We investigated the validity of seismic site response characteristics estimated from microtremors by comparing them with those of earthquake motions. For this purpose we observed microtremors as well as earthquake motions using large (5-km diameter) and small (0.5-km diameter) arrays deployed on soft sediments. Specifically, we examined four estimates from microtremors: relative site amplification factors to incident shear waves, site amplification factors by the Nakamura method, resonance frequency in horizontal-to-vertical spectral ratios, and horizontal-to-vertical spectral ratios. As a result of the comparisons, we obtained the following conclusions. The relative amplification factors can be inferred from horizontal-component ratios of microtremors to a reference site within a small area of several hundred meters. The horizontal-to-vertical spectral ratios inferred by the Nakamura method partly reflect site amplification factors, but do not agree with site amplification factors. A sharp-peak frequency in the horizontal-to-vertical spectral ratios is possibly the resonance frequency. The horizontal-to-vertical spectral ratios of microtremors either agree with those of earthquake motions at some array sites or are slightly smaller at the other sites.