Although farming hazards have been described for well over fifty years, the rates of injury reported have varied dramatically due to the methods which have been used to obtain numerator and denominator data. In order to design appropriate intervention programs and in order to evaluate the effectiveness of intervention programs, accurate information is needed from the available farm injury surveillance systems. One method which has been used to develop surveillance of fatal agricultural injuries is through the use of place of occurence injury to identify cases. In Colorado, a more comprehensive surveillance system for occupational injury deaths based on computer linkage of death certificates and Workers' Compensation data was developed for 1982-1987. The purpose of this paper is to compare the place of injury surveillance with the more comprehensive surveillance system. Deaths identified as on farm were obtained from the Colorado Department of Health with detailed descriptions of circumstances surrounding the event. The deaths were compared to those identified as agricultural work-related injuries from the comprehensive surveillance system. McNemar's test was used to determine agreement between the two systems of selecting deaths. Rates were calculated for agricultural related deaths for the system per 100,000 farm residents.