Material collected during a prospective pertussis vaccine trial in 1992-95 was examined for Bordetella pertussis (culture and serology), Bordetella parapertussis (culture), Mycoplasma pneumoniae and Chlamydia pneumoniae (PCR). From 64% (99/155) of episodes with cough for less than 100 d, 115 aetiological agents were identified in one southern and one northern subset of DT-recipients. The most common single agent was B. pertussis, representing 56%(64/115), with a median cough period of 51 d, followed by M. pneumoniae 26%(30/115), 23 d, C. pneumoniae 17% (19/115), 26 d, and B. parapertussis 2% (2/115). For co-infections, the median duration of cough was about 60 d. Spasmodic cough for 21 d or more (clinical WHO criteria for pertussis) was present in 82% (41/50) of infections with B. pertussis as single agent, 38% (17/45) with B. parapertussis, 38% (5/13) with C. pneumoniae, 26% (5/19) with M. pneumoniae and 30%(17/56) in cases where no aetiology was found. In children with cough for more than 100 d (n = 78) using all vaccine arms, B. pertussis was responsible in 83% (65/78), in 21%(16/78) together with other agents. Acellular vaccines were more efficient against serious disease than whole cell vaccine. Antibiotic treatment was more common at the southern (34%) study site than at the northern one (12%). The findings indicate that diagnosis should rely on laboratory confirmation, both for rational treatment of an individual case and for monitoring outbreaks.