Objectives: to test the hypothesis that cognitive load (low vs. high load) during a 20 min self-paced cycling exercise affects physical performance.Design: A pre-registered (https://osf.io/qept5/), randomized, within-subject design experiment.Methods: 28 trained and experienced male cyclists completed a 20 min self-paced cycling time-trial exercise in two separate sessions, corresponding to two working memory load conditions: 1-back or 2-back. We measured power output, heart rate, RPE and mental fatigue.Results: Bayes analyses revealed extreme evidence for the 2-back task being more demanding than the 1-back task, both in terms of accuracy (BF10 = 4490) and reaction time (BF =1316). The data only showed anecdotal evidence for the alternative hypothesis for the power output (BF10= 1.52), moderate evidence for the null hypothesis for the heart rate (BF10 = 0.172), anecdotal evidence for RPE (BF10 = 0.72) and anecdotal evidence for mental fatigue (BF10 = 0.588).Conclusions: Our data seem to challenge the idea that self-paced exercise is regulated by top-down processing, given that we did not show clear evidence of exercise impairment (at the physical, physiological and subjective levels) in the high cognitive load condition task with respect to the low working memory load condition. The involvement of top-down processing in self-pacing the physical effort, however, cannot be totally discarded. Factors like the duration of the physical and cognitive tasks, the potential influence of dual-tasking, and the participants’ level of expertise, should be taken into account in future attempts to investigate the role of top-down processing in self-pace exercise