Abstract

Brain activity continuously and spontaneously fluctuates during tasks of sustained attention. This spontaneous activity reflects the intrinsic dynamics of neurocognitive networks, which have been suggested to differentiate moments of externally directed task focus from episodes of mind wandering. However, the contribution of specific electrophysiological brain states and their millisecond dynamics to the experience of mind wandering is still unclear. In this study, we investigated the association between electroencephalogram microstate temporal dynamics and self-reported mind wandering. Thirty-six participants completed a sustained attention to response task in which they were asked to respond to frequently occurring upright faces (nontargets) and withhold responses to rare inverted faces (targets). Intermittently, experience sampling probes assessed whether participants were focused on the task or whether they were mind wandering (i.e., off-task). Broadband electroencephalography was recorded and segmented into a time series of brain electric microstates based on data-driven clustering of topographic voltage patterns. The strength, prevalence, and rate of occurrence of specific microstates differentiated on- versus off-task moments in the prestimulus epochs of trials preceding probes. Similar associations were also evident between microstates and variability in response times. Together, these findings demonstrate that distinct microstates and their millisecond dynamics are sensitive to the experience of mind wandering.

INTRODUCTION

It is a common experience for one's attention to wander away from the task at hand and toward internal mental events (Smallwood & Schooler, 2006, 2015). These episodes, referred to as mind wandering, are pervasive and frequently associated with poorer behavioral performance (Kane & McVay, 2012) and increased variability in behavioral responding during tasks of sustained attention (Zanesco, Denkova, Witkin, & Jha, 2020; Bastian & Sackur, 2013; Seli, Cheyne, & Smilek, 2013). The neurocognitive events corresponding with mind wandering are suggested to be multifaceted, dynamic, and reliant on a host of coordinated neural networks (Seli et al., 2018; Wang et al., 2018; Christoff, Irving, Fox, Spreng, & Andrews-Hanna, 2016; Andrews-Hanna, Smallwood, & Spreng, 2014).

Studies have demonstrated consistent patterns of activation in so-called task-positive fMRI derived networks of functionally connected regions, and corresponding deactivation in the default mode network (DMN), when attention is oriented toward the external environment (Raichle, 2015). This anticorrelated pattern of activity has led researchers to suggest that increases in DMN activity during episodes of mind wandering correspond with shifts away from externally focused to internally focused states (Turnbull et al., 2020; Andrews-Hanna, Irving, Fox, Spreng, & Christoff, 2018; Kucyi, 2018; Christoff et al., 2016; Fox, Spreng, Ellamil, Andrews-Hanna, & Christoff, 2015; Andrews-Hanna et al., 2014). Relatedly, key nodes comprising the DMN fall on one end of a macroscale functional gradient that is maximally distant from regions serving sensory–motor functions, suggesting that a principal axis of cortical organization is structured to insulate associative and internal stimulus-independent processes, such as mind wandering, from environmental sensory input (Margulies et al., 2016).

Increasingly, the dynamics of interacting networks have been investigated according to time-varying configurations of functional connectivity of distributed brain regions with fMRI (i.e., dynamic functional connectivity; Lurie et al., 2020), which demonstrate that specific functional brain states predominate temporarily before transitioning to other brain network configurations. Critical to this perspective is the notion that functional brain networks flexibly couple in a situation-dependent manner to enact cognitive functions (Turnbull et al., 2019, 2020; Mooneyham et al., 2017), making investigations of the temporal dynamics of interacting networks necessary to understand the neural bases of sustained attention and the wandering mind (Kucyi, 2018). Although studies of dynamic functional connectivity have begun to elucidate some of the temporal configurations of brain network states involved in mind wandering (Denkova, Nomi, Uddin, & Jha, 2019; Maillet, Beaty, Kucyi, & Schacter, 2019; Mooneyham et al., 2017; Kucyi & Davis, 2014), they are also limited in that dynamics of the fMRI signal can only be observed on the order of seconds to minutes given the slow time course of hemodynamic fluctuations.

Instead, the coordinated activity of the brain can be characterized with millisecond temporal precision using scalp-recorded electroencephalography (EEG). Electrophysiological studies of EEG oscillations have primarily demonstrated greater amplitude of oscillations in the alpha (8–12 Hz) frequency during moments of mind wandering compared to on-task states (Arnau et al., 2020; Wamsley, & Summer, 2020; Compton, Gearinger, & Wild, 2019; Jin, Borst, & van Vugt, 2019; Baldwin et al., 2017; Macdonald, Mathan, & Yeung, 2011; but see Broadway, Franklin, & Schooler, 2015; Braboszcz & Delorme, 2011). Importantly, however, measures of alpha power (or power in other frequency bands) from spontaneous EEG are limited in two critical ways: (1) These measures typically cannot distinguish between distinct sets of brain generators that give rise to oscillations in the scalp electric field, (2) nor do they account for the large-scale dynamics of these brain generators as the brain transitions through distinct brain states over time. These limitations obfuscate determining whether specific functional brain networks and their anticorrelated activity patterns during mind wandering also manifest in millisecond electrophysiological dynamics using noninvasive methodologies, in line with fMRI-based studies (Andrews-Hanna et al., 2014).

An alternative approach that is able to distinguish between the activity of distinct functional brain networks and characterize their ongoing dynamics involves the segmentation of the EEG time series into brain electric microstates (Michel & Koenig, 2018). Microstates refer to periods of quasi-stability in the topographic voltage configuration of the scalp electric field that result from the global synchronized activity of coordinated neuronal populations. Each microstate reflects a characteristic topographic configuration that briefly predominates in the voltage topography of the broadband EEG for roughly 40–120 msec before transitioning rapidly to other configurations. In turn, each distinct topographic configuration reflects a different functional brain state based on the coordinated activity of a different neuronal network. Any transition in the topographic configuration of the strength-normalized scalp electric field implies through physical laws a change in the distribution of active neural generators in the brain (Murray, Brunet, & Michel, 2008; Vaughan, 1982). As such, the fluctuating topographic configuration of the EEG scalp electric field can distinguish between global functional brain states resulting from predominating electrophysiological brain networks and their alternating temporal dynamics (Michel & Koenig, 2018; Murray et al., 2008).

Only a limited number of distinct microstate configurations are needed to explain most of the topographic variance of voltage maps in spontaneous EEG (Michel & Koenig, 2018). Indeed, studies have consistently identified four to seven unique microstate configurations through clustering of topographic voltage patterns, and configurations appear highly similar across studies, suggesting that microstates reflect the spontaneous activity of a common set of electrophysiological functional brain networks. Moreover, recent studies report overlap between the electrical brain generators of microstates and nodes comprising fMRI-derived functional networks, including the DMN (Bréchet et al., 2019; Custo et al., 2017; Yuan, Zotev, Phillips, Drevets, & Bodurka, 2012; Britz, Van De Ville, & Michel, 2010). For example, several microstates appear to share a set of brain generators that overlap with the main hubs of these networks (e.g., anterior and posterior cingulate cortices, superior frontal cortex, dorsal superior pFC, and insula) as well as generators that differentiate specific microstates (Michel & Koenig, 2018; Custo et al., 2017).

Examination of microstates involves categorizing the ongoing EEG time series according to these distinct topographic configurations to express the multichannel EEG as a sequence of alternating microstate occurrences. Measures of their coordinated activation strength and temporal dynamics can then be obtained. Although the cognitive relevance of microstates and their dynamics has long been of interest to researchers (Lehmann, Strik, Henggeler, Koenig, & Koukkou, 1998; Lehmann, Ozaki, & Pal, 1987), studies have primarily investigated their associations with cognition during resting-state EEG recordings. For example, differences between microstate dynamics at rest have been linked to changes in perceptual states when eyes are open or closed (Zanesco, King, Skwara, & Saron, 2020; Seitzman et al., 2017), states of alertness and arousal (Brodbeck et al., 2012), individual differences associated with cognitive abilities (Zanesco, King, et al., 2020), and psychological dysfunction (see, for a review, Rieger, Diaz Hernandez, Baenninger, & Koenig, 2016). Only a few studies have examined microstates and their associations with perception and ongoing cognitive task performance (e.g., Britz, Diaz Hernandez, Ro, & Michel, 2014; Britz, Pitts, & Michel, 2011; Britz, Landis, & Michel, 2009).

Herein, we apply a microstate-based approach to examine the electrophysiological dynamics of functional brain networks during episodes of mind wandering. We investigated whether distinct microstates and their millisecond dynamics differentiate moments of task-related focus (i.e., self-reports of being on-task) from mind wandering (i.e., self-reports of being off-task). To examine this topic, the current study relied on EEG data derived from a recent study by Denkova, Brudner, Zayan, Dunn, and Jha (2018). This study employed an experience sampling methodology, probing participants' reports of being on-task versus off-task during a sustained attention task to face stimuli. Face-sensitive event-related potentials (ERPs) were examined during the presentation of faces to compare activity during on- versus off-task episodes. Denkova et al. (2018) found that the amplitude of early face-sensitive perceptual ERP components (i.e., N170) was reduced when individuals report that they had been off-task relative to when they reported being on-task. These results are in line with prior work suggesting that mind wandering results in perceptual decoupling (Franklin, Mrazek, Broadway, & Schooler, 2013; Kam & Handy, 2013; Smallwood, 2013), in which the perceptual processing of external stimuli appears attenuated in the moments preceding experience sampling reports of mind wandering (Denkova et al., 2018; Baird, Smallwood, Lutz, & Schooler, 2014; Kam, Nagamatsu, & Handy, 2014; Kam & Handy, 2013; Kam et al., 2011; Smallwood, Beach, Schooler, & Handy, 2008).

This study extends the analyses in this prior investigation to examine prestimulus EEG microstate dynamics parsed according to participants' self-reported on-task versus off-task attentional states. Data-driven clustering identified global topographic microstates from the broadband EEG. Microstates were then examined in the prestimulus epochs of trials preceding experience sampling probes to estimate the activation strength of globally synchronized brain networks and their fine-grained temporal dynamics. Guided by fMRI studies demonstrating the contribution of anticorrelated brain networks to task-related focus and episodes of mind wandering (e.g., Turnbull et al., 2020; Fox et al., 2015), we examined whether distinct microstates show antagonistic differences in their strength and prevalence in the moments when individuals felt focused on the task (i.e., on-task reports) compared to during episodes of mind wandering (i.e., off-task reports). Furthermore, we expected these same microstates to be associated with behavioral fluctuations in response times (RTs) in a manner consistent with observed differences between on- and off-task states and in line with studies using RT variability as a behavioral indicator of attentional fluctuations and mind wandering (Zanesco, Denkova, et al., 2020; Bastian & Sackur, 2013; Seli, Cheyne, et al., 2013).

METHODS

Participants

Thirty-six undergraduate students (18 women, Mage = 18.83 years, SDage = 1.28, age range = 18–25 years) participated in this study. No participants reported a history of neurological disorders or head injury with loss of consciousness, and all had normal or corrected-to-normal vision. The study was approved by the institutional review board of the University of Miami, and participants provided written informed consent and received course credit for their participation. Data from two participants were excluded from analyses: one for incomplete data and one for poor performance (more than 4 SDs below the group accuracy mean).

Procedure

Participants sat approximately 28 in. away from a 23.5-in. LED computer monitor to complete the task in a sound-attenuating booth. E-Prime 2.0 software (Psychology Software Tools Inc.) was used for stimulus presentation and recording of behavioral responses. Participants were given detailed instructions before beginning the task and completed a 123-trial practice block with feedback to ensure that they understood the task instructions. Participants then completed three experimental blocks of the task. Each block lasted approximately 13 min with a quick break of about 2 min between blocks. Additional details regarding study procedures are reported in Denkova et al. (2018).

Sustained Attention to Response Task with Faces

Participants completed an adapted sustained attention to response task (SART; Robertson, Manly, Andrade, Baddeley, & Yiend, 1997), which utilized faces as stimuli. The face SART (F-SART) therefore consisted of a stream of successive face stimuli presented in the center of a white screen. Each face was displayed for 500 msec and followed by a fixation cross of variable duration (M = 1501 msec, SD = 117, range = 1300–1700 msec). Participants were instructed to respond via button press to frequently occurring upright faces (nontargets) and withhold their response to infrequently occurring upside-down faces (targets). They used the number pad of a computer keyboard to make their responses. Responses were recorded while the face was displayed as well as during the fixation cross after face offset. See Figure 1 for a schematic depicting several trials of the F-SART.

Figure 1. 

Schematic representation of F-SART trials. Participants were instructed to respond via button press to frequently occurring upright faces (nontargets) and withhold responses to infrequently occurring inverted faces (targets). They were also instructed to answer two experience sampling probe questions that intermittently appeared: the first question assessing if their attention was “on-task” or “off-task” (Probe 1) and the second question assessing their confidence in their Probe 1 response on a 3-point scale (Probe 2). At least six nontarget trials always preceded probes. Prestimulus epochs of these six trials and the prestimulus epoch of Probe 1 were included in EEG analyses.

Figure 1. 

Schematic representation of F-SART trials. Participants were instructed to respond via button press to frequently occurring upright faces (nontargets) and withhold responses to infrequently occurring inverted faces (targets). They were also instructed to answer two experience sampling probe questions that intermittently appeared: the first question assessing if their attention was “on-task” or “off-task” (Probe 1) and the second question assessing their confidence in their Probe 1 response on a 3-point scale (Probe 2). At least six nontarget trials always preceded probes. Prestimulus epochs of these six trials and the prestimulus epoch of Probe 1 were included in EEG analyses.

In total, there were 923 task trials, including the presentation of 833 nontargets, 45 targets, and 45 experience sampling probes. Target and probe presentation were pseudorandomized with the restriction that there were at least six consecutive nontarget trials before each target or probe trial. This allowed investigation of prestimulus brain dynamics in the seconds preceding the six nontargets occurring before the presentation of probes as a function of participants' responses to experience sampling questions (i.e., on-task vs. off-task reports). The use of six nontarget trials preceding probes is based on studies estimating the duration of mind wandering episodes (Bastian & Sackur, 2013) and the assumption that an episode of mind wandering should be ongoing in the moments directly preceding an off-task probe report. In addition, this range of trials balances the need to obtain enough data linked to each mind wandering episode without unduly sacrificing the temporal precision of the window and, as such, has been used in prior ERP studies of mind wandering (e.g., Denkova et al., 2018; Baird et al., 2014; Kam et al., 2011; Smallwood et al., 2008). Targets were not presented in the six trials preceding probes because they are presumed to disrupt any ongoing episodes of mind wandering by reengaging participants with the goals of the task (cf. Seli et al., 2018; Seli, Carriere, Levene, & Smilek, 2013; Smallwood et al., 2004).

Two experience sampling probe questions asked participants about their subjective experiences during task performance. The first question (Probe 1) asked “Where was your attention focused just before the probe?”, and participants responded that they were either (1) “on-task” or (2) “off-task.” Before the task, it was explained to participants that “on-task” referred to those instances when their attention was oriented completely and uniquely toward task performance, whereas “off-task” was explained as those instances when their attention was on something unrelated to the task. The second question (Probe 2) asked participants “How confident are you in your previous answer?” to the prior probe question. Participants rated their confidence on a 3-point scale: 1 = low, 2 = medium, and 3 = high. Probe questions were presented in succession one after the other (each for 4000 msec), and responses were recorded throughout the entire duration of probe presentation. In some cases, participants made several successive responses to the probe questions; only the first correct response option was considered here (0.9% of all probes had conflict between multiple responses). A fixation cross of variable duration (M = 1487 msec, SD = 115, range = 1313–1693 msec) followed the probes. On the basis of prior studies suggesting that a 1-min interval between probes can increase the percentage of mind wandering reports during a task (Seli, Carriere, et al., 2013), the average interval between probe trials was ∼50 sec (range = 12–83 sec).

EEG Data Acquisition and Processing

EEG data were recorded throughout the three experimental blocks of the F-SART from 64 Ag/AgCl electrodes located according to the 10–20 International System (American Electroencephalographic Society, 1991) using the BioSemi ActiveTwo system. In addition to scalp electrodes, three electrodes were placed on the outer canthi and below the left eye to record horizontal and vertical electro-oculograms, and two additional electrodes were placed on the left and right mastoids. Data were acquired at a sampling rate of 256 Hz, bandpass filtered online at 0.16–100 Hz, and subsequently processed offline using the free Cartool software toolbox Version 3.7 (Brunet, Murray, & Michel, 2011). Recordings were referenced to a common average of all 64 scalp electrodes and were bandpass filtered between 1 and 40 Hz (zero-phase, 12 dB/octave). The EEG were screened for poor signal quality, and channels with intermittent connectivity or periods of extreme amplitude were interpolated using 3-D spherical spline interpolation. The EEG were then downsampled to 128 Hz using cascaded integrator–comb and high-pass finite impulse response filters in Cartool.

For each participant, EEG data were next segmented into 1000-msec prestimulus epochs immediately preceding the presentation of each mind wandering probe (45 probe epochs) and for the six nontarget trials preceding each mind wandering probe (270 nontarget epochs). As such, prestimulus epochs began 800–1200 msec after the presentation of the previous trial stimulus to minimize stimulus and response evoked activity in the epoch.1 Epochs were included regardless of accuracy (i.e., correct and incorrect nontargets). The 315 epochs per participant were visually inspected for artifacts or excessive noise and were removed if ocular movements and blinks occurred within the window. Epochs were also removed if they contained artifacts from any scalp channels with voltage amplitude greater than ±80 μV. Participants had 213.9 (SD = 43.8, range = 116–274) epochs of EEG on average that were included in analyses. Finally, the EEG were spatially smoothed using the spatial interseptile weighted mean to reduce the influence of outliers in the electrode montage during topographic clustering (see Michel & Brunet, 2019).

Topographic Segmentation and Microstate Estimation

We conducted topographic segmentation of each of the 34 participant sets of contiguous epochs to identify periods of quasi-stability in voltage configurations of the scalp electric field. An adapted k-means clustering method was used to determine the optimal number of clusters (k) that account for the greatest global explained variance (GEV) in the voltage time series based on the smallest number of representative topographical maps (see Michel, Koenig, & Brandeis, 2009). Clustering was implemented in Cartool (Brunet et al., 2011) and occurred in two stages: clustering of maps from prestimulus epochs for individuals and a second round of clustering of subject-level centroids to produce global clusters. A schematic illustrating the major steps of this method is depicted in Figure 2.

Figure 2. 

Schematic illustrating clustering and microstate labeling of 64-channel EEG in the prestimulus epochs of trials preceding experience sampling probes. (A) Voltage maps at local maxima of GFP are identified from the EEG time series of epochs. Maps are shown as 2-D isometric projections with nasion upward. k-Means clustering for each individual identifies the optimal number of subject-level topographic clusters of maps. (B) The centroids of clusters for each individual undergo a second k-means clustering to identify the optimal k global clusters from among all individuals. Five global clusters from centroids of maps were identified. One hundred sixty-nine cluster centroids derived from k-means clustering of 34 participant recordings are shown grouped by their global cluster membership. Each global topography (A–E) is the centroid of clusters of maps. (C) Epochs are continuously labeled according to the global cluster centroid that best correlates with the voltage map at EEG samples to reexpress the EEG time series as a sequence of microstates. Maps remain unlabeled if the correlation is low (<.5). Measures of strength and temporal dynamics are subsequently derived from the categorized time series of epochs. Measures are averaged for epochs preceding each probe.

Figure 2. 

Schematic illustrating clustering and microstate labeling of 64-channel EEG in the prestimulus epochs of trials preceding experience sampling probes. (A) Voltage maps at local maxima of GFP are identified from the EEG time series of epochs. Maps are shown as 2-D isometric projections with nasion upward. k-Means clustering for each individual identifies the optimal number of subject-level topographic clusters of maps. (B) The centroids of clusters for each individual undergo a second k-means clustering to identify the optimal k global clusters from among all individuals. Five global clusters from centroids of maps were identified. One hundred sixty-nine cluster centroids derived from k-means clustering of 34 participant recordings are shown grouped by their global cluster membership. Each global topography (A–E) is the centroid of clusters of maps. (C) Epochs are continuously labeled according to the global cluster centroid that best correlates with the voltage map at EEG samples to reexpress the EEG time series as a sequence of microstates. Maps remain unlabeled if the correlation is low (<.5). Measures of strength and temporal dynamics are subsequently derived from the categorized time series of epochs. Measures are averaged for epochs preceding each probe.

Clustering of Voltage Maps

Topographic voltage maps were first generated at local maxima (peaks) in the global field power (GFP; Skrandies, 1990) time series (see Figure 2A). This was done separately for each of the 34 sets of contiguous epochs from EEG recordings. Initial maps for clustering were generated at the GFP peaks (local maxima) because these reflect optimal moments of topographic stability (Zanesco, 2020). GFP is a reference-independent measure of voltage potential (μV) that quantifies the strength of the scalp electric field and is equivalent to the spatial standard deviation of voltage amplitude over the entire average-referenced electrode montage (Skrandies, 1990).

k-Means clustering proceeded as follows for each participant set of epochs. A subset of 1–12 maps (k = [1:12]) was randomly selected from the total set of voltage maps at GFP peaks to use as initial centroids for clustering. The spatial correlation between the k centroid maps and the remaining voltage maps was then computed. Spatial correlations are based only on the topographic similarity of voltage maps and not amplitude because maps are normalized based on their GFP (Michel et al., 2009). Voltage maps were assigned to clusters with which they had the highest spatial correlation, creating k clusters of maps. Maps were only assigned to a cluster if the spatial correlation with the centroid map exceeded .5. Spatial correlations are based on the relative topographical configuration between maps but not the polarity by correcting the sign of the spatial correlation coefficients (Michel et al., 2009).2 After all initial assignments, k new centroid maps were created by combining the constituent maps assigned to a given cluster. The process was repeated, such that each voltage map was compared to the recomputed centroids and assigned again based on the correlation criterion. This process continued iteratively until the GEV between the centroids and the maps converged to a limit.

This procedure was repeated 100 times for each value of k, with a new subset of k centroids selected for each iteration. After the 100 iterations, the k set of centroids with maximal GEV was identified. This was repeated for each level of k = [1:12]. For each participant set of epochs across all levels of k, the optimal number of k clusters from the maximal GEV centroids was determined using a metacriterion defined by seven independent optimization criteria (see Bréchet et al., 2019; Custo et al., 2017). Accordingly, k-means clustering revealed four to seven topographies (M = 4.97, SD = 0.94) representing the optimal number of k clusters for individual EEG recordings (totaling 169 topographies).

Clustering of Subject-level Centroids

In a second step, we next conducted k-means clustering on the optimal centroids identified through clustering of the subject-level voltage maps (see Figure 2B). This was done to identify the optimal global clusters that best explain the subject-level cluster centroids across all participants. A set of k = [1:15] maps were randomly selected from the set of subject-level topographies and used as random centroids for clustering. For each level of k, 200 iterations were conducted, until the GEV converged to a limit and the k centroids with the maximal GEV were selected. After all iterations, the optimal number of k global clusters was determined using the optimization metacriterion, resulting in a set of k global centroids that best represent the topographic configurations of all participants. The second round of k-means clustering identified five global clusters that together explained 87.51% of the GEV among the individual subject cluster centroids. These five global clusters were selected as the optimal number based on the optimization metacriterion and appeared to be a good representation of the most common topographic patterns observed among all participants (see Figure 2B). Topographies are shown aligned in polarity with members of each cluster. Figure 2B depicts the five global cluster centroids, which we named Microstates A–E, and the 169 individual subject cluster topographies grouped according to their global cluster membership.

Parameterization of the Microstate Time Series

The global centroids were then fit back to the original EEG epochs to derive time series sequences of microstates (see Figure 2C). All samples of each 1000-msec prestimulus epoch were categorized according to the highest spatial correlation between the sample-wise voltage maps and the centroids of global clusters. Each epoch was categorized independently of other epochs. EEG samples that had low spatial correlation (i.e., <.5) with global centroids were left unassigned. Polarity was again ignored during centroid assignment by correcting the sign of the spatial correlation coefficients. Temporal smoothing was applied to the continuous microstate sequence by ignoring microstate segments that were present for less than three consecutive samples (i.e., ≤23 msec in duration) and splitting the time points between the preceding and subsequent microstates in the time series.3

The topographies of the five global cluster centroids were then assigned to all 7274 prestimulus epochs for probe and nontarget trials preceding probes. Centroids were successfully assigned (i.e., spatial correlation > .5) to 86.67% (SD = 10.2%, range = 13–100%) of samples of the continuous EEG time series for prestimulus epochs on average. When fit to the 7274 prestimulus epochs, the five global microstate topographies in total explained 62.11% (SD = 7.5%, range = 29.3–88.1%) of the GEV on average. The GEV, mean GFP, mean duration, rate of occurrence, and percent time coverage for each microstate were subsequently derived for all prestimulus epochs. Microstate measures were then averaged for the set of consecutive prestimulus epochs preceding each of the 45 experience sampling probes (up to seven epochs per probe) to obtain a single estimate of the average microstate temporal dynamics corresponding to each probe. There were 44.38 (SD = 0.95, range = 42–45) probe-averaged sets of microstate measures for each participant on average. Table 1 reports descriptive statistics of the subject-average microstate measures.

Table 1. 
Descriptive Statistics for Measures of Mean Microstate Strength and Temporal Dynamics
MeasureABCDE
GEV 10.543 (2.871) 9.077 (2.398) 31.659 (5.837) 5.229 (2.003) 5.701 (2.099) 
GFP 4.598 (0.980) 4.470 (0.965) 5.740 (1.137) 4.156 (0.837) 4.211 (0.804) 
Duration 57.169 (4.546) 56.825 (4.443) 70.474 (8.840) 55.211 (4.147) 54.842 (4.552) 
Occurrence 2.836 (0.495) 2.682 (0.542) 4.117 (0.373) 1.779 (0.511) 1.866 (0.515) 
Coverage 17.414 (3.527) 16.264 (3.319) 31.861 (5.656) 10.288 (3.112) 10.796 (3.286) 
MeasureABCDE
GEV 10.543 (2.871) 9.077 (2.398) 31.659 (5.837) 5.229 (2.003) 5.701 (2.099) 
GFP 4.598 (0.980) 4.470 (0.965) 5.740 (1.137) 4.156 (0.837) 4.211 (0.804) 
Duration 57.169 (4.546) 56.825 (4.443) 70.474 (8.840) 55.211 (4.147) 54.842 (4.552) 
Occurrence 2.836 (0.495) 2.682 (0.542) 4.117 (0.373) 1.779 (0.511) 1.866 (0.515) 
Coverage 17.414 (3.527) 16.264 (3.319) 31.861 (5.656) 10.288 (3.112) 10.796 (3.286) 

Means and standard deviations for GEV, mean GFP (in μV), microstate duration (in msec), per-second rate (Hz) of microstate occurrence, and percent time coverage (%) are provided from participant (n = 34) averages of all probe-related epochs for each microstate configuration.

Several measures describing the strength and temporal dynamics of microstates were derived from the microstate time series of each prestimulus epoch. Figure 3 depicts an example of the calculation of these measures from one 1000-msec epoch. The “GEV” is the percentage of observed topographic variance explained by a particular global topographic centroid (i.e., microstate configuration) in the continuous EEG of prestimulus epochs. “Mean GFP” is the average GFP of microstates fit separately to the GFP peaks (local maxima) and reflects the maximal field strength and degree of synchronization among the neural generators contributing to the voltage maps for each microstate configuration. “Mean microstate duration” is the average duration (in milliseconds) of contiguous samples from the continuous prestimulus time series categorized according to a specific microstate configuration. “Frequency of occurrence” represents the average rate per second a given microstate occurs in the prestimulus time series. Finally, percent time “coverage” represents the percentage of time in each epoch categorized according to a specific microstate configuration. See Khanna, Pascual-Leone, Michel, and Farzan (2015) and Michel and Koenig (2018) for more detailed neurophysiological interpretation of these measures.

Figure 3. 

Schematic illustrating the calculation of measures from the microstate time series for a single 1000-msec epoch. GEV (GEV %), mean GFP of GFP peaks (μV), mean duration (in msec), occurrence rate (in times per second, Hz), and percent time coverage (%) are calculated for each microstate configuration. The time series of microstates is shown with the GFP overlaid on top and the duration and occurrence of Microstate C visualized above. Microstate C explained 54.6% of the GEV for this epoch, occurred six times, had a mean duration of 68.4 msec when it occurred, and covered 49.6% of the epoch. The mean GFP at the GFP peaks was 13.4 μV for Microstate C. For each probe, these measures are averaged with the other epochs preceding the probe to create a set of probe-averaged microstate measures.

Figure 3. 

Schematic illustrating the calculation of measures from the microstate time series for a single 1000-msec epoch. GEV (GEV %), mean GFP of GFP peaks (μV), mean duration (in msec), occurrence rate (in times per second, Hz), and percent time coverage (%) are calculated for each microstate configuration. The time series of microstates is shown with the GFP overlaid on top and the duration and occurrence of Microstate C visualized above. Microstate C explained 54.6% of the GEV for this epoch, occurred six times, had a mean duration of 68.4 msec when it occurred, and covered 49.6% of the epoch. The mean GFP at the GFP peaks was 13.4 μV for Microstate C. For each probe, these measures are averaged with the other epochs preceding the probe to create a set of probe-averaged microstate measures.

Analyses

To reiterate, the present analyses focused on the prestimulus epochs of the six nontarget faces preceding probes as well as the prestimulus epochs of the probes. This meant there were up to seven prestimulus epochs preceding each probe (i.e., a set of probe epochs). Each set of probe epochs were categorized into “on-task” or “off-task” according to participants' subjective responses to the first probe question (Probe 1). Microstate measures for each set of probe epochs were averaged to derive mean values representing the average behavior of microstates in the time preceding experience sampling probes. Some sets of probe-averaged epochs could not be categorized as on-task or off-task because participants did not respond to the probe question. These trials were removed from analyses of subjective probe responses, leaving 43.82 (SD = 1.29, range = 41–45) probe-averaged sets of microstate measures included in analyses of probe ratings for individuals on average.

Behavioral analyses focused on the intraindividual coefficient of variation (ICV) of RTs calculated from the six nontargets preceding probes. RT ICV was calculated as the standard deviation of RT divided by the mean RT of the six nontargets preceding each probe. Some sets of trials were excluded from analyses of RT ICV because there were too few correct responses to nontarget trials to calculate reliable estimates of ICV. As such, there were 44.18 (SD = 1, range = 42–45) probe-averaged sets of microstate measures for each participant on average for trials with corresponding RT ICV. Descriptive statistics for other behavioral measures are described in Denkova et al. (2018) for a subset of the present sample of participants.

We compared differences in five microstate measures (GEV, GFP, duration, occurrence, and coverage) between probe epochs categorized as on- or off-task using multilevel mixed effects models implemented with PROC Mixed in SAS 9.4. Categorical fixed effects of probe rating (0 = on-task and 1 = off-task) reflect differences between probe ratings, and random intercepts were included to represent between-person variability. Probe-averaged measures for each probe were included in analyses as observations nested within individuals. We further examined the linear association between prestimulus microstate measures and RT ICV of the trials preceding probes. Random slopes and covariances were added for models including RT ICV, but these parameters were removed when reporting simplified models without ICV. Model parameters were estimated using restricted maximum likelihood estimation, and degrees of freedom were approximated by dividing the residual degrees of freedom into between-person and within-person components. Type III F tests of fixed effects are reported for omnibus effects, and parameter estimates are given for simplified models of significant fixed effects. To account for increased Type 1 error because of our use of five dependent measures, we set the alpha of all Type III F tests in our mixed effects models to α = .010. We further set the alpha of all follow-up comparisons of model parameters to α = .005 to account for additional tests of effects for each separate microstate (e.g., comparisons between on- and off-task probes for each microstate) and comparisons of these effects to the intercept (i.e., interaction terms).

RESULTS

Probe-related Microstate Strength and Temporal Dynamics

Participants reported being off-task on 39.0% (SD = 28%, range = 0–98%) of probes on average. Probe-averaged measures were compared between probes reported to be on- or off-task for dependent measures of microstates. We report parameter estimates and mean comparisons from mixed effects models below and in Table 2. Model estimated means and 95% confidence intervals are provided for on- and off-task conditions in Table 3 for each measure.

Table 2. 
Parameter Estimates from Analyses of Off-task Reports and Probe-related Microstate Measures
ParameterEstimate (SE)
GEV %GFP (μV)Duration (msec)Occurrence (Hz)Coverage %
Fixed effects 
 Intercept 30.892 (0.213)*** 5.685 (0.159)*** 70.415 (0.820)*** 4.066 (0.052)*** 31.197 (0.273)*** 
 Microstate A −20.341 (0.270)*** −1.110 (0.032)*** −13.261 (0.567)*** −1.227 (0.041)*** −13.782 (0.321)*** 
 Microstate B −21.736 (0.270)*** −1.222 (0.032)*** −13.570 (0.567)*** −1.392 (0.041)*** −15.000 (0.321)*** 
 Microstate D −25.628 (0.270)*** −1.522 (0.032)*** −15.233 (0.570)*** −2.279 (0.041)*** −20.948 (0.321)*** 
 Microstate E −24.788 (0.270)*** −1.435 (0.032)*** −15.549 (0.569)*** −2.107 (0.041)*** −19.726 (0.321)*** 
 Off-task 1.957 (0.313)*** 0.145 (0.037)*** – 0.133 (0.048)** 1.705 (0.375)*** 
 Off-task × A −1.953 (0.432)*** −0.092 (0.051) – −0.134 (0.065)* −1.662 (0.514)** 
 Off-task × B −2.218 (0.432)*** −0.133 (0.051)** – −0.128 (0.065) −1.618 (0.514)** 
 Off-task × D −2.045 (0.432)*** −0.168 (0.051)** – −0.149 (0.065)* −1.630 (0.514)** 
 Off-task × E −2.952 (0.432)*** −0.249 (0.051)*** – −0.366 (0.065)*** −3.394 (0.514)*** 
  
Random effects 
 Random intercept 0.275 (0.105) 0.842 (0.208) 17.373 (4.548) 0.063 (0.016) 0.731 (0.232) 
 Residual variance 33.083 (0.544) 0.453 (0.007) 239.060 (3.942) 0.760 (0.012) 46.842 (0.770) 
Participants (n34 34 34 34 34 
Observations 7450 7450 7392 7450 7450 
ParameterEstimate (SE)
GEV %GFP (μV)Duration (msec)Occurrence (Hz)Coverage %
Fixed effects 
 Intercept 30.892 (0.213)*** 5.685 (0.159)*** 70.415 (0.820)*** 4.066 (0.052)*** 31.197 (0.273)*** 
 Microstate A −20.341 (0.270)*** −1.110 (0.032)*** −13.261 (0.567)*** −1.227 (0.041)*** −13.782 (0.321)*** 
 Microstate B −21.736 (0.270)*** −1.222 (0.032)*** −13.570 (0.567)*** −1.392 (0.041)*** −15.000 (0.321)*** 
 Microstate D −25.628 (0.270)*** −1.522 (0.032)*** −15.233 (0.570)*** −2.279 (0.041)*** −20.948 (0.321)*** 
 Microstate E −24.788 (0.270)*** −1.435 (0.032)*** −15.549 (0.569)*** −2.107 (0.041)*** −19.726 (0.321)*** 
 Off-task 1.957 (0.313)*** 0.145 (0.037)*** – 0.133 (0.048)** 1.705 (0.375)*** 
 Off-task × A −1.953 (0.432)*** −0.092 (0.051) – −0.134 (0.065)* −1.662 (0.514)** 
 Off-task × B −2.218 (0.432)*** −0.133 (0.051)** – −0.128 (0.065) −1.618 (0.514)** 
 Off-task × D −2.045 (0.432)*** −0.168 (0.051)** – −0.149 (0.065)* −1.630 (0.514)** 
 Off-task × E −2.952 (0.432)*** −0.249 (0.051)*** – −0.366 (0.065)*** −3.394 (0.514)*** 
  
Random effects 
 Random intercept 0.275 (0.105) 0.842 (0.208) 17.373 (4.548) 0.063 (0.016) 0.731 (0.232) 
 Residual variance 33.083 (0.544) 0.453 (0.007) 239.060 (3.942) 0.760 (0.012) 46.842 (0.770) 
Participants (n34 34 34 34 34 
Observations 7450 7450 7392 7450 7450 

Restricted maximum likelihood estimates are reported from mixed effects models of GEV (GEV %), GFP (in μV), microstate duration (in msec), per-second rate (Hz) of microstate occurrence, and percent time coverage (%) for fixed effects of microstate configuration (Microstates A–E) and probe rating (on-task and off-task). Microstate C and on-task probe ratings serve as the reference. The number of participants (N) and observations contributing to the analyses are provided. Standard errors are reported in parentheses.

*

p < .05.

**

p < .01.

***

p < .001.

Table 3. 
Model-estimated Means and 95% Confidence Intervals for Probe-related Microstate Measures
MeasureOn-taskOff-task
GEV 
 A 10.550 [10.13, 10.97] 10.555 [10.04, 11.07] 
 B 9.156 [8.73, 9.58] 8.895 [8.38, 9.41] 
 C 30.892 [30.47, 31.31] 32.849 [32.34, 33.36] 
 D 5.264 [4.84, 5.69] 5.176 [4.66, 5.69] 
 E 6.104 [5.68, 6.52] 5.109 [4.60, 5.62] 
  
GFP 
 A 4.575 [4.26, 4.89] 4.628 [4.31, 4.94] 
 B 4.463 [4.15, 4.78] 4.474 [4.16, 4.79] 
 C 5.685 [5.37, 6.00] 5.829 [5.51, 6.15] 
 D 4.163 [3.85, 4.48] 4.139 [3.82, 4.46] 
 E 4.250 [3.94, 4.56] 4.145 [3.83, 4.46] 
  
Duration 
 A 57.030 [55.28, 58.78] 57.347 [55.42, 59.27] 
 B 56.844 [55.09, 58.60] 56.846 [54.92, 58.77] 
 C 69.732 [67.98, 71.48] 71.482 [69.56, 73.41] 
 D 54.896 [53.14, 56.65] 55.632 [53.70, 57.57] 
 E 55.378 [53.62, 57.13] 54.059 [52.13, 55.99] 
  
Occurrence 
 A 2.839 [2.74, 2.94] 2.838 [2.73, 2.95] 
 B 2.674 [2.57, 2.78] 2.679 [2.57, 2.79] 
 C 4.066 [3.96, 4.17] 4.199 [4.09, 4.31] 
 D 1.787 [1.68, 1.89] 1.771 [1.66, 1.88] 
 E 1.959 [1.86, 2.06] 1.727 [1.61, 1.84] 
  
Coverage 
 A 17.415 [16.88, 17.96] 17.458 [16.82, 18.10] 
 B 16.196 [15.66, 16.74] 16.284 [15.64, 16.93] 
 C 31.197 [30.66, 31.74] 32.902 [32.26, 33.54] 
 D 10.248 [9.71, 10.79] 10.323 [9.68, 10.97] 
 E 11.470 [10.93, 12.01] 9.782 [9.14, 10.42] 
MeasureOn-taskOff-task
GEV 
 A 10.550 [10.13, 10.97] 10.555 [10.04, 11.07] 
 B 9.156 [8.73, 9.58] 8.895 [8.38, 9.41] 
 C 30.892 [30.47, 31.31] 32.849 [32.34, 33.36] 
 D 5.264 [4.84, 5.69] 5.176 [4.66, 5.69] 
 E 6.104 [5.68, 6.52] 5.109 [4.60, 5.62] 
  
GFP 
 A 4.575 [4.26, 4.89] 4.628 [4.31, 4.94] 
 B 4.463 [4.15, 4.78] 4.474 [4.16, 4.79] 
 C 5.685 [5.37, 6.00] 5.829 [5.51, 6.15] 
 D 4.163 [3.85, 4.48] 4.139 [3.82, 4.46] 
 E 4.250 [3.94, 4.56] 4.145 [3.83, 4.46] 
  
Duration 
 A 57.030 [55.28, 58.78] 57.347 [55.42, 59.27] 
 B 56.844 [55.09, 58.60] 56.846 [54.92, 58.77] 
 C 69.732 [67.98, 71.48] 71.482 [69.56, 73.41] 
 D 54.896 [53.14, 56.65] 55.632 [53.70, 57.57] 
 E 55.378 [53.62, 57.13] 54.059 [52.13, 55.99] 
  
Occurrence 
 A 2.839 [2.74, 2.94] 2.838 [2.73, 2.95] 
 B 2.674 [2.57, 2.78] 2.679 [2.57, 2.79] 
 C 4.066 [3.96, 4.17] 4.199 [4.09, 4.31] 
 D 1.787 [1.68, 1.89] 1.771 [1.66, 1.88] 
 E 1.959 [1.86, 2.06] 1.727 [1.61, 1.84] 
  
Coverage 
 A 17.415 [16.88, 17.96] 17.458 [16.82, 18.10] 
 B 16.196 [15.66, 16.74] 16.284 [15.64, 16.93] 
 C 31.197 [30.66, 31.74] 32.902 [32.26, 33.54] 
 D 10.248 [9.71, 10.79] 10.323 [9.68, 10.97] 
 E 11.470 [10.93, 12.01] 9.782 [9.14, 10.42] 

Model-estimated means and 95% confidence intervals are provided from mixed effects models of GEV (GEV %), GFP (GFP in μV), microstate duration (in msec), per-second rate (Hz) of microstate occurrence, and percent time coverage (%) for each microstate configuration (Microstates A–E).

GEV

We first compared the mean global explained topographic variance (GEV) of the global topographies fit to the EEG time series between on- and off-task epochs. This measure reflects the amount of topographic variance explained in the probe-averaged epochs by the five global microstate configurations. We observed a significant main effect of Microstate configuration, F(4, 132) = 5270.12, p < .001, no significant main effect of Probe rating, F(1, 30) = 0.65, p = .427, and a significant interaction between Microstate configuration and Probe rating, F(4, 120) = 12.92, p < .001.

As indicated in Table 2, Microstate Configuration C (set as the reference) explained significantly more variance in the prestimulus period compared to other microstate configurations (all ps < .001). Importantly, Microstate C explained significantly more GEV (b = 1.957%, p < .001, 95% CI [1.317, 2.597]) when participants reported they were off-task on trials compared to when they were on-task. Microstates A, B, D, and E were associated differently with probe ratings than Microstate C (all ps < .001; see Table 2 for interaction terms). Accordingly, Microstate E (b = −0.995%, p = .003, 95% CI [−1.635, −0.355]) explained less GEV when participants reported they were off-task compared to when they were on-task. In contrast, Microstates A (b = 0.004%, p = .989, 95% CI [−0.636, 0.644]), B (b = −0.260%, p = .413, 95% CI [−0.900, 0.380]), and D (b = −0.088%, p = .782, 95% CI [−0.728, 0.552]) did not significantly differ on trials that were off-task compared to on-task. Figure 4 depicts the model estimated means for on- and off-task trials for Microstates C and E.

Figure 4. 

Observed data are shown on the left for probe-averaged trials reported as “on-task” in dark blue or “off-task” in light blue for dependent measures of microstates. Model-estimated means are provided as black bars. The model-estimated difference (Δ) between on-task and off-task reports and 95% confidence intervals are shown on the right for each microstate.

Figure 4. 

Observed data are shown on the left for probe-averaged trials reported as “on-task” in dark blue or “off-task” in light blue for dependent measures of microstates. Model-estimated means are provided as black bars. The model-estimated difference (Δ) between on-task and off-task reports and 95% confidence intervals are shown on the right for each microstate.

Mean GFP

We next compared the mean GFP of microstates fit to the GFP peaks for each microstate configuration. This measure reflects the maximal electric field strength of microstates occurring in the probe epochs on average. Comparisons between microstates and probe ratings revealed a significant effect of Microstate configuration, F(4, 132) = 1341.02, p < .001, no significant effect of Probe rating, F(1, 30) = 0.69, p = .414, and a significant interaction, F(4, 120) = 6.67, p < .001. Instances of Microstate C (set as the reference) had significantly stronger GFP at local maxima compared to other microstate configurations (all ps < .001; see Table 2). Importantly, Microstate C had greater GFP (b = 0.145 μV, p < .001, 95% CI [0.068, 0.221]) on off-task trials compared to on-task trials (see Figure 4). The association between probe ratings and GFP significantly differed for Microstates D and E compared to Microstate C (ps < .002; see Table 2 for interaction terms), whereas Microstates A (p = .073) and B (p = .010) did not. Microstates A (b = 0.053 μV, p = .167, 95% CI [−0.023, 0.129]), B (b = 0.011 μV, p = .764, 95% CI [−0.065, 0.088]), D (b = −0.024 μV, p = .530, 95% CI [−0.100, 0.053]), and E (b = −0.105 μV, p = .009, 95% CI [−0.181, −0.028]) did not significantly differ for trials that were off-task compared to on-task.

Mean Duration

We next compared the mean duration of occurrences of microstates. This measure reflects the average amount of time a set of neural generators remains synchronously active for each occurrence of a particular microstate configuration. On average, for participants, each occurrence of a microstate lasted between 54.72 and 70.61 msec (see Table 1). Comparison between microstates and probe ratings revealed a significant effect of Microstate configuration, F(4, 132) = 258.44, p < .001, no significant effect of Probe rating, F(1, 30) = 0.45, p = .506, and no significant interaction, F(4, 120) = 1.84, p = .126. We therefore report parameter estimates from a simplified model with only the significant effect of Configuration. As indicated in Table 2, Microstate C (set as the reference) was significantly longer in duration compared to all of the other microstate configurations (all ps < .001).

Occurrence Rate

We next compared the mean occurrence rate of microstates. This measure reflects the tendency for a set of neural generators to be synchronously active over time. Microstates A, B, D, and E all occurred between one and three times per second on average in the prestimulus period preceding probes, whereas Configuration C occurred about four times per second (see Table 1). As with GEV, we observed a significant effect of Microstate configuration, F(4, 132) = 1702.01, p < .001, no significant effect of Probe rating, F(1, 30) = 0.79, p = .381, and a significant interaction between these effects, F(4, 120) = 8.12, p < .001, on the rate of occurrence of microstates.

Microstate C (set as the reference) occurred more frequently in the prestimulus period compared to other microstate configurations (all ps < .001; see Table 2). Microstate C did not occur significantly more frequently (b = 0.133 times per second, p = .010, 95% CI [0.035, 0.232]) when participants reported they were off-task compared to when they were on-task. However, the association between probe rating and occurrence rate significantly differed for Microstate E (p < .001), compared to Microstate C (see Table 2 for interaction terms), but not for Microstate A (p = .043), B (p = .053), or D (p = .024). As such, Microstate E occurred less frequently (b = −0.232 times per second, p < .001, 95% CI [−0.331, −0.134]) when participants reported they were off-task compared to when they were on-task (see Figure 4). Microstates A (b = −0.001, p = .989, 95% CI [−0.099, 0.098]), B (b = 0.005, p = .916, 95% CI [−0.093, 0.104]), and D (b = −0.016, p = .743, 95% CI [−0.115, 0.083]) did not differ for trials that were off-task compared to on-task.

Coverage

Finally, we compared the mean percent time coverage of microstates. This measure reflects the average proportion of epochs labeled according to each microstate configuration. As with GEV and occurrence rate, we observed a significant effect of Microstate configuration, F(4, 132) = 2365.50, p < .001, no significant effect of Probe rating, F(1, 30) = 0.06, p = .815, and a significant interaction between these effects, F(4, 120) = 10.91, p < .001, for the percentage of time covered by a particular microstate.

Microstate C (set as the reference) occupied more time in the prestimulus period compared to other microstate configurations (all ps < .001; see Table 2). Importantly, Microstate C also occupied more time (b = 1.705%, p < .001, 95% CI [0.939, 2.472]) when participants reported they were off-task compared to when they were on-task (see Figure 4). The association between probe rating and coverage significantly differed for the other microstates compared to Microstate C (all ps < .003; see Table 2 for interaction terms). Microstate E occupied less time (b = −1.689%, p < .001, 95% CI [−2.455, −0.923]) in the prestimulus period when participants were off-task compared to on-task (see Figure 4). Microstates A (b = 0.043%, p = .910, 95% CI [−0.723, 0.809]), B (b = 0.087%, p = .817, 95% CI [−0.679, 0.854]), and D (b = 0.075%, p = .843, 95% CI [−0.692, 0.841]) did not differ for trials that were off-task compared to on-task.

Probe-related RT ICV and Microstate Strength and Dynamics

We next examined associations between microstate measures and RT ICV calculated from the sets of trials preceding probes. RT ICV was 0.23 (SD = 0.089, range = 0.133–0.582), and mean RT was 334.78 msec (SD = 60.65, range = 271.0–596.6 msec) for sets of probe-averaged trials for participants on average. RTs were also more variable for off-task sets of trials compared to on-task trials (b = 0.039, p < .001, 95% CI [0.019, 0.060]), confirming trial-level associations between off-task reports and RT ICV. Mean RTs did not significantly differ between on-task and off-task reports (b = 3.715 msec, p = .595, 95% CI [−10.390, 17.821]). The association between microstate-dependent measures for probe-averaged prestimulus epochs and the corresponding RT ICV of those trials was evaluated below. We report parameter estimates and mean comparisons from models below and in Table 4.

Table 4. 
Parameter Estimates from Analyses of ICV and Probe-related Microstate Measures
ParameterEstimate (SE)
GEV %GFP (μV)Duration (msec)Occurrence (Hz)Coverage %
Fixed effects 
 Intercept 30.876 (0.248)*** 5.741 (0.159)*** 70.470 (0.821)*** 4.116 (0.049)*** 31.023 (0.329)*** 
 Microstate A −20.304 (0.336)*** −1.143 (0.025)*** −13.307 (0.568)*** −1.282 (0.032)*** −13.497 (0.398)*** 
 Microstate B −21.689 (0.336)*** −1.271 (0.025)*** −13.664 (0.568)*** −1.434 (0.032)*** −14.537 (0.398)*** 
 Microstate D −25.512 (0.336)*** −1.584 (0.025)*** −15.212 (0.570)*** −2.333 (0.032)*** −20.709 (0.398)*** 
 Microstate E −25.121 (0.336)*** −1.529 (0.025)*** −15.582 (0.570)*** −2.250 (0.032)*** −20.231 (0.398)*** 
 ICV 3.249 (0.841)*** – – – 3.681 (0.952)*** 
 ICV × A −3.480 (1.157)** – – – −4.177 (1.372)** 
 ICV × B −3.753 (1.157)** – – – −4.598 (1.372)*** 
 ICV × D −3.824 (1.157)** – – – −3.616 (1.372)** 
 ICV × E −3.501 (1.157)** – – – −3.551 (1.372)** 
  
Random effects 
 Random intercept 0.112 (0.283) 0.844 (0.208) 17.433 (4.564) 0.064 (0.017) 0.994 (0.367) 
 Random covariance 0.394 (0.855) – – – −0.561 (0.387) 
 Random slope – – – 
 Residual variance 33.546 (0.549) 0.461 (0.008) 241.840 (3.972) 0.760 (0.012) 47.181 (0.772) 
Participants (n34 34 34 34 34 
Observations 7510 7510 7454 7510 7510 
ParameterEstimate (SE)
GEV %GFP (μV)Duration (msec)Occurrence (Hz)Coverage %
Fixed effects 
 Intercept 30.876 (0.248)*** 5.741 (0.159)*** 70.470 (0.821)*** 4.116 (0.049)*** 31.023 (0.329)*** 
 Microstate A −20.304 (0.336)*** −1.143 (0.025)*** −13.307 (0.568)*** −1.282 (0.032)*** −13.497 (0.398)*** 
 Microstate B −21.689 (0.336)*** −1.271 (0.025)*** −13.664 (0.568)*** −1.434 (0.032)*** −14.537 (0.398)*** 
 Microstate D −25.512 (0.336)*** −1.584 (0.025)*** −15.212 (0.570)*** −2.333 (0.032)*** −20.709 (0.398)*** 
 Microstate E −25.121 (0.336)*** −1.529 (0.025)*** −15.582 (0.570)*** −2.250 (0.032)*** −20.231 (0.398)*** 
 ICV 3.249 (0.841)*** – – – 3.681 (0.952)*** 
 ICV × A −3.480 (1.157)** – – – −4.177 (1.372)** 
 ICV × B −3.753 (1.157)** – – – −4.598 (1.372)*** 
 ICV × D −3.824 (1.157)** – – – −3.616 (1.372)** 
 ICV × E −3.501 (1.157)** – – – −3.551 (1.372)** 
  
Random effects 
 Random intercept 0.112 (0.283) 0.844 (0.208) 17.433 (4.564) 0.064 (0.017) 0.994 (0.367) 
 Random covariance 0.394 (0.855) – – – −0.561 (0.387) 
 Random slope – – – 
 Residual variance 33.546 (0.549) 0.461 (0.008) 241.840 (3.972) 0.760 (0.012) 47.181 (0.772) 
Participants (n34 34 34 34 34 
Observations 7510 7510 7454 7510 7510 

Restricted maximum likelihood estimates are reported from mixed effects models of GEV (GEV %), GFP (GFP in μV), microstate duration (in msec), per-second rate (Hz) of microstate occurrence, and percent time coverage (%) for fixed effects of microstate configuration (Microstates A–E) and RT variability (ICV). Microstate C and ICV = 0 serve as the reference. The number of participants (N) and observations contributing to the analyses are provided. Standard errors are reported in parentheses.

**

p < .01.

***

p < .001.

GEV

For associations between RT ICV and microstate GEV, we observed a significant main effect of Microstate configuration, F(4, 132) = 1991.16, p < .001, no significant main effect of RT ICV, F(1, 7467) = 0.66, p = .416, and a significant interaction between Microstate configuration and ICV, F(4, 7467) = 3.99, p = .003. Microstate C explained significantly more GEV (b = 3.249%, p < .001, 95% CI [1.600, 4.897]) for each 1-unit increase in ICV for the nontarget trials preceding probes. Microstates A, B, D, and E had significantly different associations with ICV than Microstate C (all ps < .003; see Table 4 for interaction terms). Microstates A (b = −0.231%, p = .783, 95% CI [−1.880, 1.417]), B (b = −0.504%, p = .549, 95% CI [−2.153, 1.144]), D (b = −0.575%, p = .494, 95% CI [−2.223, 1.073]), and E (b = −0.253%, p = .764, 95% CI [−1.901, 1.396]) were not associated with RT ICV. Figure 5 depicts the model-estimated association between RT ICV and GEV for Microstate C.

Figure 5. 

Observed data and model-estimated associations are provided between RT variability (ICV) for trials preceding probes and dependent measures of microstate GEV (GEV %) and time coverage (%) for Microstate C. The estimated slope (b) from mixed effects models is provided for each variable. ***p < .001.

Figure 5. 

Observed data and model-estimated associations are provided between RT variability (ICV) for trials preceding probes and dependent measures of microstate GEV (GEV %) and time coverage (%) for Microstate C. The estimated slope (b) from mixed effects models is provided for each variable. ***p < .001.

Mean GFP

For associations between RT ICV and microstate GFP, we observed a significant effect of Microstate configuration, F(4, 132) = 545.18, p < .001, no significant effect of RT ICV, F(1, 7467) = 0.00, p = .951, and no significant interaction, F(4, 7467) = 0.41, p = .798. Thus, we found no significant associations between RT ICV and microstate GFP. Only main effects of microstate configuration are reported in Table 4.

Mean Duration

For associations between RT ICV and microstate duration, we observed a significant effect of Microstate configuration, F(4, 132) = 80.94, p < .001, no significant effect of RT ICV, F(1, 7411) = 0.00, p = .969, and no significant interaction, F(4, 7411) = 3.04, p = .016. Thus, we found no significant associations between RT ICV and microstate duration. Only main effects of Microstate configuration are reported in Table 4.

Occurrence Rate

For associations between RT ICV and microstate occurrence rate, we observed a significant effect of Microstate configuration, F(4, 132) = 685.22, p < .001, but no significant effect of RT ICV, F(1, 7467) = 0.23, p = .635, and no significant interaction, F(4, 7467) = 1.18, p = .319. Thus, we found no significant associations between ICV and microstate occurrence rate. Only main effects of Microstate configuration are reported in Table 4.

Coverage

For associations between RT ICV and microstate coverage, we observed a significant effect of Microstate configuration, F(4, 132) = 884.00, p < .001, no significant effect of RT ICV, F(1, 7467) = 1.59, p = .208, and a significant interaction, F(4, 7467) = 3.57, p = .007. Microstate C occupied more time (b = 3.681%, p < .001, 95% CI [1.815, 5.548]) for each 1-unit increase in ICV (see Figure 5). Microstates A and B had significantly different associations with ICV than Microstate C (all ps < .003; see Table 4 for the interaction terms). Microstates A (b = −0.496%, p = .603, 95% CI [−2.362, 1.371]), B (b = −0.917%, p = .336, 95% CI [−2.783, 0.950]), D (b = 0.066%, p = .945, 95% CI [−1.801, 1.932]), and E (b = 0.130%, p = .891, 95% CI [−1.736, 1.997]) were not associated with RT ICV.

DISCUSSION

We investigated whether the strength and temporal dynamics of distinct functional brain states were sensitive to the self-reported experience of mind wandering and correlated with trial-by-trial fluctuations in RT variability. Microstate characteristics preceding on- versus off-task moments demonstrated opposing associations for specific microstate configurations according to their activation strength, prevalence, and rate of occurrence. Importantly, Microstate C was greater in GFP, explained more topographic variance (GEV), and covered more of the time series in the prestimulus epochs of trials preceding off-task reports relative to on-task reports. In contrast, Microstate E demonstrated the opposite pattern of associations with self-reported mind wandering and was lower in GEV, occurrence rate, and percent time coverage for off-task reports relative to on-task reports. The overall prevalence (GEV and coverage) of Microstate C was also positively associated with RT variability (ICV). These findings support the notion that distinct electrophysiological functional brain states are associated with episodes of mind wandering and provide behaviorally relevant information about one's ongoing attentional state.

The opposing patterns of association between Microstates C and E and on- versus off-task reports are reminiscent of the supposed antagonistic interplay between anticorrelated networks observed in prior studies examining fMRI-based connectivity. Studies have commonly reported reciprocity between functional networks in a variety of task conditions involving externally oriented attention compared to internally focused states (Raichle, 2015). Studies of mind wandering suggest that specific patterns of functional connectivity between the default, salience, dorsal attention, and central executive networks underlie internally focused cognition and episodes of mind wandering (Turnbull et al., 2019; Andrews-Hanna et al., 2014, 2018; Christoff et al., 2016; Fox et al., 2015). Our present findings provide complementary support for the perspective that opposing patterns of temporal dynamics between distinct electrophysiological brain networks are associated with episodes of mind wandering, and evidence for their anticorrelated dynamics can be observed in the scalp electrical field occurring at the millisecond temporal scale.

Although source localization of the electrical brain sources of microstates was not conducted herein, it is intriguing to speculate about the potential links between fMRI-derived networks and the brain generators responsible for EEG microstates at the scalp. Indeed, the brain generators of microstates have been suggested to overlap to some degree with several large-scale networks (e.g., Bréchet et al., 2019; Custo et al., 2017; Yuan et al., 2012; Britz et al., 2010). Microstate C is thought to have unique electrical brain generators in the precuneus and posterior cingulate cortex, which serve as main hubs of the DMN (Custo et al., 2017). In contrast, Microstate E has been suggested to have unique brain generators in the dorsal anterior cingulate extending to the superior frontal gyrus, middle frontal gyrus, and insula (Custo et al., 2017). These generators overlap with regions of the midcingulo-insular network (Uddin, Yeo, & Spreng, 2019), which has been implicated in the maintenance of tonic alertness (Sadaghiani & D'Esposito, 2015) and goal-directed behavior (Dosenbach et al., 2006). The midcingulo-insular network has also been suggested to have a role in regulating the DMN and its involvement in dynamic switching between large-scale networks to facilitate attention and action in response to the external environment (Rajan et al., 2019; Goulden et al., 2014; Wen, Liu, Yao, & Ding, 2013). Greater strength and prevalence of Microstate C, and corresponding decreases in Microstate E, during episodes of mind wandering, are therefore consistent with prior research and the functional roles of several large-scale networks. Future studies ought to draw on these possible interpretations to make hypothesis-driven tests of these microstates.

Prior studies of microstates have also observed similar compensatory patterns between states of directed thinking and rest (Bréchet et al., 2019; Milz et al., 2016) and at rest when individuals sit quietly with eyes closed or eyes open (Zanesco, King, et al., 2020; Seitzman et al., 2017). Microstate C occupies more of the total EEG time series when individuals are asked to rest quietly with their eyes closed compared to rest with their eyes open, whereas other microstates, specifically Microstate E, occupy less of the time series during eyes-closed rest (Zanesco, King, et al., 2020). Other microstates were also more likely to transition to Microstate E during eyes-open rest when considering their temporal sequence, suggesting that Microstate E mediates between other functional brain states to a greater degree in this perceptual state (Zanesco, King, et al., 2020). These findings suggest a potential functional correspondence between pairs of specific microstates and the global modulation of perceptual input, in line with the notion that coordinated brain networks are organized along a functional gradient that serve to insulate more abstract and associative processes from sensory–motor functions (Margulies et al., 2016).

ERP studies have also demonstrated reductions in visual perceptual processing during episodes of self-reported mind wandering compared to being on-task (e.g., Denkova et al., 2018). Consistent with the notion of reduced perceptual processing for episodes of on-task focus versus mind wandering (Smallwood, 2013), differences in microstate temporal dynamics between on- and off-task states appear to reflect network dynamics underlying shifts from externally oriented attentional states toward more internally focused states. The magnitude of observed differences in microstate dynamics between on- and off-task episodes, herein, was roughly three times smaller than prior effects comparing eyes-closed and eyes-open resting conditions (e.g., a difference of 0.13-Hz occurrence rate compared to 0.34 Hz in Zanesco, King, et al., 2020). It is perhaps unsurprising that the effects reported here are smaller, because the eyes-closed condition involves direct visual sensory disengagement from the external environment.

Microstate dynamics were also behaviorally relevant in the moments preceding probes. The ICV in RT was greater on trials for which Microstate C explained more variance and occupied more of the EEG time series. Behavioral and neural investigations have generally observed associations between variability in RTs and one's attentional focus, demonstrating that greater variability is linked to poorer performance and greater amounts of mind wandering (Zanesco, Denkova, et al., 2020; Denkova et al., 2018; Bastian & Sackur, 2013; Seli, Cheyne, et al., 2013). Associations of microstate dynamics with RT ICV across trials are therefore consistent with associations between microstates and subjective on- and off-task probe reports. It will be important for future studies to more fully examine the links between these variables, perhaps by examining whether the association of microstates with RT variability is mediated through self-reported mind wandering.

It is important to note that trial-by-trial associations between microstates and experience sampling probes appeared small. Microstate C, for example, was predicted to explain about 2% more topographic variance and occupy 2% more of the time series on average for epochs preceding probes categorized as “off-task.” This modest effect is perhaps to be expected, given the considerable dynamics of the brain at multiple spatial and temporal scales as well as the challenges inherent in experience sampling methodologies. One interesting perspective is that even modest perturbation to the ongoing dynamics of specific electrophysiological brain states can dramatically shift the overall trajectory of states through which the brain traverses. This may be particularly true of changes to Microstate C, which likely serves as a strong attractor state for the overall dynamics of the brain because it occurs most frequently, covers more of the EEG time series, and is most likely to follow all other microstates in sequence (Zanesco, King, et al., 2020).

Our findings also help inform prior studies demonstrating that alpha oscillations are greater in amplitude during episodes of mind wandering (Arnau et al., 2020; Wamsley & Summer, 2020; Compton et al., 2019; Jin et al., 2019; Baldwin et al., 2017; Macdonald et al., 2011; but see Broadway et al., 2015; Braboszcz & Delorme, 2011). An overall increase in the amplitude of alpha oscillations during mind wandering might imply that the brain generators of all spontaneous electrophysiological networks are more synchronously active at this frequency on average. Our findings showed this not to be the case. Instead, we found support for associations between distinct microstates, differentiable in the scalp electric field, and episodes of mind wandering. We found that only Microstate C demonstrated an increase in the strength of synchronized activity (i.e., GFP) during mind wandering, and other microstates did not significantly differ between on-task and off-task states. We also found that differences in microstate strength were mirrored by differences in their overall temporal prevalence. These findings imply that the opposing dynamics of distinct electrophysiological brain networks are hidden to researchers examining only average oscillatory power in certain frequency bands at the scalp.

Spectral power estimates at individual electrode sites can therefore conflate the contribution of distinct brain generators to oscillations observed at the scalp. Although microstates are a broadband oscillatory phenomenon, they have long been considered the “alpha map series” because oscillations at the alpha frequency predominantly drive their dynamics (Lehmann et al., 1987; Lehmann, 1971). Microstates inherit their periodicity from the dominant EEG frequency (von Wegner, Tagliazucchi, & Laufs, 2017), and their global topographic configuration is mostly determined by the intracortical strength and distribution of alpha generators (Milz, Pascual-Marqui, Achermann, Kochi, & Faber, 2017). However, microstate segmentation is distinct from other methods because it accounts for dynamics of global topographic patterns in the broadband multichannel EEG. This is nontrivial because it allows for differentiation of distinct electrophysiological brain networks that contribute to momentary patterns of globally synchronized activity, as is clear from the temporal progression of microstates in the EEG time series.

The experience of mind wandering is challenging to investigate not only because it relies on the uncertain but necessary process of introspection but also because of the rapid pace at which spontaneous thoughts and momentary attentional states fluctuate from one moment to the next. Investigations that pair phenomenological reporting with analyses of the dynamics of microstates nevertheless hold promise for understanding the neural correlates of attentional lapses and spontaneous thought at the millisecond temporal scale (Varela, Thompson, & Rosch, 2017; Lutz & Thompson, 2003; Varela, 1996). In turn, such neurophenomenological investigations help elucidate the functional significance of spontaneous brain activity. It is critical that future studies continue to examine the functional significance of microstates through behavioral and phenomenological methodologies as well as the links between their dynamics and those of large-scale brain networks identified through fMRI. Machine learning may be a promising approach to identify features of microstate sequences predictive of attentional states (e.g., Sikka et al., 2020). This will require replication in larger sample sizes for training and validation of predictions. Nevertheless, this study suggests the clear relevance of prestimulus brain activity and microstate temporal dynamics for differentiating the contribution of distinct functional brain states to episodes of mind wandering at the millisecond temporal scale.

Author Contributions

Anthony Paul Zanesco: Conceptualization; Data curation; Formal analysis; Investigation; Methodology; Resources; Visualization; Writing - Original Draft; Writing - Review & Editing. Ekaterina Denkova: Conceptualization; Data curation; Formal analysis; Investigation; Methodology; Project administration; Resources; Writing - Review & Editing. Amishi P. Jha: Conceptualization; Data curation; Formal analysis; Funding acquisition; Investigation; Methodology; Project administration; Resources; Writing - Review & Editing.

Acknowledgments

We thank Emily Brudner, Joseph Dunn, and Kristen Zayan for their help in collecting these data. We utilized the freely available Cartool software toolbox (cartoolcommunity.unige.ch) programmed by Denis Brunet from the Functional Brain Mapping Laboratory, Geneva, Switzerland, and supported by the Center for Biomedical Imaging of Geneva and Lausanne.

Reprint requests should be sent to Anthony Zanesco, Department of Psychology, University of Miami, 5665 Ponce de Leon, Coral Gables, FL 33146, or via e-mail: apz13@miami.edu.

Notes

1. 

The analyzed window for prestimulus epochs were therefore well outside the mean RT of all nontarget trials for participants on average (M = 331.30 msec, SD = 50.34).

2. 

Maps with the same relative dipolar configuration but inverted polarity are combined because neuroelectric oscillations (rhythmic fluctuations of excitation and inhibition in neuronal ensembles; Lopes da Silva, 1991) in the same sets of neural generators lead to inversions in polarity of the topographic configuration of the scalp electric field (see Michel & Koenig, 2018, for further discussion). Inversions in polarity of successive voltage maps with the same relative configuration (i.e., voltage minima and maxima in the same location but flipped) can be seen in Figure 2A.

3. 

Temporal smoothing is applied so that only sections of the time series demonstrating a modicum of temporal quasi-stability are labeled, whereas more transient fluctuations and maps with unstable topographies at microstate transition boundaries or discontinuities because of artifacts are not labeled as distinct occurrences of microstates. The choice of time window (i.e., >23 msec) is in line with observations that single microstates predominate for roughly 40–120 msec and other recent studies that have used similar smoothing when fitting microstates directly to the continuous EEG time series (e.g., Zanesco, King, et al., 2020; Bréchet et al., 2019).

REFERENCES

REFERENCES
Andrews-Hanna
,
J. R.
,
Irving
Z. C.
,
Fox
,
K. C. R.
,
Spreng
,
R. N.
, &
Christoff
,
K.
(
2018
).
The neuroscience of spontaneous thought: An evolving interdisciplinary field
. In
K. C. R.
Fox
&
K.
Christoff
(Eds.),
The Oxford handbook of spontaneous thought: Mind-wandering, creativity, and dreaming
(pp.
143
164
).
New York
:
Oxford University Press
.
Andrews-Hanna
,
J. R.
,
Smallwood
,
J.
, &
Spreng
,
R. N.
(
2014
).
The default network and self-generated thought: Component processes, dynamic control, and clinical relevance
.
Annals of the New York Academy of Sciences
,
1316
,
29
52
.
Arnau
,
S.
,
Löffler
,
C.
,
Rummel
,
J.
,
Hagemann
,
D.
,
Wascher
,
E.
, &
Schubert
,
A. L.
(
2020
).
Inter-trial alpha power indicates mind wandering
.
Psychophysiology
,
57
,
e13581
.
Baird
,
B.
,
Smallwood
,
J.
,
Lutz
,
A.
, &
Schooler
,
J. W.
(
2014
).
The decoupled mind: Mind-wandering disrupts cortical phase-locking to perceptual events
.
Journal of Cognitive Neuroscience
,
26
,
2596
2607
.
Baldwin
,
C. L.
,
Roberts
,
D. M.
,
Barragan
,
D.
,
Lee
,
J. D.
,
Lerner
,
N.
, &
Higgins
,
J. S.
(
2017
).
Detecting and quantifying mind wandering during simulated driving
.
Frontiers in Human Neuroscience
,
11
,
406
.
Bastian
,
M.
, &
Sackur
,
J.
(
2013
).
Mind wandering at the fingertips: Automatic parsing of subjective states based on response time variability
.
Frontiers in Psychology
,
4
,
573
.
Braboszcz
,
C.
, &
Delorme
,
A.
(
2011
).
Lost in thoughts: Neural markers of low alertness during mind wandering
.
Neuroimage
,
54
,
3040
3047
.
Bréchet
,
L.
,
Brunet
,
D.
,
Birot
,
G.
,
Gruetter
,
R.
,
Michel
,
C. M.
, &
Jorge
,
J.
(
2019
).
Capturing the spatiotemporal dynamics of self-generated, task-initiated thoughts with EEG and fMRI
.
Neuroimage
,
194
,
82
92
.
Britz
,
J.
,
Diaz Hernandez
,
L.
,
Ro
,
T.
, &
Michel
,
C. M.
(
2014
).
EEG-microstate dependent emergence of perceptual awareness
.
Frontiers in Behavioral Neuroscience
,
8
,
163
.
Britz
,
J.
,
Landis
,
T.
, &
Michel
,
C. M.
(
2009
).
Right parietal brain activity precedes perceptual alternation of bistable stimuli
.
Cerebral Cortex
,
19
,
55
65
.
Britz
,
J.
,
Pitts
,
M. A.
, &
Michel
,
C. M.
(
2011
).
Right parietal brain activity precedes perceptual alternation during binocular rivalry
.
Human Brain Mapping
,
32
,
1432
1442
.
Britz
,
J.
,
Van De Ville
,
D.
, &
Michel
,
C. M.
(
2010
).
BOLD correlates of EEG topography reveal rapid resting-state network dynamics
.
Neuroimage
,
52
,
1162
1170
.
Broadway
,
J. M.
,
Franklin
,
M. S.
, &
Schooler
,
J. W.
(
2015
).
Early event-related brain potentials and hemispheric asymmetries reveal mind-wandering while reading and predict comprehension
.
Biological Psychology
,
107
,
31
43
.
Brodbeck
,
V.
,
Kuhn
,
A.
,
von Wegner
,
F.
,
Morzelewski
,
A.
,
Tagliazucchi
,
E.
,
Borisov
,
S.
, et al
(
2012
).
EEG microstates of wakefulness and NREM sleep
.
Neuroimage
,
62
,
2129
2139
.
Brunet
,
D.
,
Murray
,
M. M.
, &
Michel
,
C. M.
(
2011
).
Spatiotemporal analysis of multichannel EEG: CARTOOL
.
Computational Intelligence and Neuroscience
,
2011
,
813870
.
Christoff
,
K.
,
Irving
,
Z. C.
,
Fox
,
K. C. R.
,
Spreng
,
R. N.
, &
Andrews-Hanna
,
J. R.
(
2016
).
Mind-wandering as spontaneous thought: A dynamic framework
.
Nature Reviews Neuroscience
,
17
,
718
731
.
Compton
,
R. J.
,
Gearinger
,
D.
, &
Wild
,
H.
(
2019
).
The wandering mind oscillates: EEG alpha power is enhanced during moments of mind-wandering
.
Cognitive, Affective & Behavioral Neuroscience
,
19
,
1184
1191
.
Custo
,
A.
,
Van De Ville
,
D.
,
Wells
,
W. M.
,
Tomescu
,
M. I.
,
Brunet
,
D.
, &
Michel
,
C. M.
(
2017
).
Electroencephalographic resting-state networks: Source localization of microstates
.
Brain Connectivity
,
7
,
671
682
.
Denkova
,
E.
,
Brudner
,
E. G.
,
Zayan
,
K.
,
Dunn
,
J.
, &
Jha
,
A. P.
(
2018
).
Attenuated face processing during mind wandering
.
Journal of Cognitive Neuroscience
,
30
,
1691
1703
.
Denkova
,
E.
,
Nomi
,
J. S.
,
Uddin
,
L. Q.
, &
Jha
,
A. P.
(
2019
).
Dynamic brain network configurations during rest and an attention task with frequent occurrence of mind wandering
.
Human Brain Mapping
,
40
,
4564
4576
.
Dosenbach
,
N. U. F.
,
Visscher
,
K. M.
,
Palmer
,
E. D.
,
Miezin
,
F. M.
,
Wenger
,
K. K.
,
Kang
,
H. C.
, et al
(
2006
).
A core system for the implementation of task sets
.
Neuron
,
50
,
799
812
.
Fox
,
K. C. R.
,
Spreng
,
R. N.
,
Ellamil
,
M.
,
Andrews-Hanna
,
J. R.
, &
Christoff
,
K.
(
2015
).
The wandering brain: Meta-analysis of functional neuroimaging studies of mind-wandering and related spontaneous thought processes
.
Neuroimage
,
111
,
611
621
.
Franklin
,
M. S.
,
Mrazek
,
M. D.
,
Broadway
,
J. M.
, &
Schooler
,
J. W.
(
2013
).
Disentangling decoupling: Comment on Smallwood (2013)
.
Psychological Bulletin
,
139
,
536
541
.
Goulden
,
N.
,
Khusnulina
,
A.
,
Davis
,
N. J.
,
Bracewell
,
R. M.
,
Bokde
,
A. L.
,
McNulty
,
J. P.
, et al
(
2014
).
The salience network is responsible for switching between the default mode network and the central executive network: Replication from DCM
.
Neuroimage
,
99
,
180
190
.
Jin
,
C. Y.
,
Borst
,
J. P.
, &
van Vugt
,
M. K.
(
2019
).
Predicting task-general mind-wandering with EEG
.
Cognitive, Affective & Behavioral Neuroscience
,
19
,
1059
1073
.
Kam
,
J. W. Y.
,
Dao
,
E.
,
Farley
,
J.
,
Fitzpatrick
,
K.
,
Smallwood
,
J.
,
Schooler
,
J. W.
, et al
(
2011
).
Slow fluctuations in attentional control of sensory cortex
.
Journal of Cognitive Neuroscience
,
23
,
460
470
.
Kam
,
J. W. Y.
, &
Handy
,
T. C.
(
2013
).
The neurocognitive consequences of the wandering mind: A mechanistic account of sensory–motor decoupling
.
Frontiers in Psychology
,
4
,
725
.
Kam
,
J. W. Y.
,
Nagamatsu
,
L. S.
, &
Handy
,
T. C.
(
2014
).
Visual asymmetry revisited: Mind wandering preferentially disrupts processing in the left visual field
.
Brain and Cognition
,
92
,
32
38
.
Kane
,
M. J.
, &
McVay
,
J. C.
(
2012
).
What mind wandering reveals about executive-control abilities and failures
.
Current Directions in Psychological Science
,
21
,
348
354
.
Khanna
,
A.
,
Pascual-Leone
,
A.
,
Michel
,
C. M.
, &
Farzan
,
F.
(
2015
).
Microstates in resting-state EEG: Current status and future directions
.
Neuroscience & Biobehavioral Reviews
,
49
,
105
113
.
Kucyi
,
A.
(
2018
).
Just a thought: How mind-wandering is represented in dynamic brain connectivity
.
Neuroimage
,
180
,
505
514
.
Kucyi
,
A.
, &
Davis
,
K. D.
(
2014
).
Dynamic functional connectivity of the default mode network tracks daydreaming
.
Neuroimage
,
100
,
471
480
.
Lehmann
,
D.
(
1971
).
Multichannel topography of human alpha EEG fields
.
Electroencephalography and Clinical Neurophysiology
,
31
,
439
449
.
Lehmann
,
D.
,
Ozaki
,
H.
, &
Pal
,
I.
(
1987
).
EEG alpha map series: Brain micro-states by space-oriented adaptive segmentation
.
Electroencephalography and Clinical Neurophysiology
,
67
,
271
288
.
Lehmann
,
D.
,
Strik
,
W. M.
,
Henggeler
,
B.
,
Koenig
,
T.
, &
Koukkou
,
M.
(
1998
).
Brain electric microstates and momentary conscious mind states as building blocks of spontaneous thinking: I. Visual imagery and abstract thoughts
.
International Journal of Psychophysiology
,
29
,
1
11
.
Lopes da Silva
,
F.
(
1991
).
Neural mechanisms underlying brain waves: From neural membranes to networks
.
Electroencephalography and Clinical Neurophysiology
,
79
,
81
93
.
Lurie
,
D. J.
,
Kessler
,
D.
Bassett
,
D. S.
,
Betzel
,
R. F.
,
Breakspear
,
M.
,
Kheilholz
,
S.
, et al
(
2020
).
Questions and controversies in the study of time-varying functional connectivity in resting fMRI
.
Network Neuroscience
,
4
,
30
69
.
Lutz
,
A.
, &
Thompson
,
E.
(
2003
).
Neurophenomenology: Integrating subjective experience and brain dynamics in the neuroscience of consciousness
.
Journal of Consciousness Studies
,
10
,
31
52
.
Macdonald
,
J. S. P.
,
Mathan
,
S.
, &
Yeung
,
N.
(
2011
).
Trial-by-trial variations in subjective attentional state are reflected in ongoing prestimulus EEG alpha oscillations
.
Frontiers in Psychology
,
2
,
82
.
Maillet
,
D.
,
Beaty
,
R. E.
,
Kucyi
,
A.
, &
Schacter
,
D. L.
(
2019
).
Large-scale network interactions involved in dividing attention between the external environment and internal thoughts to pursue two distinct goals
.
Neuroimage
,
197
,
49
59
.
Margulies
,
D. S.
,
Ghosh
,
S. S.
,
Goulas
,
A.
,
Falkiewicz
,
M.
,
Huntenburg
,
J. M.
,
Langs
,
G.
, et al
(
2016
).
Situating the default-mode network along a principal gradient of macroscale cortical organization
.
Proceedings of the National Academy of Sciences, U.S.A.
,
113
,
12574
12579
.
Michel
,
C. M.
, &
Brunet
,
D.
(
2019
).
EEG source imaging: A practical review of the analysis steps
.
Frontiers in Neurology
,
10
,
325
.
Michel
,
C. M.
, &
Koenig
,
T.
(
2018
).
EEG microstates as a tool for studying the temporal dynamics of whole-brain neuronal networks: A review
.
Neuroimage
,
180
,
577
593
.
Michel
,
C. M.
,
Koenig
,
T.
, &
Brandeis
,
D.
(
2009
).
Electrical neuroimaging in the time domain
. In
C. M.
Michel
,
T.
Koenig
,
D.
Brandeis
,
L. R. R.
Gianotti
, &
J.
Wackermann
(Eds.),
Electrical neuroimaging
(pp.
111
144
).
Cambridge
:
Cambridge University Press
.
Milz
,
P.
,
Faber
,
P. L.
,
Lehmann
,
D.
,
Koenig
,
T.
,
Kochi
,
K.
, &
Pascual-Marqui
,
R. D.
(
2016
).
The functional significance of EEG microstates—Associations with modalities of thinking
.
Neuroimage
,
125
,
643
656
.
Milz
,
P.
,
Pascual-Marqui
,
R. D.
,
Achermann
,
P.
,
Kochi
,
K.
, &
Faber
,
P. L.
(
2017
).
The EEG microstate topography is predominantly determined by intracortical sources in the alpha band
.
Neuroimage
,
162
,
353
361
.
Mooneyham
,
B. W.
,
Mrazek
,
M. D.
,
Mrazek
,
A. J.
,
Mrazek
,
K. L.
,
Phillips
,
D. T.
, &
Schooler
,
J. W.
(
2017
).
States of mind: Characterizing the neural bases of focus and mind-wandering through dynamic functional connectivity
.
Journal of Cognitive Neuroscience
,
29
,
495
506
.
Murray
,
M. M.
,
Brunet
,
D.
, &
Michel
,
C. M.
(
2008
).
Topographic ERP analyses: A step-by-step tutorial review
.
Brain Topography
,
20
,
249
264
.
Raichle
,
M. E.
(
2015
).
The brain's default mode network
.
Annual Review of Neuroscience
,
38
,
433
447
.
Rajan
,
A.
,
Meyyappan
,
S.
,
Walker
,
H.
,
Samuel
,
I. B. H.
,
Hu
,
Z.
, &
Ding
,
M.
(
2019
).
Neural mechanisms of internal distraction suppression in visual attention
.
Cortex
,
117
,
77
88
.
Rieger
,
K.
,
Diaz Hernandez
,
L.
,
Baenninger
,
A.
, &
Koenig
,
T.
(
2016
).
15 years of microstate research in schizophrenia—Where are we? A meta-analysis
.
Frontiers in Psychiatry
,
7
,
22
.
Robertson
,
I. H.
,
Manly
,
T.
,
Andrade
,
J.
,
Baddeley
,
B. T.
, &
Yiend
,
J.
(
1997
).
‘Oops!’: Performance correlations of everyday attentional failures in traumatic brain injured and normal subjects
.
Neuropsychologia
,
35
,
747
758
.
Sadaghiani
,
S.
, &
D'Esposito
,
M.
(
2015
).
Functional characterization of the cingulo-opercular network in the maintenance of tonic alertness
.
Cerebral Cortex
,
25
,
2763
2773
.
Seitzman
,
B. A.
,
Abell
,
M.
,
Bartley
,
S. C.
,
Erickson
,
M. A.
,
Bolbecker
,
A. R.
, &
Hetrick
,
W. P.
(
2017
).
Cognitive manipulation of brain electric microstates
.
Neuroimage
,
146
,
533
543
.
Seli
,
P.
,
Carriere
,
J. S. A.
,
Levene
,
M.
, &
Smilek
,
D.
(
2013
).
How few and far between? Examining the effects of probe rate on self-reported mind wandering
.
Frontiers in Psychology
,
4
,
430
.
Seli
,
P.
,
Carriere
,
J. S. A.
,
Wammes
,
J. D.
,
Risko
,
E. F.
,
Schacter
,
D. L.
, &
Smilek
,
D.
(
2018
).
On the clock: Evidence for the rapid and strategic modulation of mind wandering
.
Psychological Science
,
29
,
1247
1256
.
Seli
,
P.
,
Cheyne
,
J. A.
, &
Smilek
,
D.
(
2013
).
Wandering minds and wavering rhythms: Linking mind wandering and behavioral variability
.
Journal of Experimental Psychology: Human Perception and Performance
,
39
,
1
5
.
Seli
,
P.
,
Kane
,
M. J.
,
Smallwood
,
J.
,
Schacter
,
D. L.
,
Maillet
,
D.
,
Schooler
,
J. W.
, et al
(
2018
).
Mind-wandering as a natural kind: A family-resemblances view
.
Trends in Cognitive Sciences
,
22
,
479
490
.
Sikka
,
A.
,
Jamalabadi
,
H.
,
Krylova
,
M.
,
Alizadeh
,
S.
,
van den Meer
,
J. N.
,
Danyeli
,
L.
, et al
(
2020
).
Investigating the temporal dynamics of electroencephalogram (EEG) microstates using recurrent neural networks
.
Human Brain Mapping
,
41
,
2334
2346
.
Skrandies
,
W.
(
1990
).
Global field power and topographic similarity
.
Brain Topography
,
3
,
137
141
.
Smallwood
,
J.
(
2013
).
Distinguishing how from why the mind wanders: A process-occurrence framework for self-generated mental activity
.
Psychological Bulletin
,
139
,
519
535
.
Smallwood
,
J.
,
Beach
,
E.
,
Schooler
,
J. W.
, &
Handy
,
T. C.
(
2008
).
Going AWOL in the brain: Mind wandering reduces cortical analysis of external events
.
Journal of Cognitive Neuroscience
,
20
,
458
469
.
Smallwood
,
J.
,
Davies
,
J. B.
,
Heim
,
D.
,
Finnigan
,
F.
,
Sudberry
,
M.
,
O'Connor
,
R.
, et al
(
2004
).
Subjective experience and the attentional lapse: Task engagement and disengagement during sustained attention
.
Consciousness and Cognition
,
13
,
657
690
.
Smallwood
,
J.
, &
Schooler
,
J. W.
(
2006
).
The restless mind
.
Psychological Bulletin
,
132
,
946
958
.
Smallwood
,
J.
, &
Schooler
,
J. W.
(
2015
).
The science of mind wandering: Empirically navigating the stream of consciousness
.
Annual Review of Psychology
,
66
,
487
518
.
Turnbull
,
A.
,
Karapanagiotidis
,
T.
,
Wang
,
H.-T.
,
Bernhardt
,
B. C.
,
Leech
,
R.
,
Margulies
,
D.
, et al
(
2020
).
Reductions in task positive neural systems occur with the passage of time and are associated with changes in ongoing thought
.
Scientific Reports
,
10
,
9912
.
Turnbull
,
A.
,
Wang
,
H.-T.
,
Schooler
,
J. W.
,
Jefferies
,
E.
,
Margulies
,
D. S.
, &
Smallwood
,
J.
(
2019
).
The ebb and flow of attention: Between-subject variation in intrinsic connectivity and cognition associated with the dynamics of ongoing experience
.
Neuroimage
,
185
,
286
299
.
Uddin
,
L. Q.
,
Yeo
,
B. T. T.
,
Spreng
,
R. N.
(
2019
).
Towards a universal taxonomy of macro-scale functional human brain networks
.
Brain Topography
,
32
,
926
942
.
Varela
,
F.
(
1996
).
Neurophenomenology: A methodological remedy for the hard problem
.
Journal of Consciousness Studies
,
3
,
330
349
.
Varela
,
F.
,
Thompson
,
E.
, &
Rosch
,
E.
(
2017
).
The embodied mind: Cognitive science and human experience
(2nd ed.).
Cambridge, MA
:
MIT Press
.
Vaughan
,
H. G.
, Jr.
(
1982
).
The neural origins of human event-related potentials
.
Annals of the New York Academy of Sciences
,
388
,
125
138
.
von Wegner
,
F.
,
Tagliazucchi
,
E.
, &
Laufs
,
H.
(
2017
).
Information-theoretical analysis of resting state EEG microstate sequences—Non-Markovianity, non-stationarity and periodicities
.
Neuroimage
,
158
,
99
111
.
Wamsley
,
E. J.
, &
Summer
,
T.
(
2020
).
Spontaneous entry into an “offline” state during wakefulness: A mechanism of memory consolidation?
Journal of Cognitive Neuroscience
,
32
,
1714
1734
.
Wang
,
H.-T.
,
Poerio
,
G.
,
Murphy
,
C.
,
Bzdok
,
D.
,
Jefferies
,
E.
, &
Smallwood
,
J.
(
2018
).
Dimensions of experience: Exploring the heterogeneity of the wandering mind
.
Psychological Science
,
29
,
56
71
.
Wen
,
X.
,
Liu
,
Y.
,
Yao
,
L.
, &
Ding
,
M.
(
2013
).
Top–down regulation of default mode activity in spatial visual attention
.
Journal of Neuroscience
,
33
,
6444
6453
.
Yuan
,
J.
,
Zotev
,
V.
,
Phillips
,
R.
,
Drevets
,
W. C.
, &
Bodurka
,
J.
(
2012
).
Spatiotemporal dynamics of the brain at rest—Exploring EEG microstates as electrophysiological signatures of BOLD resting state networks
.
Neuroimage
,
60
,
2062
2072
.
Zanesco
,
A. P.
(
2020
).
EEG electric field topography is stable during moments of high field strength
.
Brain Topography
,
33
,
450
460
.
Zanesco
,
A. P.
,
Denkova
,
E.
,
Witkin
,
J. E.
, &
Jha
,
A. P.
(
2020
).
Experience sampling of the degree of mind wandering distinguishes hidden attentional states
.
Cognition
,
205
,
104380
.
Zanesco
,
A. P.
,
King
,
B. G.
,
Skwara
,
A. C.
, &
Saron
,
C. D.
(
2020
).
Within and between-person correlates of the temporal dynamics of resting EEG microstates
.
Neuroimage
,
211
,
116631
.