In vivo effects of peer modeling on drinking rate.
Four live peer models are the minimum dose to cut an adult’s drinking speed in a real bar.
01Research in Context
What this study did
Researchers visited a neighborhood tavern. They asked one woman to drink at her normal pace.
While she drank, one, two, or four nearby patrons also drank. The peers drank 50 % slower than usual.
The team flipped the set-up on and off across four nights. They counted sips per minute to see when her speed changed.
What they found
Only the four-model nights cut her drinking rate. One or two slow drinkers made no dent.
When the four peers left, her speed bounced back. Bring them back and it dropped again.
How this fits with other research
Zeiler (1969) used a film, not live people, to boost shy preschoolers’ play. One short movie worked—no crowd needed. The tavern study shows adults sometimes need a bigger live pack.
Marcucella et al. (1978) found kids learned new words after seeing a model only half the time. Again, one adult was enough. The adult tavern crowd seems to set a higher bar.
MacFarland et al. (2025) flipped the camera around. They filmed typical teens, then showed the clip to other teens so they could model friendly talk toward classmates with autism. One video, not four bodies, did the job. Together the papers trace a line: young kids and teens shift with light input; grown-ups in a bar need a fuller live chorus.
Why it matters
If you want to slow an adult’s drinking in a real social spot, one buddy won’t cut it. Aim for a group of at least four peers who model the slower pace. This head count rule can guide bar-based programs, college party teams, or any naturalistic intervention where the setting itself keeps the reinforcement flowing.
Want CEUs on This Topic?
The ABA Clubhouse has 60+ free CEUs — live every Wednesday. Ethics, supervision & clinical topics.
Join Free →Recruit four volunteers, not one, to model a slower sip rate during the next social-skills outing.
02At a glance
03Original abstract
One female subject drank beer with four female confederate models and two participant observers in a small town tavern. A single subject repeated measures reversal design was used. Condition 1 indicated subject baseline drinking rate. For the first intervention one confederate modeled at a rate 50% less than the subject's baseline rate. Interventions II and III were identical to Intervention I except that two confederates modeled at a rate 50% less than the subject's baseline rate for Intervention II and four confederates modeled at a rate 50% less than the subject's baseline rate for Intervention III. Interventions were separated by returns to baseline. The study was concluded with a final return to baseline. There was no change in subject drinking rate as a function of either one or two confederates modeling the 50% rate. However, when four models drank at the lower rate, subject drinking rate matched that of the four confederate models. Implications and suggestions for further research on modeling are presented.
Journal of applied behavior analysis, 1980 · doi:10.1901/jaba.1980.13-149