Respiratory muscle responses to sudden obstruction of learned breathing movements were studied in seven normal adults. Subjects were trained to inspire at a constant rate (0.4 liters/s), or maintain a static inspiratory effort(-10 cmH2O). On each trial these efforts were loaded unpredictably by occluding the airway or applying an opposing negative pressure at the mouth. Surface EMGs were recorded from the neck, parasternal intercostal, pectoral, diaphragmatic, and abdominal muscles. The latency and pattern of the responses to occlusion or to negative pressure were obtained from 10-trial computer-averaged records. When subjects were trained to relax in response to loading, inhibitory responses (mean latency: 32 ms) were recorded from the neck (16 out of 21 10-trial averages), diaphragm (9 out of 21), and parasternal (3 out of 21) locations. Excitatory responses (mean latency: 69 ms) were also recorded from the neck (11 out of 21 10-trial averages), diaphragm (8 out of 21), and parasternal (1 out of 21) sites. These responses were also observed in many single trial records. When subjects were trained to make an additional inspiratory effort as quickly as possible after loading, the reaction was a high-amplitude EMG burst, sometimes preceded by a brief inhibitory response. The mean reaction times for the large bursts were: 70 ms for the neck, 86 ms for the diaphragm, and 91 ms for the parasternal intercostals. Latencies in the 60- to 70-ms range were found on many 10-trial averages. Because the latencies of the excitatory responses evoked when subjects were trained to relax in response to loading were similar to those of the EMG bursts recorded when subjects were trained to react quickly in response to loading, it is not possible to distinguish long-latency reflex and learned response components on the basis of latency alone. Previous work, which has assumed that responses in the 50- to 70-ms latency range must be reflexive rather than learned, may need to be reexamined.