Muscles are the final effectors of behavior. The neural basis of behavior therefore cannot be completely understood without a description of the transfer function between neural output and muscle contraction. To this end, we have been studying muscle contraction in the well-investigated lobster pyloric system. We report here the mechanisms underlying stabilization of temporally summating contractions of the very slow dorsal dilator muscle in response to motor nerve stimulation with trains of rhythmic shock bursts at a physiological intraburst spike frequency (60 Hz), physiological cycle periods (0.5-2 s), and duty cycles from 0.1 to 0.8. For temporal summation to stabilize, the rise and relaxation amplitudes of the phasic contractions each burst induces must equalize as the rhythmic train continues. Stabilization could occur by changes in rise duration, rise slope, plateau duration, and/or relaxation slope. We demonstrate a generally applicable method for quantifying the relative contribution changes in these characteristics make to contraction stabilization. Our data show that all characteristics change as contractions stabilize, but their relative contribution differs depending on stimulation cycle period and duty cycle. The contribution of changes in rise duration did not depend on period or duty cycle for the 1-, 1.5-, and 2-s period regimes, contributing approximately 30% in all cases; but for the 0.5-s period regime, changes in rise duration increased from contributing 25% to contributing 50% as duty cycle increased from 0.1 to 0.8. At all cycle periods decreases in rise slope contributed little to stabilization at small duty cycles but increased to contributing approximately 80% at high duty cycles. The contribution of changes in plateau duration decreased in all cases as duty cycle increased; but this decrease was greater in long cycle period regimes. The contribution of changes in relaxation slope also decreased in all cases as duty cycle increased; but for this characteristic, the decrease was greatest in fast cycle period regimes, and in these regimes at high duty cycles these changes opposed contraction stabilization. Exponential fits to contraction relaxations showed that relaxation time constant increased with total contraction amplitude; this increase presumably underlies the decreased relaxation slope magnitude seen in high duty cycle, fast cycle period regimes. These data show that changes in no single contraction characteristic can account for contraction stabilization in this muscle and suggest that predicting muscle response in other systems in which slow muscles are driven by rapidly varying neuronal inputs may be similarly complex.