The dynamic strains associated with seismic waves may play a significant role in earthquake triggering, hydrological and magmatic changes, earthquake damage, and ground failure. We determine how accurately dynamic strains may be estimated from seismometer data and elastic-wave theory by comparing such estimated strains with strains measured on a three-component long-base strainmeter system at Pin??on Flat, California. We quantify the uncertainties and errors through cross-spectral analysis of data from three regional earthquakes (the M0 = 4 ?? 1017 N-m St. George, Utah; M0 = 4 ?? 1017 N-m Little Skull Mountain, Nevada; and M0 = 1 ?? 1019 N-m Northridge, California, events at distances of 470, 345, and 206 km, respectively). Our analysis indicates that in most cases the phase of the estimated strain matches that of the observed strain quite well (to within the uncertainties, which are about ?? 0.1 to ?? 0.2 cycles). However, the amplitudes are often systematically off, at levels exceeding the uncertainties (about 20%); in one case, the predicted strain amplitudes are nearly twice those observed. We also observe significant ?????? strains (?? = tangential direction), which should be zero theoretically; in the worst case, the rms ?????? strain exceeds the other nonzero components. These nonzero ?????? strains cannot be caused by deviations of the surface-wave propagation paths from the expected azimuth or by departures from the plane-wave approximation. We believe that distortion of the strain field by topography or material heterogeneities give rise to these complexities.