In order to validate an inhouse computer code, RIAM-CMEN (Computation Method for Extremely Nonlinear hydrodynamics), based on CFD techniques, the forced oscillation test in heave was conducted in a wave channel for obtaining the data of the addedmass and damping coefficients of a, box-shaped floating body. Obtained results were expected to agree with computed ones by a 2-D BEM based on the potential-flow theory. However the result was not the case, and a large discrepancy was observed. Viscous effects associated with vortex shedding from a sharp corner of the model are one reason, but the magnitude in the difference was too large to attribute only to the effect of vortex shedding. To see the reasons, we carried out observation of the flow around the model and modified the settings in the experiment. One important factor was found to be the gap between the sidewalls of the model and wave channel. In fact the gap on each side was 5 mm in the original experiment, but after attaching a thin plate on both sides of the model to lessen the gap down to 1 mm, measured results became reasonable. To understand hydrodynamic reasons in this rather drastic change, numerical computations were performed using a 3-D BEM and the 2-D version of RIAM-CMEN. We confirmed through comparisons that unnatural variation observed at higher frequencies for the gap equal to 5 mm was associated with trapped waves generated in the gap and that the difference in the damping coefficient between the experiment and the computation by BEM was associated with the effect of vortex shedding.