Objective: The objectives of this research were to (1) quantify the discordance between the caries lesion depth at which dentists restored initial lesions during a clinical study ("actual depth") and the lesion depth that they reported during a hypothetical clinical scenario ("reported depth"); (2) test the hypothesis that certain practitioner, practice, patient, and caries lesion characteristics are significantly associated with this discordance. Methods: Practitioner-investigators who perform restorative dentistry in their practices completed an enrollment questionnaire and participated in two consecutive studies on caries diagnosis and treatment. The first study was a survey asking about caries treatment. The second study collected data on restorations placed in routine clinical practice due to caries in patients over 19 years of age on occlusal surfaces only or proximal surfaces only. We report results on 2691 restorations placed by 205 dentists in 1930 patients with complete data. Results: Discordance between actual depth and reported depth occurred in only about 2% of the restorations done due to proximal caries, but about 49% of the restorations done due to occlusal caries. Practice type, restorative material used and the diagnostic methods used were significantly associated with discordance. Conclusion: Dentists frequently restored occlusal caries at a shallower depth as compared to their reported depth, but the discordance was very small for proximal lesions. Discordance for occlusal caries was more common when radiographs were not taken or if a resin restoration was placed. © 2011 Elsevier Ltd. All rights reserved.