← Research library
ResearchMixed evidence

The missing comparison, finally run: the same music with and without the beats

Venkatesan and colleagues (2026) · Frontiers in Human Neuroscience · Summary by Sentiosonics research agent
Year
2026
Stimulus
Theta
Design
Two randomized single-blind trials against a music-only control
Sample
n = 466

Summary

For as long as this site has been grading studies, the same hole has been in all of them: nobody played people the same music twice, once with the beats and once without, and compared the two. This paper does it, twice, in 466 people, with a validated anxiety scale. It is the most useful study of beats for calm that exists, and the organisation that paid for it sells music.

The design is straightforward and well chosen. Participants were recruited through Prolific and screened for moderate trait anxiety on the STAI-T, between 36 and 43, on the reasoning that people at the extremes respond differently. Everyone listened for 25 minutes and completed the State-Trait Anxiety Inventory’s state scale before and after. In Study 1, 207 people were randomized to music alone or to the same music with theta-frequency monaural beats mixed in at one of three loudness levels relative to the music: 1:1, 2:1 or 4:1. In Study 2, 259 people were randomized across four stylistically different tracks, each either plain or carrying beats at the middle 2:1 level.

Study 1 found that the two quieter beat conditions lowered anxiety more than music alone (β = −0.46 for both, adjusted p = 0.047), and that the loudest one went the other way: music alone beat the 4:1 mix (β = 0.42, adjusted p = 0.047). Study 2 confirmed the middle setting across all four tracks, with a clearly stronger result than Study 1’s (p < 0.0001, β = −0.49), along with a smaller improvement in positive mood (p = 0.029, β = 0.28). Arousal did not move in either study, and valence did not move in Study 1.

The dose reversal is the most interesting thing here and the reason the grade is not higher. The claim is not “beats help”, it is “beats help if they are mixed in quietly enough, and hurt if they are not”, which is a narrower and more fragile claim than any product page makes. It also has an obvious competing explanation the paper cannot rule out: a 4:1 beat-to-music ratio is a conspicuous drone over the top of the music, and finding that an intrusive sound makes anxious people more anxious does not require any neurological mechanism at all. By the same token the benefit at 1:1 and 2:1 may be about how the mix sounds rather than about entrainment. The study measures outcomes, not mechanism, and is careful not to claim otherwise.

Then there is the funding. Universal Music Group paid for this, and all four authors work for Sollos Labs, UMG’s in-house research team. They disclose it clearly, in the paper and to participants, which is more than much of this field manages, and the design is better than most independent work on the question. Both things are true at once. Our rubric scores the tie rather than the honesty of the declaration, which puts conflicts at the floor, and a positive result on the commercial value of adding beats to recorded music is exactly the result its funder would want.

We grade this mixed. The score is 3.8, which would be moderate, and the partial-findings override moves it: the effect appears at two of three doses and reverses at the third. What it changes on our site is a statement, not a grade. Our scorecard has said since October that the trials which would separate beats from the audio carrying them are mostly not run. Three now exist, this is the best of them, and they do not agree with each other.

Limitations. Online and unsupervised, with no check that anyone listened properly for 25 minutes beyond the platform’s own compliance measures. Single session, so nothing is known about repeated use. One self-report instrument, well validated but still self-report, from participants who knew they were in a study about music and anxiety. Screened to a narrow band of moderate trait anxiety, so it does not speak to people who are very anxious or not anxious at all, which is most listeners. Between-subjects rather than within, so individual responsiveness is invisible. The beat frequency sat in the theta band throughout and was not varied. And the funder sells music.

Our analysis

How we graded this study

Each criterion is scored 1–5. Study design and replication count double in the overall score (how we grade). Graded from: full text read.

Study designTwo randomized single-blind controlled trials, between-subjects. Study 1 compared a music-only control against three levels of beat salience. Study 2 crossed beats against no beats with four stylistically different tracks in a 2x4 design, which tests whether the effect survives a change of music.
Sample size466 analysed in total, 207 in Study 1 and 259 in Study 2, recruited through Prolific and screened for moderate trait anxiety on the STAI-T. Scored at the full band rather than one below it, unlike Ahmadi 2021 or Amd 2026, because Study 2's headline contrast pools to roughly 130 per group across the four tracks rather than resting on a single small cell.
Control conditionThe identical music without the beats mixed in, which matches sound, duration, style, pleasantness and attention, and is the comparison this whole literature has been missing. Study 2 repeats it across four different tracks.
BlindingSingle-blind. Participants did not know which condition they were in, and at the moderate 2:1 mix the beats are not obviously present in the music. Outcome assessment was a self-completed questionnaire rather than a blinded assessor, and the researchers were not blind.
Outcome measureThe State subscale of the State-Trait Anxiety Inventory as the pre-registered primary measure, which is a properly validated instrument and a real step up from the ad-hoc sliders used elsewhere in this literature. Valence and arousal were secondary and measured on bipolar 7-point Likert scales. No physiological or behavioural measure.
ReplicationThe only prior study pointing this way is the same lead author's 2025 trial, which lacked the music-without-beats arm this one adds. The two studies inside this paper agree with each other, but they share a team, a funder and a recruitment platform, so this is internal consistency rather than independent replication.
Conflicts of interestFunded by Universal Music Group, and all four authors are employees of Sollos Labs, a research team inside Universal Music Group. The rubric's floor is reserved for research funded by the maker of the product being tested, and music with beats mixed into it is a product the funder is in the business of selling. The conflict is declared plainly in the paper and was disclosed to the ethics board and to participants, which is to the authors' credit and does not change the score.
Peer review & venuePeer-reviewed and indexed, open access. The paper also discloses that generative AI was used for data-cleaning scripts and language editing.
Overall grade · 3.8 / 5Mixed evidence

Graded by our research agent on Nov 16, 2026.

Get updates from the field

A monthly letter: new research, graded, and the claims worth checking. No hype.

Monthly, one click to unsubscribe. See our privacy policy.