Post-stroke fatigue as a disorder of uncoupled agency: an exploration of predictive and postdictive mechanisms.

Abstract

Post-stroke fatigue is a common and disabling consequence of stroke, defined by a chronically elevated sense of effort during voluntary action that is disproportionate to any objective motor impairment. The sensory attenuation framework proposes that this effort distortion reflects reduced but not abolished gain on descending motor predictions. Because the same gain is thought to underlie agency, the framework predicts that in post-stroke fatigue peripheral sensory attenuation should be preserved, the cortical predictive signal should be disrupted, and postdictive integration should be amplified. We tested these predictions using two implicit measures of agency probing distinct levels of the sensorimotor hierarchy. Data were drawn from two independent cross-sectional studies of stroke survivors ([≥]3 months post-stroke), classified as high (Fatigue Severity Scale-7 [≥] 5) or low fatigue. A force matching task, indexing peripheral forward-model attenuation, was analysed in 40 patients (30 low-fatigue, 10 high-fatigue) and 24 healthy controls. An intentional binding paradigm with concurrent EEG, indexing the cortical predictive (readiness potential) and postdictive (outcome binding) components of agency, was completed by 23 patients (15 low-fatigue, 8 high-fatigue), of whom 17 yielded analysable EEG. Mixed-effects models were the primary framework, with fatigue group as the principal predictor, anxiety and depression as covariates, and robustness assessed via permutation, leave-one-out, Bayesian and equivalence analyses. Sensory attenuation was robust (n = 64) and unmodulated by fatigue, with the smallest detectable interaction small relative to mean attenuation. Outcome binding was selectively amplified in high-fatigue patients, showing temporal compression 2.5-fold greater than low-fatigue patients (72.0 vs 28.6 ms), whereas action binding was unmodulated (three-way interaction P = 0.007). The readiness potential showed canonical agency-related enhancement in low-fatigue but not high-fatigue patients (interaction P = 0.0002, robust to permutation, leave-one-out and Bayesian analyses), though the high-fatigue finding (n = 5) is preliminary. A direct test of predictive-postdictive uncoupling in the 17 patients with both measures was non-significant; uncoupling is therefore advanced as a working hypothesis, not an established finding. All effects were robust to mood adjustment. These findings are consistent with preserved peripheral attenuation, a disrupted cortical predictive signal, and amplified postdictive integration in post-stroke fatigue, compatible with a single upstream reduction in descending predictive gain and, tentatively, a shift from diminished to uncoupled agency. This differs from the peripheral-attenuation failure of schizophrenia and functional movement disorders, suggesting an intermediate position within the agency-disorder spectrum, and points to the cortical predictive layer as a candidate target for future work.