Reproducible and discriminating typing methods are required for epidemiological investigations. Numerical typing systems analyse patterns obtained in various ways by calculating similarity coefficients between isolates. In the present study, various measures of the efficiency of a numerical typing system are quantified. These include reproducibility, accuracy, and discrimination power. Three different numerical typing methods for Escherichia coli were compared using these measures: (a) Biotyping with API 50 CH system, (b) Biochemical fingerprinting with the API 50 CH system and (c) Biochemical fingerprinting with the PhP-EC system. Biotyping qualitatively measures the results of a set of biochemical reactions as + or -. Biochemical fingerprinting also uses biochemical reactions, but the tests are scored quantitatively by measuring the kinetics and intensity of each reaction. It was found that biotyping yielded poor reproducibility. When biochemical fingerprinting analysis was used with the API 50 CHE system the reproducibility and the discrimination was good. The PhP-EC system for biochemical fingerprinting showed equal reproducibility but was superior to the API 50 CH system with regard to discrimination power.