Endurance is a clinically relevant muscle parameter. It would be desirable to be able to estimate it without the need for a contraction sustained to exhaustion. The purpose of this work was to investigate the capability of the initial rate of spectral compression of the surface electromyographic (EMG) signal to predict mechanical endurance during sustained voluntary contractions of the human tibialis anterior muscle. Six healthy subjects performed voluntary isometric contractions of the tibialis anterior at 80, 70, 60, and 50% of the maximal voluntary contraction level. The contractions were sustained for 90, 120, 150, and 170 seconds, respectively. These intervals exceed the normal endurance time for this muscle and allow for a decrease of torque output. The slope of the median frequency, computed over the first 30 seconds of the contraction, was used to describe the initial spectral compression of the EMG signal. Significant correlations were found: 1) between contraction level and endurance time (p < 0.05 for each subject) and 2) between median (or mean) frequency slope and endurance time (p < 0.0001 for all subjects pooled together). The regression between median frequency slope and endurance time showed intersubject variations possibly related to the tibialis anterior muscle fiber type content. It is concluded that clinical use of the EMG spectral technique in assessing muscle fatigue may enable the clinician to estimate the endurance time without having the subject sustain a contraction until the point of contractile failure. This could be an advantage for some patient populations, such as the severely disabled, arthritic, or (rail elderly, that might not he able to tolerate long duration contractions.