- Some of the most suspicious seismic disturbances under the Comprehensive Nuclear-Test-Ban Treaty (CTBT) are likely to be those associated with mining, as they are shallow, and at least some have an explosion-like mb:Ms signature. Previous research highlighted the potential of broadband teleseismic P waves as a way of identifying large mine tremors. Broadband teleseismic P from two suspected large mine collapses, one in Germany (1302 UT, 13 March 1989, 5.4mb) and another in Wyoming (1526 UT, 3 February 1995, 5.3mb), show differences in character despite the similarity of the reported ground failure and mine types. We apply a full moment-tensor analysis to the teleseismic P waves and show that the data are inconsistent with either a shallow explosion or an earthquake (double-couple) at depth, but this method is unable to distinguish between a shallow dip-slip source and a closing-crack moment tensor. However, three-component surface-wave seismograms recorded at regional distances fit the shallow closing-crack model, but are inconsistent with a shallow earthquake source, because strong Love waves, expected from a double-couple source, are not observed at a number of stations well distributed in azimuth. Here, we restate the equivalence for shallow sources of the closing-crack model and a gravitational collapse model. We use the latter to model the broadband P waves from these mine tremors and show that, while non-unique, the differences in the observed broadband P waves from the two tremors can be attributed to the area, amount of collapse, depth, and rate of collapse. The collapse model predicts negative first-motion for all P waves in contrast to the positive polarity expected from explosions. Thus, the broadband teleseismic P waves have the potential to discriminate between large collapses and explosions.