The benefits of mind wandering on a naturalistic prospective memory task

The study investigates whether and how mind wandering (MW) during a retention interval benefits prospective memory (PM), the capacity to remember to execute intended actions at the appropriate time or in response to specific events. MW is prevalent in daily life and its definition spans spontaneous and voluntary task-unrelated, stimulus-independent thoughts with debated cognitive consequences. Despite commonly reported costs to attention and memory encoding, MW frequently exhibits a prospective bias toward future-oriented content, suggesting a potential functional role in planning and goal pursuit. Prior naturalistic work hinted that past-oriented MW may relate to improved PM execution, possibly via reactivation or reconsolidation of intentions, but lacked experimental control. The authors hypothesized that: (1) increasing MW frequency during the PM retention phase would improve PM performance; (2) spontaneous past-oriented MW would preferentially benefit the retrospective component (the “what”) of intentions encoded before the MW manipulation; and (3) spontaneous future-oriented MW (episodic future thinking, EFT) would preferentially benefit the prospective component (the “when/where”), particularly for pre-encoded intentions. They further reasoned that such effects should not apply to intentions encoded after the MW manipulation if the mechanism reflects reconsolidation. To test these predictions in an ecologically valid yet controlled setting, they used immersive virtual reality (VR) and manipulated MW frequency via working-memory load during the retention interval.

Theoretical accounts of MW vary, emphasizing either its dynamic nature or content (task-unrelated and stimulus-independent thoughts), and differentiating spontaneous versus voluntary episodes. MW has often been associated with cognitive costs (e.g., compromised attention and encoding), yet a functional perspective highlights roles in planning and goal maintenance. MW frequently shows a prospective bias: off-task thoughts are often near-future and goal-directed, implicating inner speech and planning mechanisms. However, some studies observe substantial past-oriented MW focused on recalling prior events. Past-oriented MW may refresh or reconsolidate memories via mental imagery, potentially strengthening intention representations during PM retention. PM itself involves forming an intention (prospective component: when/where) and remembering its content (retrospective component: what), and performance depends on executive and attentional resources and on the detection of cues during ongoing activities. Prior work is sparse and mixed on MW–PM links: experience sampling did not find an association between future-oriented MW and PM but did find an association with past-oriented thoughts, suggesting consolidation of programmed intentions. Task demand modulates MW frequency (harder tasks reduce MW). Immersive VR provides enhanced ecological validity for PM assessment, bridging the gap between lab tasks and real-life measures.

Design and participants: Sixty healthy French-speaking adults (mean age 22.08 ± 4.76 years; screened for normal/corrected vision, no neurological/psychiatric history, low motion-sickness risk, non-expert meditators) were randomly assigned to Low Cognitive Load (EASY; n=30) or High Cognitive Load (HARD; n=30) to manipulate MW frequency during the PM retention interval. Groups were matched on demographics, affect, executive functions, metamemory, and MW trait. Apparatus and VR: Tasks ran on a PC; MW induction administered via Python/Neuropsydia; immersive VR used HTC Vive Pro with a custom Unity 2021.1.20 Paris-like city including navigable landmarks, ambient sounds, and interactive cues. A virtual watch displayed in-environment time and a route list. MW manipulation and assessment: During the retention interval, participants performed an n-back classification task with moving item images: EASY = 1-back, HARD = 3-back. There were two blocks (Easy/Hard design between subjects), 48 trials each, with 12 pseudorandom thought probes per block. Probes queried attention state (Focus, MW, External); if MW, they probed source (Spontaneous vs Voluntary) and content: Episodic Past Thinking (EPT), Episodic Future Thinking (EFT), Planning, Imagination, or Other. MW frequency was calculated as the proportion of MW endorsements across probes. PM encoding and task: Before MW induction, participants learned 15 PM items (5 event-based semantically linked, EBL; 5 event-based non-linked, EBN; 5 time-based, TB). Each item paired a cue (place or time) with an action; cued recall immediately followed encoding with up to three learning attempts. After a 20-minute delay containing the MW task, participants navigated the VR city for ~20 minutes performing ongoing tasks and PM retrieval. Nine additional PM items (3 EBL, 3 EBN, 3 TB) were introduced during navigation via avatar interactions or virtual SMS (post-PM intentions). Ongoing tasks included: (1) navigation along a route of daily-life activities and (2) listening to a podcast with 20 comprehension questions. PM retrieval and scoring: For EB cues, participants approached a location and triggered a prompt to recall the action aloud; for TB cues, they verbally produced the action at the correct time using the virtual watch. Each PM item earned up to 2 points: 1 for the prospective component (correct cue detection/timing) and 1 for the retrospective component (correct action). After VR, delayed cued recall assessed action memory to control for retrospective encoding differences. The rater was blind to MW manipulation. Procedure: The session (~93 ± 14 min) included consent, questionnaires (MSSQ, PHQ-4), cognitive tests (Stroop, Switch, SimAct), VR training, PM encoding, MW induction with probes, VR PM task with ongoing tasks, post-task questionnaires (Presence, Cybersickness, MWQ, MPMI-s), and delayed cued recall. Data analysis: Mixed ANOVAs tested effects of Condition (Easy vs Hard) and within-subject factors: attention Orientation (Focus, MW, External), MW Content and Source, and PM Type (EBL, EBN, TB) separately for items encoded pre- and post-MW. Multiple linear regressions (backward) predicted PM performance (prospective vs retrospective components for pre/post and cue types) from MW trait, global MW frequency, MW content×source ratios, metamemory, affect, and executive function covariates. Effect sizes (partial eta squared) and Bonferroni-corrected post hocs were reported.

Manipulation checks and MW: HARD showed lower n-back performance than EASY (slower RTs to targets and lures: t≈3.49 and 3.67, p<0.001; lower hit rate t=3.325, p=0.002), confirming task difficulty. MW frequency was higher in EASY (32.8%) than HARD (17.8%), t(58)=3.323, p=0.002. MW trait (MWQ) correlated with probe-based MW frequency, r=0.232, p=0.004. Attention Orientation showed a main effect (Focus > MW > External), and Condition×Orientation interaction: HARD had more Focus and less MW than EASY. MW content and source: MW episodes were more often Planning than EFT, EPT, Imagination, or Other (F(4,232)=13.965, p<0.001). Overall MW ratio higher in EASY than HARD; no Content×Condition interaction. Source×Condition interaction: MW was more voluntary in EASY than HARD (p<0.001); spontaneous MW did not differ between groups. VR controls and ongoing tasks: No EASY/HARD differences in presence, cybersickness, or time in VR; podcast comprehension did not differ. Cued recall of actions (post-navigation): Pre-encoded items: TB actions were recalled less than EBN and EBL; no Condition effect or interaction. Post-encoded items: TB < EBL; EBN < EBL; EASY > HARD overall (p=0.033); no interaction. PM performance (sum of prospective+retrospective, max 2 per item): Pre-PM: EASY > HARD (F(1,58)=5.996, p=0.017). Cue-type effect: EBL > EBN > TB. For components, both retrospective and prospective components were higher in EASY (retro: F=4.132, p=0.047; pro: F=6.169, p=0.016). EBL outperformed EBN and TB in both components; TB consistently poorest. Post-PM: EASY > HARD overall (F(1,58)=6.184, p=0.016). Cue-type effect: EBL > EBN≈TB for total and components; TB poorest on prospective component. No interactions with condition. Regression predictors (controlling for covariates): For pre-encoded EBL items, spontaneous EFT predicted the prospective component (t=2.026, p=0.047, β=0.330), and spontaneous past-oriented MW predicted the retrospective component (t=2.327, p=0.024, β=0.282). For post-encoded EBL items, voluntary EFT predicted the retrospective component (t=2.252, p=0.028, β=0.421). Global finding: higher MW frequency (EASY condition) was associated with better overall PM performance across pre and post items, especially in EBL. Overall, results support that increased MW during the retention interval enhances PM and that MW temporal orientation and source differentially relate to PM components: spontaneous past-oriented MW benefits retrospective retrieval of previously encoded EBL intentions, whereas spontaneous EFT benefits prospective retrieval; voluntary EFT benefits retrospective retrieval for newly encoded EBL intentions.

The study demonstrates a causal link between increased MW during a PM retention interval and improved PM performance in an ecologically valid immersive VR setting. Findings address the central hypothesis by showing that MW content and source differentially support the two PM components. Spontaneous past-oriented MW during the retention phase likely reactivates and reconsolidates intention content, strengthening the retrospective component for intentions encoded prior to MW induction—consistent with the intention-superiority effect and with accounts that past-oriented MW refreshes memory through mental imagery. In contrast, spontaneous episodic future thinking aligns with planning and anticipation, facilitating detection of prospective cues (when/where) for pre-encoded intentions. Voluntary EFT predicted retrospective retrieval for newly encoded intentions, suggesting a strategic use of future-oriented imagery during or shortly after encoding that enhances associative binding of cue–action pairs, potentially via stimulus-dependent thoughts related to the VR context. The consistent advantage for EBL over EBN and TB items replicates PM literature and validates the VR paradigm. Alternative accounts (e.g., retroactive interference from the hard task) are weakened by comparable post-navigation cued recall of pre-encoded actions across conditions and prior evidence that demanding post-encoding tasks need not impair consolidation. Collectively, the results support a twofold functional role for MW: maintaining continuity of previously formed intentions through spontaneous re-encoding (past-oriented) and supporting planning/execution via future-oriented simulation (EFT).

This work shows that elevating MW during a retention interval enhances naturalistic PM performance and that MW’s temporal orientation and source map onto distinct PM components. Spontaneous past-oriented MW supports consolidation and retrieval of intention content for previously encoded, semantically linked event-based cues, whereas spontaneous EFT facilitates prospective cue detection for pre-encoded intentions. Voluntary EFT relates to improved retrospective retrieval for newly encoded intentions, consistent with strategic future simulation at encoding. Using immersive VR provided ecological validity and fine-grained PM measurement. Future research should: directly assess whether MW includes rehearsal or PM-related thoughts; disambiguate stimulus-dependent versus task-unrelated MW; parse near versus distant temporal orientation; measure MW at multiple phases (encoding, immediate post-encoding, retention, and retrieval); adopt within-subject designs; and integrate eye-tracking to capture monitoring dynamics during VR PM tasks.

Key limitations include: (1) MW content was not queried for direct rehearsal of PM intentions; the EASY group’s higher MW might have included more rehearsal, partly explaining PM advantages. (2) Between-subject design leaves room for interindividual differences; within-subject replications are desirable. (3) MW content was not tied explicitly to encoded PM cues or near/distant temporal horizons, limiting interpretation of stimulus-dependent versus task-unrelated MW. (4) The cognitive load manipulation’s nature may differentially affect MW and PM; alternative load manipulations should be tested. (5) VR setup lacked eye-tracking to quantify monitoring and attention dynamics. Despite these, VR outcomes and manipulation checks support validity of conclusions.