To evaluate the agreement between six currently available periapical radiography-based methods for measuring the root canal curvatures in mesial roots of mandibular first molars, assessed by two examiners with different proficiency levels. Non-endodontically treated mesial roots of 41 human mandibular first molars were radiographed using the parallelling technique. Two independent observers (a specialist in endodontics and radiology and a final-year dental student) assessed their root canal curvature using the methodologies described by Schneider (1971, Oral Surgery, Oral Medicine, Oral Pathology and Oral Radiology, 32, 271), Weine (1982, Endodontic therapy), Berbert and Nishiyama (1994, Revista Gaúcha de Odontología, 356), Luiten etal. (1995, Journal of Endodontics, 21, 26), Hankins and ElDeeb (1996, Journal of Endodontics, 22, 123) and Pettiette etal. (1999, Journal of Endodontics, 25, 230). Intra- and inter-examiner reliability was assessed using the intraclass correlation coefficient. The differences in curvature angle measured by the different methods were compared using the one-way anova for repeated measures test, followed by Tukey's post hoc analysis. The effect was calculated using the Cohen's d method. To determine the agreement between methods, the Bland-Altman analysis was used. The significance level was set at 5%. Agreement for the observers was excellent (>0.81) for the six methods considered. For the angle comparisons between methods, the maximum differences were for Schneider versus Weine (35.77°) and Luiten versus Hankins (35.14°), whilst the highest percentage of angles with a difference >10° were Weine versus Luiten and Berbert versus Pettiette (90%) and the comparison Weine versus Hankins presented with the lowest frequency (15%). Excellent agreement was found for five comparisons: Weine versus Luiten (0.940), Berbert versus Pettiette (0.917), Weine versus Pettiette (0.907), Luiten versus Pettiette (0.904) and Berbert versus Luiten (0.812). Compared to Schneider's method, the other methods showed a tendency of increasing difference as the angles became more acute. The other methods exhibited linear differences, remaining constant for smaller and larger angles. Reliability was excellent for all methods assessed separately. Maximum differences in curvature angles were found when comparing Schneider versus Weine and Luiten versus Hankins. Excellent agreement was found for Weine versus Luiten, Berbert versus Pettiette, Weine versus Pettiette, Luiten versus Pettiette and Berbert versus Luiten. In the presence of dilacerations, the method by Schneider was less sensitive.