Fig. 2.

(A) Mean (and SEM) subjective happiness measured at three timepoints (baseline, post intervention and post stress test) in the two groups. The breath counting group relative to the control group showed greater happiness post intervention, and an attenuated stress induced decrease in happiness post stress test. (B) Mean subjective annoyance measured at three timepoints in the two groups. The breath counting group relative to the control group showed reduced annoyance post intervention, and an attenuated stress induced increase in annoyance post stress test. (C) Mean percent choice of alcohol versus food pictures at baseline and across three time bins of the stress test, in the breath counting and control group. Both groups showed an increase in alcohol choice at stress test bin 1 vs. baseline. The groups differed thereafter. In the breath counting group, alcohol choice declined linearly back to baseline across test bins 2 and 3. In the control group, alcohol choice was stable above baseline across the stress test. (D) Moderation analysis: the effect of breath counting versus control intervention on recovery from stress induced alcohol-seeking (i.e. decline in alcohol choice across stress test trials indexed by a regression coefficient), was moderated by AUDIT scores, such that breath counting produced steeper recovery in participants with low and intermediate, but not high, AUDIT scores. (E) Equivalent moderation analysis with PROMIS alcohol use scores. Breath counting produced steeper recovery from stress induced alcohol-seeking in participants with low and intermediate, but not high, PROMIS alcohol use scores.