Only 13 of 38 hospital laboratories surveyed include a Gram stain routinely in microbiological sputum examination. In a prospective three-hospital study, 60% of over 1,200 "sputum" specimens consisted predominantly of saliva, as judged by cell composition. Compared with the results of cultures in which microorganisms presumptively identified on sputum smears were specifically sought ("directed cultures"), cultures of the same specimens processed in the routine manner missed pneumococci 61%, haemophili 23%, and yeasts 44% of the time. The findings were similar in all three hospitals despite differences in administration, staffing, primary culture media, and workload. Unless microscopic examination is routinely included, half of all microbiological information rendered on sputum specimens is meaningless and subject to dangerous misapplication. Furthermore, culture must be guided by microscopic findings, or respiratory pathogens will frequently be missed. Finally, when routine culture and smear disagree, the culture cannot be assumed to be correct. Microscopic examination should be mandatory in sputum microbiology, both for specimen evaluation and as a guide to what to look for in culture.