We apply the SPAC method to investigate site response at three sites in the Wainuiomata valley. At each site, square arrays with a fifth station at the centre were used to record 35 minutes of ambient noise. Distance between the stations on the square and the central one were 10, 20, 40, and 80 m. The data were analyzed through the computation of cross-correlation among stations. The resulting functions were inverted for the phase velocity dispersion curve of Rayleigh waves at each site. The results are reliable in the frequency band 1.1 to 3.7 Hz. The dispersion curves were inverted for the shear wave velocity profile at each site. We observe only marginal differences between sites FWP and BHP, where large amplification is given by very soft soil relative to a significantly stiffer bedrock. The third site, MCP, shows a velocity gradient with depth, without a pronounced contrast in the depth range investigated. We computed transfer functions for vertical incidence of shear waves on the inverted profiles, and compared the results with horizontal-to-vertical spectral ratios of microtremor records. The differences between the two shed some light on the complementary nature of both types of measurement and suggest that any single measurement may lead to erroneous interpretations.