A computer based rule-generation system of Inductive Learning by Logic Minimization (ILLM) was used to determine the sufficient set of biochemical reactions and necessary conditions that have to be fulfilled for correct differentiation of enterococci recovered from humans. The simplest combination of physiological tests for differentiation Enterococcus faecalis from all other enterococcal species consisted of only 3 reactions. Reactions that tested the ability of acidification D-xylose, mannitol, L-arabinose and Na-pyruvate were useful for delineation of both E. faecalis and E. faecium from all other enterococci. For differentiation of all 12 currently known clinically significant species of enterococci any one of 3 sets of nine tests suggested by ILLM could be used. The tests suggested by ILLM were applied to 153 isolates of enteroccoci recovered at our Department of Microbiology and all E. faecalis (138, 90.2%), E. faecium (13, 8.5%) and E. avium strains (2, 1.3%) were correctly differentiated.