Whether the transition in fatigue processes between "low-intensity" and "high-intensity" contractions occurs gradually, as the torque requirements are increased, or whether this transition occurs more suddenly at some identifiable "threshold", is not known. We hypothesized that the critical torque (CT; the asymptote of the torque-duration relationship) would demarcate distinct profiles of central and peripheral fatigue during intermittent isometric quadriceps contractions (3-s contraction, 2-s rest). Nine healthy men performed seven experimental trials to task failure or for up to 60 min, with maximal voluntary contractions (MVCs) performed at the end of each minute. The first five trials were performed to determine CT [~35-55% MVC, denoted severe 1 (S1) to severe 5 (S5) in ascending order], while the remaining two trials were performed 10 and 20% below the CT (denoted CT-10% and CT-20%). Dynamometer torque and the electromyogram of the right vastus lateralis were sampled continuously. Peripheral and central fatigue was determined from the fall in potentiated doublet torque and voluntary activation, respectively. Above CT, contractions progressed to task failure in ~3-18 min, at which point the MVC did not differ from the target torque (S1 target, 88.7 ± 4.3 N·m vs. MVC, 89.3 ± 8.8 N·m, P = 0.94). The potentiated doublet fell significantly in all trials, and voluntary activation was reduced in trials S1-S3, but not trials S4 and S5. Below CT, contractions could be sustained for 60 min on 17 of 18 occasions. Both central and peripheral fatigue developed, but there was a substantial reserve in MVC torque at the end of the task. The rate of global and peripheral fatigue development was four to five times greater during S1 than during CT-10% (change in MVC/change in time S1 vs. CT-10%: -7.2 ± 1.4 vs. -1.5 ± 0.4 N·m·min(-1)). These results demonstrate that CT represents a critical threshold for neuromuscular fatigue development.