Empirical Predictions and
Falsification Criteria

This document exists so that the Language of Stress framework can fail. It lists what the theory predicts, what observation would refute each prediction, and -- the part most often left out -- what the framework would lose in each case. Some failures would be local, removing a claim while leaving the argument intact. Others would be structural. The register says which is which in advance, so that a failed prediction cannot later be reclassified as peripheral.

The predictions are graded by how sharply they separate the framework from existing accounts. The first tier contains those that would discriminate: that resistance to updating a belief tracks what is staked on it rather than how confidently it is held; that a fully predicted aversive state does not lose its aversive character once prediction error has resolved; that the pull of a route to relief is laid down by what has actually relieved tension before rather than computed from current beliefs about efficacy; that two strategies rehearsed at once contaminate each other by the sign of their valence; and that no gain separable from valuation governs which demand a deliberating brain acts on.

Two failures are named in advance as fatal: if a weakly held but heavily staked belief updates readily, and if felt urgency and prioritization come apart. A third is fatal in a narrower way: if a gain separable from valuation governs human deliberation, the account fails for human deliberation, though the regress itself, a structural claim, stands. The register also says which predictions can be run now with existing paradigms, which wait on special populations or on a measure not yet specified, and which need an instrument for comparing rigidity across people that does not yet exist.