Background and study aim: Measures for competence in endoscopic retrograde cholangiopancreatography (ERCP) during training are poorly defined. Currently, various training and accreditation programs base verification of competence on performance of a minimum number of procedures. There is a general awareness that procedural competence certification should be based on objective performance criteria. Continuous self-assessment using a Rotterdam Assessment Form for ERCP (RAF-E) can provide insight into trainee performance. The study aim was to express development in ERCP competence as a learning curve. Methods: ERCP trainees at a tertiary referral center in the Netherlands were invited to participate. Performed procedures were appraised using RAF-E. Indication for each ERCP and presence of a virgin papilla were documented. Complexity was graded on a 3-point scale. The primary outcome parameter was common bile duct (CBD) cannulation success rate. Success of the intended therapeutic interventions was additionally expressed as a learning curve. Results: 15 trainees were included. 1541 ERCPs (624 procedures in native papillary anatomy) were assessed through RAF-E. Unassisted CBD cannulation success rate improved from 36 % at baseline to 85 % after 200 procedures (P < 0.001), and in 624 patients with a virgin papilla from 22 % at baseline to 68 % after 180 procedures (P < 0.001). Learning curves for therapeutic interventions showed significant improvements for successful sphincterotomy (P = 0.01) and stent placement (P < 0.001). Conclusions: Learning curves are a valuable means for assessing competence in ERCP. Differences in learning curves can be shown with RAF-E. Verification of competence should be based on actual performance, instead of minimum numbers.