Occlusal caries lesions may progress into the dentin without this resulting in a macroscopic breakdown of the enamel surface. Imaging methods may therefore be needed to aid in the visual detection of occlusal caries. It was the aim of this study to evaluate diagnostic accuracy in a laboratory set-up of visual inspection (VI), fiber-optic-transillumination (FOTI), conventional radiography (CR), and two digital radiographic image modalities (DRm and DRr) for detection of occlusal caries in clinically non-cavitated teeth. Eighty-one extracted third molars from 18-20-year-old males were assessed by four observers on a five-rank confidence scale by the five methods. Ground sections (500-600 microns) served as validation for true state of disease: 1 = no caries in dentin, 2 = caries just beyond dentino-enamel junction, or 3 = deep dentinal caries, halfway or more to the pulp. ROC analysis was performed on the basis of the confidence rank scale data on two diagnostic thresholds, T1 = caries in dentin (disease state 2+3) and T2 = caries deep in dentin (state 3). On the T1 level, use of the FOTI method gave on average the most accurate diagnosis, closely followed by VI, both performing better than use of radiography. On the T2 level, all five diagnostic methods performed equally well.