AIM--To develop an internal quality assessment (IQA) scheme in a clinical bacteriology laboratory. METHODS--Over 24 months, 1230 diagnostic specimens, representing 0.42% of laboratory workload, were anonymised and resubmitted for analysis. Six hundred and twenty one (48.7%) of these gave positive culture results; 44 fecal and upper respiratory specimens were "spiked" (artificially inoculated) to increase the proportion of positive samples. RESULTS--Discrepancies between IQA and clinical sample results occurred in 188 cases (14.8%): 76.6% of these were in culture results, 13.3% in microscopy performance, and 10.1% in clerical recording. The culture discrepancy rate for each positive sample was lowest for wound (17.5%) and urine (18.1%) specimens, and highest for faeces (34.9%) and upper respiratory (37.7%) samples. Discrepancies in several areas responded to staff training and improvement in technical methods. CONCLUSIONS--An IQA programme of this type assesses the reproducibility of tests within a diagnostic laboratory when analysing common specimen types and organisms. It permits blind assessment of many areas of diagnostic work that are not readily amenable to other quality assurance methods, and it raises the awareness of all staff to the importance of quality in every aspect of specimen and data processing.