Geophysical investigation and Bayesian inversion have been performed to investigate the cavity structure. The targeted cavity is a lava cave in Cheju Island, Korea and the geometry of the cavity is already known. For the cavity detection, we have used the both dipole-dipole array resistivity and gravity method. Although the two methods generally have different objectives, common characteristics may be appeared in the cavity structure. That is, the resistivity value is high but the gravity one is low in the presence of the cavity. We choose the dipole-dipole array resistivity data as prior information for the Bayesian inversion of gravity data. We applied the indicator variogram modeling to transform the result of resistivity inversion to density information, as higher resistivity block has lower density one. This assumption is valid in Cheju Island where most of the rocks are composed of basalt. Generally, cavity investigation by single geophysical exploration is difficult to get effective result. However, this coupled interpretation proposes a new way to provide remarkable improvement and stability when compared to the result of single method.