Auditing Constraint Asymmetry

This Methods paper specifies a latency-curve diagnostic that distinguishes neutrality-induced delay from confound-induced delay. When outcome resolution slows as competing alternatives near balance, the protocol tests whether that delay reflects genuine constraint neutrality or an unmodeled confound — boundary expansion, reduced effective noise, starting-point bias, hidden asymmetry, or non-decision-time shift.

The diagnostic compares the observed latency curve against the closed-form drift-diffusion prediction across five independently checkable axes: the ceiling at zero asymmetry, near-ceiling curvature, symmetry, full-curve shape, and a non-decision-time check — with the ceiling and the drift link estimated from channels that do not use the latency feature being tested. Each named confound produces a distinct, predictable deviation pattern, and the protocol explicitly catches the circular-fit failure mode in which per-condition drift is fit from the same response times it is meant to explain. Its formal pair, TN-S₂, proves the underlying monotone first-passage-time bound; real-data execution against perceptual-decision data is carried by a separate Tier 1.6 demonstration.

What acceptance commits you to. Acceptance does not require adopting Universal Collapse Theory. It requires only that the conditional lemma is correct and that the audit improves discrimination over standard drift-diffusion goodness-of-fit.

Keywords: latency curve; drift-diffusion; confound discrimination; circular fit; S₂ signature.