The bacterial degradation of organic compounds plays a crucial role in the biogeochemical cycles of the earth and in the clean-up of contaminated soils. The processes are carried out by bacterial consortia, rather than isolated strains, which are usually modelled by phenomenological kinetic equations which describe a fictitious, homogeneous bacterial species which mimics the behaviour of the consortium. An alternative modelling framework is presented here, where the bacterial consortia are considered as networks of genes interacting with other genes as well as with chemicals, which may be either introduced from outside or produced by bacterial metabolism. The model is based on an extension of the random Boolean network model of genetic networks, which makes use of continuous dynamical variables. Three different models are introduced, which differ in the way how they account for the existence of different species: (i) a single supercell model, where all the genes can interact strongly with each other; (ii) a graded interaction model, where genes interact strongly within a species, and weakly among different species; and (iii) a separate subsets model, where genes interact only within species. It is shown how this modelling framework is sound, as it is able to reproduce some of the generic behaviours of bacterial consortia, describing experimentally observed phenomena like population changes induced by contamination, and prey-predator dynamics.Copyright 1997 Academic Press Limited Copyright 1997 Academic Press Limited