This paper demonstrates that cognitive hysteresis consists of two distinct mechanisms: persistence (state bias) and adaptive recalibration (control adjustment). Using multiple datasets (anchoring, Stroop, bandit tasks), the study shows that these mechanisms are empirically separable, statistically significant, and task-specific. Null results in unrelated paradigms confirm that the effects are not universal artifacts. The work provides a minimal, falsifiable framework for understanding how past experience shapes cognitive behavior.
Thomas S. Mitchell (Wed,) studied this question.