This study measured validity and reproducibility of unaided clinical diagnosis fibre-optic transillumination (FOTI), and bite-wing radiology in the diagnosis of approximal caries. Sixty models were made using extracted premolars and molars, each containing four teeth with six contacting approximal surfaces. The teeth were examined first using unaided clinical examination and then using FOTI. Bite-wing radiographs were then taken of the teeth set in the models and examined. The three examinations were repeated after lweek. Histological sections of the undecalcified teeth were prepared following their removal from the models, and those showing signs of caries were examined to give the valid state of disease in each surface. The diagnostic threshold was caries penetrating into dentine. The reproducibility of all three methods was acceptable with kappa values exceeding 0.6. All specificity values exceeded 0.95. Statistically significant differences were seen between sensitivities for clinical (0.38) and bite-wing (0.59) diagnosis and between clinical and FOTI (0.67) examination, but not between bite-wing and FOTI. It is concluded that the validity of FOTI is at least as high as that of bite-wing radiology, and both are superior to unaided clinical diagnosis.