This study aimed to assess the capacity of recurrence quantification analysis (RQA) to detect potentiation and to determine the fatigue components to which RQA is sensitive. Fifteen men were divided in two groups [8 endurance-trained athletes (END) and 7 power-trained athletes (POW)]. They performed a 10-min intermittent (5 s contraction, 5 s rest) knee extension exercise at 50% of their maximal voluntary isometric contraction. Muscular fatigue and potentiation were evaluated with neurostimulation technique. Mechanical (peak torque, Pt) and electrophysiological (M-wave) responses following electrical stimulation of the femoral nerve were measured at rest and every 10 s throughout exercise. Vastus lateralis muscle activity (root mean square, RMS) was recorded during each contraction, and RMS was normalized to M-wave area (RMS/M). During contraction, muscle activity was analyzed with RQA to obtain the percentage of determinism (%Det). At the beginning of exercise, a significant Pt increase (+52%, P < 0.001) was observed in both groups, indicating potentiation. At this time, %Det remained constant in both groups, indicating that RQA did not detect potentiation. Thereafter, Pt decreased in POW from 5 min 30 s of exercise (−30%, P < 0.001), reflecting impairment in excitation–contraction coupling, and %Det increased from 3 min 30 s (P < 0.01). In END, Pt remained high and %Det was unchanged. These two results indicated that RQA detected the peripheral component of fatigue. Conversely, RQA did not detect central adaptation to fatigue since %Det remained constant when a significant increase in RMS/M (P < 0.01) appeared in END.