Traditionally, joint interpretation of seismic refraction and wide-angle reflection data and gravity data is based upon a well-known correlation between seismic P-wave velocity and density proved by numerous laboratory investigations of elastic properties of crustal rocks. One of the problems connected with this approach is that rocks with high content of calcium-reach plagioclase have higher P-wave velocity and do not satisfy the common density-Vp relationship. That is why joint interpretation based upon any conventional relationship between density and P-wave velocity cannot be applied to wide-angle profiles across large anorogenic rapakivi-gabbro-anorthosite massifs composed of rocks with high content of plagioclase. The problem can be solved if both P- and S-wave velocities are used to calculate the density model. The results of laboratory studies of rock properties demonstrate strong correlation between density and S-wave velocity. Moreover, the isotropic S-wave velocity seems to be generally more correlated to density than the P-wave velocity and less affected by high content of plagioclase. In spite of that, the known relationships connecting density to S-wave velocity or to both P- and S-wave velocities are very seldom used for joint interpretation of seismic and gravity data. The main reason for this is a lower quality of S-wave arrivals in explosion seismology, which makes it difficult to obtain reliable S-wave velocity models. In our paper we present the results of joint interpretation of seismic and gravity data collected along the EUROBRIDGE'97 wide-angle reflection and refraction profile in the Ukrainian Shield, where the absence of thick sediments made it possible to obtain both P- and S-wave velocity models. To calculate the density model along the EUROBRIDGE'97 profile we used a method of gravity data inversion, in which the density model was parameterised by the relationship connecting density to both P- and S-wave velocity models. Such a parameterisation makes it possible to obtain the relationship between density and seismic velocities by inverting the gravity data. As a result, non-linear and scattered relationship between density and seismic velocities was obtained for the EUROBRIDGE'97 profile. Analysis of the relationship demonstrated that the reason for this scattering is difference in density-velocity relationships for large-scale geological units crossed by the profile. In order to explain this difference, we compared the relationship between seismic velocities and density in three major geological units crossed by the EUROBRIDGE'97 profile with the petro-physical data from the Ukrainian Shield and other selected Precambrian areas. We demonstrated that the deviations from the averaged density-velocity relationships can be explained by specific mineral composition of rocks resulting from different age and conditions of crust formation. We showed how the analysis of density-velocity diagrams can be used to restrict the composition of the crust and, in particular, the composition and metamorphic grade of the lower crust.