Dissociated networks of neurons typically exhibit bursting behavior, whose features are strongly influenced by the age of the culture, by chemical/electrical stimulation or by environmental conditions. To help the experimenter in identifying the changes possibly induced by specific protocols, we developed a self-adapting method for detecting both bursts and network bursts from electrophysiological activity recorded by means of micro-electrode arrays. The algorithm is based on the computation of the logarithmic inter-spike interval histogram and automatically detects the best threshold to distinguish between inter- and intra-burst inter-spike intervals for each recording channel of the array. An analogous procedure is followed for the detection of network bursts, looking for sequences of closely spaced single-channel bursts. We tested our algorithm on recordings of spontaneous as well as chemically stimulated activity, comparing its performance to other methods available in the literature.