Description: A combined 3-D physical oceanographic model and a field sampling program was performed in July and August 1994 to investigate the potential drift of larval Baltic cod from the center of spawning effort in the Bornholm Basin, Baltic Sea. The goal of this exercise was to predict the drift trajectories of cod larvae in the Bornholm Basin, thereby aiding in the development of future sampling programs as well as the identification of processes influencing larval retention/dispersion in the Bornholm Basin. Distributions of variables (T, S and larval distribution) were obtained utilizing a three-dimensional eddy-resolving baroclinic model of the Baltic Sea based on the Bryan-Cox-Semtner code. Larval drift was simulated by the incorporation of a passive tracer into the model utilized to represent individual cod larvae. Additionally, simulated Lagrangian drift trajectories are presented. For model purposes, initial fields of temperature, salinity and cod larvae concentration for the Bornholm Basin were constructed by objective analysis using observations taken during a research survey in early July, 1994. Outside the Bornholm Basin generalized hydrographic features of the Baltic Sea were utilized with the baroclinic model forced by wind data for the whole Baltic taken from the Europa-Modell (EM) of the German weather service, Offenbach. Verification of simulations was performed by comparison with field measurements of hydrographic variables and ADCP derived current measurements taken during the surveys. In general, most of the hydrographic features observed during the second research cruise are correctly simulated, with variations mainly attributed to the prescribed initial conditions outside the Bornholm Basin. Results from larval sampling during the second cruise could not entirely confirm the modeled larval distributions due to the low numbers of larvae captured. However, the modeled results based on the agreement of the flow fields and hydrographic properties with observed features suggest that predictions of larval distributions can be made with a high degree of confidence if appropriate larval behaviours are included in the simulations.