We have installed borehole tiltmeters at five sites in Yellowstone National Park, Wyoming, and have used these instruments to measure the spatial variation of the amplitude and phase of the principal semidiurnal tide. The measured tides vary both with position and azimuth and differ from the sum of the body tide and the ocean load by up to 50%. The difference predicted by a finite element model constructed from seismic, refraction, and gravity data has a maximum value of only 12%, although the discrepancy between our observations and the model is only marginally significant at some sites. The disagreement between the model and our observations is much larger than we observed using the same instruments at other sites and cannot be attributed to an instrumental effect. We have been unable to modify the model to explain our results while keeping it consistent with the previous observations.