The authors describe and catalog 280 observations of a seismic phase taken at the single station SLK, NW of Long Valley caldera. These observations are observed to precede S by a more or less constant 1.6 sec in the distance range 30 to 90 km, and are closely fit by the least-squares line TSLK = (0.2853 ± 0.0009)Δ − (0.9029 ± 0.052). These observations can only be explained in terms of laterally heterogeneous velocity structure near the caldera. The model proposed here—that these so-called “SLK phases” result from an S-to-P conversion across a dipping planar structure NW of the caldera in the vicinity of Inyo Craters—fits not only these observations, but is consistent with the data presented in two recent papers (Luetgert and Mooney, 1985; Zucca et al., 1987). On the other hand, the data presented here are inconsistent with the models proposed by those authors involving deep reflections from magma bodies associated with the caldera. Furthermore, these observations are not related to the magma bodies within the caldera proposed by Sanders (1984), as suggested in previous abstracts. The original vertical-component observations at SLK are supplemented by three-component observations obtained on a small (several hundred meters aperture) array near SLK, which identify the SLK phase unambiguously as having longitudinal particle motion, consistent with the proposal that it is a S-to-P conversion occurring NW of the caldera boundary.