The amplitude of locked instabilities, likely magnetic islands, seen as precursors to disruptions has been studied using data from the JET, ASDEX Upgrade and COMPASS tokamaks. It was found that the thermal quench, that often initiates the disruption, is triggered when the amplitude has reached a distinct level. This information can be used to determine thresholds for simple disruption prediction schemes. The measured amplitude in part depends on the distance of the perturbation to the measurement coils. Hence the threshold for the measured amplitude depends on the mode location (i.e. the rational q-surface) and thus indirectly on parameters such as the edge safety factor, q95, and the internal inductance, li(3), that determine the shape of the q-profile. These dependencies can be used to set the disruption thresholds more precisely. For the ITER baseline scenario, with typically q95 = 3.2, li(3) = 0.9 and taking into account the position of the measurement coils on ITER, the maximum allowable measured locked mode amplitude normalized to engineering parameters was estimated to be a·BML(rc)/Ip = 0.92 m mT/MA, or directly as a fraction edge poloidal magnetic field: BML(rc)/Bθ(a) = 5 · 10−3. But these values decrease for operation at higher q95 or lower li(3). The analysis found furthermore that the above empirical criterion to trigger a thermal quench is more consistent with a criterion derived with the concept of a critical island size, i.e. the thermal quench seemed to be triggered at a distinct island width.