PURPOSE: To assess the accuracy of residents' record review, using trained abstractors as a gold standard comparison. METHOD: In 2005, the authors asked 74 residents to review their own charts (n = 392) after they received brief instruction on both how to locate data on the medical record and how to use a data abstraction form. Trained abstractors then re-reviewed these charts to assess performance of preventive health care measures in medicine (smoking screening, smoking cessation advice, mammography, colon cancer screening, lipid screening, and pneumonia vaccination) and pediatrics (parent smoking screening, parent smoking cessation advice, car seat safety, car restraint use, eye alignment, and immunizations up to date). The authors then quantified agreement between the two record reviews and assessed the sensitivity and specificity of the residents versus the trained abstractors. RESULTS: Overall resident-measured performance was similar (within 5%) to that of the trained abstractor for five of six measures in medicine and four of six in pediatrics. For the various measures, sensitivity of resident-measured performance ranged from 100% to 15% and specificity from 100% to 33% compared with the trained abstractors. Relative to the trained abstractor record review, residents did not overestimate their performance. Most residents' (81%) relative performance rankings did not change when the basis for the ranking was resident measured versus trained abstractor measured. CONCLUSIONS: Residents' self-abstraction can be an alternative to costly trained abstractors. Appropriate use of these data should be carefully considered, acknowledging the limitations. © 2009 Association of American Medical Colleges.