The study aimed to assess the accuracy and reproducibility of occlusal caries depth measurements obtained from different imaging modalities. The study comprised 21 human mandibular molar teeth with occlusal caries. Teeth were imaged using film, CCD, two different cone-beam computerized tomography (CBCT) units and a microcomputer tomography (micro-CT). Thereafter, each tooth was serially sectioned, and the section with the deepest carious lesion was scanned using a high-resolution scanner. Each image set was separately viewed by three oral radiologists. Images were viewed randomly, and each set was viewed twice. Lesion depth was measured on film images using a digital caliper, on CCD and CBCT images using built-in measurement tools, on micro-CT images using the Mimics software program, and on histological images using AxioVision Rel. 4.7. Intra- and inter-rater reliabilities were assessed according to the Bland/Altman method by calculating Intraclass Correlation Coefficients (ICCs). Mean/median values obtained with intraoral systems were lower than those obtained with 3-D and histological images for all observers and both readings. Intra-observer ICC values for all observers were highest for histology and micro-CT. In addition, intra-observer ICC values were higher for histology and CBCT than for histology and intra-oral methods. Inter-observer ICC values for first and second readings were high for all observers. No differences in repeatability were found between Accuitomo and Iluma CBCT images or between intra-oral film and CCD images. Micro-CT was found to be the best imaging method for the ex vivo measurement of occlusal caries depth. In addition, both CBCT units performed similarly and better than intra-oral modalities.