Roadside screening by police in Canada is a technique designed to improve the probability of detecting an impaired driver by providing an immediate indication of the blood alcohol concentration (BAC). Recent amendments to the Criminal Code of Canada provide for the police use of approved roadside screening devices and the compulsory participation of drivers suspected of having alcohol in their bodies. Laboratory evaluation of the ALERT was conducted by comparing ALERT direct blood, and Breathalyzer measurements in the post-absorptive stage of 32 human subjects with 10 to 180 mg percent consumption of alcoholic beverages of their choice. The results obtained with actual breath and blood samples were consistent with the manufacturers' claim of + 20 percent accuracy at 100 mg percent and above. A field evaluation of ALERT models set to 'fail' at 100 mg percent was conducted by 16 recently trained Breathalyzer operators in the Calgary City Police for 2 months commencing in November 1973. A total of 72 requests for breath samples were made to suspected impaired drivers; and Breathalyzer tests obtained the following 'fail' responses on the ALERT from 29 drivers resulted in a mean BAC of 143 mg percent. (The mean BAC of the remaining drivers was significantly greater at 184 mg percent). Additional field trials held in a rural area of Alberta for a 2-month period commencing in March 1976 resulted in 13 of 28 impaired driver suspects 'failing' an ALERT test. However, these 13 drivers who 'failed' were not offered Breathalyzer tests, nor were they charged--apparently because clinical signs of impairment necessary to cause arrest were not observed. Seven of the 29 drivers in the Calgary project and 2 of the 15 who 'failed' in the rural trial and who were subsequently measured on a Breathalyzer had BAC's of less than 100 mg percent. It is recommended that, in order to lessen discrepancies between ALERT and subsequent Breathalyzer tests, the ALERT device should be calibrated to 'fail' at a concentration 10 or 20 mg percent greater than the level at which formal charges for drunken driving are made. Two tables and 17 references are provided.