Mind wandering is a natural, transient state wherein our neurocognitive systems become temporarily decoupled from the external sensory environment as our thoughts drift away from the current task at hand. Yet despite the ubiquity of mind wandering in everyday human life, we rarely seem impaired in our ability to adaptively respond to the external environment when mind wandering. This suggests that despite widespread neurocognitive decoupling during mind wandering states, we may nevertheless retain some capacity to attentionally monitor external events. But what specific capacities? In Experiment 1, using traditional performance measures, we found that both volitional and automatic forms of visual–spatial attentional orienting were significantly attenuated when mind wandering. In Experiment 2, however, ERPs revealed that, during mind wandering states, there was a relative preservation of sensitivity to deviant or unexpected sensory events, as measured via the auditory N1 component. Taken together, our findings suggest that, although some selective attentional processes may be subject to down-regulation during mind wandering, we may adaptively compensate for these neurocognitively decoupled states by maintaining automatic deviance–detection functions.
One of the odd quirks of human cognition is that we frequently get lost in our own trains of thought, even when doing attention-demanding tasks. When driving, for example, many people have had that unsettling experience of suddenly realizing that they've been completely tuned out for the past few miles, with little recollection of the traffic and terrain that's been navigated in the interim. But this raises a striking question regarding our natural propensity to have our thoughts drift off-task—how is it that our minds can regularly wander like this during on-going tasks, yet we still seem to retain some capacity to monitor and respond to the external environment? Is some ability to selectively attend to salient events in the outside world actually preserved when in mind wandering states?
The question is all the more perplexing given what we know about the effect of mind wandering on stimulus processing in cortex. When in mind wandering states, there is a significant reduction in the extent to which we cognitively analyze or process task-relevant events, relative to when in “on-task” attentional states (e.g., Barron, Riby, Greer, & Smallwood, 2011; O'Connell et al., 2009; Smallwood, Beach, Schooler, & Handy, 2008). Likewise, the initial sensory-evoked cortical activity engendered by task-irrelevant events also decreases, an effect observed in both the visual and auditory domains (e.g., Braboszcz & Delorme, 2011; Kam et al., 2011). Such evidence has suggested that mind wandering facilitates the production and maintenance of internal trains of thought by transiently “decoupling” neurocognitive systems from external stimulus inputs (e.g., Barron et al., 2011; Schooler et al., 2011; Smallwood et al., 2011; Smallwood, Obonsawin, & Heim, 2003). But if our thoughts become decoupled when mind wandering, do our attentional systems decouple as well?
Given this question, the goal of our study was to examine whether controlled or more volitional attentional functions change as we drift in and out of mind wandering states, and if so, how this compares to the possible effect of mind wandering on more automatic or reflexive attentional functions. In our first experiment, we thus asked participants to perform two different visual–spatial cuing tasks that required making manual responses to lateralized targets. One task involved volitional spatial orienting (e.g., Posner 1980), and the other task involved reflexive spatial orienting (e.g., Tipper, Handy, Giesbrecht, & Kingstone, 2008; Friesen & Kingstone, 1998). The volitional orienting of spatial attention involves individuals voluntarily shifting their attention elsewhere, whereas reflexive spatial orienting involves an external stimulus that attracts one's attention involuntarily. At unpredictable intervals during task performance, we stopped participants and asked them to report on their task-related attentional state-either “on-task” or “mind wandering.” We then examined the RTs to targets as a function of whether they were in cued or uncued spatial locations and whether they immediately preceded an “on-task” versus “mind wandering” report. If mind wandering disrupts visual selective attention, it is predicted that RTs should be faster for cued versus uncued targets just before “on-task” reports, but not just before reports of “mind wandering.” We predicted that these effects would be observed in both volitional and reflexive spatial orienting.
A total of 32 individuals participated (17 women; M = 20.83 years, SD = 1.42 years), with 17 participants completing the volitional spatial orienting task and 15 participants completing the reflexive spatial orienting task. All were right-handed and had normal or corrected-to-normal vision. They all gave written informed consent and were given course extracredit for their participation. This study was approved by the UBC Behavioral Review Ethics Board.
Stimuli and Procedure
Stimuli were presented on an 18-in. colored monitor, placed 80 cm away from the participants. In both tasks, a fixation dot first appeared in the center of the screen for 2000 msec. Following this, a cue was presented at fixation and remained on the screen for 1300 msec. The target, which was an X (1.1° × 0.9°), appeared either in the left or right visual field (5.2° from the left/right edge of the screen). It was presented 800 msec after the onset of the cue and lasted for 100 msec. The intertrial interval was randomly jittered between 1500 and 1700 msec, during which a response was made. For both tasks, participants were instructed to press a designated button as quickly and accurately as possible when the target appeared, regardless of its location.
In the volitional spatial orienting task, participants were instructed to attend to the left visual field if the inner circles were green and attend to the right visual field if the inner circles were red. The cue on each trial was predictive of the upcoming target location (left vs. right visual field) with 0.8 probability, thereby providing the incentive to volitionally orient attention to the cued location. The cue was two circles (2° × 2°) stacked vertically on top of each other, with two smaller inner circles (0.7° × 0.7°) colored either in green or red. In the reflexive spatial orienting task, an eye-gaze cue was made of two circles (1.9° × 1.9°) presented horizontally next to each other with small, black inner circles (0.9° × 0.9°), to mimic eyes. These “eyes” would either look to the left or to the right (see Figure 1). The direction of eye gaze was nonpredictive of the upcoming target location, in that the target was only presented at the gazed-at location with .5 probability (across the trial block). Our eye gaze paradigm is a canonical one that has been used to elicit reflexive orienting of attention (e.g., Tipper et al., 2008; Friesen & Kingstone, 1998, etc.). Participants were not given instructions as to where to attend, as previous evidence suggests that individuals reflexively orient their attention to the direction of others' eye gaze (Friesen & Kingstone, 1998).
To measure task-related attention, participants were instructed to report their attention state as either being “on-task” or “mind wandering” at the end of each trial block (cf. Kam et al., 2011; Smallwood et al., 2008). Participants were provided with definitions of these two attention states before starting the testing session. “On-task” states were defined as when one's attention was firmly directed toward the task, whereas “mind wandering” states were described as when one's attention was not focused on the task. At the conclusion of each trial block, attentional reports were recorded by the investigator, who was in the room with the participant throughout the experiment. These reports were then used to sort the RT data based on “on-task” versus “mind wandering” states. The block duration itself was randomly varied between 30 and 90 sec to minimize predictability of block completion and maximize variability of attentional state at the time of block completion. Given one trial lasted approximately 2900 msec, each block would have consisted between 10 and 30 trials.
We conducted an omnibus ANOVA that had Orienting Condition (volitional vs. reflexive) as a between-subject factor and both Selective Attention (cued vs. uncued) and Task-related Attention (on-task vs. mind wandering) as within-subject factors. The behavioral data for both cued and uncued conditions were based on averaging together the RT to the four targets preceding each of the two attentional state reports (on-task vs. mind wandering). Our analyses were based on the assumption that the 12 sec before each report would on average reliably capture the reported attentional state (cf. Kam et al., 2011; Smallwood et al., 2008), given recent evidence suggesting that the time course of off-task thinking approximates this time window (e.g., Christoff, Gordon, Smallwood, Smith, & Schooler, 2009; Sonuga-Barke & Castellanos, 2007). The number of events included in the averages was an attempt to maximize the number of events per each average without extending the window so far back in time as to consistently capture the preceding attentional state.
Participants completed an average of 30 trial blocks, of which 43.2% ended with an “on-task” report and 56.8% ended with a “mind wandering” report (SD = 19.37).
The RT data are shown in Figure 2 and suggest that attentional orienting effects were indeed attenuated when in “mind wandering” states in both orienting conditions. This pattern was confirmed statistically as we found significant main effects of Selective Attention, F(1, 30) = 15.38, p < .001, and Task-related Attention, F(1, 30) = 5.87, p = .02, and a significant interaction between the two, F(1, 30) = 6.33, p = .02. There were no significant interactions involving Orienting Condition nor a main effect of this factor (all p > .10).
Toward understanding precisely how attentional orienting was affected by mind wandering, we conducted two planned follow-up analyses for the significant Selective Attention × Task-related Attention interaction. The first examined whether selective attention effects were individually present under the “on-task” and “mind wandering” states. This analysis revealed a significant main effect of Selective Attention in “on-task” states (t(31) = 6.65; p < .001) but not in “mind wandering” states (t(31) = 0.93, p = .36). The second planned analyses examined whether RTs changed between “on-task” versus “mind wandering” states for cued trials, uncued trials, or both. This analysis revealed that RTs were significantly faster during “on-task” states versus “mind wandering” states for cued trials (t(31) = −4.33; p < .001) but not in uncued trials (t(31) = −0.78; p = .44).
We also examined the participants' anticipatory responses of targets preceding an on-task versus mind wandering report. Across both tasks, we found that responses made within the cue–target interval during on-task periods (M = 3.56) did not significantly differ from those made during mind wandering periods (M = 2.62; t(31) = 1.02, p = .32). Similarly, responses made within 150 msec of target onset during on-task periods (M = 8.66) did not significantly differ from those made during mind wandering periods (M = 7.94; t(31) = 0.58, p = .57).
Our findings from Experiment 1 indicate that visual–spatial attentional orienting attenuates during periods of mind wandering. This effect was observed in both volitional orienting elicited by arbitrary stimulus–response associations, as well as reflexive orienting elicited by social eyes-mimicking stimuli. Of relevance, this raises the question of how exactly mind wandering might disrupt attentional orienting, and whether this effect could be driven by sensory attenuation of the cue itself—a point to which we return in the General Discussion. Importantly, a key aspect of the data pattern suggest that this finding cannot simply be dismissed as participants having a reduced will or motivation to orient their visual–spatial attention during mind wandering attentional states. That is, the attenuation in orienting during mind wandering states was observed for both volitional and reflexive orienting conditions. If the effect of mind wandering we found was solely an issue of decreased motivation, it is difficult to understand how that would affect reflexive orienting, which is presumably not under volitional control. As such, our data support the hypothesis that both volitional and reflexive forms of top–down visual spatial orienting are diminished when mind wandering.
That mind wandering attenuated volitional visual–spatial orienting is certainly consistent with what is known about the cortical regions involved in both attentional control and mind wandering. In particular, volitional spatial orienting engages left dorsolateral pFC (e.g., Hopfinger, Buonocore, & Mangun, 2000; Corbetta, Miezin, Dobmeyer, Shulman, & Petersen, 1991), one of the key brain regions that has been shown to up-regulate activity during periods of mind wandering (e.g., Christoff et al., 2009; Mason et al., 2007). Likewise, neural areas involved in reflexive spatial orienting to eye gaze cues, including temporaparietal junction and STS (Tipper et al., 2008; Hooker et al., 2003), are also activated during mind wandering (Christoff et al., 2009; Mason et al., 2007). If mind wandering and spatial orienting do indeed engage a common set of neural regions, then this would explain the absence of spatial orienting during off-task attentional states—when mind wandering, the cortical areas necessary for visual spatial orienting are unavailable to support that function.
Although these data suggest that mind wandering does in fact disrupt visual-attentional orienting, one of the questions driving our study remained unanswered: Are there any attentional functions that may be preserved during mind wandering, functions that might allow us to adaptively respond to the outside environment despite our “cognitively decoupled” state? In Experiment 2, we thus examined the impact of mind wandering on a second form of attention that is qualitatively distinct from visual orienting-that of deviance detection.
The occurrence of a stimulus that deviates from the prevailing situational context unavoidably captures our attention. For example, when we hear a physically deviant sound embedded in a sequence of repetitive, standard sounds, it automatically triggers a deviance detection mechanism the activity of which is indexed by an ERP component known as the MMN (Escera & Corral, 2007; Escera, Alho, Winkler, & Näätänen, 1998; Näätänen, 1990; Näätänen, Gaillard, & Mantysalo, 1978). Importantly, the major region that contributes to the generation of the MMN, namely bilateral supratemporal cortex (Giard et al., 1995; Giard, Perrin, Pernier, & Bouchet, 1990; Scherg, Vajsar, & Picton, 1989), has not been implicated in mind wandering nor stimulus-independent thoughts (Christoff et al., 2009; Mason et al., 2007). Accordingly, in Experiment 2, we examined the ERP responses to standard and deviant auditory tones as a function of whether or not participants were in a mind wandering state. If deviance detection is indeed preserved during mind wandering, it predicted that mind wandering should differentially affect the ERP responses to the standard versus deviant tones.
Twenty participants (13 women, 7 men; M = 24.6 years, SD = 7.1 years) completed the experiment in exchange for $20 (Canadian dollars). They were all right handed, with no history of neurological problems and had normal or corrected-to-normal vision. Participants provided written informed consent to the experimental procedure, according to the guidelines of the UBC Behavioral Review Ethics Board.
Stimuli and Paradigm
Participants were presented auditory stimuli while they read a book, which is a part of the MMN protocol commonly used as a control condition (Sussman, Winkler, & Wang, 2003; Escera et al., 1998; Näätänen, Paavilainen, Tiitinen, Jiang, & Alho, 1993). The book that each participant read was Francis Bacon's Essays. This book presented the author's theory on various topics, and participants were allowed to read whichever chapter interested them. They were told to concentrate on reading the book and to ignore the tones presented throughout the experimental session.
Each trial consisted of an auditory stimulus that was either a standard or deviant tone with duration of 200 msec and presented at 80 dB SPL through headphones. The intertrial interval was randomly jittered between 500 and 700 msec, thus SOA was approximately 800 msec. Standard tones were 600 Hz, and deviant tones were 800 Hz, both presented in a random order with probabilities of 0.8 and 0.2, respectively.
Our measure of task-related attention is identical to the methods used in Experiment 1. Specifically, participants were asked to report their attention state as either being “on-task” or “mind wandering” at the end of each trial block. Given that each trial lasted approximately 800 msec and that each block lasted 30–90 sec, each trial block consisted between 38 and 112 trials. Each participant was tested in a single session lasting approximately 2 hr in total. The EEG setup and the experimental task each lasted about 60 min. There were no official breaks during the task; however, participants were allowed to rest briefly if they requested.
Electrophysiological Recording and Analysis
During task performance, EEGs were recorded from 64 active electrodes using a Biosemi Active-Two amplifier system (Amsterdam, The Netherlands). Two additional electrodes located over medial–parietal cortex (Common Mode Sense and Driven Right Leg) were used as ground electrodes. All EEG activities were amplified with a band-pass filter of 0.1–30 Hz, digitized on-line at a sampling rate of 256 samples per second. To ensure proper eye fixation and allow for the removal of events associated with eye movement artifacts, vertical and horizontal EOGs were also recorded—the vertical EOGs from an electrode inferior to the right eye and the horizontal EOGs from two electrodes on the right and left outer canthus. Offline, computerized artifact rejection was used to eliminate trials during which detectable eye movements and blinks occurred. These eye artifacts were detected by identifying the minimum and maximum voltage values on all recorded EOG channels from −50 to 600 msec poststimulus for each event epoch and then removing the trial from subsequent signal averaging if that value exceeded 150 μV, a value calibrated to capture all blinks, saccades, and other eye movements exceeding approximately 1 degree of visual angle. An average of 17% of the total number of trials across participants were rejected because of these signal artifacts. The percentage of trials rejected in the on-task versus mind wandering condition did not significantly differ from each other (t(19) = 0.53, p = .70). For each participant, ERPs for each condition of interest were averaged into 3000-msec epochs, beginning 1500 msec before stimulus onset. Subsequently, all ERPs were algebraically rereferenced to the average of the left and right mastoid signals and filtered with a low-pass Gaussian filter (25.6 Hz half-amplitude cutoff) to eliminate any residual high-frequency artifacts in the waveforms. The resulting ERPs were used to generate grand-averaged waveforms.
All ERP data analyses were based on mean amplitude measures using repeated-measures ANOVAs, with specific time windows of analyses identified below as per each reported ANOVA. These analysis time windows were centered on the peak of the relevant component as identified at each electrode site in the grand-averaged waveform. Statistical quantification of ERP data was based on mean amplitude measures relative to a −200 to 0 msec prestimulus baseline. The ERP waveforms for each condition of interest were based on averaging together the EEG epochs for the 15 tones preceding each of the two attentional state reports (on-task vs. mind wandering). Our analyses were based on the same assumption we made in Experiment 1 that the 12 sec before each report would reliably capture the given attentional state (cf. Kam et al., 2011; Smallwood et al., 2008).
Participants completed an average of 41 trial blocks, of which 61% ended with an “on-task” report and 39% ended with a “mind wandering” report (SD = 10.1).
Analysis of the ERP data focused on two issues a priori. First, we wanted to determine how, if at all, mind wandering affected the automatic detection of deviant auditory signals, as measured via the MMN. Second, to assess whether participants were in fact reliably reporting their on-task versus mind wandering states, we wanted to examine the amplitude of the sensory-evoked auditory N1 component, a component known to attenuate in amplitude during mind wandering states (Kam et al., 2011). To address these issues, we conducted repeated-measures ANOVAs that included factors of Attention State (on-task vs. mind wandering) as well as Electrodes and Participants; however, for brevity, no main effects or interactions involving Electrode or Participants are reported. The ANOVA examining the N1 component also included a within-subject factor of Tones (standard vs. deviant).
The main focus of ERP data analysis was to assess the effects of task-related attention on deviance detection, as indexed by the MMN elicited by deviant tones. ERP waveforms for “on-task” and “mind wandering” states were based on averaging standard/deviant tones presented within the last 12 sec of each trial block, as described above. As such, we first derived difference waveforms by subtracting the standard tones averaged waveforms from the deviant tones averaged waveforms (e.g., Escera et al., 1998; Sams, Paavilainen, Alho, & Näätänen, 1985; Näätänen et al., 1993) for the on-task and mind wandering conditions. Next, the MMN mean amplitude of the difference waveforms was then statistically compared between on-task and mind wandering conditions. The MMN of the difference waveforms as a function of attentional state was examined at midline frontocentral scalp electrode sites, Fz and Cz, where the amplitude of the MMN is typically maximal (e.g., Escera et al., 1998; Sams et al., 1985). In examining the difference waveform, we observed that the peak of the MMN (90–150 msec) coincided with the peak of N1 (95–115 msec). That our MMN peaked earlier than the typical MMN at 150–250 msec is consistent with past studies that suggested peak latency does get shorter with greater magnitude of stimulus change (Amanedo & Escera, 2000; Tiitinen, May, Reinikainen, & Näätänen, 1994; Näätänen, Paavilainen, Alho, Reinikainen, & Sams, 1989). That is the case for our standard (600 Hz) and deviant (800 Hz) stimuli, whereas in other studies, the difference in tones is around 100 Hz (e.g., Escera et al., 1998). Nevertheless, given the overlap in time window and owing to the fact that in subtraction waveforms like the MMN, the variance of the waveform is the sum of the variance in the two parent waveforms (e.g., Picton et al., 2000), we elected to focus our analysis on the N1 waveforms.
We wanted to examine whether there was a normal sensory attenuation of the auditory stimuli during mind wandering periods, as would be predicted by our previous finding (Kam et al., 2011). As such, we compared the N1 component to both tones during on-task versus mind wandering states. Specifically, we conducted repeated-measures ANOVAs that included factors of Attentional State (on-task vs. mind wandering) and Tones (standard vs. deviant) to examine the interaction between the two. The N1 elicited by both standard and deviant tones as a function of attentional state are shown in Figure 3A and B and was examined at midline frontocentral scalp electrode sites, Fz and Cz, where the amplitude of the N1 is typically maximal (e.g., Woldorff & Hillyard, 1991). The N1 mean amplitudes and standard errors of the mean are shown in Table 1. Mean amplitude measures were taken across a 95–115 msec poststimulus time window. Although the main effect of Attention was not significant (F < 1.00), the main effect of Tones was significant, F(1, 19) = 38.24, p < .001. Importantly, we found a significant interaction between Attention State and Tones, F(1, 19) = 5.06, p = .037. Separate analyses revealed a significant main effect of Attention State for standard tones, F(1, 19) = 10.84, p < .005, but not deviant tones, F(1, 19) = 0.19, p = .669. Specifically, although the N1 elicited by standard tones was significantly greater during “on-task” states relative to mind wandering states, this difference was absent for deviant tones.
|Standard tones||Fz||−1.93 (0.20)||−1.20 (0.35)|
|Cz||−1.87 (0.16)||−1.30 (0.19)|
|Deviant tones||Fz||−2.91 (0.32)||−3.11 (0.46)|
|Cz||−2.44 (0.28)||−2.61 (0.29)|
|Standard tones||Fz||−1.93 (0.20)||−1.20 (0.35)|
|Cz||−1.87 (0.16)||−1.30 (0.19)|
|Deviant tones||Fz||−2.91 (0.32)||−3.11 (0.46)|
|Cz||−2.44 (0.28)||−2.61 (0.29)|
Amplitudes and standard errors of each component at electrode sites Fz and Cz are presented as a function of attention states (on-task vs. mind wandering).
Given the results reported above, we wanted to examine an additional control issue concerning our findings. Specifically, we observed residual noise in the pre-N1 portion of the ERP waveforms, and the waveforms for the deviant tones in particular, which may have impacted the reliability of our results. As such, we bootstrapped our current findings with 20 participants to empirically create more samples to establish the reliability of our data. Bootstrapping is a nonparametric approach to analyzing ERP data without assuming normality of the sampling distribution (e.g., Keselman, Wilcox, & Lix, 2003; Wasserman & Bockenholt, 1989). This procedure requires resampling of data with replacement and leads to more accurate inferences (Fox, 2002).
First, we determined the average N1 amplitude at both electrodes (i.e., Fz and Cz) in the specified time window used in our analyses for each subject, for both standard and deviant tones during both on-task and mind wandering conditions. Subsequently, we computed the difference between on-task and mind wandering states to represent the attentional difference for both standard and deviant tones. We then performed a bootstrap simulation that involves resampling our participants' data with replacement for both the standard and deviant tones difference scores. This process involves creating a large number of “bootstrap samples” of 20 data points, with each data point chosen randomly and independently from the original set of 20 data points with replacement. Each participant's data have an equal chance of being chosen at each random draw, and each data point can be selected more than once in each bootstrap sample data set (Wilcox, 2001).
In generating 4999 bootstrap samples (Fox, 2002) and computing a mean of the scores from each sample, a bootstrapped sampling distribution is formed and the 95th percentile confidence intervals allow one to make inferences about the statistic at hand. In our case, we wanted to determine whether the difference between on-task and mind wandering states is significant for the standard tone, as well as the deviant tone. Because we created bootstrap samples on a difference score, a confidence interval that includes 0 suggests that the difference between the two attention states can be 0, thereby leading to the conclusion of retaining the null hypothesis. On the other hand, if the confidence interval does not include 0, then the two attention states is significantly different from each other, and thus one can reject the null hypothesis.
The 95th percentile confidence interval for standard tones is [−1.2236, −0.1407] at Fz and [−1.0417, −0.3600] at Cz, and the 95th percentile confidence interval for deviant tones is [−0.9383, 1.2428] at Fz and [−0.4845, 0.8545] at Cz. That the confidence intervals at both electrodes for standard tones do not include 0 suggest there is a significant difference in the N1 amplitude between on-task and mind wandering conditions. On the other hand, the confidence intervals for deviant tones do include 0, suggesting the difference in N1 between the two attentional states was not significant at either Fz or Cz. Both conclusions are consistent with our omnibus N1 ANOVA and follow-up analyses.
In Experiment 2, we found that there was an increase in the relative saliency of deviant events regardless of attention state and a decrease in sensory-related processing of standard events during mind wandering, as measured in the N1 ERP component. This suggests that despite the characteristic down-regulation of sensory processing in auditory cortex during mind wandering (Kam et al., 2011), there is a preserved ability to detect deviant events. We support and expand on these conclusions in the General Discussion.
This study examined whether some ongoing attentional functions are preserved while we are mind wandering. In Experiment 1, using traditional performance measures, we found that mind wandering appears to disrupt the strategic orienting of visual spatial attention, both volitionally and reflexively. In Experiment 2, however, using ERP-based measures, we found that mind wandering maintained the salience of unexpected or deviant auditory events but decreased the sensory responses to standard tones, as measured via the N1 ERP component. Taken together, what our data suggest is that our attentional systems adaptively respond to mind wandering, such that under conditions leading to a down-regulation of strategic spatial orienting, there is a preservation of more automatic deviance detection. Given our conclusion, several key issues and questions arise.
First, to what extent is the task-related attention effect in Experiment 1 because of sensory and/or cognitive attenuation of the cue during mind wandering, as opposed to a direct impairment of attentional orienting per se, as we would prefer to conclude? Specifically, if mind wandering did attenuate sensory processing of the cue, the disruption in attentional orienting could be a result of the cue not being processed sufficiently at a sensory (e.g., Kam et al., 2011) and/or cognitive (e.g., Smallwood et al., 2008) level. Two lines of evidence suggest, however, that this may be a less than complete account of our behavioral findings. For one, the fMRI-based findings of Christoff et al. (2009) and Mason et al. (2007) indicate that periods of mind wandering are also associated with a down-regulation of activity in the same left prefrontal cortical regions consistently associated with the top–down control of selective visual attention (e.g., Hopfinger et al., 2000; Corbetta, Miezing, Shulman, & Petersen, 1993). This indicates that mind wandering has the capacity to directly impact the top–down control of attention itself, regardless of any concomitant effects on the sensory and/or cognitive processing of external stimuli. Second, the sensory attenuation reported in Kam et al. (2011) was found specifically for task-irrelevant stimuli in the upper visual periphery, whereas Smallwood et al. (2008) found no sensory attenuation for task-relevant stimuli at fixation. Given that our attention-directing cues were presented at fixation, this would appear to mitigate at least purely sensory-level accounts of our Experiment 1 findings. Regardless of these specific possibilities though, the broader implication is that mind wandering is likely able to exert an impact on attentional orienting in both ways—via reducing the extent to which attentionally imperative stimuli are processed, as well as directly down-regulating activity in attentional control regions of left pFC.
Second, our finding of an attenuation of automatic attention orienting during mind wandering may be limited to the specific stimulus set used in our study. To facilitate automatic spatial orienting, we used cues that mimic eyes. This particular set of orienting stimulus have successfully elicited eye gaze orienting in past studies (e.g., Tipper et al., 2008; Friesen & Kingstone, 1998). Notably, such reflexive orienting can arise along a number of other dimensions and can be elicited by various types of stimuli, whether it be visual (e.g., arrows) or auditory (e.g., loud sounds). Therefore, future studies are needed to elucidate the effects of mind wandering of these other types of stimuli.
A third issue concerns what our findings specifically reveal about the adaptive nature of attentional control during mind wandering. Our data may shed light on why our minds can frequently wander, and yet we are still capable of monitoring and responding to the external environment. In particular, that deviance detection is preserved during mind wandering suggests that we seem to be designed to operate on autopilot during mind wandering but become more automatically vigilant for things out of the ordinary. This would imply, for example, that when mind wandering while driving, we would continue to mind wander so long as the traffic patterns and behaviors of other drivers remain somewhat normal and expected. However, if something unusual were to happen, such as a driver suddenly veering ahead, our data predict that such events should be readily detected, thereby allowing us to emerge out of our state of automaticity and instead adaptively respond to the deviant event.
Fourth, we found that the N1 to deviant events was preserved during mind wandering, whereas Braboszcz and Delorme (2011) reported an attenuation of the MMN to deviant events during mind wandering compared with breath focus. Although this appears to stand in contrast with our finding of a preserved N1 during mind wandering, there are several reasons that can account for this difference. First, various differences between our studies may have contributed to the potential differences observed. For example, although innovative, the authors considered data occurring both before and after an attention report. Following the interruption of the actual report of one's attention, it would be difficult to determine whether the participant is still in the reported attention state or whether that interruption would put the participant back in focus even if temporarily. In contrast, we only examined the data immediately preceding an attention report. Furthermore, the condition of mind wandering was compared with the condition of “breath focus.” Participants were instructed to count their breath cycles with their eyes closed and to ignore the auditory tones. On the other hand, we were comparing between conditions of attention directed toward the main task, reading, and attention away from the task. In other words, our main task required attention to an external, visual stimulus, whereas the “breath focus” condition required attention to an internal, nonvisual stimulus.
A fifth issue concerns whether mind wandering differentially modulates processing of stimuli presented in different modalities. In Experiment 1, mind wandering disrupted spatial orienting to the visual stimuli, whereas in Experiment 2, mind wandering preserved the detection of the deviant auditory stimuli. At first glance, this may seem to suggest visual and auditory processing are differentially altered by mind wandering. Nevertheless, previous evidence has suggested that mind wandering attenuates sensory processing in the same manner whether the stimuli were presented in the visual or auditory modality (Kam et al., 2011). Perhaps the crucial factor modulating stimulus processing does not lie in the modality of stimulus, but whether the external stimulus was more worthy of our attention than our internal thoughts at any given time. In particular, it has been suggested that our minds generally shield us from mundane sensory events to facilitate internal thoughts (Barron et al., 2011; Schooler et al., 2011). This is consistent with findings from Experiment 1, where processing of the anticipated stimuli was attenuated. However, when a change occurs in the environment, for example, an unexpected stimulus that is potentially harmful or dangerous, we may automatically shift attention from our internal thoughts to the external environment, as observed in Experiment 2 in our response to the deviant stimuli. Together, this suggests that our minds may engage in this ongoing automatic evaluation of the importance of both external and internal stimuli, at which point our attention is then allocated accordingly.
Lastly, our findings from Experiment 1 raise an important issue concerning the interaction between mind wandering and selective attention. Canonical models of both spatial- (e.g., Corbetta & Shulman, 2002; Posner 1980) and object-based (e.g., Duncan, Humphreys, & Ward, 1997; Desimone & Duncan, 1995) attentional selection implicitly assume that we are always selecting something from the external environment for higher levels of cognitive analysis. But if mind wandering disrupts selective attention, it would suggest that there are systematic periods of time when we select nothing from the external environment. This is consistent with our previous finding that mind wandering attenuates initial level sensory processing of stimuli (Kam et al., 2011). Our study furthers this idea in two ways. First, we demonstrated that we don't simply shut down sensory processing per se, but we do in fact shut down the spotlight. Second, we also showed that we compensate for this dampened spotlight by preserving our sensitivity to the unusual events. That is, we don't deliberately select for anything to process, but we maintain vigilance instead.
Reprint requests should be sent to Julia W. Y. Kam, Department of Psychology, 2136 West Mall, Vancouver, British Columbia, Canada, V6T 1Z4, or via e-mail: Julia@psych.ubc.ca.