In the aftermath of two destructive urban earthquakes in 1999 in Turkey, empirical models of strong motion attenuation relationships that have been previously developed for North American and European earthquakes have been utilized in a number of national seismic hazard studies.. However, comparison of empirical evidence and estimates present significant differences. For that reason, a data set created from a suite of 100 vertical strong ground motion records from 47 national earthquakes that occurred between 1976 and 2002 has been used to develop attenuation relationships for strong ground motion in Turkey. A consistent set of empirical attenuation relationships was derived for predicting vertical peak and pseudo-absolute vertical acceleration spectral ordinates in terms of magnitude, source-to-site distance, and local geological conditions. The study manifests the strong dependence of vertical to horizontal (V/H) acceleration ratio on spectral periods and relatively weaker dependence on site geology, magnitude, and distance. The V/H ratio is found to be particularly significant at the higher frequency end of the spectrum, reaching values as high as 0.9 at short distances on soil sites. The largest long-period spectral ratios are observed to occur on rock sites where they can reach values in excess of 0.5. These results raise misgivings concerning the practice of assigning the V/H ratio a standard value of two-thirds. Hence, nonconservatism of this value at short periods and its conservatism at long periods underline the need for its revision, at least for practice in Turkey.