We compare site-specific response characteristics inferred from microtremors with those from seismic shear waves in terms of (1) predominant frequencies (PP), (2) horizontal component spectral ratios relative to a reference station (HH), (3) horizontal-to-vertical spectral ratios (HV). We furthermore compare the ratios of horizontal-to-vertical ratios of microtremors between stations (RHV) with the spectral ratios of seismic motions to examine the validity of the method proposed by Nakamura. The predominant frequency of seismic motions is compared with two kinds of predominant frequencies which are estimated from the power spectra and the horizontal-to-vertical spectral ratios of microtremors. To examine microtremor source effects we utilized data from two different-size arrays. The diameters of these arrays are 0.5km and 6km, respectively. We obtained the following conclusions from the small-array data: (1) The predominant frequencies estimated from seismic motions are coincident with the two kinds of predominant frequencies only if their spectral peaks are conspicuous. In comparison with the power spectra the spectral ratios (HV) is superior as a method for inferring the predominant frequency in that they give fewer spurious predominant frequencies. (2) The spectral ratios HH of microtremors are in agreement with those of seismic motion below 5Hz. (3) The spectral ratios RHV of microtremors do not agree with the spectral ratios HH of seismic motions. (4) The spectral ratios HV of microtremors are in agreement with those of seismic motions. For the large-array data, we obtained the same conclusions, except for conclusion (2). This indicates that spectral ratios HH can be used within an area in which microtremor source characteristics are spatially uniform.