Conflict Delays, Load Creates Tail Risk: Distributional Signatures Across Two Cognitive Paradigms

We analyze reaction-time (RT) distributions across two cognitive paradigms—Eriksen Flanker (response conflict) and Sternberg working memory (load)—to determine how different constraints reshape behavioral structure. Results show two distinct distributional regimes. In the Flanker task, incongruent trials produce a near-uniform shift across quantiles, consistent with additive delay. In contrast, increasing working-memory load (5→7 items) produces a selective expansion of the upper tail (P90), with P90 shifts (~+234 ms) far exceeding lower-quantile shifts (~+47 ms at P10). This upper-tail emergence is robust to outlier removal, confirmed by bootstrap confidence intervals, and captured by a significant increase in the Tail Risk Index (TRI = P90 − P50). Subjects exhibiting stronger tail expansion show lower accuracy (p = 0.043), indicating that tail risk identifies performance-relevant structure. Pupil-linked physiology shows no detectable association with these distributional metrics. These findings demonstrate that cognitive constraint operates in at least two regimes: translation under response conflict and upper-tail risk emergence under memory load. Behavioral distribution shape provides a low-cost, physiology-independent marker of cognitive constraint beyond mean RT or variance. Data sources: OpenNeuro ds000102 (Flanker) and ds004117 (Sternberg). Cross-task generalization beyond these paradigms remains to be tested.