Cross-borehole electrical resistivity tomography was used to detect and image a concealed air-filled mineshaft at a greenfield test site. The measurement configurations and panel combinations were selected using a two-stage optimization process. An optimal set of array configurations was selected for each cross-borehole panel on the basis of the model resolution matrix. Subsequently, various combinations of panels were tested with synthetic and field data to determine the effects of coverage and data density on the resulting tomographic image. In the field trials, complicating factors were introduced by the use of resistive cement linings in the boreholes. A resistive feature was detected between the boreholes using a single panel and a 2.5D inversion, but the image quality was too poor to identify this as a mineshaft. A much-improved image was obtained using eight boreholes and eight panels with a full 3D inversion. Only four of these panels intersected the shaft. Crucially, the other panels provided coverage of outlying regions of the model, enabling the inversion algorithm to distinguish between the resistive effects of the borehole linings and the mineshaft.