Due to the real need for careful staff training in age assessment, in order to improve capacity, consistency and competence, new research on the reliability and repeatability of methods frequently used in age assessment are required. The aim of this study was two-fold: first, to test the accuracy of this method for age estimation; second, to obtain data on the reliability of this technique. A sample of 81 peri-apical radiographs of upper canines (44 men and 37 women), aged between 19 and 74 years, was used; the teeth were taken from the osteological collection of Sassari (Sardinia, Italy). Three blinded observers used the technique in order to perform the age estimation. The mean real age of the 81 observations was 37.21 (CI95% 34.37 40.05), and estimated ages ranged from 36.65 to 38.99 (CI95%-Ex1 35.42;41.28; CI95%-Ex2 33.89;39.41; CI95%-Ex3 35.92;42.06). The module differences found by the three observers were 3.43, 4.24 and 4.45, respectively for Ex1 x Ex2, Ex1 x Ex3 and Ex2 x Ex3. The module differences observed among real and observed ages were 2.55 (CI95% 1.90; 3.20), 2.22 (CI95% 1.65;2.78) and 4.39 (CI95% 3.80;5.75), respectively for Ex1, Ex2 and Ex3. No differences were observed among measurements. This technique can be reproduced and repeated after proper training, since it was found high reliability and accuracy.