A survey of 42 hospitals revealed that the frequency of synovial fluid analysis was low (mean 3.9 per month). The mean was skewed upward by a higher frequency in 3 teaching hospitals, such that the overall median was 1.5 per month, with 15 hospitals studying 1 or less per month. There did not appear to be any consensus regarding what constitutes a routine synovial fluid analysis. Technical errors occurred, including dilution with acetic acid for cell counts and inappropriate use of automated cell counters. Twenty-six laboratories studied 4 "unknown" fluids for crystals. In 50 opportunities to detect each crystal type, calcium pyrophosphate was detected only 6 times, and sodium urate was detected 39 times. The white blood cell count reported for a single fluid ranged from 2,467-12,000/mm3 (mean +/- SD 6,683 +/- 1,992). The proportion of neutrophils identified varied from 0-30% (mean +/- SD 13.8 +/- 8.6). The majority of respondents believed that training and reference materials related to synovial fluid analysis should be improved.