time window allowed us to disentangle stimulus- three EEG measures. Stimulus-evoked activity as captured by P3 ERP com-
evoked, task-dependent responses captured by ERP compo- ponents was examined during the 0- to 0.6-s poststimulus time window.
nents in the earlier window and stimulus-independent activity Alpha power and variability index intrinsic neural activity not impacted by
likely associated with the subject’s ongoing thoughts as captured an external stimulus examined after the offset of P3s during the 0.6- to 1.8-s
by alpha power in the later time window (as illustrated in Fig. 1). poststimulus time window.

2 of 9 | PNAS Kam et al.


https://doi.org/10.1073/pnas.2011796118 Distinct electrophysiological signatures of task-unrelated and dynamic thoughts
Table 1. Percentage of attention reports by thought types
Thought types

Task relatedness, % Task-related: 24.5 (3.8) Task-unrelated: 66.4 (4.2)*


Freely moving, % Not free: 41.9 (3.4) Free: 46.5 (4.1)
Deliberate constraints, % Deliberate: 34.4 (3.5) Not deliberate: 57.5 (3.9)*
Automatic constraints, % Automatic: 17.0 (2.4) Not automatic: 76.1 (2.8)*

Mean (and SE) percentages of each condition are shown.


*Indicates significant difference between thought types at P ≤ 0.001.

to which they co-occur. In accordance with the dynamic model Frontal P3 amplitudes were larger for deliberately constrained
which specifically contrasts task-unrelated thoughts with the other thoughts relative to thoughts that were not deliberately constrained
three dynamic thought types, we correlated the occurrence of task- (in the poststimulus window of 369 to 424 ms). Both of these results
unrelated thoughts with freely moving thoughts and deliberately remained significant after statistically controlling for other thought
and automatically constrained thoughts within individuals. These types, suggesting the observed patterns cannot be attributed to
intraindividual correlations were then tested at the group level other thought types. Details of these post hoc control analyses are
against 0 using the sign test, a nonparametric version of the single- reported in SI Appendix. The remaining two thought types did not
sample t test. Task-unrelated thoughts correlated positively with reveal significant conditional differences for either ERP compo-
freely moving thoughts (mean ± SE = 0.51 ± 0.05; Z = 4.29, P < nent. In addition to comparisons within each thought type, we
0.001) and automatically constrained thoughts (mean ± SE = implemented post hoc analyses to compare the ERP components
0.27 ± 0.04; Z = 4.03, P < 0.001), and negatively with deliberately between task-unrelated thought and the other three thought types.
constrained thoughts (mean ± SE = −0.26 ± 0.07; Z = −2.91,

PSYCHOLOGICAL AND
Details of these analyses are also reported in SI Appendix.

COGNITIVE SCIENCES
P = 0.004).
Time-Frequency Power and Variability Results. In order to deter-
Behavioral Results. In order to identify the behavioral correlates
mine the neural correlates of thought types independent of
of each thought type, we examined mean accuracy, reaction time,
and reaction-time variability as a function of thought types. The
reaction time measures are shown in Fig. 2. Mean accuracy was A response time mean
high across all conditions (M = 0.95, SE = 0.012). There were no
significant differences in accuracy that survived the Bonferroni 600 * * *
correction between any of the thought types, likely due to the
ceiling effect in accuracy. 500
Mean reaction time across conditions was 500 ms (SE =
400
Time (ms)

12 ms). Reaction time was slower during task-unrelated periods


than during task-related periods (P = 0.018). Freely moving 300
periods were associated with slower reaction time than periods
that were not freely moving (P = 0.013). Reaction time was faster 200
during deliberately constrained periods relative to periods not
deliberately constrained (P = 0.021). However, none of these 100
effects survived when a Bonferroni correction was applied. No
significant difference in reaction time was observed for auto- 0
TR TU F NF D ND A NA
matically constrained thought (P = 0.069).
Reaction-time variability across conditions was 156 ms (SE = Thought Dimensions
14 ms). Increased reaction-time variability was observed during
task-unrelated periods (M = 161 ms, SE = 16 ms) relative to task- B response time variability
related periods (M = 114 ms, SE = 16 ms; P < 0.001; Fig. 2B). 300 ** ** * **
There was more reaction-time variability during freely moving
periods (M = 162 ms, SE = 19 ms) than during periods that were 250
not freely moving (M = 119 ms, SE = 15 ms; P = 0.006). De-
creased reaction time variability was observed during deliber- 200
Time (ms)

ately constrained periods relative to periods not deliberately


constrained (P = 0.017), but the effect did not survive the 150
Bonferroni correction. There was less reaction-time variability 100
during automatically constrained periods (M = 102 ms, SE =
13 ms) than during periods that were not automatically con- 50
strained (M = 152 ms, SE = 16 ms; P = 0.003).
0
ERP Results. To ascertain the impact of thought types on the pro- TR TU F NF D ND A NA
cessing of external stimuli, we examined stimulus-evoked activity as Thought Dimensions
captured by ERPs using time point-by-time point permutation tests.
Fig. 3 presents the ERP waveforms at the predetermined channels Fig. 2. Reaction time. Mean reaction time (A) and reaction-time variability
Downloaded at Bolivia: PNAS Sponsored on January 21, 2021

(B) are shown separately for each thought type: A, automatically con-
relevant for the corresponding ERPs. ERP data analyses focused
strained; D, deliberately constrained; F, freely moving; NA, not automatically
on the frontal P3 and parietal P3 mean amplitude measures. We constrained; ND, not deliberately constrained; NF, not freely moving; TR,
statistically compared the mean amplitude for each ERP compo- task-related thought; TU, task-unrelated thought. Reaction-time variability
nent separately for each thought type. Parietal P3 amplitudes were was higher during task-unrelated thoughts, freely moving thoughts, and
larger for task-related thoughts compared with task-unrelated thoughts not automatically constrained. Error bars indicate SEM. *P < 0.05
thoughts (in the poststimulus time window of 367 to 432 ms). and **P < 0.0125.

Kam et al. PNAS | 3 of 9


Distinct electrophysiological signatures of task-unrelated and dynamic thoughts https://doi.org/10.1073/pnas.2011796118
A task-relatedness B freely moving C deliberate constraints D automatic constraints
2 Fz 2 2 2

1 1 1 1

0 0 0 0

-1 -1 -1 -1

-2 -2 -2 -2
-0.2 0 0.2 0.4 0.6 -0.2 0 0.2 0.4 0.6 -0.2 0 0.2 0.4 0.6 -0.2 0 0.2 0.4 0.6
Time (s) Time (s) Time (s) Time (s)

3 Pz TR 3 F 3 D 3 A
TU NF ND NA
2 2 2 2

1 1 1 1

0 0 0 0

-1 -1 -1 -1

-2 -2 -2 -2
-0.2 0 0.2 0.4 0.6 -0.2 0 0.2 0.4 0.6 -0.2 0 0.2 0.4 0.6 -0.2 0 0.2 0.4 0.6
Time (s) Time (s) Time (s) Time (s)

Fig. 3. Stimulus-evoked P3 event-related potentials. The ERP waveforms at frontal site Fz (Top) and parietal site Pz (Bottom) are shown separately for the
following thought types: (A) task-related thoughts, (B) freely moving thoughts, (C) deliberately constrained thoughts, and (D) automatically constrained
thoughts. Parietal P3 was reduced during task-unrelated thoughts, whereas frontal P3 was enhanced during deliberately constrained thoughts. Solid black
lines indicate the time points of significant conditional differences. Topographic map insets are shown for the P3 during these significant time windows. TR,
task-related; TU, task-unrelated; F, freely moving; NF, not freely moving; D, deliberately constrained; ND, not deliberately constrained; A, automatically
constrained; NA, not automatically constrained.

stimulus-evoked activity, we statistically examined spectral power alpha during task-unrelated thought using a traditional approach
in the alpha band between 600 and 1,800 ms poststimulus using with a priori electrode selection, instead of the data-driven ap-
cluster-based permutation tests. Fig. 4 shows the time-frequency proach of cluster-based permutation tests implemented on all
plots for the significant thought types, along with the alpha- electrodes. Specifically, we conducted permutation tests at the
power time series extracted from the cluster of significant elec- single-electrode level to examine conditional differences in posterior
trodes. Our analyses focused on alpha power given its role in alpha power. The aforementioned studies converged on maximal
internal attention and divergent thinking. There was a significant alpha power at midline posterior site Oz for task-unrelated thought
cluster around 1,100 to 1,350 ms over the frontal sites that relative to task-related thought (23, 25), where alpha power is also
showed greater alpha power during freely moving thoughts than maximal in our data (as shown in SI Appendix, Fig. S1). Accordingly,
during thoughts that were not freely moving (P = 0.023). This we conducted a permutation test to compare the alpha power at
effect did not survive the conservative Bonferroni correction for Oz in two time windows (600 to 1,200 and 1,200 to 1,800 ms).
multiple comparisons, but remained significant after statistically Posterior alpha power was increased during task-unrelated thought
controlling for other thought types. relative to task-related thought during the 1,200- to 1,800-ms time
There were no significant clusters over posterior sites that dif- window (P = 0.037). There were no differences in posterior alpha
ferentiated alpha power between task-related and task-unrelated during the 600- to 1,200-ms time window (P = 0.255).
thoughts. Given that enhanced posterior alpha power has been Next, we statistically examined variability in whole-brain alpha
consistently reported during internal attention (23, 25) and spon- activity across time and trials between 600 and 1,800 ms using
taneous brain activity at rest (26, 27), we implemented post hoc permutation tests. Fig. 5 illustrates alpha variability as a function of
analysis to directly test our a priori hypothesis of increased posterior thought types. All four thought types showed significant differences.

freely moving not freely moving


50 50 1.5 2
Frequency (Hz)
Frequency (Hz)

Alpha Power (µV2)

40 40

30 30
1
20 20
F
10 10 NF
Downloaded at Bolivia: PNAS Sponsored on January 21, 2021

0 0
0 0.5 1 1.5 0 0.5 1 1.5 0 0.5 1 1.5
Time (ms) Time (ms) Time (ms)

Fig. 4. Stimulus-independent alpha power. Alpha power averaged across significant electrodes emerging from a cluster-based permutation test for freely
moving thought. Time-frequency representations of low-frequency power are shown (Left and Middle), with the location of the significant cluster of
electrodes and black box outlining time windows of significance. Mean alpha power (shading indicates SEM) for each condition (freely moving and not freely
moving) is displayed (Right), with the gray box highlighting the time window of significance. No significant clusters emerged for other thought types.

4 of 9 | PNAS Kam et al.


https://doi.org/10.1073/pnas.2011796118 Distinct electrophysiological signatures of task-unrelated and dynamic thoughts
In particular, task-unrelated thought (P = 0.001), freely moving Our results are also consistent with recent resting-state func-
thought (P = 0.006), thought that was not deliberately constrained tional MRI (fMRI) findings suggesting that functional connec-
(P = 0.003), and thought that was not automatically constrained tivity between resting-state networks fluctuates over time (4).
(P = 0.009) showed greater alpha variability than task-related Given that freely moving thought is conceptually independent
thought, thought that was not freely moving, deliberately con- from task-relatedness, it is unsurprising that the former did not
strained thought, and automatically constrained thought, respec- predict differences in the magnitude of task-evoked parietal P3.
tively. All these conditional differences remained significant after Our finding that freely moving thought is associated with in-
statistically controlling for other thought types. creased frontal alpha power should be interpreted with caution;
although this was a planned comparison that remained signifi-
Discussion cant after controlling for other thought types, this effect did not
One of the most notable discoveries in human neuroscience in survive our conservative method of correction for multiple com-
the past few decades is that the brain is active during rest (1). parisons. This pattern is consistent with previous studies that have
Early studies primarily associated this resting state with the implicated frontal alpha activity in creativity (28). For example,
brain’s “default network.” Yet it has become clear the brain’s increased frontal alpha was observed during creative idea gener-
intrinsic activity and connectivity are heterogeneous: They dy- ation (30) and in highly creative individuals relative to less creative
namically fluctuate over time (3, 4) and across contexts (5). We individuals (31). Furthermore, noninvasive stimulation of frontal
developed self-report questions to study the heterogeneity of our alpha enhanced performance on a creativity task (29). Creativity
trains of thought, and established the electrophysiological sig- involves the generation of novel ideas and associations (28)
natures for four types of thought–task-unrelatedness, free move- through a divergent thought process that is theoretically similar to
ment, deliberate constraints, and automatic constraints. Our first set the free movement of thoughts going from topic to topic (12, 32).
of results concerns stimulus-evoked activity. During task-unrelated Increased frontal alpha during freely moving thought and during
thought, stimulus-evoked parietal P3 was attenuated relative to creativity may therefore provide a signature of relatively uncon-
task-related thought. Deliberately constrained thought led to an strained thinking. Taken together, these findings suggest that the

PSYCHOLOGICAL AND
COGNITIVE SCIENCES
increased stimulus-evoked frontal P3, relative to thought that dynamic characteristics of our thoughts may be directly reflected
lacked deliberate constraints. Our second set of results concerns in our electrophysiological measures.
stimulus-independent activity that was measured during a time In our study, task-unrelated thoughts reduced stimulus-evoked
window that is unlikely to be impacted by stimulus-evoked re- parietal P3. This is consistent with our prediction that task-
sponses. This allowed us to plausibly isolate electrophysiological unrelated thoughts would uniquely disrupt stimulus-evoked ac-
activity associated with a subject’s ongoing thoughts from task- tivity associated with task performance. This prediction flows from
relevant response. We observed that freely moving thought was the decoupling model, which proposes the attenuation of external
linked to greater intrinsic frontal alpha power than thoughts that processing is necessary for supporting task-unrelated thoughts
are not freely moving. We also observed increased variability in (33). In particular, task-unrelated thought recruits executive re-
intrinsic alpha power during task-unrelated thought, freely moving sources, which enables our disengagement from an external task
thought, and unconstrained thoughts. Taken together, our results via top–down control, resulting in the disruption of responses to
suggest that there are distinct neural signatures of task-unrelated external stimuli. Accordingly, our finding of a decreased parietal
and dynamic thoughts. P3 evoked by visual stimuli thus validates our self-reported mea-
Freely moving thought was associated with increased frontal sure of task-unrelated thought, and is consistent with previous
alpha power and alpha-power variability. Both measures were reports of a reduction in parietal P3 amplitude during task-
obtained during a time window unlikely to be impacted by external unrelated thoughts (21, 22, 34).
stimuli, which suggests that these measures reflect intrinsic activity Given that posterior alpha power has been shown to increase
associated with freely moving thoughts. Freely moving thought during cognitive states similar to task-unrelated thoughts, such as
showed increased variability in alpha power. This suggests that the attentional lapses (23), internally oriented attention (25), or rest
cognitive variability of freely moving thought parallels neural (26), we predicted an increase in posterior alpha power during
variability, particularly in the form of alpha-power fluctuations. task-unrelated thoughts. These past findings suggest that poste-
rior alpha not only reflects top–down inhibitory control of ex-
ternal inputs (35) as shown in traditional studies of selective
** ** * ** attention but also serves as an electrophysiological signature of
4 internally oriented task-unrelated thoughts. In our study, the
Alpha Power Variability

cluster-based permutation tests did not yield significant clusters;


3 however, our planned post hoc analysis focusing on the posterior
midline site in which alpha power is typically maximal (23)
revealed increased alpha power during task-unrelated thought
2 relative to task-related thought. Notably, task-unrelated thoughts
also led to an increase in alpha-power variability. This is likely
because task-related thought in the current study is highly con-
1
strained by the restrictive nature of our experimental task (i.e., the
left arrow versus the right arrow), whereas thoughts unrelated to
0 the ongoing task are not. Due to this lack of experimental con-
TR TU F NF D ND A NA straints, task-unrelated thoughts can move to different topics and
Thought Dimensions contents, which may have contributed to increased alpha-power
variability.
Fig. 5. Stimulus-independent alpha-power variability. Alpha-power vari- Although there was a positive correlation between occurrence
Downloaded at Bolivia: PNAS Sponsored on January 21, 2021

ability across the 600- to 1,800-ms time window is shown separately for each
of task-unrelated thoughts and freely moving thoughts, we ob-
thought type. Alpha-power variability was higher during task-unrelated
thoughts, freely moving thoughts, and unconstrained thoughts. TR, task-
served a number of differences between these two thought types.
related; TU, task-unrelated; F, freely moving; NF, not freely moving; D, de- For example, task-unrelated and freely moving thoughts occurred
liberately constrained; ND, not deliberately constrained; A, automatically at different rates (66 vs. 47%). Furthermore, our results revealed
constrained; NA, not automatically constrained. Error bars indicate SEM. different electrophysiological patterns between these two thought
*P < 0.05, and **P < 0.0125. types. Whereas task-unrelated thinking was primarily associated

Kam et al. PNAS | 5 of 9


Distinct electrophysiological signatures of task-unrelated and dynamic thoughts https://doi.org/10.1073/pnas.2011796118
with changes in stimulus-evoked ERPs, freely moving thought was Our methods for measuring the dynamics of thought are po-
primarily associated with changes in intrinsic alpha power. These tentially applicable to many fields of psychology. The dynamic
measures also remained significant after controlling for other thought-sampling methods could be used to test theoretical links
thought types. Therefore, our findings indicate that task-unrelated between various psychological phenomena and dynamic thought
thought and freely moving thought are uniquely characterized by a types. Clinical conditions have been hypothesized to involve high
different set of electrophysiological features. These results provide rates of freely moving (e.g., attention deficit hyperactivity disor-
support for the dynamic model, which proposed that task- der) or automatically constrained thought (e.g., depression and
unrelated and freely moving thoughts are distinct (12). obsessive compulsive disorder) (12). Developmental psychologists
Deliberately and automatically constrained thoughts were ac- have argued that children may learn better than adults in certain
companied by some overlapping patterns of electrophysiological contexts, due to their free and associative style of thinking (38, 39).
activity and some patterns that differed. Later frontal P3 was All these fields may benefit from a better understanding of, and a
enhanced during deliberately constrained thoughts compared with better ability to measure, the dynamics of thinking.
thoughts that were not deliberately constrained. Previous studies William James said that “the natural tendency of attention
have found that cognitive control engages midline frontocentral when left to itself is to wander to ever new things” (40). We
regions (36), suggesting this enhanced frontal P3 during deliber- prefer to think that attention has multiple “natural tendencies”:
ately constrained thoughts may subserve cognitive control. This Sometimes it moves freely, sometimes we direct it, and sometimes
interpretation is consistent with the dynamic model’s conceptual- it gets “stuck.” Our study shows that those thought types have
ization of deliberately constrained thoughts as goal-oriented and distinct behavioral and electrophysiological signatures. These
implemented through top–down control (12). We did not observe findings suggest that the heterogeneity of thought is reflected in
any differences in stimulus-evoked P3 amplitudes as a function of the brain.
automatically constrained thoughts.
In terms of stimulus-independent activity, both deliberately and Materials and Methods
automatically constrained thoughts were accompanied by reduced Participants. Forty-five individuals (13 females and 32 males; M = 20.1 y, SD =
alpha-power variability. In other words, alpha-power variability 1.83) participated in the study. Three participants had missing data due to
decreased for both types of constrained thought and increased for technical difficulties (with two subjects missing EEG data and one missing
freely moving thoughts. Similarly, reaction-time variability de- behavioral data). Another three participants’ EEG data files were corrupted.
These six individuals were excluded from analysis, resulting in a sample of
creased for both types of constrained thought and increased for
39. All participants had normal or corrected-to-normal vision and did not
freely moving thought. Both results support the dynamic model’s have neurological disorders. They provided written informed consent, and
prediction that constrained thought is the opposite of freely were paid for their participation. The study was approved by the Institu-
moving thought (12). These findings suggest that variability in our tional Review Board at the University of California, Berkeley.
cognitive states manifests in the variability of both behavioral and
electrophysiological measures. Future work may explore whether Task Paradigm and Experimental Procedure. Participants performed a simple
other measures of neural dynamics differentiate between delib- attention task, in which they were required to press the left and right arrow
erate and automatic constraints. keys on the keyboard when they saw left and right arrows, respectively. Both
The primary contribution of our paper is the identification of arrows were made of black solid lines and presented in a randomized order
electrophysiological signatures of four thought types: task-unrelated, with equal probability. Each stimulus was presented for 2,000 ms, with a
randomly jittered interstimulus interval between 200 and 800 ms during
freely moving, deliberately constrained, and automatically con-
which a fixation cross was presented. All stimuli (i.e., arrows and fixation
strained. An important question for future research is how these cross) were displayed at the center of a cathode-ray tube monitor, where
dynamic categories relate to task-unrelated thought. Our behav- participants were asked to keep their eyes fixated at all times. Stimulus
ioral results provide initial evidence about the relationship be- presentation was controlled by E-Prime 2.0 (Psychology Software Tools). The
tween task-unrelated thought and dynamic categories. We found task consisted of 840 trials, and lasted ∼40 min.
that occurrence of task-unrelated thought positively correlated Throughout the task, thought probes were presented at the end of each
with freely moving and automatically constrained thoughts, block, in which participants answered several questions about their thoughts
whereas deliberately constrained thought showed the opposite occurring within the last 10 to 15 s preceding the probe. As mentioned, some
pattern. Notably, task unrelatedness did not show a one-to-one of these questions were designed to capture subtypes of thoughts proposed
mapping with any of the dynamic categories, explaining less than in the dynamic framework of spontaneous thought (12, 14). Participants
were asked to categorize their thoughts by responding on a 7-point Likert
26% variance in any of these categories. These observed patterns
scale for the following questions: 1) Were your thoughts related to the task?
are consistent with predictions of the dynamic framework (12) for
; 2) were your thoughts freely moving?; 3) were your thoughts deliberately
two reasons. First, task-unrelated thoughts can be freely moving, constrained?; and 4) were your thoughts automatically constrained? Only
deliberately constrained, or automatically constrained. Second, the first two questions have been asked in previous studies (8, 19).
although task-related thoughts are often deliberately constrained, To ensure that participants understood the meaning of these questions,
they can also be automatically constrained (e.g., when you obsess they completed a training session at the beginning of the experiment. We
over a problem at work) or even freely moving (e.g., when you provided definitions and example scenarios of each thought type, as shown in
freely associate ideas for a new research project). Interestingly, we Table 2. Following this, participants were given new example scenarios that
found that the correlation between task-unrelated and freely they were likely to encounter during the experiment and asked to answer
moving thought is stronger than in previous studies conducted in the thought probe questions based on the scenario. If they answered any
everyday life (8, 11). This may be because experimental tasks are questions incorrectly, experimenters would provide the correct answer and
explain the reason. Experimenters proceeded to the next question only after
more structured and therefore more likely to constrain free
the subjects showed full understanding of the question and correct re-
movement (37). For instance, thoughts about the arrow task in our sponse. One example scenario is as follows:
study are less conducive to free movement; in contrast, everyday
tasks are more likely to require creative idea generation, allowing While performing the attention task, a student can’t stop thinking
for our thoughts to move freely between disparate ideas (12, 32). about whether they got a good grade on their test yesterday. “Did I
get question 3 wrong?” they think, picturing their answer. “What if I
Downloaded at Bolivia: PNAS Sponsored on January 21, 2021

Future work could overcome limitations of the current study by


didn’t pass? I knew I should have studied more!!”
building a larger dataset with sufficient data points per thought
type, which enables examining the interactions between task-
unrelated thoughts and the three subtypes of dynamic thoughts, In this example, the student would categorize their experience as task-
including more fine-grained analyses such as contrasting task- unrelated and automatically constrained because they cannot stop thinking
unrelated thoughts that are freely moving and not freely moving. about the test but neither freely moving nor deliberately constrained.

6 of 9 | PNAS Kam et al.


https://doi.org/10.1073/pnas.2011796118 Distinct electrophysiological signatures of task-unrelated and dynamic thoughts
At the end of each block, participants provided a response to each of these thought types: task-related versus task-unrelated, freely moving versus not
four questions characterizing the type of thoughts they experienced in the freely moving, deliberately constrained versus not deliberately constrained,
prior 10 to 15 s. Their responses were dichotomized into two groups, sep- and automatically constrained versus not automatically constrained. All ERPs
arating responses on the lower end (1 to 3) and upper end (5 to 7) of the Likert were quantified by the mean amplitude measure relative to a 200- to 0-ms
scale and discarding the middle response (4). For instance, their response to prestimulus baseline. The frontal P3 mean amplitude was measured over
the first question would be categorized as task-related (responses 1 to 3), midline frontal sites (Fz, F1, F2), whereas the parietal P3b mean amplitude
task-unrelated (responses 5 to 7), or discarded if they responded 4 on the was measured over midline parietal sites (Pz, P1, P2). The time window of
Likert scale. This is consistent with previous studies that also used a 7-point analysis spanned from 0 to 600 ms poststimulus, which marks the offset of
Likert scale for thought-sampling questions (41, 42). We used their responses the stimulus-evoked response.
to each thought probe to label the six trials (∼15 s) preceding the probe. This Alpha power: Non–stimulus-evoked activity. To assess neural activity not influ-
15-s window was chosen for several reasons. First, this time window has enced by stimulus-evoked responses, we examined mean spectral power as
been consistently used in neuroimaging studies involving experience sam- well as variability in spectral power as a function of thought types in a later
pling (20, 41, 43, 44). Second, theoretical frameworks have purported that time window after the offset of the stimulus-evoked response (as shown in
attention can fluctuate at a slow timescale on the order of tens of seconds Fig. 1). Spectral decomposition was conducted across a −1-s prestimulus to
(45). This is supported by fMRI and intracranial EEG studies reporting slow 3-s poststimulus time window for each trial, using fast Fourier transforms
fluctuations in spontaneous rest activity [<0.1 Hz (46, 47)] as well as EEG using a Hanning window for frequencies between 2 and 30 Hz at 1-Hz steps.
studies reporting signatures of task-unrelated thoughts are observed for up Each trial was then baseline-corrected by subtracting and dividing by
to 20 s (23). Finally, this time window allowed us to maximize the number of the average power in the prestimulus interval from −300 to −100 ms. As
trials included to create a reliable average while maintaining a reasonable stimulus-evoked activity subsided ∼600 ms after stimulus presentation and
validity of the participants’ report. There was a total of 35 blocks, and each the shortest trial duration is 1,800 ms, our analysis focused on 600 to
block consisted of 18 to 30 trials (mean = 24, equivalent to ∼1 min per block). 1,800 ms following stimulus presentation. Conditional differences observed
in this poststimulus-evoked time window measure spontaneous activity as-
EEG Data Acquisition and Preprocessing. EEG was recorded continuously from sociated with thought types that are not impacted by stimulus-evoked re-
64 active, preamplified Ag/AgCl electrodes mounted on a cap according to sponses. Given the role of the alpha band in internal attention and creative
the extended 10-20 layout using the BioSemi ActiveTwo System. Vertical and thinking as previously mentioned, our analysis focused on the alpha band

PSYCHOLOGICAL AND
COGNITIVE SCIENCES
horizontal eye movements were recorded from electrodes above and below (8 to 14 Hz).
the right eye and two electrodes placed at the right and left outer canthi. To examine mean spectral power, we extracted alpha power across the
Data were amplified and digitized at 512 Hz. poststimulus-evoked activity window (600 to 1,800 ms) for all artifact-free
EEG data were bandpass-filtered between 1 and 50 Hz, and referenced and correct trials for each thought type. We used the cluster-based per-
offline to the average of the two mastoid electrodes. Independent compo- mutation test to determine whether alpha power differed within thought
nent analysis was used to correct for ocular and muscle artifacts. Data de- types and, if so, to depict the time course of these differences. This data-
composition was performed using the fastICA toolbox in EEGLAB, and driven approach has the advantage of not having to presume the region or
artifactual components were manually detected based on component time time window in which conditional differences may emerge. For each elec-
course, topography, and power spectral density. Electrodes with excessively trode, we averaged alpha power across trials yielding one time series per
noisy signals were interpolated from neighboring electrodes using spherical thought type. Each subject’s alpha-power time series for all electrodes were
spline interpolation (48). Continuous EEG data were segmented into 4,000- then submitted to a cluster-based permutation test to determine differences
ms epochs, beginning at 1,000 ms prior to stimulus onset. Each trial was between thought types.
visually inspected for remaining artifacts, which were removed from sub- To examine variability within participants, we focused on change over time
sequent analyses. Common average reference was then applied to the data and across trials. We extracted the relative change in alpha power across the
prior to analysis. EEG data preprocessing and analysis were performed using poststimulus-evoked activity window (600 to 1,800 ms) for all electrodes for
EEGLAB (49) and FieldTrip (50) within Matlab (MathWorks). each participant. In keeping with the cluster-based permutation tests that
considered all electrodes, we did not restrict our analyses to focal regions
EEG Data Quantification. and instead considered all electrodes. For each electrode, we assessed fluc-
ERP: Stimulus-evoked activity. For stimulus-evoked activity, we examined the tuations in spectral power over the designated time window by computing
P3 event-related potentials in a time window immediately following stim- the SD in alpha power across the time window for that electrode, yielding
ulus onset (as shown in Fig. 1). EEG signals were bandpass-filtered at 1 to 20 one variability value per trial. We then computed the SD of these values
Hz. Only artifact-free and correct trials were included in the analysis. These across trials. This variability across time and trial, averaged across electrodes,
trials were averaged separately within participants as a function of their was compared between thought types.

Table 2. Definitions and examples of each thought type


Thought types Definitions and examples

Task unrelatedness Definition: Thoughts unrelated to an ongoing task.


Example: You might think about plans for the weekend as you are
completing an assignment.
Freely moving Definition: Thoughts that drift from one thing to another without
focusing on anything for an extended period of time.
Example: You imagine yourself having dinner this evening, then
wonder if you’ve been eating too much fast food recently, then
notice a smudge on the computer screen, then remember that you
have to do laundry tonight.
Deliberately constrained Definition: Thoughts that are focused on an overarching goal; when
distracted from that goal, you bring your thoughts back on track.
Example: You might be actively focusing on the experimental task
and ensuring that nothing else is going through your mind.
Downloaded at Bolivia: PNAS Sponsored on January 21, 2021

Automatically constrained Definition: Thoughts that are drawn to something and difficult to
disengage from, whether or not you’re actively thinking about
them.
Example: You might be thinking over and over again about a fight
you just had with your housemate.

Kam et al. PNAS | 7 of 9


Distinct electrophysiological signatures of task-unrelated and dynamic thoughts https://doi.org/10.1073/pnas.2011796118
Statistical Analyses. Given that no experimental manipulations were placed was performed separately for the frontal and parietal P3 ERP components.
on the subjects to facilitate specific thought types, there is a wide range of Similar to the behavioral measures, we applied the same Bonferroni-
responses across individuals on each thought type. In order to implement the corrected critical alpha on the P3 ERP analyses. These analyses reflect the
repeated-measures design of the experiment, all statistical analyses included impact of engaging in certain thought types on the cognitive processing of
only subjects who reported experiencing all four thought types, and spe- external visual stimuli.
cifically both ends of the spectrum of each thought type (i.e., task-related or For mean spectral power, we submitted each subject’s alpha-power time
task-unrelated; freely moving or not; deliberately constrained or not; au- series for all electrodes to a cluster-based permutation test to determine
tomatically constrained or not). This yielded a final sample of 24 partici- differences between thought types. This was identified by means of dependent-
pants. This approach allowed us to examine differences across these thought samples t tests thresholded at an alpha of 0.05. A null distribution was estimated
types within an individual.
by randomly shuffling trials 1,000 times between conditions before averaging,
For attentional measures, we first conducted nonparametric pairwise
followed by the same clustering procedure. Using the Monte Carlo method with
comparisons to determine if occurrence rates differed within each thought
1,000 random permutations, P values of the observed clusters were calculated
type. In accordance with the dynamic model which specifically contrasts task-
as the proportion of random partitions that yielded a larger effect than the
unrelated thoughts with the other three dynamic thought types, we also
observed experimental effect (52). We adopted a two-tailed test, which sets
implemented post hoc analyses to examine the extent to which task-
critical alpha at 0.025. The cluster-based test approach corrects for the multiple
unrelated thoughts and the three subtypes of dynamic thought types co-
occurred. To do so, we implemented Pearson correlations to assess the co- comparisons of time points and electrodes. Similar to the above analyses, we
occurrence of task-unrelated thoughts with freely moving thoughts and applied the same Bonferroni-corrected critical alpha on the alpha-power
deliberately and automatically constrained thoughts within individuals. analyses. These analyses reveal the non–stimulus-evoked spectral correlates of
These intraindividual correlations were converted to the standardized rho each thought type.
values and tested at the group level. To determine whether they were sig- We also assessed the within-subject variability of alpha power across time
nificantly different from 0, we used the sign test, a nonparametric version of and trials. In particular, we implemented permutation tests to statistically
the single-sample t test. compare variability between thought types. Similar to the above analyses,
For behavioral measures, we focused on accuracy as well as reaction time variability values for artifact-free and correct trials were shuffled between
mean and variability for accurate trials only. Each measure was compared two conditions (e.g., task-related and task-unrelated), and tested using a
between each of the four thought types using separate Wilcoxon signed-rank dependent-samples t test. Based on 1,000 random permutations, P values of
tests, the nonparametric version of the paired-samples t test. Although these the observed difference were calculated as the proportion of random per-
are all planned comparisons, we implemented the conservative approach of mutations that yielded a larger effect than the observed experimental ef-
Bonferroni correction on the resulting P values of these analyses to correct fect. We applied the same Bonferroni-corrected critical alpha above on these
for multiple comparisons. Given the comparisons of four thought types in variability analyses. This variability measure reveals neural activity unfolding
our study, the critical alpha was set to 0.05/4 = 0.0125. These analyses reflect across time and trials outside of the stimulus-evoked time window. All sta-
differences in behavioral markers of different thought types. tistical analyses were performed using Matlab.
For ERP measures, we examined differences between each of the four
thought types in the amplitude of the frontal P3 and parietal P3 ERP com-
Data Availability. Data and code reported in this article have been deposited
ponents using the time point-by-time point permutation test. In particular,
in the Collaborative Research in Computational Neuroscience database
for each time point in the 0- to 600-ms poststimulus time window, trials
(https://crcns.org; accession no. K0N8780K) (53).
containing the ERP waveform time series were shuffled between two con-
ditions reflecting opposite ends of the same thought type (e.g., task-related
ACKNOWLEDGMENTS. Many thanks to our participants for their time and
and task-unrelated), and tested using a dependent-samples t test. Based on
effort in our study. We also appreciate useful feedback from Kalina
1,000 random permutations, P values of the observed difference were cal-
Christoff, Alan Shen, and Chandra Sripada during study conceptualization.
culated as the proportion of random permutations that yielded a larger Z.C.I., C.M., and J.W.Y.K. were supported by the Templeton Foundation.
effect than the observed experimental effect. Conditional differences were J.W.Y.K. was also supported by the James McDonnell Foundation and the
considered as significant if the P values that were below 0.05 lasted for a Natural Sciences and Engineering Research Council of Canada. A.G. was
minimum of 50 consecutive milliseconds. This duration threshold far exceeds supported by DARPA Machine Common Sense, the John Templeton Founda-
the minimum number of consecutive time points necessary for a time series tion, and the Bezos Foundation. R.T.K. was supported by National Institute of
to be considered as significantly different from 0 (51). This permutation test Neurological Disorders and Stroke Grant NS21135.

1. M. E. Raichle et al., A default mode of brain function. Proc. Natl. Acad. Sci. U.S.A. 98, 14. Z. C. Irving, Mind-wandering is unguided attention: Accounting for the “purposeful”
676–682 (2001). wanderer. Philos. Stud. 173, 547–571 (2016).
2. H. T. Wang et al., Dimensions of experience: Exploring the heterogeneity of the 15. C. Sripada, “An exploration/exploitation tradeoff between mind wandering and task-
wandering mind. Psychol. Sci. 29, 56–71 (2018). directed thinking” in Oxford Handbook of Spontaneous Thought and Creativity,
3. M. G. Preti, T. A. Bolton, D. Van De Ville, The dynamic functional connectome: K. C. R. Fox, K. Christoff, Eds. (Oxford University Press, 2018), pp. 23–34.
State-of-the-art and perspectives. Neuroimage 160, 41–54 (2017). 16. K. Christoff, Undirected thought: Neural determinants and correlates. Brain Res.
4. J. Gonzalez-Castillo et al., The spatial structure of resting state connectivity stability 1428, 51–59 (2012).
on the scale of minutes. Front. Neurosci. 8, 138 (2014). 17. K. Christoff et al., The ethics of belief. Psychol. Bull. 4, 29–52 (2011).
5. M. L. Dixon et al., Interactions between the default network and dorsal attention 18. J. W. Y. Kam et al., Slow fluctuations in attentional control of sensory cortex. J. Cogn.
network vary across default subsystems, time, and cognitive states. Neuroimage 147, Neurosci. 23, 460–470 (2011).
632–649 (2017). 19. C. Braboszcz, A. Delorme, Lost in thoughts: Neural markers of low alertness during
6. A. Turnbull et al., The ebb and flow of attention: Between-subject variation in in- mind wandering. Neuroimage 54, 3040–3047 (2011).
trinsic connectivity and cognition associated with the dynamics of ongoing experi- 20. B. Baird, J. Smallwood, A. Lutz, J. W. Schooler, The decoupled mind: Mind-wandering
ence. Neuroimage 185, 286–299 (2019).
disrupts cortical phase-locking to perceptual events. J. Cogn. Neurosci. 26, 2596–2607
7. F. Callard, J. Smallwood, J. Golchert, D. S. Margulies, The era of the wandering mind?
(2014).
Twenty-first century research on self-generated mental activity. Front. Psychol. 4, 891
21. J. W. Y. Kam et al., Mind wandering and motor control: Off-task thinking disrupts the
(2013).
online adjustment of behavior. Front. Hum. Neurosci. 6, 329 (2012).
8. C. Mills, Q. Raffaelli, Z. C. Irving, D. Stan, K. Christoff, Is an off-task mind a freely-
22. J. Smallwood, E. Beach, J. W. Schooler, T. C. Handy, Going AWOL in the brain: Mind
moving mind? Examining the relationship between different dimensions of thought.
wandering reduces cortical analysis of external events. J. Cogn. Neurosci. 20, 458–469
Conscious. Cogn. 58, 20–33 (2018).
9. J. Smallwood, J. W. Schooler, The restless mind. Psychol. Bull. 132, 946–958 (2006). (2008).
10. J. Smallwood, J. W. Schooler, The science of mind wandering: Empirically navigating 23. R. G. O’Connell et al., Uncovering the neural signature of lapsing attention: Elec-
the stream of consciousness. Annu. Rev. Psychol. 66, 487–518 (2015). trophysiological signals predict errors up to 20 s before they occur. J. Neurosci. 29,
11. G. K. Smith, C. Mills, A. Paxton, K. Christoff, Mind-wandering rates fluctuate across 8604–8611 (2009).
Downloaded at Bolivia: PNAS Sponsored on January 21, 2021

the day: Evidence from an experience-sampling study. Cogn. Res. Princ. Implic. 3, 54 24. J. Xu, D. Friedman, J. Metcalfe, Attenuation of deep semantic processing during mind
(2018). wandering: An event-related potential study. Neuroreport 29, 380–384 (2018).
12. K. Christoff, Z. C. Irving, K. C. R. Fox, R. N. Spreng, J. R. Andrews-Hanna, Mind- 25. J. W. Y. Kam, A. K. Solbakk, T. Endestad, T. R. Meling, R. T. Knight, Lateral prefrontal
wandering as spontaneous thought: A dynamic framework. Nat. Rev. Neurosci. 17, cortex lesion impairs regulation of internally and externally directed attention.
718–731 (2016). Neuroimage 175, 91–99 (2018).
13. P. Seli et al., Mind-wandering as a natural kind: A family-resemblances view. Trends 26. H. Laufs et al., EEG-correlated fMRI of human alpha activity. Neuroimage 19,
Cogn. Sci. 22, 479–490 (2018). 1463–1476 (2003).

8 of 9 | PNAS Kam et al.


https://doi.org/10.1073/pnas.2011796118 Distinct electrophysiological signatures of task-unrelated and dynamic thoughts
27. H. Laufs et al., Where the BOLD signal goes when alpha EEG leaves. Neuroimage 31, 41. K. Christoff, A. M. Gordon, J. Smallwood, R. Smith, J. W. Schooler, Experience sam-
1408–1418 (2006). pling during fMRI reveals default network and executive system contributions to
28. A. Fink, M. Benedek, EEG alpha power and creative ideation. Neurosci. Biobehav. Rev. mind wandering. Proc. Natl. Acad. Sci. U.S.A. 106, 8719–8724 (2009).
44, 111–123 (2014). 42. A. Kirschner, J. W. Y. Kam, T. C. Handy, L. M. Ward, Differential synchronization in
29. C. Lustenberger, M. R. Boyle, A. A. Foulser, J. M. Mellin, F. Fröhlich, Functional role of default and task-specific networks of the human brain. Front. Hum. Neurosci. 6, 139
frontal alpha oscillations in creativity. Cortex 67, 74–82 (2015). (2012).
30. A. Fink, R. H. Grabner, M. Benedek, A. C. Neubauer, Divergent thinking training is 43. J. W. Y. Kam, E. Dao, M. Stanciulescu, H. Tildesley, T. C. Handy, Mind wandering and
the adaptive control of attentional resources. J. Cogn. Neurosci. 25, 952–960 (2013).
related to frontal electroencephalogram alpha synchronization. Eur. J. Neurosci. 23,
44. D. Stawarczyk, S. Majerus, M. Maj, M. Van der Linden, A. D’Argembeau, Mind-wan-
2241–2246 (2006).
dering: Phenomenology and function as assessed with a novel experience sampling
31. O. M. Razumnikova, Creativity related cortex activity in the remote associates task.
method. Acta Psychol. (Amst.) 136, 370–381 (2011).
Brain Res. Bull. 73, 96–102 (2007). 45. E. J. S. Sonuga-Barke, F. X. Castellanos, Spontaneous attentional fluctuations in im-
32. G. Manesh, C. Mills, L. Roseman, R. Carhart-Harris, K. Christoff, Updating the dynamic paired states and pathological conditions: A neurobiological hypothesis. Neurosci.
framework of thought: Creativity and psychadelics. Neuroimage 213, 116726 (2020). Biobehav. Rev. 31, 977–986 (2007).
33. J. Smallwood, Distinguishing how from why the mind wanders: A process-occurrence 46. P. Fransson, Spontaneous low-frequency BOLD signal fluctuations: An fMRI investi-
framework for self-generated mental activity. Psychol. Bull. 139, 519–535 (2013). gation of the resting-state default mode of brain function hypothesis. Hum. Brain
34. E. Barron, L. M. Riby, J. Greer, J. Smallwood, Absorbed in thought: The effect of mind Mapp. 26, 15–29 (2005).
wandering on the processing of relevant and irrelevant events. Psychol. Sci. 22, 47. B. L. Foster, V. Rangarajan, W. R. Shirer, J. Parvizi, Intrinsic and task-dependent coupling
596–601 (2011). of neuronal population activity in human parietal cortex. Neuron 86, 578–590 (2015).
35. O. Jensen, A. Mazaheri, Shaping functional architecture by oscillatory alpha activity: 48. F. Perrin, J. Pernier, O. Bertrand, J. F. Echallier, Spherical splines for scalp potential and
Gating by inhibition. Front. Hum. Neurosci. 4, 186 (2010). current density mapping. Electroencephalogr. Clin. Neurophysiol. 72, 184–187 (1989).
36. J. F. Cavanagh, M. J. Frank, Frontal theta as a mechanism for cognitive control. Trends 49. A. Delorme, S. Makeig, EEGLAB: An open source toolbox for analysis of single-trial
Cogn. Sci. 18, 414–421 (2014). EEG dynamics including independent component analysis. J. Neurosci. Methods 134,
37. Z. C. Irving, Drifting and directed minds: The significance of mind-wandering for 9–21 (2004).
50. R. Oostenveld, P. Fries, E. Maris, J. M. Schoffelen, FieldTrip: Open source software for
mental agency. PsyArXiv:10.31234/osf.io/3spnd (20 November 2019).
advanced analysis of MEG, EEG, and invasive electrophysiological data. Comput. In-
38. A. Gopnik et al., Changes in cognitive flexibility and hypothesis search across human
tell. Neurosci. 2011, 156869 (2011).
life history from childhood to adolescence to adulthood. Proc. Natl. Acad. Sci. U.S.A.
51. D. Guthrie, J. S. Buchwald, Significance testing of difference potentials. Psychophys-
114, 7892–7899 (2017).

PSYCHOLOGICAL AND
iology 28, 240–244 (1991).

COGNITIVE SCIENCES
39. C. G. Lucas, S. Bridgers, T. L. Griffiths, A. Gopnik, When children are better (or at least 52. E. Maris, R. Oostenveld, Nonparametric statistical testing of EEG- and MEG-data.
more open-minded) learners than adults: Developmental differences in learning the J. Neurosci. Methods 164, 177–190 (2007).
forms of causal relationships. Cognition 131, 284–299 (2014). 53. J. Kam, 64-channel human scalp EEG from 24 subjects examining spontaneous
40. W. James, F. Burkhardt, F. Bowers, I. K. Skrupskelis, The Principles of Psychology thought using experience sampling. Collaborative Research in Computational Neu-
(Harvard University Press, Cambridge, MA, 1981). roscience. https://crcns.org/data-sets/methods/eeg-1. Deposited 31 December, 2020.
Downloaded at Bolivia: PNAS Sponsored on January 21, 2021

Kam et al. PNAS | 9 of 9


Distinct electrophysiological signatures of task-unrelated and dynamic thoughts https://doi.org/10.1073/pnas.2011796118

You might also like