In 439 consecutive patients with reversible bronchial obstruction, commercial RAST and provocation inhalation tests were performed, resulting in 985 paired observations. In 405 of these, obtained before November 1977 when the old RAST technique was applied, RAST showed a high specificity (87%) but low sensitivity (29%) in predicting the result of provocation tests. However, in 580 subsequent paired observations RAST, using D epsilon 2 specific anti-IgE antibodies (modified technique), was both specific (78%) and sensitive (77%). In this way the mean of specificity and sensitivity, which was called "mean recognition", was better with the modified technique (78%) than with the old one (58%). RAST was equally indicative in patients showing either early, biphasic or late bronchial reactions after inhalation of the allergen. The predictive properties of the patients' histories, skin tests and RAST with respect to inhalation-provocation studies were compared. In general, histories were fairly specific but poorly sensitive, and skin tests were sensitive but not specific. RAST (modified version) was both specific and sensitive, but did not exceed a mean recognition of 78%.