Fifty-nine community-acquired pneumonias were treated in a randomized double blind trial with cefamandole or ceftazidime. A prospective scoring system was used to define severity. This made use of basic clinical data, associated diseases, white blood count, blood gases and chest radiographs. There were no serious side-effects from the drugs. There were two deaths and six failed treatment. The scoring system which defined an 'ill group' showed as good a response of these ill patients to the new cephalosporin, ceftazidime as to cefamandole.