Objective Development of a rational and enforceable basis for controlling the impact of cannabis use on traffic safety. Methods An international working group of experts on issues related to drug use and traffic safety evaluated evidence from experimental and epidemiological research and discussed potential approaches to developing per se limits for cannabis. Results In analogy to alcohol, finite (non-zero) per se limits for delta-9-tetrahydrocannabinol (THC) in blood appear to be the most effective approach to separating drivers who are impaired by cannabis use from those who are no longer under the influence. Limited epidemiological studies indicate that serum concentrations of THC below 10 ng/ml are not associated with an elevated accident risk. A comparison of meta-analyses of experimental studies on the impairment of driving-relevant skills by alcohol or cannabis suggests that a THC concentration in the serum of 7-10 ng/ml is correlated with an impairment comparable to that caused by a blood alcohol concentration (BAC) of 0.05%. Thus, a suitable numerical limit for THC in serum may fall in that range. Conclusions This analysis offers an empirical basis for a per se limit for THC that allows identification of drivers impaired by cannabis. The limited epidemiological data render this limit preliminary.