Estimating the strength, especially compressive strength, of rocks is one of the major problems in many civil engineering applications. Compressive strength of a rock is usually measured in a laboratory test of rock specimens obtained in boreholes drilled in the investigation site. If seismic velocity can be used for estimating rock strength, the seismic method can be employed effectively for profiling strength of a large rock mass because it is capable of determining subsurface seismic properties over a large area. Rock strength is often estimated from seismic velocity using correlations between data of each property, measured in the specific rock mass, or by using existing empirical relations. These techniques, however, have problems in accuracy and reliability of their estimates, because the correlations between measured data generally are made with a small number of data, and empirical relations are restricted for applicable rock types. In this study, a rock physics model is studied to estimate the compressive strength from seismic velocity more accurately and reliably. The confined compressive strength-seismic velocity relationship is modelled by combining two effective-medium models for (1) the confined compressive strength v. porosity and (2) seismic velocity v. porosity. The model is applied to S-wave velocity log data in soft sedimentary rocks, and the model predictions are compared with confined compressive strength, measured with the triaxial compression test on rock cores sampled in the same borehole as that used for the velocity logging. The model is also applied to ultrasonic P-wave velocity and confined compressive strength data measured in the laboratory, using core samples of sedimentary rocks collected from various sites in Japan. Good agreement between model-calculated and measured data in the confined compressive strength-seismic velocity relationships in these cases reveals that the confined compressive strength can be estimated from seismic velocity using a rock physics model.