The aims of this study were to determine the diagnostic accuracy and reproducibility of the laser fluorescence device (LF), and the relationship between laser fluorescence readings taken at the entrance of the fissure, ICDAS visual examination caries detection system and the histological depth of the lesion. Two hundred and forty teeth (122 human third molars and 118 bicuspids) were selected from 62 patients enrolled in the study. Visual and LF examinations (Diagnodent, Kavo, Biberach, Germany) of the occlusal surfaces were performed in vivo. After tooth extraction, histological sections were evaluated by stereomicroscopy in vitro. Statistical analysis was performed using multiple statistical methods (SPSS ver. 17). Intra-examiner reproducibility for the LF measurements was excellent: intra-class-correlation coefficient (ICC) for LF was 0.957. Kappa values for each examiner's reproducibility were 0.74-0.82. The diagnostic performance of the LF device gave a good overall diagnostic accuracy according to ICDAS II codes and histological values as indicated by the area under the ROC curve of 0.707 and 0.709 respectively. The results of the study showed acceptable diagnostic accuracy for the laser fluorescence device. This supports the view that dentists can be site specific in applying fluorescence-based devices to multiple discrete sites within the same surface. In conclusion, these diagnostic methods have different characteristics, indications and limitations for use. In order to detect caries on occlusal surfaces thoroughly, a combination of methods would be the best practice moderated by clinical knowledge and experience.