Objective: To compare the accuracy of cone-beam CT (CBCT) with conventional horizontal and vertical parallax for the localization of ectopic maxillary canines. Design: Observational study. Methods: A typodont was constructed using human teeth embedded in radio-dense impression material, containing a simulated ectopic canine. Image sets (vertical/horizontal parallax and CBCT scans) were obtained for nine positions. A photograph confirmed the true position of the canine. Six observers used horizontal (HP) and vertical (VP) parallax techniques with conventional radiographs and the axial, coronal and sagittal CBCT views to locate the canine tip. Images were manipulated as required using proprietary software (http://www.carestream.com and http://www.i-cat.com). Intra-observer reproducibility was calculated by one observer re-evaluating the image sets after 2 months. Individual observer validity was calculated in relation to the photographic position of the canine using weighted Kappa. Differences in the proportion of correct locations between CBCT and vertical/horizontal parallax were tested using McNemar tests (P<0·05). Results: Intra- and inter-observer agreements were excellent (0·8985) and substantial (0·7528), respectively. Individual observer validity was substantial-excellent (0·7368–0·900). The canine position was correctly identified in 94% of cases located using CBCT, 83% using HP and 65% using VP. The differences between CBCT and vertical and horizontal parallax were highly statistically significant (P<0·01). Conclusion: CBCT is more accurate than either horizontal or vertical parallax for the localization of ectopic maxillary canine teeth.