Efficient, adaptive behavior relies on the ability to flexibly move between internally focused (IF) and externally focused (EF) attentional states. Despite evidence that IF cognitive processes such as event imagination comprise a significant amount of awake cognition, the consequences of internal absorption on the subsequent recruitment of brain networks during EF tasks are unknown. The present functional magnetic resonance imaging (fMRI) study employed a novel attentional state switching task. Subjects imagined positive and negative events (IF task) or performed a working memory task (EF task) before switching to a target detection (TD) task also requiring attention to external information, allowing for the investigation of neural functioning during external attention based on prior attentional state. There was a robust increase of activity in frontal, parietal, and temporal regions during TD when subjects were previously performing the EF compared with IF task, an effect that was most pronounced following negative IF. Additionally, dorsolateral prefrontal cortex was less negatively coupled with ventromedial prefrontal and posterior cingulate cortices during TD following IF compared with EF. These findings reveal the striking consequences for brain activity following immersion in an IF attentional state, which have strong implications for psychiatric disorders characterized by excessive internal focus.

Introduction

Imagine that you are sitting in a presentation attending to the words of the speaker, when you receive an email message informing you that a recent manuscript has been rejected. Suddenly, you have trouble focusing on the once engaging talk, as you remember the weekends spent working on the paper, imagine how you will have to revise it, and consider how the rejection reflects on your personal qualities as a researcher. Depending on how upset you feel, you may find it difficult or even impossible to disengage from these thoughts to focus back on the speaker's words. The ability to switch attention flexibly between an internally focused (IF), perceptually decoupled state and an externally focused (EF), perceptually driven state is critical for efficient functioning. Estimates suggest that IF cognitive processes such as event imagination, autobiographical memory, introspection, and mentalizing account for approximately 30% of all waking cognition (Kane et al. 2007; Smallwood et al. 2012). Although individuals spontaneously move between IF and EF states (Smallwood and Schooler 2006; Vanhaudenhuyse et al. 2011), these transitions can also be goal-directed, such as when attempting to consciously divert attention away from thoughts about a manuscript rejection to focus on a presentation. The capacity to disengage from IF cognition in a goal-directed manner is not only important for understanding the neural mechanisms of everyday cognitive functioning but is also highly relevant for psychiatric disorders characterized by excessive internal focus, such as obsessive-compulsive disorder, depression, and anxiety.

Both spontaneous and goal-directed IF cognition activate the default mode network (DMN), a large-scale system including medial frontal and parietal cortex, posterior inferior parietal cortex, anterior temporal cortex, parahippocampal gyrus, and hippocampus (Addis et al. 2007; Spreng et al. 2010; Stawarczyk et al. 2011). Whereas increases in DMN activity are associated with improved performance on IF tasks of episodic memory (Rugg and Vilberg 2013), decreases (or deactivations) occur during tasks requiring EF cognition (Shulman et al. 1997), likely reflecting the suspension of IF cognition when attention is directed to external information (Buckner et al. 2008; Andrews-Hanna et al. 2010). In contrast, EF cognition activates networks centered in lateral frontal and parietal cortex. A dorsal attention network (DAN) consisting of precentral gyrus, superior parietal cortex, and posterior temporal/occipital cortex has been linked to visuospatial attention and motor planning. A fronto-parietal “control” network (FPN) involved in acting on or manipulating information includes dorsal frontal regions, anterior insula, inferior parietal cortex, posterior temporal cortex, and the striatum (Corbetta and Shulman 2002; Vincent et al. 2008; Spreng et al. 2010). FPN positively couples with DMN or DAN based on whether the source of the information being manipulated is internal or external, respectively, whereas DMN and DAN exhibit an antagonistic relationship (Spreng et al. 2010; Gao and Lin 2012) possibly providing the neural basis for competition between internal and external attention.

Despite the prevalence of IF cognition, it is unknown whether immersion in an IF state has consequences on subsequent responses to external information, a critical question for understanding how the brain transitions between modes of processing. The current study examined network activation and functional connectivity during an EF task (target detection) based upon whether subjects previously performed an IF task (event imagination) or another EF task (spatial working memory). In addition, the impact of prior emotion on network activity was probed by comparing negative and positive events.

Materials and Methods

Subjects

Seventeen subjects were recruited to participate in the experiment using flyers and internet advertisements. Following an initial phone screen, subjects were assessed in person using the Mini International Neuropsychiatric Interview (MINI; Sheehan et al. 1998). All subjects had normal or corrected-to-normal vision, were free of psychiatric or neurological disorders, and were not taking any psychotropic medications. Three subjects were excluded for excessive movement (n = 1), anxiety during the session (n = 1), and incidental finding on scan (n = 1), leaving 14 subjects for analysis (9 women; 12 right-handed; mean age: 31.6 years, range: 19–46 years; mean education: 15.6 years, range: 13–20 years). All subjects provided written informed consent as approved by the IRB of the Icahn School of Medicine at Mount Sinai.

Task Overview

The attentional state switching task (see Fig. 1) employed a block fMRI design containing 48 sequences, each consisting of 3 phases. In the “first phase” of a sequence, subjects performed 1 of 3 “initial” tasks: internally focused event imagination (IF block, including both positive and negative events), externally focused spatial working memory (EF block), or rest. The duration of each initial task was 15 s on average, jittered in 1.5 s increments between 12 and 18 s. Following performance of 1 of these 3 tasks, subjects switched to an EF target detection task (TD block) for 15 s, which constituted the “second phase” of the sequence. All TD blocks were identical regardless of whether the IF, EF, or rest block preceded it. Following the TD block, the “third phase” of the sequence had subjects rate how easy it was for them to perform the initial task from the first phase (e.g., how easy it was for them to imagine the event, to perform the working memory task, or to rest, from very difficult to very easy) and their emotion during the initial task (from very negative to very positive) on 5-point Likert scales (rating block). The length of the rating block depended on how quickly subjects made their ratings, with a maximum of 4 s allowed per question. For rating blocks taking less than 8 s, the remaining time was added to an inter-trial interval where subjects fixated on a crosshair. Following the crosshair, a new sequence began.

Figure 1.

Attentional state switching task. In the task, subjects performed 1 of 3 tasks prior to switching to target detection (TD): (A) internally focused (IF) event imagination, (B) externally focused (EF) spatial working memory, and (C) rest. After TD, subjects rated the ease and emotion associated with IF, EF, or rest.

Figure 1.

Attentional state switching task. In the task, subjects performed 1 of 3 tasks prior to switching to target detection (TD): (A) internally focused (IF) event imagination, (B) externally focused (EF) spatial working memory, and (C) rest. After TD, subjects rated the ease and emotion associated with IF, EF, or rest.

Subjects were instructed to keep their eyes open and looking at the screen during the entire task (6 runs), which was confirmed by a camera focused on the eyes. The order of sequence presentation within each run was pseudorandom to break colinearity between sequences. Despite signal overlap due to the sluggish hemodynamic response between the first and second phases of the sequence (initial tasks and TD blocks), block durations of 15 s on average were optimized for the identification of unique brain activity attributable to IF, EF, rest, and TD blocks following IF (TD-IF), TD blocks following EF (TD-EF), and TD blocks following rest (TD-rest). Twenty-four IF, twelve EF, and twelve rest sequences were presented over 6 runs. Twice as many IF sequences were used in order to probe for effects of imagined event valence within IF (there were 12 positive and 12 negative event sequences).

Following performance of the task, subjects used 5-point Likert scales to rate their level of distraction/engagement with event imagination, the vividness of their visual imagery, and their person perspective for each IF block.

Internal Focus (IF) Block (First Phase)

During IF sequences, subjects imagined 1 of 24 different event scenarios during the IF block prior to switching to perform TD. Each IF block started with a cue sentence describing the event to be imagined (3 s), followed by the word “Imagine…” at the top of the screen for 12–18 s. Subjects were instructed to immerse themselves in the imagined scenario as much as possible and to simulate the events as if they were actually happening.

Development of Event Scenarios for IF Block

A total of 24 events to be imagined in the scanner were selected from a screening session where subjects rated the valence of 48 novel event scenarios (24 positive and 24 negative). These 48 scenarios covered a range of potential events and were selected from a larger pool of 80 scenarios. Based on pilot ratings of these 80 scenarios, 48 events were selected that were both moderately arousing (e.g., “Your boss tells you that everyone in your office is getting an extra day of vacation” and “Your train breaks down in the morning, and you end up being late for an important meeting”) and highly arousing (e.g., “You watch the evening lottery draw, and find out you have the winning ticket for 5 million dollars” and “On a trip to a foreign nation, the bus you are on is hijacked and you are held at gunpoint”). To control for personal variability in event appraisal, the 24 events rated the most positive (n = 12) and most negative (n = 12) for each individual subject during the screening session were selected for use in the fMRI task. In the scanner, a one-sentence description was provided of the event to be imagined, and positive and negative event descriptions were matched on sentence complexity (Szmrecsanyi 2004) for each subject.

External Focus (EF) Block (First Phase)

During EF sequences (n = 12), instead of imagining an event, subjects performed a 2-back visuospatial working memory task (EF block) prior to switching to perform TD. During the EF block, subjects tracked and remembered the location of a white circle on the screen while fixating on a center crosshair. The EF block was preceded by a 3-s instruction screen reminding subjects of the task, followed by the presentation of the circle, which moved between 8 locations surrounding the crosshair (each location shown for 1250 ms with 250-ms crosshair in-between) for 12–18 s. After the circle moved between 3 and 5 locations, it disappeared and subjects received the instruction “LOOK” at the center of the screen, at which time they moved their eyes to the area on the screen where the circle was located 2 screens back. Eye movements rather than button press responses were employed to match the behavioral output required for EF and IF blocks, and subjects were told that their eye movements were being tracked to obtain measures of accuracy. Although equipment malfunction made us unable to actually track eye movements, an in-scanner video camera confirmed that subjects' eyes were open and moving at the appropriate times during the block.

Rest Block (First Phase)

During rest sequences (n = 12), subjects simply rested while keeping their eyes on the screen prior to performing TD. Similar to other blocks, a 3-s instruction screen was followed by the word “Relax…” for 12–18 s at the top of the screen for the duration of the block. Much data indicate that rest periods are characterized by “mind wandering”, which has been shown to include a variety of IF cognitive experiences, including spontaneous event imagination, episodic memory, planning, and random thoughts about subjective emotional, cognitive, and somatic states (Ingvar 1985; Fransson 2006; Buckner and Vincent 2007; Andrews-Hanna 2012). To assess the presence of IF cognition during rest, subjects filled out a detailed questionnaire at the end of the experiment where they rated the amount of time spent on different activities during rest (1: none of the time, 5: all of the time). As expected, subjects reported spending time engaged in perceptually decoupled IF cognition during rest blocks, including thoughts about future (mean rating: 2.1), thoughts about the past (2.0), thoughts about their body (3.2), thoughts about their emotional or cognitive state (2.9), thoughts about another person (2.4), thoughts about the task scenarios (2.5), thoughts about other aspects of the task (3.6), and visual imagery (2.3). As such, comparisons of IF and EF sequences with rest sequences served to isolate activity related to goal-directed IF and EF cognition, respectively, over and above brain activity associated with spontaneous IF thought during rest.

Target Detection (TD) Block (Second Phase)

After performing the IF, EF, or rest blocks, subjects switched to perform the TD task. This task presented 15 sequential letters for a total of 15 s, during which time subjects pressed one button for the target letter “a” and another button for all other letters. The target letter was presented ∼30% of the time (4–5 presentations across the 15-letter string). TD blocks were identical regardless of the initial task preceding it. This allowed for the critical comparison of TD following IF (TD-IF) with TD following EF (TD-EF) to investigate network activity during external cognition based upon prior attentional state.

Rating Block (Third Phase)

In the final phase of the sequence (rating block), subjects rated ease of performance and emotion associated with the task from the first phase. The order of questions was counterbalanced across subjects (ease before emotion: n = 9, emotion before ease: n = 5). Emotion ratings were important not only to probe inter-subject variability in negative and positive emotionality during IF, EF, and rest blocks but also served to validate that subjects were actually imagining target events during the IF block, under the assumption that subjects who did not vary their emotion ratings based on scenario valence or who made ratings that were the opposite of the intended valence (e.g., rating negative emotion when imagining a positive scenario) were not performing the task as instructed. In general, it is not possible to guarantee compliance in a task without an objective behavioral measure. However, in order to make the experience of event imagination in the scanner as ecological valid as possible, we did not wish to fundamentally alter the process by using external stimuli or requiring responses during the IF block. Indeed, multiple prior studies of event imagination have successfully employed approaches similar to the current study (Addis et al. 2007; Hassabis et al. 2007; D'Argembeau et al. 2008; D'Argembeau et al. 2010; Addis et al. 2011). Analysis of emotion ratings for all the initial tasks (IF, EF, and rest) suggested that subjects were indeed performing the tasks as instructed, which was confirmed by subject reports during a post-task debriefing session.

Neuroimaging Data Acquisition and Preprocessing

MRI scanning occurred on a Siemens Allegra 3T scanner following standard protocols. After sagittal localization, functional images were acquired with a T2*-weighted, gradient echo planar sequence (repetition time [TR] = 2000 ms, echo time [TE] = 30 ms, 36 slices of 3 mm thick, skip = 0, flip angle [FA] = 90°, field of view [FOV] = 210 mm, matrix size = 64 × 64). Subjects performed 6 runs, each consisting of 188 volumes plus 4 initial discarded volumes to allow for thermal equilibration of scanner signal, for a total of 1128 volumes. A high-resolution T1-weighted anatomical image (MP-RAGE) was also acquired. Task stimuli were presented to the subject using back projection (800 × 600 resolution).

Preprocessing of functional data was performed using Statistical Parametric Mapping software (SPM8) and included (in order): slice-time correction, realignment of functional images, coregistration of functional images to anatomical image, normalization to MNI152 template (an average of 152 T1 images from the Montreal Neurological Institute), and spatial smoothing of functional images with a 5-mm Gaussian kernel. Due to the intrinsic spatial smoothness of the data, the total average smoothness for contrasts of interest was 10-mm FWHM.

Data Analysis

Behavioral analysis examined ratings of ease, ratings of emotion, reaction time (RT) during TD, and percent errors during TD in 4 separate repeated-measures ANOVAs with sequence type (IF, EF, rest) as within-subjects factor. Significant main effects were followed up with post-hoc comparisons using paired-samples t-tests. Statistical significance was set to an α-level of 0.05.

For analysis of neuroimaging data, the primary model (Model 1) used regressors at the first-level specifying onset times for IF, EF, and rest blocks (at time of cue) and for TD blocks following each block type (TD-IF, TD-EF, and TD-rest). These blocks were modeled as epochs with durations set to block length (between 12 and 18 s), thus capturing neural activity related to processing “mode” rather than a discrete event (i.e., at the time of the switch from initial task to TD block). Regressors were also specified for the rating period (epochs with duration equal to time to make response) to reduce error variance but were not analyzed further. All regressors were convolved with the canonical hemodynamic response function using the general linear model as implemented in SPM8. The main contrasts of interest were: (1) IF >rest, (2) EF >rest, (3) IF >EF, (4) EF >IF, (5) TD-IF >TD-rest, (6) TD-EF >TD-rest, (7) TD-IF >TD-EF, and (8) TD-EF >TD-IF. Block designs are highly efficient (Friston et al. 1999) and do not require intervening rest or baseline intervals in order to identify activity unique to each condition. Although there would be overlapping activations between sequential blocks in our task (i.e., EF and TD-EF) due to the hemodynamic delay, the current block lengths of 12–18 s provided sufficient amount of non-overlapping signal for the unique estimation of each block. To confirm the statistical independence between sequential blocks, we computed the average and maximum colinearity between regressors in the model by measuring the correlation as the normalized dot product between convolved time-courses for all combinations of first- and second-phase regressors (IF, EF, rest, TD-IF, TD-EF, and TD-rest). The average correlation between regressors was 0.004, and the maximum correlation for any individual subject for any individual combination of 2 regressors was 0.04, confirming very low colinearity between sequential blocks.

As described in Results section, mean RT during TD and neural activity during TD differed significantly based on prior attentional focus. In order to interrogate the relationship between RT and brain differences, we ran another first-level model examining the effect of trial-to-trial variability in RT on brain activity during the TD block. In this secondary model (Model 2), 1 regressor specified the onset and durations for all TD blocks (regardless of prior block type). This regressor was parametrically modulated first by RT (a continuous predictor) and then with categorical predictors coding for prior block type (dummy codes for TD-IF and TD-EF, with TD-rest block serving as reference level). This model allowed us to: (1) examine the effect of trial-to-trial variations in RT on brain activity during TD and (2) investigate whether activation differences between TD-IF and TD-EF found in Model 1 remained after accounting for variance due to RT.

Unless otherwise noted, group effects of neuroimaging data were analyzed using one-sample t-tests, with α = 0.05, corrected for multiple comparisons across the whole brain at the cluster level using topological FDR as implemented in SPM8 (voxel-wise threshold of P < 0.001).

Aligning Task Activations with Intrinsic Networks

To identify the overlap between activation patterns identified during task and the intrinsic topography of DAN, DMN, and FPN, we performed a functional connectivity analysis on resting-state data (360 s, 180 volumes) obtained in a separate cohort of 17 healthy individuals [details of MRI acquisition parameters and cohort characteristics can be found in Stern et al. (2012)]. We placed 5-mm-radius seeds in regions that have dissociated DAN, DMN, and FPN in previous resting-state functional connectivity experiments (Vincent et al. 2008). Four seeds were selected for each of the 3 networks: bilateral seeds in motion-sensitive middle temporal area (MT+) and superior parietal lobule (SPL) were used to identify DAN; bilateral seeds in hippocampal formation and posterior inferior parietal lobule (pIPL) were used to identify DMN; and bilateral seeds in anterior prefrontal cortex (aPFC) and anterior inferior parietal lobule (aIPL) were used to identify FPN (coordinates taken from Vincent et al. 2008). Following our previously published protocol (Stern et al. 2012), we used the “conn” toolbox (www.nitrc.org/projects/conn) to examine connectivity between each seed and whole-brain gray matter, controlling for several factors. First, the top 5 principal components associated with segmented white matter and cerebrospinal fluid images for each individual subject were identified using the “CompCor” method (Behzadi et al. 2007) and regressed out of whole-brain gray matter activity. Critically, this method corrects for positivity biases arising from “noise correlations” related to non-neural sources (such as respiration or cardiac activity) without regressing out the global signal, which has been shown to lead to spurious negative correlations (Murphy et al. 2009). In addition, 12 motion regressors (6 realignment parameters and first derivatives) were included, and data were filtered between 0.01 and 0.10 Hz. At the first level, these partial correlation coefficient images between each seed and the whole brain were z-transformed. Connectivity maps for each of the 4 seeds for each network were then averaged and thresholded at P < 0.001 voxel-wise with cluster level correction for whole-brain analyses at P < 0.05 using FDR. Although this analysis yielded largely distinct networks, there were some areas of overlap between DAN, DMN, and FPN (see Supplementary Fig. 1); as such, we created masks that contained only those regions unique to each network in order to compare with task activations.

Psychophysiological Interaction (PPI) Analysis

To determine whether activation differences found during TD based on prior attentional focus (see Results, below) were also associated with differences in connectivity during these blocks, we performed a PPI analysis (Friston et al. 1997) comparing connectivity during TD-EF with connectivity during TD-IF. We selected bilateral seed regions consisting of 5-mm-radius spheres centered in DLPFC coordinates that were identified in the main comparison of TD-EF >TD-IF and which overlapped with our FPN mask (x = −50, y = 44, z = 0 and x = 48, y = 36, z = 22). At the first (subject) level, PPI contrasts were created for each subject to examine connectivity between each DLPFC seed and all gray matter voxels during TD-EF >TD-IF, TD-EF >fixation (baseline), and TD-IF >fixation (baseline). Second (group)-level analyses used one-sample t-tests to compare DLPFC connectivity between TD-EF and TD-IF.

Relationship Between Age and Task Effects

As we had a considerable age range in our sample (19–46 years), secondary analyses investigated whether there were main effects of age or interactions between age and task conditions for the behavioral and brain effects identified in primary comparisons. The results from these analyses are presented in Supplementary Materials; generally, brain activations were reduced in older participants during IF, EF, and during TD following IF.

Results

Behavioral

Ratings of Ease and Emotion

Mean ratings of emotion experienced during performance of initial tasks (1 = “very negative”, 5 = “very positive”) were 3.1 (standard deviation, SD: 0.39) for IF, 3.6 (SD: 0.70) for EF, and 3.9 (SD: 0.66) for rest blocks. There was a main effect of block type (IF, EF, rest) on ratings of emotion (F2,26 = 11.1, P < 0.001), with post-hoc paired t-tests indicating that rest was associated with more positive emotion than EF (t(13) = −2.4, P = 0.03), which itself was rated as more positive than IF (t(13) = −2.5, P = 0.03, Fig. 2A). Unpacking emotion ratings within IF, positive IF (i.e., imagining positive events) was rated as more positive (mean rating: 4.1, SD: 0.58) than negative IF (i.e., imagining negative events; mean rating: 2.0, SD: 0.66) (t(13) = 8.2, P < 0.001), as would be expected, Positive IF also elicited more positive emotion than EF (t(13) = −2.9, P = 0.01), with no difference from rest. Negative IF was associated with more negative emotion than both EF (t(13) = 5.4, P < 0.001) and rest (t(13) = 6.6, P < 0.001).

Figure 2.

Ratings of emotion and RT during TD. (A) Ratings of emotion were significantly different between all conditions except for positive IF vs. rest. (B) Reaction times were faster for TD following positive IF (TD-positive IF) than for TD following EF (TD-EF), with no significant differences in other pairwise comparisons. *P < 0.05, **P < 0.001.

Figure 2.

Ratings of emotion and RT during TD. (A) Ratings of emotion were significantly different between all conditions except for positive IF vs. rest. (B) Reaction times were faster for TD following positive IF (TD-positive IF) than for TD following EF (TD-EF), with no significant differences in other pairwise comparisons. *P < 0.05, **P < 0.001.

Mean ratings of ease of performance of initial tasks (1 = “very difficult”, 5 = “very easy”) were 3.8 (SD: 0.68) for IF, 4.2 (SD: 0.72) for EF, and 4.2 (SD: 0.69) for rest blocks (see Supplementary Fig. 2). Despite IF being rated as slightly harder on average than EF or rest blocks, there was neither a significant main effect of block type or significant condition differences in post-hoc t-tests between IF, EF and rest blocks. Effects did not become significant after removing 1 subject who exhibited IF ratings greater than 3 SD lower than the group mean for this condition (see Supplementary Fig. 2).

Reaction Time and Accuracy During TD

Mean RTs for correct trials were 542.0 ms (SD: 53.4) for TD-IF, 551.6 ms (SD: 53.6) for TD-EF, and 546.3 ms (SD: 56.8) for TD-rest. There was a main effect of prior block type on correct trial RT during TD (F2,26 = 3.6, P = 0.04), with post-hoc paired t-tests indicating that responses were faster for TD-IF than TD-EF (t(13) = −2.8, P = 0.02), with no differences with TD-rest (Fig. 2B). Breaking down the results further, TD following positive IF (TD-positive IF: 539.9 ms, SD: 52.6) was significantly faster than TD-EF (t(13) = 2.4, P = 0.03), but no different from TD-rest. There were no significant differences between TD following negative IF (TD-negative IF: 543.9 ms, SD: 55.8) and TD-EF (although there was a trend for the former to be faster, t(13) = 2.1, P = 0.06), TD-negative IF and TD-rest, or TD-negative IF and TD-positive IF.

Overall accuracy during TD was very high (on average, the percentage of correct responses was 97.6). Mean percentage errors were 2.2 (SD: 2.1) for IF, 2.8 (SD: 2.5) for EF, and 2.3 (SD: 2.3) for rest blocks (see Supplementary Fig. 2). Although there were a smaller mean number of errors for IF and rest blocks than EF blocks, there was neither a significant main effect of block type or significant condition differences in post-hoc t-tests between IF, EF, and rest blocks.

Post-Task Ratings of Engagement, Vividness, and Perspective During IF

Overall, subjects reported being highly engaged in the event imagination block (mean: 3.8, SD: 0.7, 1 = “complete distraction”, 5 = “complete engagement”). In addition, subjects reported a high degree of imagery vividness (mean: 3.7, SD: 0.8, 1 = “no image at all”, 5 = “image is as vivid as normal vision”), and most scenarios were imagined from a first-person perspective (mean: 3.9, SD: 0.9, 1 = “entirely third person”, 5 = “entirely first-person”).

Functional Magnetic Resonance Imaging

Internal vs. External Attentional State

Prior to TD, the EF spatial working memory block activated regions of FPN and DAN more than the IF event imagination block, including areas of lateral frontal cortex (orbitofrontal cortex [OFC], inferior frontal gyrus [IFG], dorsolateral prefrontal cortex [DLPFC], and precentral gyrus), anterior insula, dorsal medial frontal cortex including anterior cingulate cortex (ACC) and supplementary motor area (SMA), mid-cingulate cortex, lateral parietal cortex (inferior parietal lobule [IPL] and SPL), posterior temporal cortex, occipital cortex, striatum, and thalamus (including pulvinar, ventral nuclei, and medial dorsal nucleus) (Fig. 3A, regions in blue, Table 1). As would be expected, IF blocks elicited more activation in DMN regions including ventromedial prefrontal cortex (VMPFC), posterior cingulate cortex (PCC), anterior MTG/temporal pole, and medial temporal lobe structures (hippocampus and parahippocampal gyrus) than EF (Fig. 3A, regions in red, Table 1).

Table 1

Comparisons between EF and IF blocks

Contrast/region (side) BA k x y z Max z Network 
EF >IF 
 Frontal 
  Precentral/postcentral (L) 3, 4, 6 839 −40 −8 50 4.72 DAN, FPN 
  DLPFC (L) 9, 10, 46 300 −44 52 10 5.16 FPN 
  Mid-insula (L) N/A 75 −42 4.84 None 
  Anterior insula (L) N/A 185 −34 16 4.72 FPN 
  Precentral (L) 6, 9 319 −54 34 4.47 DAN 
  OFC (R) 10 73 36 58 −4 3.91 FPN 
  IFG/precentral/anterior insula/putamen (R) 6, 9, 44, 45, 47 2333 30 16 5.02 DAN, FPN 
  DLPFC (R) 9, 10, 46 808 48 38 18 4.56 FPN 
  Mid/anterior cingulate (B) 24, 33 102 −2 30 4.53 FPN 
  Anterior cingulate/SMA (B) 6, 32 188 54 4.78 DAN 
 Parietal 
  IPL/STG (L) 40, 42 259 −64 −32 24 5.20 None 
  Mid/posterior cingulate (B) 31 56 10 −30 38 4.28 DMN, FPN 
 Temporal/occipital 
  Cuneus (R) 17, 18 54 −92 4.11 DAN 
  Cuneus/precuneus/IPL/SPL (B) 7, 18, 19, 37, 40 10 632 −28 −82 24 5.91 DAN, DMN, FPN 
 Subcortical 
  Thal (L) N/A 121 −20 −32 4.45 None 
  Caudate/putamen (L) N/A 86 −10 10 3.94 FPN 
  Thal/putamen/caudate/midbrain/hippocampus (R) N/A 974 22 −12 5.23 DMN, FPN 
IF >EF 
 Frontal 
  DMPFC (L) 90 −10 50 40 4.00 FPN 
  SMA (L) 62 −8 18 68 3.91 FPN 
  OFC/IFG (R) 11, 47 67 40 32 −12 4.67 None 
  VMPFC (B) 11 124 −4 42 −22 4.70 DMN 
 Parietal 
  Precuneus/PCC (B) 23, 31 307 −6 −62 22 4.34 DMN 
 Temporal 
  MTG/ITG/temporal pole (L) 20, 21, 38 1038 −42 16 −30 4.93 DAN, DMN, FPN 
  MTG/ITG (R) 20, 21 101 58 −12 −24 3.90 DMN 
 Occipital 
  PG/fusiform gyrus (R) N/A 66 34 −52 −4 4.21 None 
 Subcortical 
  PG/hippocampus/amygdala (L) 28, 34, 35 369 −26 −10 −24 5.11 DMN 
  PG/hippocampus/amygdala (R) 28 67 26 −10 −24 4.42 DMN 
Contrast/region (side) BA k x y z Max z Network 
EF >IF 
 Frontal 
  Precentral/postcentral (L) 3, 4, 6 839 −40 −8 50 4.72 DAN, FPN 
  DLPFC (L) 9, 10, 46 300 −44 52 10 5.16 FPN 
  Mid-insula (L) N/A 75 −42 4.84 None 
  Anterior insula (L) N/A 185 −34 16 4.72 FPN 
  Precentral (L) 6, 9 319 −54 34 4.47 DAN 
  OFC (R) 10 73 36 58 −4 3.91 FPN 
  IFG/precentral/anterior insula/putamen (R) 6, 9, 44, 45, 47 2333 30 16 5.02 DAN, FPN 
  DLPFC (R) 9, 10, 46 808 48 38 18 4.56 FPN 
  Mid/anterior cingulate (B) 24, 33 102 −2 30 4.53 FPN 
  Anterior cingulate/SMA (B) 6, 32 188 54 4.78 DAN 
 Parietal 
  IPL/STG (L) 40, 42 259 −64 −32 24 5.20 None 
  Mid/posterior cingulate (B) 31 56 10 −30 38 4.28 DMN, FPN 
 Temporal/occipital 
  Cuneus (R) 17, 18 54 −92 4.11 DAN 
  Cuneus/precuneus/IPL/SPL (B) 7, 18, 19, 37, 40 10 632 −28 −82 24 5.91 DAN, DMN, FPN 
 Subcortical 
  Thal (L) N/A 121 −20 −32 4.45 None 
  Caudate/putamen (L) N/A 86 −10 10 3.94 FPN 
  Thal/putamen/caudate/midbrain/hippocampus (R) N/A 974 22 −12 5.23 DMN, FPN 
IF >EF 
 Frontal 
  DMPFC (L) 90 −10 50 40 4.00 FPN 
  SMA (L) 62 −8 18 68 3.91 FPN 
  OFC/IFG (R) 11, 47 67 40 32 −12 4.67 None 
  VMPFC (B) 11 124 −4 42 −22 4.70 DMN 
 Parietal 
  Precuneus/PCC (B) 23, 31 307 −6 −62 22 4.34 DMN 
 Temporal 
  MTG/ITG/temporal pole (L) 20, 21, 38 1038 −42 16 −30 4.93 DAN, DMN, FPN 
  MTG/ITG (R) 20, 21 101 58 −12 −24 3.90 DMN 
 Occipital 
  PG/fusiform gyrus (R) N/A 66 34 −52 −4 4.21 None 
 Subcortical 
  PG/hippocampus/amygdala (L) 28, 34, 35 369 −26 −10 −24 5.11 DMN 
  PG/hippocampus/amygdala (R) 28 67 26 −10 −24 4.42 DMN 

BA, Brodmann's areas; k, number of voxels; L, left; R, right; B, both hemispheres; coordinates are in MNI space. “Network” column indicates to which of 3 resting-state networks each cluster belongs, listed in alphabetical order (DAN, dorsal attention network; DMN, default mode network; FPN, fronto-parietal network). DLPFC, dorsolateral prefrontal cortex; DMPFC, dorsomedial prefrontal cortex; IFG, inferior frontal gyrus; IPL, inferior parietal lobule; ITG, inferior temporal gyrus; MTG, middle temporal gyrus; OFC, orbitofrontal cortex; PG, parahippocamal gyrus; PCC, posterior cingulate cortex; SMA, supplementary motor area; SPL, superior parietal lobule; STG, superior temporal gyrus; Thal, thalamus; VMPFC, ventromedial prefrontal cortex.

P < 0.05 corrected for multiple comparisons across the whole brain.

Figure 3.

Comparisons between IF, EF, and rest blocks. (A) IF showed greater activity than EF in DMN regions including VMPFC and temporal regions (areas in red), whereas EF exhibited more widespread fronto-parietal and DAN activation than IF (areas in blue). (B) IF and EF blocks both activated several fronto-parietal and striatal regions when compared with rest (IF >rest: red, EF >rest: blue, overlap: pink). Color bar represents t-scale. Activations are significant at P < 0.05, corrected for whole-brain comparisons.

Figure 3.

Comparisons between IF, EF, and rest blocks. (A) IF showed greater activity than EF in DMN regions including VMPFC and temporal regions (areas in red), whereas EF exhibited more widespread fronto-parietal and DAN activation than IF (areas in blue). (B) IF and EF blocks both activated several fronto-parietal and striatal regions when compared with rest (IF >rest: red, EF >rest: blue, overlap: pink). Color bar represents t-scale. Activations are significant at P < 0.05, corrected for whole-brain comparisons.

When compared with rest, overlapping activations for EF >rest and IF >rest were found in OFC, IFG, DLPFC, precentral gyrus, anterior insula, SMA, IPL, posterior middle temporal gyrus (MTG), putamen, and thalamus (Fig. 3B, EF >rest in blue, IF >rest in red, overlap in pink). The overall extent of activation was greater for EF >rest, particularly evident in parietal, temporal, and occipital cortices, and subcortical regions of thalamus and midbrain. The IF >rest comparison revealed unique activation in VMPFC, anterior DMPFC, anterior MTG/temporal pole, and medial temporal lobe structures (hippocampus and parahippocampal gyrus). Overall, these data support and extend prior studies and indicate that the present task engaged the expected brain networks underlying EF and IF cognition.

There is reported evidence of greater activation in some DMN regions for imagining positive as opposed to negative events (Sharot et al. 2007). In order to investigate whether DMN activity during IF differed based on the valence of the imagined event, parameter estimates were extracted from regions-of-interest (ROIs) for the IF >EF contrast (Table 1) separately for negative and positive event scenarios and compared using paired-samples t-tests. There were no significant differences in activation in DMN ROIs during IF based on whether the imagined events were positive or negative. For completeness, we also examined whether there were differences based on event valence during IF in those regions found for the EF >IF contrast. Comparison of parameter estimates for negative and positive IF from these ROIs (Table 1) revealed no significant differences.

Target Detection Based upon Prior Attentional State

For the key comparison of activation during TD based upon the prior attentional state, TD following external focus exhibited significantly greater activation than TD following internal focus (TD-EF >TD-IF) in multiple areas of lateral frontal cortex (OFC, IFG, DLPFC, precentral gyrus), anterior insula, SMA, mid- and posterior cingulate, IPL and precuneus, posterior MTG/STG extending into occipital cortex, thalamus, striatum, and cerebellum (Fig. 4A, Table 2). Overlap was found between these regions and all 3 resting-state networks (DAN, DMN, and FPN) (Fig. 4B, Table 2), with the majority of overlap occurring with FPN.

Table 2

Comparisons between TD following EF (TD-EF) and TD following IF (TD-IF)

Contrast/region (side) BA k x y z Max z Effect of IF valence* EF vs. IF block+ Network 
TD-EF >TD-IF 
 Frontal 
  DLPFC/IFG (L) 10, 45, 46 38a −50 44 4.56 t = −2.8, P = 0.02  FPN 
  OFC/anterior insula (L) 11, 47 74a −28 32 −4 4.98 t = −2.4, P = 0.03  FPN 
  OFC/IFG/anterior insula (L) 44, 45, 47 490 −52 26 −8 4.29 t = −2.5, P = 0.03  FPN 
  Precentral/DLPFC (L) 6, 9 337 −32 30 36 4.34  t = −2.2, P = 0.05 DAN, FPN 
  Precentral/IFG (L) 44 94 −64 10 3.94 t = −2.9, P = 0.01 t = −2.9, P = 0.01 FPN 
  OFC/IFG/anterior insula (R) 44, 47 271 42 12 4.21 t = −2.4, P = 0.03 t = −2.6, P = 0.02 FPN 
  DLPFC (R) 10, 46 66b 48 36 22 4.56 t = −4.2, P = 0.001 t = −6.1, P < 0.001 FPN 
  DLPFC (R) 8, 9 115b 38 28 38 4.74  t = −3.3, P = 0.006 DMN, FPN 
  DLPFC/SMA (R) 83b 14 12 58 4.46  t = −3.7, P = 0.002 DAN, FPN 
  SMA (B) 200 −10 66 3.79   None 
 Parietal 
  IPL/STG (L) 40, 42 336 −60 −36 20 3.89  t = −6.0, P < 0.001 FPN 
  IPL/STG (R) 40 241d 46 −52 26 4.74  t = −4.5, P = 0.004 DMN, FPN 
  Precuneus/cuneus (R) 31 37c 12 −64 24 4.45   None 
  Mid/posterior cingulate (B) 23, 24, 31 435c −30 42 4.69 t = −2.2, P = 0.05  DMN, FPN 
 Temporal/occipital 
  MTG (L) 21 160 −54 −8 −8 4.32  t = 6.4, P < 0.001 DAN, DMN 
  MTG/STG (L) 21, 39 325 −46 −54 4.40 t = −2.1, P = 0.05  DAN, DMN 
  MTG/STG (L) 22, 39 379 −52 −66 22 5.58 t = −2.6, P = 0.02 t = 2.8, P = 0.01 DAN, DMN, FPN 
  Middle occipital/ITG (L) 19, 37 123 −56 −62 −12 4.69  t = −6.6, P < 0.001 DAN 
  MTG (R) 21 26d 62 −32 −12 4.45   DMN 
  MTG (R) 22 49d 54 −40 −2 4.39   None 
  STG (R) 22 30d 64 −44 10 4.52  t = −5.7, P < 0.001 None 
STG (R) 39 31d 42 −48 12 4.31  t = −3.1, P = 0.009 None 
 Subcortical 
  Putamen/caudate/thal (R) N/A 147 28 −10 16 4.84 t = −2.3, P = 0.04 t = −3.9, P = 0.002 None 
  Cerebellum (B) N/A 58c (4216) −40 −6 4.90   DMN 
Contrast/region (side) BA k x y z Max z Effect of IF valence* EF vs. IF block+ Network 
TD-EF >TD-IF 
 Frontal 
  DLPFC/IFG (L) 10, 45, 46 38a −50 44 4.56 t = −2.8, P = 0.02  FPN 
  OFC/anterior insula (L) 11, 47 74a −28 32 −4 4.98 t = −2.4, P = 0.03  FPN 
  OFC/IFG/anterior insula (L) 44, 45, 47 490 −52 26 −8 4.29 t = −2.5, P = 0.03  FPN 
  Precentral/DLPFC (L) 6, 9 337 −32 30 36 4.34  t = −2.2, P = 0.05 DAN, FPN 
  Precentral/IFG (L) 44 94 −64 10 3.94 t = −2.9, P = 0.01 t = −2.9, P = 0.01 FPN 
  OFC/IFG/anterior insula (R) 44, 47 271 42 12 4.21 t = −2.4, P = 0.03 t = −2.6, P = 0.02 FPN 
  DLPFC (R) 10, 46 66b 48 36 22 4.56 t = −4.2, P = 0.001 t = −6.1, P < 0.001 FPN 
  DLPFC (R) 8, 9 115b 38 28 38 4.74  t = −3.3, P = 0.006 DMN, FPN 
  DLPFC/SMA (R) 83b 14 12 58 4.46  t = −3.7, P = 0.002 DAN, FPN 
  SMA (B) 200 −10 66 3.79   None 
 Parietal 
  IPL/STG (L) 40, 42 336 −60 −36 20 3.89  t = −6.0, P < 0.001 FPN 
  IPL/STG (R) 40 241d 46 −52 26 4.74  t = −4.5, P = 0.004 DMN, FPN 
  Precuneus/cuneus (R) 31 37c 12 −64 24 4.45   None 
  Mid/posterior cingulate (B) 23, 24, 31 435c −30 42 4.69 t = −2.2, P = 0.05  DMN, FPN 
 Temporal/occipital 
  MTG (L) 21 160 −54 −8 −8 4.32  t = 6.4, P < 0.001 DAN, DMN 
  MTG/STG (L) 21, 39 325 −46 −54 4.40 t = −2.1, P = 0.05  DAN, DMN 
  MTG/STG (L) 22, 39 379 −52 −66 22 5.58 t = −2.6, P = 0.02 t = 2.8, P = 0.01 DAN, DMN, FPN 
  Middle occipital/ITG (L) 19, 37 123 −56 −62 −12 4.69  t = −6.6, P < 0.001 DAN 
  MTG (R) 21 26d 62 −32 −12 4.45   DMN 
  MTG (R) 22 49d 54 −40 −2 4.39   None 
  STG (R) 22 30d 64 −44 10 4.52  t = −5.7, P < 0.001 None 
STG (R) 39 31d 42 −48 12 4.31  t = −3.1, P = 0.009 None 
 Subcortical 
  Putamen/caudate/thal (R) N/A 147 28 −10 16 4.84 t = −2.3, P = 0.04 t = −3.9, P = 0.002 None 
  Cerebellum (B) N/A 58c (4216) −40 −6 4.90   DMN 

BA, Brodmann's areas; L, left; R, right; B, both hemispheres; k, number of voxels; coordinates are in MNI space. “Network” column indicates to which of 3 resting-state networks each cluster belongs, listed in alphabetical order (DAN, dorsal attention network; DMN, default mode network; FPN, fronto-parietal network). DLPFC, dorsolateral prefrontal cortex; IFG, inferior frontal gyrus; IPL, inferior parietal lobule; ITG, inferior temporal gyrus; MTG, middle temporal gyrus; OFC, orbitofrontal cortex; SMA, supplementary motor area; STG, superior temporal gyrus; thal, thalamus.

P < 0.05 corrected for multiple comparisons across the whole brain.

a–dCoordinates represent subclusters within 4 larger clusters (derived from thresholding at P < 0.0001) containing over 500 voxels and spanning multiple regions of cortex.

*Significant differences reflect less activation during TD following negative IF (TD-negative IF) as compared with TD following positive IF (TD-positive IF). +Negative t-values indicate EF >IF, positive t-values indicate IF >EF. T-tests are not corrected for multiple comparisons.

Figure 4.

Activation during TD based on prior focus. (A) TD-EF showed greater activation than TD-IF in several regions of frontal, parietal, and temporal cortex. Activations are significant at P < 0.05, corrected for whole-brain comparisons. Color bar represents t-scale. There were no regions showing significantly greater activity during TD-IF compared with TD-EF. (B) Areas of greater activation during TD-EF than TD-IF that overlapped with the 3 resting-state networks identified from a different data set. DAN, dorsal attention network; DMN, default mode network; FPN, fronto-parietal network.

Figure 4.

Activation during TD based on prior focus. (A) TD-EF showed greater activation than TD-IF in several regions of frontal, parietal, and temporal cortex. Activations are significant at P < 0.05, corrected for whole-brain comparisons. Color bar represents t-scale. There were no regions showing significantly greater activity during TD-IF compared with TD-EF. (B) Areas of greater activation during TD-EF than TD-IF that overlapped with the 3 resting-state networks identified from a different data set. DAN, dorsal attention network; DMN, default mode network; FPN, fronto-parietal network.

Comparisons with TD-rest allowed us to unpack these effects. Similar to differences with TD-IF, TD-EF showed significantly greater activation than TD-rest in lateral frontal cortex (OFC, IFG, DLPFC, and precentral gyrus), insula, dorsal medial frontal cortex including ACC and SMA, mid- and PCC, IPL and SPL, posterior and anterior MTG/STG, occipital cortex, parahippocampal gyrus, thalamus, midbrain, and cerebellum. There were fewer differences between TD-IF and TD-rest, with TD-IF showing more activity than TD-rest only in right frontal cortex (precentral gyrus and SMA). There were no regions where TD-IF showed more activation than TD-EF, or where TD-rest showed more activity than either TD-EF or TD-IF.

In order to investigate whether differences during TD were related to the valence of the previously imagined event, parameter estimates from ROIs where TD-EF >TD-IF (Table 2) were extracted separately for TD following positive (TD-positive IF) and TD following negative (TD-negative IF) events. Paired-samples t-tests revealed several regions showing more activity during TD-positive IF (Fig. 5, areas in red and dark gray bars) than TD-negative IF (Fig. 5, areas in blue and light gray bars), including OFC, IFG, DLPFC, precentral gyrus, anterior insula, mid- and PCC, posterior MTG/STG, and striatum/thalamus (see Table 2). No regions showed greater activity for TD-negative IF than TD-positive IF.

Figure 5.

Activation differences during TD based on valence of prior internal focus. (A) There was more activation in several regions during TD-EF than TD-negative IF (areas shown in red), whereas differences between TD-EF and TD-positive IF were weaker (areas shown in blue, overlap in pink). Activations are significant at P < 0.05, corrected for whole-brain comparisons. Color bar represents t-scale. (B) Regions of interest from Table 2 showed differences between TD-EF and TD-negative IF (dark gray bars) that were significantly greater than differences between TD-EF and TD-positive IF (light gray bars). Error bars represent standard error of the mean.

Figure 5.

Activation differences during TD based on valence of prior internal focus. (A) There was more activation in several regions during TD-EF than TD-negative IF (areas shown in red), whereas differences between TD-EF and TD-positive IF were weaker (areas shown in blue, overlap in pink). Activations are significant at P < 0.05, corrected for whole-brain comparisons. Color bar represents t-scale. (B) Regions of interest from Table 2 showed differences between TD-EF and TD-negative IF (dark gray bars) that were significantly greater than differences between TD-EF and TD-positive IF (light gray bars). Error bars represent standard error of the mean.

Psychophysiological Interaction During Target Detection

For PPI analyses, we selected seeds in left and right DLPFC regions that were found for TD-EF >TD-IF and which overlapped with our FPN mask. Based on results showing increased activation of FPN/DAN and decreased activation of DMN during EF compared with IF blocks, we hypothesized that DLPFC would show increased correlations with other FPN/DAN regions and decreased correlations with DMN, during TD following EF as compared with TD following IF. Although PPI effects did not pass whole-brain-corrected thresholds for significance, we probed for the presence of these hypothesized effects at a threshold of P < 0.005/k >20.

There was increased connectivity between right DLPFC and a region of right anterior insula that overlapped with our FPN mask (x = 42, y = 26, z = −8, k = 33, z = 3.5) during TD-EF compared with TD-IF. In contrast, there was decreased functional connectivity between right DLPFC and regions of VMPFC/subgenual cingulate (x = 10, y = 26, z = −6, k = 22, z = 3.5) and PCC (x = 8, y = −52, z = 24, k = 25, z = 3.6) that overlapped with our DMN mask during TD-EF compared with TD-IF (Fig. 6). Examination of parameter estimates of PPI contrasts between each TD condition vs. baseline indicated that these differences were driven by less positive connectivity between DLPFC and anterior insula, as well as less negative connectivity between DLPFC and VMPFC and posterior cingulate cortex, during TD-IF than during TD-EF (Fig. 6). For the left DLPFC seed, there was increased connectivity during TD-EF compared with TD-IF with a region of left parietal cortex (x = −40, y = −68, z = 16, k = 26, z = 3.4) that overlapped with our DAN mask. Unlike the right DLPFC seed, the left DLPFC seed did not show decreased connectivity with DMN for TD-EF compared with TD-IF.

Figure 6.

Psychophysiological interaction (PPI) during TD based on prior attentional focus. There was positive connectivity between a seed in right dorsolateral prefrontal cortex (DLPFC, shown in yellow) and right anterior insula (shown in green) during TD-EF, and negative connectivity between these regions during TD-IF. There was negative connectivity between right DLPFC and regions of VMPFC and PCC (shown in blue) during TD-EF, and less negative connectivity (DLPFC with VMPFC) or even positive connectivity (DLPFC with PCC) during TD-IF. Results are displayed at P < 0.005/k >20. Error bars represent standard error of the mean.

Figure 6.

Psychophysiological interaction (PPI) during TD based on prior attentional focus. There was positive connectivity between a seed in right dorsolateral prefrontal cortex (DLPFC, shown in yellow) and right anterior insula (shown in green) during TD-EF, and negative connectivity between these regions during TD-IF. There was negative connectivity between right DLPFC and regions of VMPFC and PCC (shown in blue) during TD-EF, and less negative connectivity (DLPFC with VMPFC) or even positive connectivity (DLPFC with PCC) during TD-IF. Results are displayed at P < 0.005/k >20. Error bars represent standard error of the mean.

Relationship to Reaction Time

Given that RT differed between TD-IF and TD-EF, we sought to investigate whether brain differences in these conditions were related to response times. As described in Methods, we ran a secondary model at the first level where activity during TD was parametrically modulated by RT (entered first), followed by regressors capturing variance based on prior attentional state (IF, EF, and rest). Group analyses revealed no brain regions that were significantly correlated with RT during TD. Furthermore, network differences between TD-IF and TD-EF remained highly significant after accounting for effects of RT (Fig. 7A). To confirm this latter effect, we extracted parameter estimates from the regions differing in the primary model (i.e., TD-EF >TD-IF ROIs, Table 2) for the contrast comparing TD-IF and TD-EF after controlling for RT. Paired-samples t-tests indicated that all regions were significantly more active during TD-EF as compared with TD-IF after accounting for effects of RT (Fig. 7B, light gray bars, all significantly negative at P < 0.005). In many cases, these effects were larger than for the primary model that did not account for RT (Fig. 7B, dark gray bars).

Figure 7.

Activation during TD when controlling for effects of RT. (A) After controlling for differences in RT, TD-EF showed widespread increases compared with TD-IF. Activations are significant at P < 0.05, corrected for whole-brain comparisons. Color bar represents t-scale. (B) All differences between TD-IF and TD-EF in the primary model (dark gray bars) remained significant after controlling for RT (light gray bars), in many cases exhibiting effects of even larger magnitude. Regions of interest are shown on the x-axis in the order in which they are listed in Table 2. Error bars represent standard error of the mean.

Figure 7.

Activation during TD when controlling for effects of RT. (A) After controlling for differences in RT, TD-EF showed widespread increases compared with TD-IF. Activations are significant at P < 0.05, corrected for whole-brain comparisons. Color bar represents t-scale. (B) All differences between TD-IF and TD-EF in the primary model (dark gray bars) remained significant after controlling for RT (light gray bars), in many cases exhibiting effects of even larger magnitude. Regions of interest are shown on the x-axis in the order in which they are listed in Table 2. Error bars represent standard error of the mean.

Relationship to Activation During Internal and External Focus Blocks

In order to investigate whether any of the regions that were greater for TD-EF compared with TD-IF were also greater for EF compared with IF, paired sample t-tests compared parameter estimates for IF >rest and EF >rest in those ROIs where TD-EF >TD-IF (regions in Table 2). Results indicated that half of the ROIs (including areas of DLPFC, precentral gyrus, anterior insula, IPL, occipital cortex, and putamen, regions with negative t-scores in “EF vs. IF block” column in Table 2) had significantly greater activation during the EF block than the IF block. Note that these common effects are not due to “spillover” of signal between the blocks due to the hemodynamic delay (i.e., similar effects for TD-EF >TD-IF and EF >IF contrasts are not due to signal overlap between EF and TD-EF blocks) because the model estimates unique activity for each block type, as explained in Methods section. Two ROIs in left MTG that were more active for TD-EF compared with TD-IF were actually more active during IF than EF blocks (regions with positive t-scores in Table 2), and several frontal and parietal ROIs were not significantly different between IF and EF blocks (regions with no t-score listed).

Discussion

Results from the present study identified a robust difference in the activation of several areas of frontal, parietal, and temporal cortex during an EF target detection (TD) task based on the prior attentional focus. Several regions were less active during TD when attention was previously IF as compared with when it had already been EF. Brain networks associated with IF and EF cognition are fairly well delineated (Corbetta and Shulman 2002; Addis et al. 2007; Vincent et al. 2008; Smith et al. 2009; Spreng et al. 2010; Andrews-Hanna 2012; Schacter et al. 2012; Smallwood et al. 2012). Spontaneous bouts of IF cognition (i.e., “mind wandering”) have consequences for both brain activity and behavior during EF tasks of attention and working memory (Mason et al. 2007; Stawarczyk et al. 2011), but it is unknown whether the engagement of IF cognition is associated with a more sustained effect on the subsequent recruitment of brain networks. The ability to flexibly and consciously disengage from a behaviorally irrelevant IF state to focus on external information is a common and necessary cognitive function, the dysregulation of which may contribute to several forms of psychopathology. To our knowledge, this is the first study to investigate these processes, with results indicating that network activation and connectivity during EF cognition were strongly dependent on the preceding attentional state. Although effects during TD were found in several areas, the majority of regions that differed based on prior attentional state are associated with FPN and DAN.

What could be causing the differences in FPN/DAN activation during TD following an internally as compared with EF attentional state? Greater activation during TD following EF could be due to increased difficulty of switching between 2 EF tasks, compared with switching between internally and externally focused tasks. This interpretation would be consistent with the finding of slower RT during TD following EF than TD following IF. However, there are several ways in which the RT data do match the neuroimaging results, which weakens the strength of this explanation. First, when breaking up TD-IF trials based upon whether the prior imagined event was positive or negative, the largest neural differences were between conditions showing the smallest RT differences (i.e., TD following negative IF and TD following EF did not have significantly different RTs but showed widespread network differences). Similarly, there were no significant RT differences between TD following EF and TD following rest, yet network differences between these conditions were robust and very similar to the TD-EF >TD-IF comparison. Finally, in our supplemental analyses examining the effect of age on activation during TD (see Supplementary Materials), we found that older subjects had relatively slower RT in addition to reduced activation in FPN during TD-IF compared with younger subjects; in this analysis, RT and brain activation exhibited a relationship at the level of the subject (slower RT and reduced brain activity) that is of the opposite direction to the relationship found at the level of the condition (faster RT and reduced brain activity). These pieces of evidence suggest that RT is not well linked to neural effects, and thus, interpretations of switching difficulty that are based on RT effects should be made with caution.

Another possibility is that greater activation during TD following EF compared with TD following IF is due to the prior activation of these same regions during the EF spatial working memory task. Post-hoc analyses confirmed that many of the regions that were more active for EF than IF were also more active for TD-EF than TD-IF. Note that these effects are not due to BOLD signal overlap between the EF and TD-EF blocks related to hemodynamic delay, as the model estimates unique activity for each block (i.e., variance that is not shared, explained in Methods section) and block colinearity was extremely low. Given the general absence of a brain–behavior relationship during TD, however, it is not possible to know whether greater activity during TD following EF reflects “improved” functioning of brain systems associated with EF cognition, or a less “efficient” implementation of those systems. In the former scenario, engagement of FPN and DAN during the EF spatial working memory task could lead to increased activity during TD because of greater readiness of these regions to respond (Winkielman et al. 2007). In the latter scenario, fatigue of the system due to prolonged activation during EF may result in greater inefficiency and increased activation during TD as a compensatory mechanism. Although the present results cannot definitively distinguish between these 2 explanations, there is some evidence to argue against the latter explanation. PPI results showed greater negative correlations between regions of FPN (DLPFC) and DMN (ventromedial prefrontal and posterior cingulate cortices) during TD following EF than during TD following IF, and negative correlations between FPN and DMN typically reflect a more, not less, efficient system (Kelly et al. 2008; Anticevic et al. 2012; Wen et al. 2013). Furthermore, as described earlier, our secondary analyses examining the effects of age on brain activation during TD found that greater age was associated with reduced network activation during TD following IF, further arguing against an explanation relating reduced activation during TD-IF to greater neural efficiency.

However, if there is in fact “improved” FPN/DAN functioning during TD following EF, we would expect there to be better performance during TD following EF as well, which was not found. One reason for this discrepancy may be due to the TD task itself, which was relatively simple (average accuracy was 97.6%). Although prior studies examining more complex tasks of working memory and sustained attention have linked poorer task performance with both reduced (Paus et al. 1997; Coull et al. 1998; Lim et al. 2010; Yun et al. 2010; Nagel et al. 2011) and increased (Rypma et al. 2002) FPN/DAN activity, simple TD does not typically elicit widespread activation of FPN and DAN (Linden et al. 1999; Kirino et al. 2000). We chose the minimally demanding TD task based on our prior work identifying performance decrements and network abnormalities in patients with obsessive-compulsive disorder specifically during conditions of low cognitive demand (Stern et al. 2013), yet it is possible that, among healthy individuals, reduced network activation following IF would only be associated with impaired performance in a more difficult task. Future studies interrogating the relationship between prior attentional state, brain activity, and behavior using more demanding tasks will be necessary to disentangle these potential explanations. Nevertheless, this study represents an important first step toward investigating neural mechanisms of attentional state switching, which is critical for understanding efficient cognitive functioning among both healthy individuals and those with brain-based disorders.

The present findings are interesting in light of a previous investigation examining the influence of prior cognitive load on network functioning during working memory (Yun et al. 2010). In that study, FPN and DAN activation were increased during a difficult 4-back task as compared with an easier 1-back task, as would be expected. However, during a subsequent 2-back task, FPN and DAN were reduced following the 4-back compared with the 1-back task, which was interpreted as reflecting impaired network recruitment due to prior cognitive overloading. Curiously, the pattern of activation differences due to prior working memory load in that study is strikingly similar to the present differences during TD based on prior attentional state. This suggests that network recruitment during an external cognition task is influenced both by prior cognitive load and prior attentional state, but in opposing ways: FPN/DAN activation during conditions of high compared with low cognitive load is associated with decreases in subsequent network activation, but FPN/DAN activation during conditions of externally compared with IF attention is associated with increases in subsequent network activation.

In addition to the impact of prior attentional state, emotion experienced during IF cognition significantly affected brain activation during TD. Differences between TD-IF and TD-EF were extremely robust when the prior imagined event was negative, with several regions showing significantly less activity during TD-negative IF than TD-positive IF, indicating that engaging with IF negative information strongly impacts subsequent network recruitment. This raises the intriguing possibility that altered functioning in patients with internalizing disorders may be partially attributable to abnormal network activation following a negatively valenced internal state associated with DMN activation. For example, obsessive-compulsive disorder is characterized by excessive focus on imagined negative events and is associated with increased DMN activity (Fitzgerald et al. 2005, 2010; Stern et al. 2011; Harrison et al. 2012; Stern et al. 2013) and impaired negative connectivity between networks processing internal and external states (Stern et al. 2012). Self-focused negative rumination, a hallmark of major depression, is also linked to DMN activation (Ray et al. 2005; Hamilton et al. 2011). Perhaps not surprisingly, obsessive-compulsive and depressed patients exhibit deficits in neuropsychological tests of EF cognition (Kuelz et al. 2004; Murrough et al. 2011). The current findings offer a potential explanation for impaired EF cognition in these disorders, namely that excessive time spent in a DMN-based IF state significantly alters the recruitment of FPN and DAN following this state. Further research is necessary in order to test this theory in patient populations.

Interestingly, reduced network activation was also evident for TD-rest compared with TD-EF, and TD-IF showed only minor increases in activity over TD-rest. It has been previously shown that in the absence of external stimulation, rest periods are characterized by “mind wandering”, which consists of a variety of IF cognitive experiences including spontaneous event imagination, episodic memory, planning and random thoughts about subjective emotional, cognitive and somatic states (Ingvar 1985; Fransson 2006; Buckner and Vincent 2007; Andrews-Hanna et al. 2010). Consistent with this notion, subjects reported having a variety of IF cognitive experiences during the rest block in the post-task debriefing. The similarity between TD-IF and TD-rest indicates that goal-directed IF cognition and spontaneous mind wandering have similar consequences for subsequent network engagement during EF cognition. Note that this relative lack of difference between TD-IF and TD-rest occurred despite the fact that the IF block elicited significantly more activity in both DMN and FPN/DAN than the rest block (prior to TD), indicating that activity during TD is not merely a function of activity during the prior blocks.

Although the current findings serve to elucidate the neural consequences of prior attentional state, limitations and questions remain for future research. The lack of brain–behavior relationship could be due to certain aspects of the study design. As mentioned earlier, the TD task may have been too simple for performance to depend on FPN and DAN activation. In addition, the sample size was relatively small, which could have contributed to the inability to identify a correlation between behavior and brain activity. Although our main effects (IF vs. EF, TD-IF vs. TD-EF) were quite robust, these findings should be replicated with a larger sample of subjects. Another limitation is that the imagined scenarios were not the same for each subject. Different scenarios were selected for each subject based on individual ratings in order to equate the emotional strength of the event (positive and negative) across subjects; however, there could have been other sources of variability introduced by this method, such as how much a given scenario resonated with subjects, which influenced network activation. Future studies would benefit from obtaining ratings of personal relevance for imagined events, or to have subjects imagine scenarios that they personally identify, in order to investigate the impact of this factor on brain activity and behavior. Finally, it was not possible to match the internal and external focus tasks for cognitive load or effort. Although ratings of ease, which served as a proxy for subjective effort, were not significantly different between IF and EF conditions, it is possible that cognitive load effects that were not assessed were influencing results. Despite these limitations, we have used a novel approach to identify robust and widespread changes in brain functioning based on prior attentional state. These results are significant for our understanding of the neural mechanisms associated with modes of processing, with particular relevance for psychopathological states characterized by excessive engagement in negative, IF cognition.

Supplementary Material

Supplementary material can be found at: http://www.cercor.oxfordjournals.org/.

Funding

The project described was supported by Grant UL1 TR000067/KL2 TR00069 from the National Center for Advancing Translational Sciences (NCATS), a component of the National Institutes of Health (NIH) to Mount Sinai/ERS.

Notes

Conflict of Interest: None declared.

References

Addis
DR
Cheng
T
Roberts
RP
Schacter
DL
2011
.
Hippocampal contributions to the episodic simulation of specific and general future events
.
Hippocampus
 .
21
:
1045
1052
.
Addis
DR
Wong
AT
Schacter
DL
2007
.
Remembering the past and imagining the future: common and distinct neural substrates during event construction and elaboration
.
Neuropsychologia
 .
45
:
1363
1377
.
Andrews-Hanna
JR
2012
.
The brain's default network and its adaptive role in internal mentation
.
Neuroscientist
 .
18
:
251
270
.
Andrews-Hanna
JR
Reidler
JS
Sepulcre
J
Poulin
R
Buckner
RL
2010
.
Functional-anatomic fractionation of the brain's default network
.
Neuron
 .
65
:
550
562
.
Anticevic
A
Cole
MW
Murray
JD
Corlett
PR
Wang
XJ
Krystal
JH
2012
.
The role of default network deactivation in cognition and disease
.
Trends Cogn Sci
 .
16
:
584
592
.
Behzadi
Y
Restom
K
Liau
J
Liu
TT
2007
.
A component based noise correction method (CompCor) for BOLD and perfusion based fMRI
.
Neuroimage
 .
37
:
90
101
.
Buckner
RL
Andrews-Hanna
JR
Schacter
DL
2008
.
The brain's default network: anatomy, function, and relevance to disease
.
Ann N Y Acad Sci
 .
1124
:
1
38
.
Buckner
RL
Vincent
JL
2007
.
Unrest at rest: default activity and spontaneous network correlations
.
Neuroimage
 .
37
:
1091
1096
.
Corbetta
M
Shulman
GL
2002
.
Control of goal-directed and stimulus-driven attention in the brain
.
Nat Rev Neurosci
 .
3
:
201
215
.
Coull
JT
Frackowiak
RS
Frith
CD
1998
.
Monitoring for target objects: activation of right frontal and parietal cortices with increasing time on task
.
Neuropsychologia
 .
36
:
1325
1334
.
D'Argembeau
A
Stawarczyk
D
Majerus
S
Collette
F
Van der Linden
M
Feyers
D
Maquet
P
Salmon
E
2010
.
The neural basis of personal goal processing when envisioning future events
.
J Cogn Neurosci
 .
22
:
1701
1713
.
D'Argembeau
A
Xue
G
Lu
ZL
Van der Linden
M
Bechara
A
2008
.
Neural correlates of envisioning emotional events in the near and far future
.
Neuroimage
 .
40
:
398
407
.
Fitzgerald
KD
Stern
ER
Angstadt
M
Nicholson-Muth
KC
Maynor
MR
Welsh
RC
Hanna
GL
Taylor
SF
2010
.
Altered function and connectivity of the medial frontal cortex in pediatric obsessive-compulsive disorder
.
Biol Psychiatry
 .
68
:
1039
1047
.
Fitzgerald
KD
Welsh
RC
Gehring
WJ
Abelson
JL
Himle
JA
Liberzon
I
Taylor
SF
2005
.
Error-related hyperactivity of the anterior cingulate cortex in obsessive-compulsive disorder
.
Biol Psychiatry
 .
57
:
287
294
.
Fransson
P
2006
.
How default is the default mode of brain function? Further evidence from intrinsic BOLD signal fluctuations
.
Neuropsychologia
 .
44
:
2836
2845
.
Friston
KJ
Buechel
C
Fink
GR
Morris
J
Rolls
E
Dolan
RJ
1997
.
Psychophysiological and modulatory interactions in neuroimaging
.
Neuroimage
 .
6
:
218
229
.
Friston
KJ
Zarahn
E
Josephs
O
Henson
RNA
Dale
AM
1999
.
Stochastic designs in event-related fMRI
.
Neuroimage
 .
10
:
607
619
.
Gao
W
Lin
W
2012
.
Frontal parietal control network regulates the anti-correlated default and dorsal attention networks
.
Human Brain Map
 .
33
:
192
202
.
Hamilton
JP
Furman
DJ
Chang
C
Thomason
ME
Dennis
E
Gotlib
IH
2011
.
Default-mode and task-positive network activity in major depressive disorder: implications for adaptive and maladaptive rumination
.
Biol Psychiat
 .
70
:
327
333
.
Harrison
BJ
Pujol
J
Soriano-Mas
C
Hernandez-Ribas
R
Lopez-Sola
M
Ortiz
H
Alonso
P
Deus
J
Menchon
JM
Real
E
et al
2012
.
Neural correlates of moral sensitivity in obsessive-compulsive disorder
.
Arch Gen Psychiatry
 .
69
:
741
749
.
Hassabis
D
Kumaran
D
Maguire
EA
2007
.
Using imagination to understand the neural basis of episodic memory
.
J Neurosci
 .
27
:
14365
14374
.
Ingvar
DH
1985
.
Memory of the future - an essay on the temporal organization of conscious awareness
.
Hum Neurobiol
 .
4
:
127
136
.
Kane
MJ
Brown
LH
McVay
JC
Silvia
PJ
Myin-Germeys
I
Kwapil
TR
2007
.
For whom the mind wanders, and when: an experience-sampling study of working memory and executive control in daily life
.
Psychol Sci
 .
18
:
614
621
.
Kelly
AM
Uddin
LQ
Biswal
BB
Castellanos
FX
Milham
MP
2008
.
Competition between functional brain networks mediates behavioral variability
.
Neuroimage
 .
39
:
527
537
.
Kirino
E
Belger
A
Goldman-Rakic
P
McCarthy
G
2000
.
Prefrontal activation evoked by infrequent target and novel stimuli in a visual target detection task: an event-related functional magnetic resonance imaging study
.
J Neurosci
 .
20
:
6612
6618
.
Kuelz
AK
Hohagen
F
Voderholzer
U
2004
.
Neuropsychological performance in obsessive-compulsive disorder: a critical review
.
Biol Psychol
 .
65
:
185
236
.
Lim
J
Wu
WC
Wang
J
Detre
JA
Dinges
DF
Rao
H
2010
.
Imaging brain fatigue from sustained mental workload: an ASL perfusion study of the time-on-task effect
.
Neuroimage
 .
49
:
3426
3435
.
Linden
DE
Prvulovic
D
Formisano
E
Vollinger
M
Zanella
FE
Goebel
R
Dierks
T
1999
.
The functional neuroanatomy of target detection: an fMRI study of visual and auditory oddball tasks
.
Cereb Cortex
 .
9
:
815
823
.
Mason
MF
Norton
MI
Van Horn
JD
Wegner
DM
Grafton
ST
Macrae
CN
2007
.
Wandering minds: the default network and stimulus-independent thought
.
Science
 .
315
:
393
395
.
Murphy
K
Birn
RM
Handwerker
DA
Jones
TB
Bandettini
PA
2009
.
The impact of global signal regression on resting state correlations: are anti-correlated networks introduced?
Neuroimage
 .
44
:
893
905
.
Murrough
JW
Iacoviello
B
Neumeister
A
Charney
DS
Iosifescu
DV
2011
.
Cognitive dysfunction in depression: neurocircuitry and new therapeutic strategies
.
Neurobiol Learn Mem
 .
96
:
553
563
.
Nagel
IE
Preuschhof
C
Li
SC
Nyberg
L
Backman
L
Lindenberger
U
Heekeren
HR
2011
.
Load modulation of BOLD response and connectivity predicts working memory performance in younger and older adults
.
J Cogn Neurosci
 .
23
:
2030
2045
.
Paus
T
Zatorre
RJ
Hofle
N
Caramanos
Z
Gotman
J
Petrides
M
Evans
AC
1997
.
Time-related changes in neural systems underlying attention and arousal during the performance of an auditory vigilance task
.
J Cogn Neurosci
 .
9
:
392
408
.
Ray
RD
Ochsner
KN
Cooper
JC
Robertson
ER
Gabrieli
JDE
Gross
JJ
2005
.
Individual differences in trait rumination and the neural systems supporting cognitive reappraisal
.
Cogn Affect Behav Ne
 .
5
:
156
168
.
Rugg
MD
Vilberg
KL
2013
.
Brain networks underlying episodic memory retrieval
.
Curr Opin Neurobiol
 .
23
:
255
260
.
Rypma
B
Berger
JS
D'Esposito
M
2002
.
The influence of working-memory demand and subject performance on prefrontal cortical activity
.
J Cogn Neurosci
 .
14
:
721
731
.
Schacter
DL
Addis
DR
Hassabis
D
Martin
VC
Spreng
RN
Szpunar
KK
2012
.
The future of memory: remembering, imagining, and the brain
.
Neuron
 .
76
:
677
694
.
Sharot
T
Riccardi
AM
Raio
CM
Phelps
EA
2007
.
Neural mechanisms mediating optimism bias
.
Nature
 .
450
:
102
105
.
Sheehan
DV
Lecrubier
Y
Sheehan
KH
Amorim
P
Janavs
J
Weiller
E
Hergueta
T
Baker
R
Dunbar
GC
1998
.
The Mini-International Neuropsychiatric Interview (M.I.N.I.): the development and validation of a structured diagnostic psychiatric interview for DSM-IV and ICD-10
.
J Clin Psychiatry
 .
59
(Suppl 20)
:
22
33
;
quiz 34–57
.
Shulman
GL
Fiez
JA
Corbetta
M
Buckner
RL
Miezin
FM
Raichle
ME
Petersen
SE
1997
.
Common blood flow changes across visual tasks: II. Decreases in cerebral cortex
.
J Cogn Neurosci
 .
9
:
648
663
.
Smallwood
J
Brown
K
Baird
B
Schooler
JW
2012
.
Cooperation between the default mode network and the frontal-parietal network in the production of an internal train of thought
.
Brain Res
 .
1428
:
60
70
.
Smallwood
J
Schooler
JW
2006
.
The restless mind
.
Psychol Bull
 .
132
:
946
958
.
Smith
SM
Fox
PT
Miller
KL
Glahn
DC
Fox
PM
Mackay
CE
Filippini
N
Watkins
KE
Toro
R
Laird
AR
Beckmann
CF.
2009
.
Correspondence of the brain's functional architecture during activation and rest
.
Proc Natl Acad Sci USA
 .
106
:
13040
13045
.
Spreng
RN
Stevens
WD
Chamberlain
JP
Gilmore
AW
Schacter
DL
2010
.
Default network activity, coupled with the frontoparietal control network, supports goal-directed cognition
.
Neuroimage
 .
53
:
303
317
.
Stawarczyk
D
Majerus
S
Maquet
P
D'Argembeau
A
2011
.
Neural correlates of ongoing conscious experience: both task-unrelatedness and stimulus-independence are related to default network activity
.
PLoS One
 .
6
:
e16997
.
Stern
ER
Fitzgerald
KD
Welsh
RC
Abelson
JL
Taylor
SF
2012
.
Resting state functional connectivity between fronto-parietal and default mode networks in obsessive-compulsive disorder
.
PLoS One
 .
7
:
e36356
.
Stern
ER
Welsh
RC
Fitzgerald
KD
Gehring
WJ
Lister
JJ
Himle
JA
Abelson
JL
Taylor
SF
2011
.
Hyperactive error responses and altered connectivity in ventromedial and frontoinsular cortices in obsessive-compulsive disorder
.
Biol Psychiatry
 .
69
:
583
591
.
Stern
ER
Welsh
RC
Gonzalez
R
Fitzgerald
KD
Abelson
JL
Taylor
SF
2013
.
Subjective uncertainty and limbic hyperactivation in obsessive-compulsive disorder
.
Human Brain Map
 .
34
:
1956
1970
.
Szmrecsanyi
B
2004
.
On operationalizing syntactic complexity
.
Journees Internationales D'Analyse Statistique des Donnees Textuelles
 .
2
:
1032
1039
.
Vanhaudenhuyse
A
Demertzi
A
Schabus
M
Noirhomme
Q
Bredart
S
Boly
M
Phillips
C
Soddu
A
Luxen
A
Moonen
G
et al
2011
.
Two distinct neuronal networks mediate the awareness of environment and of self
.
J Cogn Neurosci
 .
23
:
570
578
.
Vincent
JL
Kahn
I
Snyder
AZ
Raichle
ME
Buckner
RL
2008
.
Evidence for a frontoparietal control system revealed by intrinsic functional connectivity
.
J Neurophysiol
 .
100
:
3328
3342
.
Wen
X
Liu
Y
Yao
L
Ding
M
2013
.
Top-down regulation of default mode activity in spatial visual attention
.
J Neurosci
 .
33
:
6444
6453
.
Winkielman
P
Knutson
B
Paulus
M
Trujillo
JL
2007
.
Affective influence on judgments and decisions: Moving towards core mechanisms
.
Rev Gen Psychol
 .
11
:
179
192
.
Yun
RJ
Krystal
JH
Mathalon
DH
2010
.
Working memory overload: fronto-limbic interactions and effects on subsequent working memory function
.
Brain Imaging Behav
 .
4
:
96
108
.