The radioallergosorbent test (RAST) has been used for in vitro diagnosis of allergy in a selected group of children with asthma and allergic rhinoconjunctivitis. The results were compared with those of provocation tests and also with the results of skin tests. Where the results of RAST and the provocation tests differed, repeated tests were performed, initially in vitro and secondly in vivo. Selected commercial allergens of good quality were used and, for the most part, the same allergen batches were used in RAST, in the provocation tests and in the skin tests. A 76% correlation between RAST and provocation tests was found on preliminary investigation before divergent results were repeated. This degree of correlation is similar to that obtained previously in unselected patients using RAST as a part of the routine diagnostic. On repeated examination, agreement was obtained in an additional 16% of the cases. The major part of the improvement (13%) was due to conversion of preliminary RAST results from negative to positive in children with positive provocation tests. To a minor extent, the improvement was due to the fact that, in some children with positive RAST, negative provocation tests became positive when repeated. Finally, in one case an originally positive RAST in a child with a negative provocation test was found to be negative on re-examination. Comparing the final revised results of RAST and provocation tests there were no cases where a negative provocation test was combined with a positive RAST. Nearly all the children with positive provocation tests in spite of a negative RAST were shown to have low degree allergy. Although circulating reaginic antibodies are not always present, it seems that, on repeated investigations, the great majority of children with asthma and hay fever can be shown to have such antibodies even though the serum concentration may be low or very low. In doubtful cases repeated RAST examinations may be of value.