The objective of this study was to assess the predictive value of signs, symptoms, and rapidly available laboratory parameters for pneumococci in community-acquired pneumonia (CAP). A prospective study on patients with CAP who were admitted to hospital was conducted. Clinical and laboratory data were collected according to a protocol. Two hundred sixty-eight patients aged 18 years or older, not living in a nursing home or not admitted to hospital within one week of this admission, with a new infiltrate on the chest radiograph consistent with pneumonia were included. According to microbiological and serological tests, patients were allocated to one of two aetiological groups, Streptococcus pneumoniae or "other pathogens". Seventy-three variables were examined for a correlation with one of the aetiological categories by means of univariate and multivariate analysis. The resulting discriminant function was considered a clinical test for which posttest probabilities for pneumococcal pneumonia were calculated. Streptococcus pneumoniae was demonstrated in 79 patients and other pathogens in 83; no pathogens were detectable in 106 patients. The variables "cardiovascular disease", "acute onset", "pleuritic pain", "gram-positive bacteria in the sputum Gram stain", and "leucocyte count" correctly predicted the cause of CAP in 80% of all cases in both groups. Depending on the prevalence of Streptococcus pneumoniae, posttest probabilities for pneumococcal pneumonia were up to 90%. It is concluded that data on history, together with the result of the Gram stain of sputum and the leucocyte count, can help to distinguish Streptococcus pneumoniae from other pathogens causing CAP.