A k(L)a correlation has been developed with the aid of a neural network model. The neural network model has served as a non-linear relationship performer correlating the volumetric mass transfer coefficient k(L)a in stirred tank reactors with the operating conditions, reactor geometry and material properties. In order to achieve an optimum correlation, experimental data taken from different sources have been used to train the network. The correlation obtained in this way is able to predict k(L)a in stirrer tanks reasonably well, if the operating conditions, reactor geometry and material properties fall into the trained ranges. Although the experimental data were widely spread and could only be fitted to individual correlations, the neural network is in a position to give a general correlation for all the data.