Mind wandering in sensory cortices

Abstract

The recent task-free approach in Cognitive Neuroscience has sparked interest in understanding the brain's default mode network (DMN). One particular mental activity that has been identified to recruit such a network is mind-wandering, which points to the functional aspect of mind-wandering as a default system. However, the phenomenological aspect of mind-wandering has been missing in the literature on brain imaging. To tackle this issue, we adopted online thought sampling while participants underwent a simple fixation task over multiple sessions in the scanner. During 10 h of scanning of each participant, over 200 mind wander episodes were labelled in each participant. With linear support vector machine classification on mind-wandering episodes with exclusive sensory content, we found that decoding accuracy in content-corresponding sensory cortices was significantly higher, indicating the neural bases of the phenomenology of mind-wandering. Unique patterns in classification were revealed in different individuals, pointing to individual variances in our phenomenal experiences.

Keywords: Mind-wandering; Phenomenology; Sensory cortices; Support vector machine.

Fig. 1

Experiment procedure. Each participant underwent 5 main mind-wandering experiment sessions and 1 localizer scan. Each lasted 2 h. In each session, approximately 10 runs were collected. With each run, 9 TRs precedent to the probe was categorized as a non-mind-wandering or a wandering event and labelled with respective sensory content. Five regions including Auditory Cortex (AC, in pink), striate cortex (V1, in purple), FFA (in orange), LOC (in blue), and PPA (in green) were functionally localized in each individual in the separate localizer scan. The multivoxel classification analyses were performed in these sensory ROIs. (For interpretation of the references to color in this figure legend, the reader is referred to the Web version of this article.)

Fig. 2

Online content sampling procedure. The participant underwent a fixation task in each 8-min run. Every 45–90 s, adjusted according to each individual’s mind-wandering frequency, a question popped out and asked Q1: How focused were you on the task? If the answer was 5–8, this probe was categorized as a non-mind-wandering event, and the participant resumed the fixation task immediately. If the answer was 1–4, the participant proceeded to report a series of questions regarding the mind-wandering content. In Q6, if either of the three answers (visual/auditory/emotional) was chosen, a corresponding subsequent subcategory question would be asked to gather the details of sensory content. Participants were allowed to report their mind-wandering event if they caught it voluntarily, however, no such event took place in our study.

Fig. 3

Decoding accuracy (ACC) of three individuals in sensory ROIs. The decoding ACC was compared against the baseline where the labels were randomly shuffled. The top half of the figure denotes decoding ACC while the bottom half shows the decoding ACC differences between correctly versus randomly labelled in bar plot. Top. Decoding ACC of visual-only vs. auditory-only sensory events in 5 respective sensory regions. Bottom. Decoding ACC of face-only vs. object-only sensory events in 3 respective sensory regions. Asterisk denotes significance (all paired t, p < 0.00001, significant after Bonferroni correction). Error bars denote 99.9% confidence intervals.