Following the work of Phillips (1985), we have computed site amplification factors for coda waves at many sites in the Long Valley region in the eastern Sierra Nevada. We computed ratios of coda amplitudes measured at 15 stations in and around Long Valley caldera relative to a granitic site, MMPM, for six frequency bands centered at 1.5, 3.0, 6.0, 9.0, 12.0, and 15.0 Hz. All station sites located within the caldera experienced large ground motion amplification at 1.5 and 3.0 Hz, ranging between five and 17 times that of the reference site. However, at higher frequencies, these same sites exhibited significantly less amplification than the reference granite site. This is attributed to the competing effects of an impedance contrast between the basement rock and caldera fill and very high absorption in the caldera fill at high frequencies. Station MMLM, located on top of a volcanic plug, displayed the largest amplitudes of all the sites studied for frequencies between 9.0 and 15.0 Hz. A dike structure attached to the plug couples the basement rock to the surface. At high frequencies, the resulting large amplitudes at MMLM are not due to amplification resulting from a strong impedance contrast; rather, the absorption under this site is very low, perhaps lower than at the reference site, MMPM. Outside the caldera, another hard-rock site located at Devil's Postpile, MDPM, generally behaved like the reference site for all frequencies. The lowest amplifications observed came from a site outside the caldera, MDCM, located on thin pyroclastic ash deposits overlying granitic basement. This can be attributed to a dominance of absorption over the amplification caused by lower impedance of this layer. Variations among sites on similar surface geology may be due to small local variations in impedance and absorption under and adjacent to the site. The range in the spectral decay parameter, κ, between caldera and rock sites are comparable to results of Anderson and Hough (1984) for sites on alluvium and rock in the San Fernando region. These surprisingly different amplifications support the need for additional site-specific studies. Amplifications determined in this study for the frequency range 1.5 and 3.0 Hz correlate remarkably well with Eaton's (1990) residuals for duration magnitude, FMAG, and amplitude magnitude, XMAG, for the USGS northern California seismic array, further supporting the use of coda waves in determining site-specific amplification.