The aim of the present study was to test the reproducibility, sensitivity, and specificity of cone-beam computed tomography (CBCT) in detecting incipient furcation involvement. Fifteen macerated pig mandibles, with intact second molar teeth and preserved adjacent cortical areas, were used. Simulated lesions were created in the furcation region of these teeth by applying 70% perchloric acid in up to four possible buccal/lingual sites in the right/left sides of each mandible. The mandibles were then submitted to a CBCT scan. Two blinded and calibrated experienced oral and maxillofacial radiologists interpreted the exams. Furcation involvement was also assessed in the regions without simulated lesions. CBCT showed high levels of accuracy, ranging from 78% to 88%. The variations in Kappa values for intra- and inter-observer agreement (0.41-0.59) were considered moderate. CBCT can be considered a reliable and accurate method for detecting incipient furcation involvement.