Four college students were exposed to a Sidman avoidance procedure to determine if an avoidance contingency involving 20% carbon dioxide-enriched air (CO2) would produce and maintain responding. In Phase 1, two conditions (contingent and noncontingent) were conducted each day. These conditions were distinguished by the presence or absence of a blue or green box on a computer screen. In the contingent condition, CO2 presentation were delivered every 3 s unless a subject pulled a plunger. Each plunger pull postponed CO2 presentations for 10 s. In the noncontingent condition, CO2 presentations occurred on the average of every 5 min independent of responding. Following stable responding in Phase 1, condition-correlated stimuli were reversed. In both conditions, plunger response rate was high during the contingent condition and low or zero during the noncontingent condition. Furthermore, subjects avoided most CO2 presentations. However, CO2 presentations did not increase verbal reports of fear. Overall, the results from the present study suggest that CO2 can be used effectively in basic studies of aversive control and in laboratory analogues of response patterns commonly referred to as anxiety.