Objective: Off-label prescription use can be a valuable source of innovation and is often supported by considerable evidence. However, many uses are not supported by evidence and unregulated off-label use may expose patients to serious harm. We sought to examine physicians' knowledge, attitudes, and experiences regarding off-label use. Methods: Ongoing nationally representative random-sample mail survey of general internists and psychiatrists. Knowledge of approved indications was assessed using a list of 15 specific drug-indication pairs for each specialty that varied in their clinical applications and levels of supporting evidence. Results: There were 166 respondents at the time of abstract submission (14% preliminary response rate); 67% were psychiatrists and 33% internists. Of the drug-indication pairs examined among general internists, use of the drug for the specified off-label indication varied from 24% (venlafaxine for adjustment disorder) to 87% (lisinopril for diabetic proteinuria) (mean 59%). Of these instances, physicians accurately identified the off-label status in less than half of the cases, ranging from 22% correct identification that escitalopram was off-label for panic disorder to 71% comfirmation that trazadone was off-label for insomnia (mean 36%). Results were substantively similar among psychiatrists; there was common use of specific drugs for off-label indications, despite poor to moderate knowledge of the labeling status of the use. Overall, the most important factors associated with off-label were practice guidelines or references (26%), successful off-label use in other patients (24%), and the absence of a labeled alternative (14%). Conclusion: These preliminary findings suggest that practicing internists and psychiatrists may not have precise knowledge of the FDA approved indications of the drugs they prescribe. Given how commonly off-label use occurs, further work is needed to better delineate and communicate the levels of evidence that should be incorporated in clinical decision-making to guide optimal prescribing.