1. The activity of 40 triceps brachii motor units was recorded from the dominant arms of 9 healthy adult volunteers (age 27.8 +/- 4.4 yr, mean +/- SD) during a fatigue task that included both isometric and anisometric contractions. The fatigue task lasted 8.3 min and consisted of 50 extension and 50 flexion movements of the elbow. Each movement (40 degrees in 0.8s) was separated by an isometric contraction. A constant load resisting extension of 17.7 +/- 3.0% of maximal voluntary contractions (MVC) was applied throughout the task. This paradigm enabled the direct contrast of motor-unit discharge behavior during the different types of fatiguing contractions. 2. Motor-unit behavior was examined to determine the relative contribution of two mechanisms for optimizing force production under fatiguing conditions: recruitment of motor units and modulation of motor-unit discharge following recruitment. Threshold torques for motor-unit recruitment thresholds were determined by ramp-and-hold isometric contractions. Motor-unit discharge was evaluated during the fatigue task by contrasting the number of motor-unit potentials (spikes) per contraction for concentric eccentric, and isometric contractions. 3. The fatigue task resulted in a 30 +/- 12% decline in the mean MVC of elbow extension. Recruitment of nine new motor units (23%) was evident during the fatiguing extension movements, often within five to seven movements (i.e., within 25-35 s). Each newly recruited motor unit had the largest recruitment threshold torque in that experiment. 4. Analysis of the motor units that were active from the beginning of the fatigue task revealed that the mean number of motor-unit spikes per contraction increased, or remained constant as fatigue ensued, yet for the majority of motor units it increased or remained constant. None of the newly recruited motor units demonstrated decreased number of mean spikes per contraction after recruitment. Further, concurrently active motor units displayed different discharge behavior in two-thirds of the subjects. It is proposed that if the neural drive to the muscle is distributed uniformly upon the motoneuron pool, peripheral feedback from the exercising muscle may modulate specific motoneuron discharge levels during fatigue.