As a support service for a pertussis vaccine efficacy trial, a central diagnostic laboratory was established. Physicians in the geographic areas of the planned study were encouraged to send nasopharyngeal specimens from children and household contacts with cough illnesses whether or not the illnesses were typical of pertussis. From April, 1991, to February, 1992, 3629 specimens were received and in 601 instances (16.6%) Bordetella pertussis was isolated. Only 3.3% of patients with positive cultures had received pertussis vaccine whereas 16.1% of culture-negative patients had received vaccine (P < 0.0001). Fever was more common (12.2%) in patients with negative cultures compared with those with positive cultures (5.4%) (P < 0.0001). B. pertussis isolation rates fell markedly after 21 days of cough. Significantly more patients with negative cultures compared with those with positive cultures had been treated with erythromycin (8.5 vs. 2.9%; P < 0.0001). Patients with cough for greater than 4 weeks and specimen collection within 2 weeks of cough onset had a B. pertussis isolation rate of 59%. Similarly if whoop occurred under the same circumstances the isolation rate was 80%. In this study 25.5% of patients with culture confirmed pertussis had illnesses with cough of less than 21 days duration. This finding suggests to us that a pertussis case definition in efficacy trials that requires cough of 21 days is excessively restrictive.