Movement unpredictability and temporal constraints affect the integration of muscle fatigue information into forward models

Abstract

When performing self-paced movements in a fatigued state, internal models can predict the mechanical effects of muscle fatigue. Yet, it is unclear if this is still true when movements are submitted to additional constraints. The purpose of the present study was to investigate the Central Nervous System’s (CNS) capacity to integrate fatigue signals into forward models’ prediction processes when the movement to perform is unpredictable and temporally constrained. Participants had to perform two different focal movements, depending on the random presentation of color word-color associations (Stroop-like task). They had to trigger fast arm extensions and flexions in a standing posture at the presentation of congruent and incongruent colors, respectively. These movements involve the implementation of predictive mechanisms of control, namely Anticipatory Postural Adjustments (APAs), which have been shown to depend on arm peak acceleration. APA timing and magnitude were measured using surface electromyography. The experimental task was performed before and after a fatiguing protocol involving the prime mover of the flexion movement. According to the assumption that APAs are constructed upstream from the primary motor cortex (M1), up to 1.4s before movement onset on the basis of the processing of the relevant contextual-sensory information, the Stroop-like task aimed to prevent participants from predicting the movement to perform and to reduce movement preparation time. While the fatigue protocol resulted in significant alterations of arm flexion peak accelerations, APAs were not modified post-fatigue as compared to control trials. It is proposed that with unpredictable and temporally constrained movements, the CNS cannot incorporate fatigue signals in internal models’ prediction processes to reweight the motor information contained in the efference copy. It is also suggested that APA implementation is based on predictive processes occurring in internal models located upstream from M1.