These pressure anomaly data and the failures of rock associated with these anomalies are suggestive that shield-leg pressures and/or pillar pressure may be used as an indicator of impending ground hazards in underground coal mines, especially in the critical areas close to the face. Data for the minor face spalling events, the first major roof cave and floor heave failures discussed above indicate the time durations for the pressure changes associated with these events are approximately 1 h, 3 h, and 5 h, respectively. Estimates of the effective length of rock fractured by these failures discussed above are approximately 3 m, 60 m, and 30 m. The effective rupture length of the minor bumping (spalling in the immediate vicinity of the cutter) at the face is taken to be the height of the face, 3 m. Estimates of the effective lengths of rock fractured by the roof cave and floor heave failures are taken to be 60 m and 30 m, respectively. The length of the cave failure is estimated to run from the shield line to the startup room. Most likely, this value is an upper limit. The 30 m for the floor heave event is based on visual floor and pillar damage along the affected portion of the tailgate. This value is most likely underestimated, as damage from the fracturing along the tailgate beyond the observed floor heave is difficult to determine without additional geophysical and geotechnical data. Figure 6 shows the plot of these failure data along with other published data on anomalies observed prior to catastrophic failure. Preparation or precursor time (tau(0)) is interpreted as the time interval that the anomaly persists. In this report, tau(0) is taken to be the time interval from the initiation of the pressure anomaly to the occurrence of the failure. Focal region length, L, is taken to be the total observed (or estimated) length of the rock mass; in our application the L is assumed to be the length of roof or floor rock that is fractured by the event. In figure 6, various mine failures, laboratory studies of precursors to small (approximately 10 cm) failures of rock specimens, and low magnitude earthquakes [Blue Mountain Lake] are shown for comparison with observations reported in this paper. Note that we assume that a precursor to failure, such as shown in figure 6, indicates that the failure process has begun and is self sustaining. This condition presumes that the far-field boundary conditions do not change during this time interval. Clearly defined rules for defending what constitutes a valid precursor, assuming that one exists, are not available. Precursory failure data from different classes of failure (laboratory, earthquake), should be interpreted carefully in this context. These data are suggestive that laboratory sized rock specimens have a precursor time of several hundred microseconds prior to catastrophic failure, mine failures, such as rock bursts and coal bumps, typically from several minutes to hours, and earthquakes, a few days to years prior to the occurrence of the failure. Of particular importance to the mining industry, however, these results are indicative that anticipation of failures, particularly those on the mine size scale may be possible since the failure event and the phenomena that precede the event may be scale invariant, and that the time interval (approximately a few minutes to hours) is of sufficient length to be of practical benefit to the miner once the anomaly is detected. It must be noted, however, that pressure anomalies are not observed prior to all mine failures with the GCMS. The statistics of anomaly-burst and anomaly-no burst are not known at this time.