Dose as a design variable in a sound intervention

Abstract

The literature on sound interventions records small to moderate effects without having fixed the exposure parameters that produce them. A systematic review published in Acta Neuropsychiatrica in 2026, covering ten clinical trials and around 970 participants aged nineteen to twenty-four, documents effect sizes of d between 0.34 and 0.59 for binaural beats and d between −0.47 and −0.36 for music, together with a methodological heterogeneity that prevented a definitive meta-analysis: the exposures studied range from a single five-minute session to five weeks of nightly listening. A randomized trial published in PLOS Mental Health on January 21, 2026, with 144 participants with moderate trait anxiety and four arms, provides an example of the dose-response modeling the review calls for: music with theta-band auditory stimulation outperformed the pink noise control on the anxiety measures, with medium effect sizes, and the 36-minute session reduced negative affect significantly more than the 12-minute one, with a large effect size, while on anxiety the differences between durations did not reach statistical significance. Dose constitutes a declarable protocol parameter and its optimal value depends on the outcome variable being measured.

Introduction

Subjective improvement after a listening session carries no evidential weight on the efficacy of the intervention as long as the exact stimulus, the duration of the exposure, the person’s starting state, the comparator and the measurement instrument remain unrecorded. Accumulating evidence on sound requires those parameters to be declared before the first data point is collected.

Dose is the first of them. Fixing it in writing turns a playlist into a protocol, and a protocol is the object another team can replicate or refute.

Exposure heterogeneity in the available literature

A systematic review published in Acta Neuropsychiatrica in 2026 gathered ten clinical trials on music and binaural beats in young adults, with around 970 participants aged nineteen to twenty-four. The effects exist and run from small to moderate: d between 0.34 and 0.59 for binaural beats, with the theta band standing out on physiological stress reduction, and d between −0.47 and −0.36 for music, at its strongest in clinical anxiety settings.

Methodological heterogeneity prevented a definitive meta-analysis, and the reason sits in the exposure parameters: the interventions included range from a single five-minute session to nightly listening programs lasting five weeks. Across that range, the question of whether sound works has no available answer, because each study tested a different exposure.

The authors themselves put it as a recommendation for the field’s future: preregistered, adequately powered trials with standardized acoustic parameters and dose-response modeling.

Dose-response evidence in a randomized trial

That modeling has an example. A randomized trial published in PLOS Mental Health on January 21, 2026 recruited 144 participants with moderate trait anxiety and ongoing anxiolytic medication, mean age 37.6, selected from 1,310 pre-screened people.

The design compared four conditions. Three consisted of instrumental music with nature sounds combined with theta-band auditory beat stimulation, lasting 12, 24 and 36 minutes. The fourth served as a control with 24 minutes of pink noise. State anxiety was measured with the STICSA scale, which separates cognitive from somatic symptoms, and affect with the PANAS scale.

Music with auditory stimulation outperformed pink noise on the anxiety measures, with medium effect sizes. On negative affect the dose relationship appeared: the 36-minute session produced a significantly greater reduction than the 12-minute one, with a large effect size, and the authors identify it as the most effective of the three doses.

On the anxiety measures, the 24-minute session obtained the greatest numerical reduction, and the differences between the three durations did not reach statistical significance. The study’s press release presents those 24 minutes as a balance point, on an argument of practical convenience that differs from what the paper concludes, which declares the 36-minute session the most effective dose on negative affect.

The discrepancy between the two readings carries a design consequence: the optimal dose depends on the outcome variable being measured. One duration can be the most effective for negative affect and indistinguishable from the others on anxiety. If more exposure does not automatically produce more effect, a curve exists, and a curve defines parameters to estimate, a model another team can test with its own sample, and a prediction that can fail.

Declarable parameters of a sound intervention

Duration is the most visible parameter of a larger set. The critical review of the evidence on binaural beats that yeshcube published documents the methodological limitations this literature carries, and nearly all of them come down to parameters left undeclared.

• The stimulus. Frequency band, delivery format, presence and type of music, criteria for curating the content.
• The schedule. Session length, number of sessions and spacing between them.
• The moment. Time of day and position relative to the activity being supported.
• The starting state. Trait anxiety, concurrent treatment, prior experience with the intervention.
• The measure. Validated instrument, assessment points and separation of cognitive from somatic symptoms.
• The comparator. Pink noise, silence, waiting, or an established alternative intervention.

All six parameters can be declared before the first data point is collected. A study that declares them lets another repeat it; one that omits them produces a result that cannot be accumulated with any other.

Validating an adaptive intervention

An adaptive sound intervention stops fixing dose and stimulus and adjusts them to context signals. That approach responds to the data above: if the optimal dose varies with the outcome pursued, it is reasonable to expect it also varies between people and between moments.

The approach creates a methodological problem at the same time. When each participant receives a different sequence, the intervention stops being a fixed object, and what has to be validated becomes the adaptation rule rather than a particular playlist. A classic two-arm trial is insufficient for that object, and the design has to be sought in other families, such as n-of-1 series or micro-randomized trials.

Limitations of the available evidence

The systematic review corresponds to a population aged nineteen to twenty-four, with small to moderate effects and a heterogeneity of exposures that prevented meta-analysis. Its conclusions describe that age group and that literature.

The dose-response evidence comes from a single randomized trial, in a population with moderate trait anxiety and ongoing anxiolytic medication. Transferring its results to other populations and other clinical states requires specific confirmation.

On the trial’s anxiety measures, the differences between the three durations did not reach statistical significance: the superiority of one specific duration is established only for negative affect. The reading of the 24 minutes as a balance point comes from the press release and rests on an argument of practical convenience, without support in the paper’s conclusions.

Conclusions

Dose constitutes a declarable design variable in a sound intervention, together with the stimulus, the schedule, the moment, the starting state, the measure and the comparator. The available evidence indicates that music with theta-band auditory stimulation outperformed pink noise on anxiety with medium effect sizes in the population studied, and that the 36-minute session reduced negative affect significantly more than the 12-minute one. The optimal dose depends on the outcome variable being measured. The question of whether sound works has no answer while each study tests a different exposure; the question with an empirical way out is how much, of what, for whom and measured with which instrument, and its answer takes the form of a replicable protocol. Evaluating an adaptive intervention additionally requires designs that validate the adaptation rule, such as n-of-1 series or micro-randomized trials.

Open research lines: Atlas Project

The Atlas Project, Audio Therapeutic Library & Adaptive System, is yeshcube’s applied research program devoted to organizing, evaluating and adapting sound content for wellbeing and emotional regulation experiences. It combines a structured library with an adaptation system based on artificial intelligence, and sits in a grounding and methodological design phase, with evidence review, taxonomy and the preparation of validation protocols as its initial work.

That order of work follows from the evidence reviewed: validating an adaptation rule is slower than publishing a library of sounds, and for that reason protocols are prepared before products. The time-bounded sessions VVAVVE works with answer to that criterion of defined exposure.

Collaborating on the validation of sound interventions

yeshcube develops this line within Allies, its scientific collaboration system, with four partner types and three principles: value for value, traceability and independence. No partner can veto a publication.

Modeling a dose-response relationship takes samples and repetitions no single team assembles alone. The work is of interest to psychoacoustics and clinical psychology groups, to sleep laboratories, and to teams experienced in n-of-1 and micro-randomized designs, which are the ones that allow an intervention changing with the person to be evaluated.

Discover our Allies program →

Fill in our contact form →

References

• “Music and binaural beat interventions for young adults: a systematic review of effects on anxiety, sleep and cognition”. Acta Neuropsychiatrica, 2026. Systematic review of ten clinical trials, around 970 participants aged nineteen to twenty-four.
• Randomized trial of three listening durations against a pink noise control. PLOS Mental Health, January 21, 2026. 144 participants with moderate trait anxiety and anxiolytic medication, four arms, STICSA and PANAS scales.