The precursory swarm hypothesis as developed in New Zealand has been generalised on the basis of a systematic study of multiple earthquake events at moderate to large magnitudes in the Japan region. In the generalised hypothesis the precursory seismic sequence consists of one or more swarms followed by one or more mainshock events in the same sub-region. The predictor is the characteristic swarm magnitude, defined as the average magnitude of the three largest earthquakes in the swarm or swarms. The observed value of the predictor yields estimates of the magnitude and time of occurrence of the largest mainshock in the sequence; these parameters are estimated as soon as the first swarm occurs, and are revised on the occurrence of successive swarms in the sequence. Further statistical analysis of secondary mainshock events is required, after which a "score-sheet" of past successes, failures and false-alarms can be compiled, and the hypothesis can then be rigorously tested against future earthquake activity in the region.