Pernasal aspirate (PNA) was obtained from 543 children during a 6-month period when whooping cough was prevalent. Three tests for diagnosing pertussis were performed on the PNA: (a) examination of direct smears by immunofluorescence (IF) for Bordetella pertussis; (b) culture; and (c) estimation of B. pertussis-specific immunoglobulin-A antibody (P-IgA) by an enzyme-linked immunosorbent assay (ELISA). On clinical review, 395 children were assessed to have had pertussis (P children) and 148 children not to have had pertussis (non-P children). The non-P children comprised 66 admitted to hospital for acute respiratory infections and 82 outpatients suspected of having pertussis. Analysis of the results of the tests on the PNAs of the non-P children helped to assess the P-IgA test. The analysis showed that artificial immunisation against pertussis did not affect the antibody results, but that non-specific positive results occur requiring the labelling of many P-IgA results as "doubtful". Among the 395 P children, 36% yielded positive cultures and more than half of these also had positive IF tests. The ELISA for P-IgA was positive in 24% of all the P children, equivalent to nearly 40% of the culture-negative P children. For the 148 non-P children, IF and culture-negative by definition, the P-IgA test was positive in 9%. The antibody test result was doubtful in 28% of the P children and in 40% of the non-P children. Estimation of P-IgA antibodies in PNA is a useful and economic complement to culture and IF in the diagnosis of pertussis.(ABSTRACT TRUNCATED AT 250 WORDS)