The degree and long-term variability of airway responsiveness to eucapneic hyperventilation with cold air was assessed in 287 children and young adults (less than 23 yr of age) (735 challenge tests) enrolled in a longitudinal study of pulmonary function, 179 of whom underwent two to five cold-air challenge tests between 1982 and 1986. Survey-to-survey variability in airway responsiveness was assessed with computation of adjusted within-subject correlation coefficients for continuous measures of response (percent decrease in FEV1 after challenge) and with adjusted odds ratios for dichotomous outcomes ("positive" if percent decrease FEV1 greater than or equal to 0.13). The relationships between variability in responsiveness and the following variables were examined: age, baseline level of pulmonary function, presence of respiratory symptoms and illnesses, smoking exposures, season, level of ventilation achieved during testing, and temperature of the expired air. Airway hyperresponsiveness was demonstrated consistently over all surveys in only six of 49 (12%) subjects who had ever demonstrated a "positive" airway response. The odds relating positive responses between surveys was 12.1 for airway hyperresponsiveness, 21.5 for "persistent" wheeze, and 6.7 for "any" wheeze. Significant predictors of airway hyperresponsiveness were wheeze symptoms (OR = 2.3), hay fever (OR = 1.6), and a chest illness requiring bed rest (OR = 2.5). Adjustment for survey-to-survey differences in respiratory symptoms, or for differences in testing conditions between surveys, did not alter the observed variability in airway responsiveness. Adjustment for "random error" (assessed with replicate measures of airway responsiveness in a subsample of 21 subjects) improved the observed survey-to-survey variability by 30%.(ABSTRACT TRUNCATED AT 250 WORDS)