Colima city is the capital of the Mexican federal state of the same name. It is located close to the Pacific coast and is subjected to a large seismic risk. We present a microzonation study in this city, based on microtremors using single-station and array measurements. We applied horizontal-to-vertical spectral ratios (hvsr) analysis to single-station measurements at 310 sites within the city, concentrating measurements in zones that were damaged by the January 2003 (M 7.4) earthquake. The results show that a seismic zonation based exclusively on single-station microtremor measurements is not a reliable alternative when the local geology is complex and site effects are not the result of a single-impedance contrast. For this reason, we applied two independent analysis techniques to array measurements of microtremors: the spatial autocorrelation (spac) method and the refraction microtremor (ReMi) method. We used linear arrays to record 25-sec microtremor windows at eight sites within the city, which were analyzed with those two techniques. The result of both techniques of analysis is a phase-velocity dispersion curve, which can be inverted to obtain a shallow S-wave velocity profile. Two of the sites were the location of shallow (50 m) boreholes, where P- and S-wave velocity profiles were measured using a P-S suspension log. The phase-velocity dispersion curves obtained from the ReMi and spac analyses of the microtremor records showed very good agreement. The velocity profiles inverted from the phase-velocity dispersion curves showed good agreement with the suspension logging measurements at one of the two sites where they were available and poor agreement at the other site. The transfer functions computed from the inverted soil profiles are in good agreement with previous estimates of local amplification from spectral ratios analysis of earthquake records. Our results are compatible with previous indications of site effects and explain the failure of single-station microtremor measurements when the concept of dominant frequency loses its meaning. Finally, we propose an estimate of local site amplification at the city of Colima, which will be useful for future predictions of ground motion at this city.