Aim of this in vitro study was to assess the reliability of measurements by a laser fluorescence technique for occlusal caries detection. Four dentists using four DIAGNOdent devices (Dds) according to manufacturer's instructions examined 80 carefully cleaned extracted molars with non-cavitated occlusal lesions. Inter- and intra-examiner reproducibility were calculated using Lin's intra-class correlation coefficient (ICCLin) and the limits of agreement by Bland and Altman (Comp Biol Med 1990;20:337). An excellent intra-examiner reproducibility was found for all dentists (ICCLin 0.93-0.98). The inter-examiner reliability was proved to be good and excellent (ICCLin 0.74-0.98). By reason of the registered lower mean fluorescence values of dentist 3 compared with his/her colleagues all these measurements were excluded from further calculations. Differences between Dd instruments were not evident. When the values of the Dd were categorized according to the treatment related cut-off limits a substantial reproducibility was found in the range of 0-14 (ICCLin 0.66; DIFF 4.0/-3.5) and >30 (ICC(Lin) 0.6% of all Dd measurements were reproduced in the range 15-30 (ICC(Lin) 0.16; DIFF 8.1/-7.3). The results suggest that dentists should be trained before using Dd. The low reliability in the interval of 15-30 could indicate a limited use for a longitudinal caries monitoring on pits and fissures; further research should be focused on this clinically important interval.