The laboratory approach brings a significant simplification compared to the actual conditions in situ. Laboratory experiments represent the only chance to control the physical conditions under which the investigated physical phenomena occur. Acoustic emission (AE) is the process accompanying the brittle fracturing of solid body and simultaneously an indispensable tool for its study. Laboratory experiments under controlled loading conditions make it possible to differentiate the effect of important factors like material structure, stress field, crack occurrence, etc. on fracture initiation and development, and allow simulate the nature in situ. Microearthquakes detected during AE can be analyzed by methods developed in earthquake seismology. We apply the shear-tensile crack (STC) to describe the source mechanism of the AE events with the aim to detect the mode of rock fracturing, in particular to distinguish between a shear slip and tensile crack, the latter both in the phase of its opening and closing. The benefit of discerning between shear and tensile fracturing is an insight into changes of the permeability of the rock massif both in space and time. By contrast to natural seismology, tens of thousands AE events occur in laboratory during the experiment. Expecting to process large volumes of data, an urgent demand was to make the non-linear STC search together with the estimate of the errors involved as fast as possible. To assess the reliability of the STC solution, the confidence regions of source model parameters are constructed. The misfit function is converted into the probability density function which is integrated over a trial volume of low misfit until requested probability content is achieved. For individual microearthquakes, we display confidence regions both for the mechanism orientation and its decomposition. Aiming to process a large bulk of AE data, a method of assessing of these zones needs to be proposed, which describes them by estimates of their extreme size. This allows us select for subsequent interpretation from all solutions only those that are stable and reliable. We have applied this approach to the experimental data obtained from a couple of uniaxial loading tests performed on a Westerly Granite and Liberec Granite specimen using a 14 channel AE monitoring system.