← All posts

Thoughts / The craft

Music therapy in the incubator: does it really work for preterm babies?

August 7, 2026 · Thoughts / The craft

A neonatal intensive care unit is not a quiet place: alarms, pumps, ventilators, staff voices handing over the shift. The first soundscape of a baby born at twenty-eight weeks is made of machines. Then his mother arrives, sits down next to the incubator, looks at that nine-hundred-gram body threaded with lines, and doesn't know what to do with her own voice: they've told her that talking to him is good for him, but it feels ridiculous, or maybe just too intimate for a room full of people. So she stays silent.

There it is: music therapy in a neonatal intensive care unit is there, first of all, to break that silence.

Does it really work?

Yes — but for what it does, not for what we'd like it to do. And it's worth being precise, because in this field promises come cheap.

The largest study ever run is called LongSTEP: five countries, level III and IV neonatal units, parents singing to their own baby with a music therapist beside them. The question was a big one — does it improve parent-infant bonding? The answer, at six and at twelve months, was no, not significantly (Ghetti and colleagues, 2023); and a follow-up analysis of neurodevelopment at twenty-four months found nothing there either (Bieleninik and colleagues, 2024). And it's the same researchers who, in a paper published this year, explain why the evidence on those outcomes stays so elusive: a preterm ear is not a term baby's ear, the acoustics of the unit work against you, study protocols vary wildly, and the babies' medical condition forces you to change plans halfway through (Giordano and colleagues, 2026).

A study that honest is worth ten enthusiasms. And it says something we should tell each other more often: singing to a premature baby will not raise his IQ, and anyone who promises you that is selling you something.

What does get measured, and measured well, sits much closer: in the present — procedural pain, the stability of vital signs, the parents' anxiety.

The study that opened the road is from 2013: two hundred and seventy-two babies under thirty-two weeks, three live interventions — a lullaby chosen by the parents and sung live, and two instruments picked for what they imitate: a slit drum, two notes a minor third apart, and an ocean drum, which reproduces the sound of the fluids of the womb. The singing and the drum lowered heart rate, the ocean drum, matched to the baby's breathing, improved sleep, and all three reduced the stress the parents reported (Loewy and colleagues, 2013). A paper from this year describes something else: the pacifier that starts a lullaby recorded by the baby's own mother whenever he sucks, used during eye examinations for retinopathy of prematurity. In twenty-five babies, heart rate, oxygen saturation, breathing and pain scores came back towards baseline after five minutes (Crouse and colleagues, 2026).

Five minutes is not a small thing, because in that room pain adds up one needle at a time: between seven and seventeen invasive procedures a day, which for some babies go on for months (Cruz and colleagues, 2016).

So why do it at all?

Because the goal here isn't an "improved" baby: it's the present — less pain, more stability, and parents who stop being spectators. It holds true in all our settings, but here it's out in the open: you state your goals first. If I say I want to reduce stress during a procedure, I also know how to measure it. If I say I want to "create a bond", I've said something beautiful that nobody can check — and indeed, when somebody measured it properly, the result was sober.

Who does it, and with whom

Not a music therapist turning up with a guitar and improvising. An Italian experience published this year shows it well: in a level III neonatal unit, live music therapy didn't arrive as an isolated novelty, but was grafted onto a programme that was already there — positive touch, kangaroo care, babywearing — with a music therapist certified in neonatal work alongside a nursing consultant, and a parents' association funding the project (Buonomo and colleagues, 2026). Everything is already in place: the team, the specific training, the shared goals.

And in Italy this is no longer an isolated case: at the Miulli hospital in Acquaviva, mum's and dad's voices went into the incubators this summer; the neonatal unit of the Del Ponte hospital in Varese has been doing music therapy for fifteen years and is now an international reference; in Rimini a parents' association brought music therapy onto the ward through a public fundraiser, and both there and in Parma the projects are funded over the long term — five and six years, which in this field is a rarity; at the Moscati hospital in Avellino a randomised pilot study on parental humming during painful procedures grew out of a research idea from music therapist Marina Elizabeth Caputo, taken up by the head of the unit and then supported by the whole ward. And it was Varese, Rimini, Parma and Avellino who brought the first Italian round table on the subject to the world congress in Bologna. That Italy has a voice in this field is also shown by the international volume on music therapy in the NICU, where the Italian chapter is written by someone who works on that ward (Sgobbi & Shoemark, 2020).

A voice you can't delegate

In Sweden a study is about to start that will measure the brain activity of a mother and her baby at the same time, during a blood draw, while she holds him skin to skin and sings to him: for now it's only the protocol, there are no results yet (Eriksson and colleagues, 2026), but the hypothesis is already a stance — that whatever happens does not happen inside the baby, nor inside the mother, but between the two of them.

And if it lives in the between, you can't hand it over to a loudspeaker. No playlist, however well chosen, has a voice that trembles.

This isn't just a nice image, either. We've known it since 1980: a baby a few days old, given a pacifier wired to a tape recorder, sucks in the pattern that brings back his mother's voice and not another woman's (DeCasper and Fifer, 1980). Forty-five years later, MRI confirms it from the inside: when a newborn is played sung music rather than instrumental music, the right temporal areas light up more for the voice (Loukas and colleagues, 2024) — forty-five babies, thirty-five of them born preterm.

The human voice, for a brain that has only just arrived, is not one sound among many.

That's why in the incubator we don't bring music to the baby: we give the parents their own sound back. In a place where everything is done by somebody else — the machines, the doctors, the tubes keeping him alive — it's the one thing only they can do.

Sources

Related

Read also

Episode

#07 Does music therapy work?

Episode

#18 Music therapy in a team: can it be done?

Article

Does music therapy work for autism? What the research really says