Human subjects generated a specified level of isometric torque with elbow flexor muscles of one arm, the reference arm, under visual feedback. They were then asked to generate what they perceived to be the same level, with the other arm, the indicator, but with no visual feedback. A number of torque levels, between 2% and 30% of maximum were used in the matching trials. Elbow flexors of one arm were then exercised eccentrically on a dynamometer. Immediately after the exercise, there was a large (40%) drop in maximum voluntary torque, as well as some soreness and swelling 24 h later, indicative of muscle damage. When the torque-matching experiment was repeated after the indicator arm had been exercised, the indicator signalled torque levels significantly below the reference level (P<0.05). When the reference arm was exercised, errors were in the opposite direction. Over the 4 days of testing post-exercise, errors became less as torque levels returned to normal. When errors were expressed in terms of maximum torque post-exercise, they were significantly reduced. This suggested that subjects were using as a matching cue the perceived effort required to generate a given level of torque rather than the level of torque itself. Persisting matching errors, from 24 h onwards after the eccentric contractions, were proposed to include a component attributable to the muscle soreness. Changes in electromyogram recorded after eccentric exercise were consistent with the effort-matching hypothesis. The muscle's torque-angle relationship was used to estimate matching ability in the absence of fatigue. One forearm was placed at various angles and its reference torque was matched by the other, the indicator, always at 90 degrees. Again, matching errors were consistent with an interpretation based on a match of effort rather than torque.