PL is a long-period (20 s or more) wave train beginning just after the P arrival in seismograms and continuing until the S arrival or the Rayleigh wave. The wave train can be observed at epicentral distances of about 5¿-20¿ with instruments sensitive to these low frequencies. PL propagates as a partially trapped P-SV wave in the crust; S wave energy is lost to the mantle during propagation making PL a ''leaky mode. We study PL propagation for a variety of earth models using the synthetic seismogram algorithm wave number integration and find that the vertical time in the crust is the most important parameter controlling PL's oscillation period. This period can vary by more than a factor of two between oceanic and continental paths. P-SV leaky mode propagation includes many different modes; the low-frequency motion terms ''PL is only the first, or fundamental mode, in this family. A second, higher-frequency mode roughly equal in amplitude to the fundamental appears for models without a net positive velocity gradient in the crust. We use these results to match an observed PL wave train whose propagation path consisted almost entirely of the Tibetan Plateu. By considering first and second PL mode behavior we find that the Tibetan crust is about 85 km thick, with an uncertainty of about 20 km, and possesses a significant (≤0.01 s-1) positive velocity gradient. The presence of a large, low-velocity zone in the lower Tibetan crust thus seems unlikely. Much higher-frequency PL modes also appear in the Green's functions for layer over half space models and appear to be responsible for the high-frequency Pg phase, making PL and Pg different members of the same type of leaky mode propagation.