The interpolated twitch technique (ITT) is the gold standard to assess voluntary activation and central fatigue. Yet, its validity has been questioned. Here we studied how peripheral fatigue can affect the ITT. Repeated contractions at submaximal frequencies were produced by supramaximal electrical stimulations of the human adductor pollicis muscle in vivo and of isolated rat soleus fiber bundles; an extra stimulation pulse was given during contractions to induce a superimposed twitch. Human muscles fatigued by repeated 30-Hz stimulation trains (3 s on-1 s off) showed an ~80% reduction in the superimposed twitch force accompanied by a severely reduced EMG response (M-wave amplitude), which implies action potential failure. Subsequent experiments combined a less intense stimulation protocol (1.5 s on-3 s off) with ischemia to cause muscle fatigue, but which preserved M-wave amplitude. However, the superimposed twitch force still decreased markedly more than the potentiated twitch force; with ITT this would reflect increased "voluntary activation." In contrast, the superimposed twitch force was relatively spared when a similar protocol was performed in rat soleus bundles. Force relaxation was slowed by >150% in fatigued human muscles, whereas it was unchanged in rat soleus bundles. Accordingly, results similar to those in the human muscle were obtained when relaxation was slowed by cooling the rat soleus muscles. In conclusion, our data demonstrate that muscle fatigue can confound the quantification of central fatigue using the ITT.