Decreasing response to stress has been one goal of interventions aimed at reducing relapse to substances of abuse. A laboratory stress test that can be repeated would be helpful in testing the efficacy of interventions in decreasing the response to stress before more extensive trials are begun. The effects of two types of psychological stress tests, the Trier Social Stress Test (TSST) and a stress imagery test, on psychological, physiological, and hormonal responses (salivary cortisol and DHEA) were examined when each test was given twice to cocaine- or methamphetamine-dependent human subjects, 24 of whom completed at least one session. The stress imagery test produced significant changes in several of the subjective response measures in both first and second sessions, including several measures of negative affect and a craving measure. The TSST produced significant changes only in the second session. The stress imagery protocol showed better replicability across two sessions. Cocaine users and methamphetamine users did not respond similarly in their craving responses. Reported craving for methamphetamine after stress testing showed decreases or much smaller increases compared to that for cocaine. Neither stress test significantly increased salivary cortisol or DHEA, and changes in hormone concentrations were not related to subjective responses. These results suggest that stress imagery testing procedures may be useful as provocative tests of stress-induced affect and stimulant drug craving. Although less convincing because of the heterogeneity of the subjects, they also suggest that HPA axis responsivity is not clearly linked to acute stress-induced stimulant craving or affective response.