OBJECTIVE: To determine whether general practitioners can make accurate self assessments of their knowledge in specific areas. DESIGN: 67 general practitioners completed a self assessment of their level of knowledge over a variety of topics using a nine point semantic differential scale. An objective assessment of their knowledge was then made by administering true-false tests on two of the topics: thyroid disorders and non-insulin dependent diabetes. The study was repeated with another group of 60 general practitioners, using sexually transmitted diseases as the topic. SETTING: General practices in New Zealand. SUBJECTS: Random sample of 67 general practitioners in Auckland. MAIN OUTCOME MEASURE: Test scores for self assessment and for actual knowledge. RESULTS: Correlations between self assessments and test scores were poor for all three topics studied (r = 0.19 for thyroid disorders, 0.21 for non-insulin dependent diabetes, 0.19 for sexually transmitted diseases). CONCLUSIONS: As general practitioners cannot accurately assess their own level of knowledge on a given topic, professional development programmes that rely on the doctors' self perceptions to assess their needs are likely to be seriously flawed.