Subjects in a control group were given feedback training for two unidentified visceral responses (increases and decreases in heart rate) and were then asked to provide a verbal report describing what they had done to control the feedback displays. Judges were given these reports and asked to determine training condition (increase heart rate on A trials and decrease heart rate on B trials, or the reverse) from them. Subjects in an experimental group received the same procedure but were also asked for a verbal report prior to receiving their first feedback display and thenceforth after each of the first 10 trials of feedback training. The results showed that: a) success at biofeedback learning is accompanied by verbal awareness of activities contributing to response production, and b) learning can be predicted (r= .80) by probing the subject's problem space before he has seen his first feedback trial. Extensive verbal probing of experimental subjects did not eventuate in superior learning in this group. The nature and role of problem-solving activity in biofeedback are discussed.