The Steel Helmet project was meant to assess the feasibility of the chemical warfare agent (CWA) detector dog concept. A relatively benign organophosphate pesticide called dichlorvos was used as a surrogate for CWAs. Using conventional training techniques, U.S. Department of Defense military working dogs were taught to discriminate scent boxes containing dichlorvos from “vehicle” scent boxes. Experiment 1 appeared to show that two out of three subjects were capable of criterion accuracy (0.95 or better) at the lowest test concentrations of dichlorvos— 3 and 1 parts per billion by volume (ppbv). An additional manipulation showed that, when differential contamination of the scent boxes in Experiment 1 was accounted for, all three subjects fell short of criterion accuracy when tested at 1 ppbv. The canine dichlorvos detection "threshold" was therefore estimated at equal to or less than 3 ppbv, but not so low as 1 ppbv. Experiment 2 demonstrated that detection responding was specifically controlled by dichlorvos, rather than concomitant odors, and that the subjects were not merely reacting to the novelty or salience of dichlorvos vapor. The implications of these results for the feasibility of the CWA detector dog concept are discussed in terms of safe canine CWA exposure levels.