A new method for classifying bacteria is presented and applied to a la
rge set of biochemical data for the Enterobacteriaceae. The method min
imizes the bits needed to encode the classes and the items or, equival
ently, maximizes the information content of the classification. The re
sulting taxonomy of Enterobacteriaceae corresponds well to the general
structure of earlier classifications. Minimization of stochastic comp
lexity can be considered as a useful tool to create bacterial classifi
cations that are optimal from the point of view of information theory.