Precision Experience Sampling Tracks Idiosyncratic Human Thoughts Across Daily Activities

Researchers investigating human cognition have demonstrated that repeated naturalistic monitoring of the same individuals across multiple sessions uncovers deeply idiosyncratic patterns of thought that traditional group averages fail to capture. Published in Cognitive Research: Principles and Implications, the study applied precision experience sampling—an intensive idiographic methodology that tracks internal mental experiences across time, context, and activities within single individuals—to determine how everyday computer tasks modulate ongoing thoughts. By examining seven individuals across seven distinct eighty-minute experimental sessions alongside a comparative cohort of forty-nine participants completing a single session, the researchers discovered that the way ongoing activities shape mental states varies dramatically from person to person. Group-level aggregations regularly masked or contradicted these individual patterns, showing that population averages do not reliably represent personal cognitive tendencies. Furthermore, machine learning classifiers trained on reports of current activities predicted an individual’s momentary thoughts across nine psychological dimensions with significantly higher accuracy when using personal longitudinal data rather than group-derived data.

The study addresses a longstanding tension in psychological research between nomothetic approaches, which aggregate data across large groups to extract general population patterns, and idiographic approaches, which track intra-individual variation over time. Nomothetic research relies on the assumption of homogeneity across individuals, presuming that group averages accurately describe any single participant’s inner life. However, ongoing thoughts fluctuate continuously throughout waking hours across content, dynamics, and sensory modality. Prior meta-analyses indicate that off-task thoughts occupy roughly thirty-five percent of daily life, impairing performance and provoking negative affect when unintentional, while freely moving thoughts account for thirty to fifty percent of waking hours. Although past studies confirmed that external contexts modulate thoughts at the group level, they left open whether specific tasks reliably evoke the same mental states in different individuals. Precision experience sampling resolves this limitation by gathering dense longitudinal observations within individuals across varied sessions, enabling empirical tests of whether population findings truly characterize individual experiences.

To test these dynamics, the researchers recruited two distinct cohorts at the University of Calgary, creating an idiographic dataset alongside a nomothetic validation dataset with an identical total number of experimental sessions and thought probes. The idiographic cohort comprised seven undergraduate and graduate students, including four females and three males, with an average age of 24.57 years and 16.43 years of education. Each idiographic participant completed seven separate eighty-minute laboratory sessions alongside electroencephalography and eye-tracking preparation, generating forty-nine datasets and 245 probes per person. The nomothetic validation cohort included forty-nine undergraduate students, comprising thirty-eight females, nine males, one transgender male, and one non-binary participant, with an average age of 19.67 years and 13.41 years of education, who each completed a single eighty-minute session. Participants in both groups had normal or corrected-to-normal vision and reported no neurological disorders. This design yielded exactly forty-nine sessions and 1,715 probe-level data points in each cohort, balancing total observations between intensive individual sampling and broad group sampling.

In each session, participants engaged in self-selected activities on a laboratory desktop computer for eighty minutes, ensuring naturalistic engagement and high ecological validity. The authors classified typed activity reports into reading or studying, writing or editing, watching videos, web browsing, playing games like Sudoku or racing, and other demanding tasks such as programming or creating presentations; passive behaviors like doing nothing accounted for only 0.1 percent of instances and were excluded. Throughout each session, thirty-five thought probes appeared every 1.5 to 2.5 minutes, alerted by a 200-millisecond tone at 1000 Hertz. Participants identified their current activity and rated their immediate mental states on seven-point Likert scales across nine thought dimensions. These dimensions evaluated thought content (internally oriented, off-task, self-oriented, others-oriented), thought dynamics (freely moving, goal-oriented, sticky), and sensory form (visual and auditory modalities), providing a multidimensional assessment of cognitive fluctuations across tasks.

Preliminary analyses compared the idiographic and nomothetic cohorts to confirm that the smaller sample produced valid, representative thought patterns. Linear mixed effects models controlling for age revealed group differences in mean ratings across seven dimensions: idiographic participants reported significantly higher levels of off-task, freely moving, goal-oriented, sticky, self-oriented, and auditory thoughts, but lower others-oriented thoughts, while internally oriented and visual ratings did not differ significantly. Rating variability across sessions was comparable between groups, with only sticky thoughts showing significantly greater standard deviations among nomothetic participants. A Mantel test comparing the grand average correlation matrices of the nine thought dimensions demonstrated significant structural alignment between cohorts at r = 0.63. Established psychological relationships replicated across both datasets: task-unrelated thoughts correlated positively with freely moving thoughts in both idiographic and nomothetic participants, while freely moving thoughts correlated negatively with goal-oriented thoughts, verifying that the idiographic dataset reflected consistent relational dynamics.

The authors next examined measurement reliability by calculating intraclass correlation coefficients, which capture between-participant variance relative to total variance. Overall reliability across thought dimensions was numerically higher in the idiographic cohort (mean 0.47) than in the nomothetic cohort (mean 0.37). However, this benefit applied selectively rather than universally. The idiographic approach increased reliability for six dimensions reflecting thought content and dynamics: internally oriented, off-task, freely moving, goal-oriented, self-oriented, and others-oriented thoughts. Conversely, auditory thought reliability was higher in the nomothetic group, while sticky and visual thoughts yielded comparable reliability across cohorts. The authors suggested that thought content and dynamics constitute more stable cognitive traits that become clearer with multi-session sampling across varied contexts. In contrast, sensory form and sticky thoughts depend heavily on immediate environmental factors, such as task modalities or salient life events; consequently, repeated naturalistic sampling exposes shifting situational influences that counteract simple statistical reliability gains.

Turning to primary research questions, the team implemented linear mixed effects models with participant-mean centering to evaluate how ongoing tasks modulated thought dimensions within each idiographic participant. Applying a Bonferroni-corrected alpha of 0.006 across the nine dimensions revealed considerable individual variability. Significant task modulations emerged for off-task and visual thoughts in five participants, for goal-oriented and others-oriented thoughts in four, and for sticky thoughts in only one participant, with other dimensions showing significant effects in two to five participants. Pairwise task contrasts further exposed individualized responses: not a single pairwise contrast held across all seven individuals. For example, Participant 2 exhibited their highest off-task thoughts during web browsing compared to all other activities, whereas Participant 6 maintained uniform off-task levels across all tasks. The most widely shared contrasts—such as higher visual thoughts during web surfing versus writing, or elevated off-task thoughts during reading versus writing—occurred in at most four participants.

When researchers analyzed idiographic data at the group level, substantial divergences appeared between aggregate findings and individual realities. Aggregated models indicated significant main effects of task across eight thought dimensions, with only sticky thoughts failing to reach significance. However, these aggregate effects frequently contradicted individual patterns. At the group level, writing predicted significantly lower off-task thoughts than web surfing, gaming, or demanding tasks, yet this contrast appeared in only one or two individual participants. Video watching was associated with the lowest off-task ratings at the group level, but this held for only a single participant. Playing games yielded the highest numerical off-task ratings overall, but predicted relatively lower off-task thoughts in three participants when compared to other activities.

The second objective evaluated whether precision experience sampling enhances machine learning predictions of ongoing thoughts from task activities. Using a k-nearest neighbor classifier with k set to twenty-five, three-fold stratified within-subject cross-validation over one hundred iterations, and synthetic minority over-sampling for training balance, the researchers predicted exact seven-point ratings for each thought dimension. Within-participant idiographic models achieved superior classification performance across all nine dimensions compared to the unaugmented nomothetic cohort, producing mean Matthew’s correlation coefficients between 0.21 and 0.38, whereas nomothetic models remained at or below chance. To verify that performance differences did not stem solely from sample volume, researchers augmented nomothetic data sevenfold using synthetic minority over-sampling to match the 245 data points per participant; yet idiographic models remained significantly superior across all dimensions.

The authors outlined several methodological constraints, noting that while the nine queried dimensions represent major constructs in cognitive literature, they do not capture the exhaustive scope of human internal experience. The exploratory design did not strictly control for all family-wise error inflation, and the cohorts involved small university samples recruited within a Western, educated, industrialized, rich, and democratic context. Furthermore, the idiographic and nomothetic cohorts comprised different individuals rather than tracking the same participants under both paradigms, and unconstrained computer activities sacrificed experimental standardization over task duration and complexity to maximize ecological validity. Future investigations could combine precision sampling with standardized cognitive tasks across larger, more diverse cohorts. Nevertheless, by showing that repeated sampling within individuals captures idiosyncratic cognitive profiles and dramatically improves machine learning predictions of ongoing thoughts, precision experience sampling offers a powerful foundation for personalizing cognitive interventions in education and psychiatric care.

Subject of Research: Individual variations in naturalistic thoughts captured through precision experience sampling

Article Title: Capturing naturalistic thoughts using a precision experience sampling idiographic approach

Article References: Kam, J. W. Y., Jothiraj, S. N., Beauchemin, E., Al Nahin Ch, N., Allen, L. K., Wormwood, J. B., & Mills, C. (2026). Capturing naturalistic thoughts using a precision experience sampling idiographic approach. Cognitive Research: Principles and Implications, 11(1), Article 33. https://doi.org/10.1186/s41235-026-00728-8

Image Credits: AI Generated

DOI: 10.1186/s41235-026-00728-8

Keywords: precision experience sampling, idiographic approach, naturalistic thoughts, cognitive psychology, machine learning, task modulation, mind wandering, intraclass correlation, Capturing, naturalistic, thoughts, precision