Ambient drinking cues are easy to underestimate - the stuff in the environment that shapes your expectations before you've consciously decided anything. Music is an obvious one. What made me stop on this paper: the link between party music and alcohol-themed lyrics is predictable from the sound alone. No lyric-reading, no genre labels. Just the audio profile.

What they did

Researchers analyzed 6,110 songs from the Billboard Hot 100 released between 1959 and 2020. For each track, they extracted Spotify audio features: danceability, speechiness, energy, valence, acousticness, liveness, instrumentalness, loudness, tempo, and mode. Logistic regression models then tested which features predicted whether a song's lyrics contained alcohol references.

The theoretical frame is Uses and Gratifications Theory, which holds that people actively choose media to satisfy psychological and social needs. The authors' premise: party contexts pull toward both a specific sound profile and alcohol-themed lyrics simultaneously - not by design, but because both serve the same social function. If that's right, the audio features and the lyric content should co-occur in a systematic pattern.

They do.

What they found

Alcohol references appeared in 16.1% of songs, with prevalence rising significantly across the 60-year window.

Audio features predicting higher odds:

  • Danceability (OR = 1.35) - stronger beat and rhythmic feel
  • Speechiness (OR = 1.39) - the strongest single predictor; vocal-heavy, rap-adjacent tracks
  • Liveness (OR = 1.09) - recorded with audible crowd energy
  • Positive valence (OR = 1.11) - emotionally upbeat
  • Major key (OR = 1.18) - minor-key songs skew away from it

Higher acousticness (OR = 0.82) and instrumentalness (OR = 0.85) went the other way. Loudness, tempo, and raw energy weren't significant.

What it means

Put those features together and you get the party track: danceable, crowd-sounding, emotionally bright, vocal-heavy, major key. That matches intuition. But odds ratios across 6,000 hits are harder to argue with than a feeling.

The more surprising result: loudness, tempo, and raw energy all came back insignificant. Intensity doesn't discriminate. Social texture does: vocal character, rhythmic feel, live energy.

Speechiness is the number I kept coming back to. Hip-hop and rap are the genres most likely to name specific drinks, and they also top Spotify's speechiness scale. Genre isn't in the model directly, but speechiness is doing that work - and at OR = 1.39, it's the strongest predictor in the whole model.

Liveness is worth a moment too. Songs that sound like live recordings - crowd noise, performance energy - are more likely to mention alcohol. That makes sense: live venues and drinking co-occur constantly. The music that captures that social energy apparently also captures the language around it.

The rising prevalence over six decades probably reflects genre shifts more than broad cultural drift. The window runs from 1959 to 2020, which is exactly the period hip-hop moved from the cultural margins to the dominant commercial form.

For an individual listener, this is background noise, not a warning: a playlist built around high danceability and positive energy is probably carrying more alcohol references than you'd guess while you're focused on the conversation.

The researchers' practical goal is a surveillance tool - a way to scan large music libraries and flag songs for lyric review without manual checking. Useful for public health monitoring, though the model isn't validated for that purpose yet, and the Billboard sample skews toward commercial mainstream.

One limit matters: the study doesn't distinguish between songs that glorify drinking and songs that mention it in passing. Those are different kinds of exposure. This study doesn't make any behavior-change claim - it's correlational throughout - so that distinction stays unresolved.

Source: Alcohol and Alcoholism, 10.1093/alcalc/agag041