Using records from the EURO-SEISTEST area near Thessaloniki (Greece), we studied empirically the amplification effects of a sedimentary valley. One of the main advantages of this site is the very detailed knowledge of the near-subsurface structure due to exhaustive geophysical and geotechnical measurements and data analyses (Jongmans et al., 1998). The uniqueness of the seismological data set is the very dense network: 31 stations were installed perpendicular and parallel to the valley axis, with minimum interstation distances of 250 m. The mean amplifications along both axes, estimated from 13 local events, were imaged using the traditional spectral ratio technique. Variations of the mean amplifications due to the choice of different data subsets were tested by the use of up to 89 local events, supplemented by four teleseismic events offering good low-frequency resolution. Results obtained using the traditional spectral ratio technique were compared with results from a generalized inversion technique, the H/V ratio technique, a coda wave technique, and Nakamura's technique. Advantages and disadvantages of each method are outlined, especially from a practical point of view.