We examined the results of tests with 22 substrates for their ability to discriminate a series of 917 strains of Escherichia coli collected from different sources. The tests with three of the substrates were discarded because of difficulties in performance or interpretation, and another nine substrates because they provided little discrimination. The tests used to obtain biotype profiles for strains were those for the fermentation of dulcitol, D-raffinose or sucrose or both, L-rhamnose and L-sorbose, the decarboxylation of L-lysine and L-ornithine, the hydrolysis of aesculin, motility, and prototrophy. Observations on several series of cultures from different sources showed that biotype characters were stable in vivo and after storage on non-selective medium. The biotype profiles obtained were as reliable as partial O serotyping for the routine subtyping of strains of E. coli isolated from the urine of patients with long-term urinary-tract infections and those from other sources in different patients. Biotyping and O serotyping used in conjuction offered a very fine degree of strain discrimination.