In order to assess and correlate the bacteriology of neonatal septic arthritis with its clinical presentation, the records of nine infants with this disease that were diagnosed at Edmonton hospitals between 1964 and 1981 were reviewed and 92 other cases reported in the English-language literature since 1960 were evaluated. All Edmonton cases developed outside of the hospital in previously healthy infants. A Streptococcus species was isolated in four of six patients from whom joint fluid was obtained before antibiotic therapy. Analysis of cases from the literature revealed a more variable bacteriology that seemed dependent on whether the case was hospital or community acquired. In 52 hospital-acquired cases, staphylococci were the predominant isolates (62%); next in frequency were Candida species (17%) and gram-negative enteric bacilli (15%). Community-acquired neonatal septic arthritis was most often caused by streptococci (52% of cases), staphylococci (26%), and gonococci (17%). Since 1970 the relative imbalance between staphylococcal (5%) and streptococcal (75%) isolates in community-acquired neonatal septic arthritis is even more pronounced. This pattern emphasizes the importance of ensuring optimal coverage against penicillin-sensitive organisms in community-acquired neonatal septic arthritis; this is in contrast to the situation with hospital-acquired arthritis, where wider-spectrum coverage against staphylococci and gram-negative enteric bacilli remains mandatory.