**Rationale:** DAF disrupts reliable auditory feedback; AWS may shift control to somatosensory/proprioceptive cues or stronger internal models. **Prediction/test:** If AWS rely more on somatosensory under DAF, then perturbing somatosensory input (e.g., light mechanical lip/tongue perturbation) during DAF should disproportionately affect AWS’ stability. # Error-threshold (gating) hypothesis **Rationale:** AWS may apply a nonlinear gate to error processing: small errors trigger corrective updates, very large errors are downweighted or ignored. **Prediction/test:** Fit nonlinear error-update models to trial-by-trial data. Look for threshold-like behavior in AWS (strong updates below threshold, diminished updates above).