Beyond aggregate error rate: time-on-task and history dependent dynamics in antisaccade performance

Im Fokus Onkologie · Springer Nature · 2026-11-17 · DOI: 10.1007/s00221-026-07433-4

Аннотация

Aggregate accuracy and reaction-time remain standard outcomes in cognitive tasks, but they compress behavior into static summaries and obscure the temporal dynamics through which performance unfolds. We examined whether antisaccade errors show time-on-task and history dependent dynamics, and whether these dynamics are modulated by task parameters and participant characteristics. We analyzed 147,744 antisaccade trials from 2006 young army recruits. Error probability was modeled at single-trial resolution using logistic mixed-effects models. Time-on-task effects were estimated with a piecewise model selected by grid search over candidate breakpoints. Performance-history effects were modeled using an exponentially weighted moving average (EWMA) history state, representing prior outcomes as a continuously updated signal with graded decay. Mean error rate was analyzed separately to test whether temporal parameters captured structure hidden by block-level measures. Error probability followed a biphasic within-task trajectory. The optimal breakpoint occurred at trial 18. Errors declined during the early phase and then increased gradually during the late phase. The optimal EWMA decay parameter was λ = .83, corresponding to an effective history timescale of 8–10 trials; across the state range, error odds increased by 137%. Task parameters modulated both temporal and history effects. Individual-difference associations were dissociable: novelty seeking and self-transcendence modulated time-on-task trajectories, while age and impulsivity modulated EWMA-state sensitivity. Conversely, aggregate error rate was associated with education and Raven score. This approach reframes antisaccade performance from aggregate indices into dynamic behavioral signatures that may map onto separable mechanisms such as attentional fluctuation, therefore providing interpretable parameters for future neuroimaging and clinical-neuroscience studies.

Читать у издателя →

Похожие статьи