Classical 3-D refraction traveltime tomography algorithms may suffer from computational limitations due to the large datasets that come from current seismic acquisition surveys. To overcome these issues, we suggest a 3-D refraction tomography algorithm based on adjoint state techniques to derive the gradient of the traveltime misfit function. We use the Eikonal equation for the forward modelling, and iterate with a conjugate gradient method. A 3-D synthetic example with a realistic size acquisition demonstrates the efficiency and the great potential of the adjoint state method for 3-D applications of refraction tomography.