Using particle motion, frequency-wavenumber analysis, and planewave fitting, we observe large anomalies in the bearing and apparent velocity of 0.5 to 1.0 Hz teleseismic P-waves incident at stations near the resurgent dome of Long Valley caldera, California. Our 3-component regional and teleseismic data come from three 2-mo deployments: (1) a 10 station, 2.9-km aperture array, (2) a 5 station, 1.9-km aperture array, and (3) a 3 station network, spanning about 5 km. We find stacked particle motion to be the most informative polarization method when the signal-to-noise ratio (SNR) is high and wavefronts are nonplanar. However, the insensitivity of the broadband frequency-wavenumber method (BBFK) to noise makes it more useful as SNR declines. Numerical simulations and data analyses using the BBFK method show that for a 10 station, 3-km aperture array, precision of roughly 8° in bearing and 0.008 sec/km in apparent slowness can be obtained with average teleseismic data in regions having complex structure. Bearing anomalies at Long Valley caldera reach up to 124°, whereas apparent slowness differ from expected values by up to a factor of 2. The anomalies are most likely caused by strong lateral velocity variations in the subcaldera crust between depths of 6 to 35 km.