The main strength of this study was the large sample size, representing 13 166 primary care physicians in Canada. Hence, the data are representative and can be generalized to Canadian primary care physicians because health care delivery systems are similar. Furthermore, these findings can also be generalized to other countries with similar provision of health care services and practice settings. However, despite the large sample there were some areas where data may have been sparse. For example, few physicians would have had less than 5 years of practice experience and been older than 55 years. Hence, although statistical significance may have been reported for some associations, these may occasionally have been due to artificially produced gaps in data. The main limitation of this study was the inherent effect of self-reporting, with no means of concurrent validation. Validating physician self-reporting of counseling and prescription behaviors would be extremely time-consuming but could be performed in smaller cohorts through medical record audits, interviews with patients after visits, or the use of administrative databases if coding included preventive health activity, such as physical activity. Note that self-report may, in fact, reflect intention rather than real activity behavior, and it is not clear what is actually being provided. For example, one study29 of physicians' counseling practices had physicians keep a log of what they did during each session. These authors found that physicians who reported counseling more also reported more frequent counseling on the questionnaire they had filled out later. Similarly, the US National Ambulatory Medical Care Survey30 also reported low rates of healthy lifestyle counseling in patients with cardiovascular risk factors, but whether reported counseling as collected in this administrative database reflects real behavior is not known. Further studies should be designed to measure practice activity rather than perceived or intended behavior. A significant limitation in the present study was the absence of specific information regarding the types of physical activity advice, assessment tests, or the content discussed or prescribed to patients. This information would have been important in the further definition of which attributes could be considered in future physical activity counseling training programs for primary care physicians. Finally, this study observed a response rate of 51%, which was lower than the rates of 54% to 76% seen in other physical activity surveys. However, these other studies used considerably smaller and less representative sample sizes (ie, 63-78 respondents),4,7,17- 19,22 and, hence, the generalizability and potential impact of these previous studies compared with the present study should be considered.