A chemical multisensor array is used in combination with an artificial neural network to estimate the biomass concentration and specific growth rate in a recombination Escherichia coli batch cultivation. It is shown that by providing sufficient information to the artificial neural network, an accuracy comparable to that of an established dry weight method can be achieved. The obtained prediction error (1 sigma) of 0.043 g l-1 for biomass compares well with the error of the dry weight method in this low biomass concentration range (0.1-3 g l-1). The prediction for the specific growth rate is accurate during important parts of the cell growth (1 sigma = 0.025 h-1). The results show that this non-invasive method is potentially useful for estimating biomass and specific growth rate on-line in bioprocesses.