EVIDENCE LIBRARY
Alcohol and the Cocktail Party Effect: What an 81-Person Listening Study Found
Alcohol did not significantly change detection of a person’s name in ignored speech. Evidence for other listening outcomes was less decisive, and the task did not establish intact driving attention.
Alcohol did not significantly change whether people noticed their own name in an ignored speech stream in this experiment. The result supports a specific null finding about selective listening. It does not mean that a person who follows a conversation has normal attention for driving or that alcohol leaves every auditory function unchanged.
What the cocktail party task tested
Harvey and Beaman recruited 81 undergraduates with normal hearing and randomized them to alcohol or placebo. Before drinking, participants completed an operation-span task measuring working-memory capacity.
After beverage absorption, participants repeated words spoken by a female voice into one ear while ignoring a male voice in the other. Their own name was secretly inserted into the ignored stream. The investigators later asked what they remembered or had noticed.
This was a controlled headphone task using synchronized word lists, not a real party conversation or a driving simulation. Its outcome concerned detection of personally meaningful information in a channel participants were told to ignore.
Exposure was estimated from breath
The alcohol group’s mean BAC estimate was approximately 0.08% before the listening task and 0.07% later. These were converted from breath measurements using a fixed 2300:1 ratio, rather than measured directly in blood.
The alcohol group also reported greater subjective intoxication than placebo. Demonstrating exposure or perceived intoxication does not require every selected performance endpoint to change significantly.
Name detection showed no alcohol effect
Overall, 42 of 81 participants noticed their own name. Detection was approximately half in each beverage condition. Neither alcohol nor the working-memory classification produced the predicted difference in name detection.
The authors’ Bayesian analyses supported the null interpretation for the central name-detection findings. Such support is specific to the tested models, sample and exposure; it is not a probability that a particular listener is unimpaired.
Working memory was measured while participants were sober and then divided into higher and lower groups. This did not directly measure how alcohol changed each person’s working-memory capacity.
Not every null result was equally informative
The overall alcohol comparison for speech-shadowing errors had p = 0.065. Its Bayesian evidence was approximately balanced between models, rather than clearly demonstrating no effect. The working-memory comparison for overall errors was also inconclusive.
Participants who noticed their name made more shadowing errors around its presentation, consistent with a brief shift of attention. Alcohol did not significantly alter that local pattern.
Alcohol did not significantly affect incidental recall of the speech. Lower working-memory participants recalled fewer recent words from the attended stream, but the overall working-memory main effect and the position-specific interaction required separate interpretation. The paper’s broad abstract should not flatten all these statistical distinctions into one uniformly established null result.
Why conversational ability cannot establish driving fitness
The experiment did not test detection of sirens, localization of sound, comprehension of complex instructions, divided visual–auditory attention or vehicle control. Spoken words and simple auditory tones also need not respond to alcohol in the same way.
The between-person design, one exposure range and possible differences in participants’ listening strategies limit generalization. The study cannot reconstruct an earlier BAC or diagnose impairment from whether a person answered a question coherently.
The useful conclusion is that alcohol effects depend on the task and that normal performance on one listening measure is not a comprehensive functional examination. The site’s laboratory-to-driving prediction analysis explains that distinction. Its simulated commute study assesses a different, directly driving-related combination of demands.
Primary source
Harvey AJ, Beaman CP. Acute alcohol intoxication and the cocktail party problem: do “mocktails” help or hinder? Psychopharmacology. 2021;238:3083–3093. https://doi.org/10.1007/s00213-021-05924-6.