Abstract Most of the seismicity characterising volcanic and geothermal areas is represented by relatively small events, that reflect the local answer to the global tectonic setting. Due to the strong seismic noise encountered in these areas, the source mechanisms of these events are usually very difficult to be studied by standard techniques based on first arrivals, unless a very dense network of seismometers is available. When studying induced seismicity, a well-known problem for geothermal areas, or volcanic events, it is highly desirable to treat a seismic source in a form not a priori restricted to a double couple because the mechanism may reflect the local conditions such as small-scale tectonics, fluid motion and man-made factors. The decomposition of the full (or unconstrained) moment tensor solution into a double-couple component (DC), a volumetric (V) component, and a “compensated linear vector dipole” (CLVD) component can indicate a measure of the relative influence of global tectonics and local effects. In orogenic areas seismic events can be very strong, but the largest part of the earthquakes, which very likely reflect small scale complexities of the tectonic structure, is characterized by very small magnitudes. Therefore, the study of weak events is very important for many different purposes and we apply here the inversion scheme of high-frequency seismograms, recorded by a local network, based upon the unconstrained moment tensor description. A good knowledge of the structural model is a definite prerequisite to obtain reliable source information (moment tensor and time history not biased by propagation effects, even though not unique). For this reason we have inverted signals recorded in the Pozzuoli (Italy) area, using as input structure the model determined from a set of independent high-accuracy measurements. The inversions shows a first episode with the main peak centred around 0.6 s, followed by another significant peak, with duration of 0.4 s, centred at 2 s, and by a final energy release extending for about 1.0 s. The procedure has also been applied to vertical component seismograms recorded in the Friuli (northeastern Italy) orogenic area in correspondence to the M = 2.9 event of December 27, 1987, and two events of February 01, 1988, with M = 3.2 and M = 3.6, respectively. The retrieved source mechanisms are generally in agreement with the distribution of the few first arrivals polarities available. The source time functions indicate possible multiple rupture processes also for relatively small magnitude events.