Predicting task-general mind-wandering with EEG

Abstract

Mind-wandering refers to the process of thinking task-unrelated thoughts while performing a task. The dynamics of mind-wandering remain elusive because it is difficult to track when someone's mind is wandering based only on behavior. The goal of this study is to develop a machine-learning classifier that can determine someone's mind-wandering state online using electroencephalography (EEG) in a way that generalizes across tasks. In particular, we trained machine-learning models on EEG markers to classify the participants' current state as either mind-wandering or on-task. To be able to examine the task generality of the classifier, two different paradigms were adopted in this study: a sustained attention to response task (SART) and a visual search task. In both tasks, probe questions asking for a self-report of the thoughts at that moment were inserted at random moments, and participants' responses to the probes were used to create labels for the classifier. The 6 trials preceding an off-task response were labeled as mind-wandering, whereas the 6 trials predicting an on-task response were labeled as on-task. The EEG markers used as features for the classifier included single-trial P1, N1, and P3, the power and coherence in the theta (4-8 Hz) and alpha (8.5-12 Hz) bands at PO7, Pz, PO8, and Fz. We used a support vector machine as the training algorithm to learn the connection between EEG markers and the current mind-wandering state. We were able to distinguish between on-task and off-task thinking with an accuracy ranging from 0.50 to 0.85. Moreover, the classifiers were task-general: The average accuracy in across-task prediction was 60%, which was above chance level. Among all the extracted EEG markers, alpha power was most predictive of mind-wandering.

Keywords: Alpha oscillations; EEG; Mind-wandering; Single-trial ERP; Spontaneous thought; Support vector machine; Sustained attention to response task.

Fig. 1

Experimental procedure. a In the SART, every trial started with a fixation cross, followed by a word for 300 ms and a mask for 900 ms. There was a 3-s blank as the intertrial interval (ITI). Two types of stimuli are illustrated: a lowercase word (tea) as the go stimuli, and an uppercase word (OFTEN) as the no-go stimulus, which was the target. Probes always occurred after a no-go trial. b In the visual search task, every trial started with a fixation cross, followed by a search panel for 3 s. Two consecutive probes were separated by 7 ~ 24 trials. A visual search target was present on half of the trials and absent on the other half. NT = nontarget; SP = search panel. (Color figure online)

Fig. 2

a Mexican-hat wavelet (t = 446, s = 1,188), using the parameters detected as the local extreme in c. b An example of EEG epoch time-locked to stimulus onset. c The resulting W-value matrix shown in a contour map when doing the template matching using the trial in b. The local extreme detected in the time window of 250 ms ~ 600 ms indicates the single-trial P3. (Color figure online)

Fig. 3

Behavioral results by task. Bars show the behavioral difference (MW minus OT) between conditions. Error bars indicate the 95% confidence interval. ACC = accuracy; RT = response time; MW = mind-wandering; OT = on-task; SART = sustained attention to response task; VS = visual search task

Fig. 4

Classifier performance for each participant shown by (a) prediction accuracy obtained from the within-task leave-one-out cross validation (LOOCV) and across-task predictions, and (b) accuracy, sensitivity, and specificity. Maroon horizontal dashed line in a indicates chance level. (Color figure online)

Fig. 5

Correlation between mind-wandering rate, sensitivity, and specificity. Shaded area indicates the 95% confidence interval. SART = sustained attention to response task; VS = visual search task. A mind-wandering rate of 1 indicates the participant is mind-wandering every time a probe is presented, whereas a mind-wandering rate of 0 indicates the participant is never mind-wandering when the probe is presented

Fig. 6

a Performance of single-marker classifiers shown as mean accuracy across individuals. Whole model at the bottom refers to the modeling performance with all the EEG markers listed above as predictors. Error bars indicate 95% confidence interval. Black vertical dashed line indicates the chance level. b Selected channels to examine in the 128-channel Biosemi system in the upper panel and their approximate locations in the 10–20 system in the lower panel. (Color figure online)

Fig. 7

Visualization of the EEG markers in the mind-wandering (MW) and on-task (OT) state. a Group averaged ERP wave graph computed by both the single-trial algorithm and the traditional averaging method. Shaded area in the waveform shows the standard error. b Group mean of the normalized power and intersite-phase clustering (ISPC) of baseline (−400 ms ~ 0 ms) and after-stimulus onset (ASO, 0 ms ~ 600 ms). Error bar indicates one standard error of the mean. (Color figure online)