Male albino rats were trained to traverse a straight alley for a reward of five intravenous injections of cocaine (0.75 mg/kg/injection in a volume of 0.1 ml/injection delivered over 4 s). Animals were tested one trial per day with the following dependent measures assessed on each trial: start latency, running time, the number of retreats, and the location within the alley where each retreat occurred. While start latencies remained short and stable, running times tended to increase over days. This effect was apparently related to a concomitant increase in the number of retreats occurring in the alley (r = 0.896). Retreats tended to occur in very close proximity to the goal box, suggesting that animals working for IV cocaine come to exhibit a form of conflict behavior (i.e., retreats) putatively stemming from the drug's well documented rewarding and anxiogenic properties. Consistent with this hypothesis was the demonstration that diazepam (0.5, 1.0, 2.0 mg/kg IP) pretreatment dose-dependently reduced the incidence of retreat behaviors in the alley. In addition, the rewarding efficacy of the cocaine dosing parameters was subsequently confirmed in the runway subjects by conditioned place preference. The present paradigm, therefore, provides a useful method for investigating the anxiogenic effects of self-administered cocaine in laboratory animals.