Pupil baseline trends on ~ 10-s timescales track near-term task performance
Abstract Human task performance fluctuates on infraslow timescales, including a putative ~ 20-s cycle of task-positive/default-mode switching, yet a continuously computable biomarker remains scarce. Task-evoked pupil dilation tracks momentary facilitation of performance, but links between slow baseline dynamics and performance are inconsistent. Because prior research used demanding vigilance and working memory tasks, we tested whether upward 10-s baseline trends were associated with faster responses in a relatively simple, low-error go/no-go task. We quantified baseline “pupil trends” as slope-like changes at predefined timescales using two independent pipelines: moving-average smoothing and wavelet-based scale isolation. Upward trends were associated with faster responses across several smoothing windows, including 10 s, while the scale-isolated analysis identified the prespecified ~ 8-s dyadic component; downward trends were linked to slower responses (fastest vs. slowest quartile). These associations were supported by the strictly pre-target 10-s precursor and by trial-level analysis of the ~ 8-s component controlling for canonical pupil–behavior covariates. Timescale-specific pupil dynamics may provide a continuously computable index of local fluctuations in near-term task performance, with potential utility for monitoring in ecologically valid settings.
Authors
- Jumpei Yamashita (ORCID: https://orcid.org/0000-0001-9411-0523)
- Takatsune Kumada (ORCID: https://orcid.org/0000-0002-0198-0501)
- Makoto Yoneya (ORCID: https://orcid.org/0000-0002-4727-2498)
- Kazushi Maruya (ORCID: https://orcid.org/0009-0004-5219-7535)
- Hiroki Terashima (ORCID: https://orcid.org/0000-0002-8825-5299)
Institutions
- Kyoto University (JP)
- NTT (Japan) (JP)
- NTT Basic Research Laboratories (JP)
Publication Details
- Journal
- Scientific Reports
- Published
- 2026-11-13
- DOI
- https://doi.org/10.1038/s41598-026-73046-0
- Primary Topic
- Mind wandering and attention
- Type
- article
- Field-Weighted Citation Impact
- 0.00