We examined whether forelimb and hindlimb phasic afferent input is a prerequisite for the production of avian locomotor patterns. We eliminated phasic afferent feedback through paralysis of a decerebrate animal. The term "fictive" has been used to describe the neural activity associated with spontaneous or evoked motor output during neuromuscular paralysis. We observed that a paralysed decerebrate bird is capable of producing similar locomotor activity patterns as an unparalysed preparation, regardless of whether the "fictive" locomotion is generated spontaneously, or in response to focal electrical and/or neurochemical stimulation of discrete brainstem locomotor regions. Not all aspects of "fictive" locomotor patterns were identical to the locomotion elicited prior to paralysis. The stimulus current threshold necessary to evoke hindlimb locomotion increased from 69 +/- 22 mu A (mean +/- S.D.) prior to paralysis to 185 +/- 87 mu A for "fictive" stepping. For wing activity, the threshold increased from 84 +/- 46 mu A during wing flapping to 228 +/- 148 mu A for "fictive" flight. In addition, the frequency of "fictive" efferent locomotor activity from the leg nerve (1.04 +/- 0.44 Hz) decreased relative to the frequency of leg activity prior to paralysis (1.55 +/- 0.70 Hz). Similarly, the frequency of wing activity decreased from 2.73 +/- 0.73 Hz before paralysis to 1.8 +/- 0.69 Hz after paralysis. Finally flexor burst duration remained constant during treadmill and "fictive" walking while the extensor burst duration was markedly increased during "fictive" walking. Thus, the relative contributions of leg flexor activity to the overall step cycle (burst proportion = burst duration/cycle duration) decreased during evoked "fictive" stepping, while the burst proportion of the leg extensor increased. Afferent feedback therefore appears to modulate leg extensor burst duration more than leg flexor duration. For the wings, the burst proportion of the major wing depressors remained constant before and after paralysis.