This short note proposes that certain forms of engagement are not primarily determined by stimulus content, but by the regime under which experience unfolds. The core distinction is between a permissive regime and an evaluative regime. In permissive conditions, perceptual and affective signals are not required to close evaluatively. Recognition does not automatically recruit preference, and discrepancies can remain unresolved without being treated as problems to fix, interpret, or optimize. In evaluative conditions, experience is recruited into comparison, narrative framing, goal orientation, or explicit measurement. Under these demands, specific modes of engagement may collapse. The framework identifies two sources of evaluative collapse: contextual evaluation (ratings, performance framing, time pressure, observation, explicit measurement) and endogenous evaluation (self-monitoring, anticipatory testing, self-narration, optimization tendencies). Minimizing external measurement is therefore necessary but not always sufficient. The note proposes regime-sensitive operational indicators that do not rely on intensity ratings, including engagement continuity, recognition without preference, resistance to interruption, and absence of spontaneous evaluative narration. Finally, it introduces a three-component architecture of evaluative load (“Z”): accumulated evaluative exposure (Z_acc), discrete structural transition into anticipatory monitoring (Z_shift), and contextual evaluative demands (Z_ctx). Regime stability is treated as an interaction between these components. The central claim is structural and falsifiable: the same experience can function differently depending on whether it must be evaluated, and evaluative operations should reliably reduce engagement probability under otherwise matched conditions.