Shear-wave velocities (V5 \ which are widely used for earthquake ground-motion site characterization studies, are now largely obtained using borehole methods. Drilling holes, however, is expensive. Surface methods are less expensive for obtaining V5 information, but not many comparisons with direct borehole measurements have been made. Because different assumptions are used in data interpretation of each surface method, and because public safety is involved in site characterization for engineering structures, it is important to validate the surface methods by additional comparisons with borehole measurements. We compare results obtained from a particular surface method (array measurement of surface waves associated with microtremor) with results obtained from borehole methods. Using a ten-element nested-triangular array of 100-m aperture, we measured surface-wave phase velocities at two California sites, Garner Valley near Hemet and Hollister Municipal Airport. The Garner Valley site is located at an ancient lake bed where water-saturated sediment overlies decomposed granite on top of granite bedrock. Our array was deployed at a location where seismic velocities had been determined to a depth of 500 m by borehole methods. At Hollister, where the near-surface sediment consists of clay, sand, and gravel, we determined phase velocities using an array located close to a 60-m deep borehole where downhole velocity logs already exist. Because we want to assess the measurements uncomplicated by uncertainties introduced by the inversion process, we compare our phase-velocity results with the borehole V5 depth profile by calculating fundamental-mode Rayleigh-wave phase velocities from an earth model constructed from the borehole data. For wavelengths <~2 times of the array aperture at Garner Valley, phase-velocity results from array measurements agree with the calculated Rayleigh-wave velocities to better than 11%. Measurement errors become larger for wavelengths >2 times of the array aperture. At Hollister, the measured phase velocity at 3.9 Hz (near the upper edge of the microtremor frequency band) is within 20% of the calculated Rayleigh-wave velocity. Because shear-wave velocity is the predominant factor controlling Rayleigh-wave phase velocities, these comparisons suggest that this non-intrusive method can provide Vs information adequate for ground motion estimation provided two conditions are met. These conditions are: (1) the site velocity structure can be approximated by a horizontally-layered structure at least on the size of the seismic array, and (2) when the surface \\ avelength is <~2 times of the array aperture.