The thresholds used to define a positive result for bronchial provocation challenges (BPC) are arbitrary. Requiring smaller decrements in expired flow to define a positive study would capture more cases of reactive airways (increased sensitivity) but would include some "normal" responses (decreased specificity). To examine the relationship between threshold definition and the ability to correctly classify subjects as either normal or as having airways hyperresponsiveness (AHR), four different BPC tests were administered on different days to 20 patients with a clinical diagnosis of exercise-induced bronchospasm (EIB) and 20 control subjects. The four BPC tests were indoor exercise on a cycle ergometer, methacholine inhalation challenge (MIC), eucapnic voluntary hyperventilation (EVH) with dry gas, and EVH with cold gas. Our results indicate that the thresholds which best separate the two groups are different for each of the four BPC techniques. For methacholine inhalation (MIC), a fall in FEV1 (d%FEV1) of 15 percent or greater at 188 cumulative breath units was 100 percent specific for AHR but had a sensitivity of only 55 percent. Eucapnic voluntary hyperventilation (EVH) with room temperature dry gas was 100 percent specific at a d%FEV1 of 11 percent, but, at that threshold, sensitivity was only 50 percent. EVH with cold air was 100 percent specific at a d%FEV1 of 12 percent but sensitivity was only 35 percent. The bicycle ergometer challenge was far too insensitive to be of value in evaluating AHR. Based on their respective receiver operating characteristic curves, the best separation of the two subject groups occurred at a d%FEV1 of 5 percent and 12 percent for the two EVH techniques and MIC, respectively. An individual's response to one test was highly correlated with the response to either of the other two (r = 0.66, p less than 0.001 for dry vs cold gas EVH; r = 0.56, p less than 0.001 for dry gas EVH vs methacholine; and r = 0.69, p less than 0.001 for cold gas EVH vs methacholine). Thus, MIC and EVH techniques are equally useful in defining AHR and each has its optimal threshold for a positive test result.