OBJECTIVE: To compare the accuracy and repeatability of three diagnostic systems (DIAGNOdent, visual and radiographic) for occlusal caries diagnosis in primary molars. DESIGN: Two examiner, in vitro, blinded study. Histological gold standard. MATERIALS AND METHODS: 58 occlusal surfaces of primary molars examined in turn by two examiners using each of three diagnostic systems (DIAGNOdent, visual and radiographic). These results were compared with a histological gold standard. Sensitivity and specificity were calculated for each diagnostic system at a range of thresholds. Inter- and intra- examiner repeatability were calculated for each diagnostic system using the kappa statistic. MAIN OUTCOME MEASURES: Sensitivity, specificity, inter and intra examiner repeatability for each diagnostic system. RESULTS: The highest sensitivity values were provided by DIAGNOdent (0.77 and 0.80, examiners 1 and 2 respectively) however this was offset by a lower specificity (0.82 and 0.85) than all other systems with the exception of examiner 1 at V1 visual threshold. The DIAGNOdent gave the highest values of kappa for intra- and inter-examiner repeatability with the exception of intra-examiner repeatability for examiner 2 where visual diagnosis had the highest value of kappa. CONCLUSION: The DIAGNOdent was the most accurate system tested for the detection of occlusal dentine caries in primary molars. The performance of the DIAGNOdent systems was not statistically significantly better than that achieved using visual examination for non-cavitated teeth (V1 and V2 threshold). DIAGNOdent may prove useful as a predictive clinical tool, however with appropriate training visual examination may offer similar results without the need for additional equipment.