The response of biofilters to varying periods of starvation and to changes in inlet concentrations of a mixture of toluene and xylene have been tested in laboratory-scale biofilters packed with a mixture of Perlite and compost. These results have been mathematically modelled taking the emission of carbon dioxide as a proxy for kinetics measurements. The use of CO2 is a more practical approach than that of kinetics based upon batch experiments on pure cultures. A simplification of Zarook's method, our model produced good outlet predictions given small changes in the inlet concentration of toluene and xylene. But for more stressful situations, like the resumption of the feed after periods of starvation, the use of carbon dioxide proved to be inappropriate as an indicator of the biomass activity, greatly overestimating biofilter performance. This suggested either the occurrence of cryptic growth (as a result of the stress inflicted on the biomass) or perhaps the utilisation of the compost as a carbon source.