Data analyses. Because questionnaires from an individ- ual subject were non-independent, the data from multiple administrations were aggregated. Furthermore, because previous analyses had revealed no systematic difference in craving level as a function of mode of cue manipula- tion (e.g., Drobes & Tiffany, 1997), data from imagery and in vivo trials were collapsed based on urge content. Therefore, for each participant, an average QSU-Brief score was determined from the four urge trials (two urge–imagery, two urge–in vivo) and the four neutral trials (two neutral–imagery, two neutral–in vivo).