How massage actually works

Ask why massage helps and you will usually be told something about lactic acid, toxins, circulation or fascia. All four explanations fail when they are tested directly, and each has its own page in this section.

This page is the replacement. It is the account that does survive the research, and it is a better story than the ones it retires.

The short version

• Massage works largely through the nervous system, not by changing tissue.¹
• Touch reduces how much pain signal reaches the brain. That is the gate control theory, published in 1965 and still the best supported account of why touch helps.²
• Massage triggers the body's own pain-damping response, at a strength comparable to the standard laboratory test of that response.³
• There is a real tissue effect, measured in muscle biopsies: less inflammatory signalling, more mitochondrial growth, and no effect on lactate.⁴
• The specific technique matters less than practitioners think. For mobilisation, 88% of the immediate effect on pain was not attributable to the technique itself.⁵

The gate

In 1965 Ronald Melzack and Patrick Wall proposed that the spinal cord does not simply relay pain signals to the brain. It regulates them. Large nerve fibres carrying touch and pressure information reach the same part of the spinal cord as the smaller fibres carrying pain, and when the touch fibres are active they reduce how much of the pain signal gets through.²

This is why rubbing a knee you have just banged genuinely helps, and it is not a placebo. It is also the reason a massage can reduce pain in an area where nothing about the tissue has changed.

Gate control has been refined many times since 1965, but nothing has replaced it, and it remains the mechanism that most directly explains the thing people notice on the table.

The cascade

The fuller modern model comes from Joel Bialosky and colleagues, first in 2009 and updated in 2018. Their account is that a mechanical force applied to the body initiates a cascade of neurophysiological responses across the peripheral nervous system, the spinal cord and the brain, and that it is those responses, rather than any change to the tissue, which produce the clinical result.⁶¹

Two things follow from that, and both are uncomfortable for how massage is usually sold.

The first is that the specific technique matters less than the profession assumes. A 2024 meta-analysis found that for joint mobilisation, 88% of the immediate effect on pain intensity was not attributable to the specific technique used.⁵ The context, the expectation, the therapeutic relationship and the touch itself did most of the work. There is no reason to think massage is different in kind.

The second is that this is not a lesser explanation. Contextual and neurological effects are real effects. They change how much pain a person is in. Calling them "just" anything mistakes the mechanism for the outcome.

The body's own brake

The nervous system has a built-in pain-inhibition system: pain in one part of the body reduces pain sensitivity elsewhere. Researchers test it with a cold pressor task, where a hand is held in ice water.

A 2021 study compared massage against that test directly and found massage produced a reduction in pain sensitivity of a comparable size.³ Massage appears to recruit the same descending inhibition pathway the body already uses, which is a mechanism with a name, a measurement and a plausible route from the hands on the table to the change in the person.

What actually happens in the tissue

The tissue story is not empty, it is just different from the marketing version. In 2012 a study in Science Translational Medicine took muscle biopsies from people after damaging exercise, with and without massage. Massage reduced inflammatory signalling, including TNF-alpha and IL-6, and promoted mitochondrial biogenesis.⁴

The same study measured lactate and found no effect at all. The paper most often cited as proof that massage does something biological is also the paper that removes the most popular explanation for it.

Blood flow to the skin does rise. That is the warmth and the pinkness, and it is honest to describe. Blood flow to the muscle does not.^shoemaker-1997] Those two things get conflated constantly, and the [circulation page sets out why.

Why the honest version is more useful

If massage worked by mechanically changing tissue, pressure would be the active ingredient and more of it would be better. It does not, and it is not.

If it works through the nervous system, then the things that actually matter are the ones a good therapist already does. Whether you feel safe enough to stop guarding. Whether the pressure is one you can stay relaxed under rather than brace against. Whether you come back often enough for the effect to accumulate, which the dose evidence supports.

That is a more demanding standard than "we flushed your lactic acid", and it happens to be true.

Frequently asked questions

How does massage reduce pain?

Mainly through the nervous system. Touch and pressure signals reaching the spinal cord reduce how much pain signal passes through to the brain, which is the gate control theory. Massage also appears to recruit the body's own descending pain-inhibition system, producing an effect comparable to the standard laboratory test of that response.

Does massage change anything physical in the muscle?

Yes, but not what is usually claimed. Muscle biopsies after damaging exercise showed massage reduced inflammatory signalling and encouraged mitochondrial growth. The same study found no effect on lactate at all.

Does the type of massage matter?

Less than the industry suggests. For joint mobilisation, 88% of the immediate effect on pain was not attributable to the specific technique. Context, expectation, the therapeutic relationship and touch itself account for most of it.

Does deeper pressure work better?

There is no evidence that it does. If the mechanism is neurological rather than mechanical, pressure hard enough to make you brace is working against the response you are trying to produce.

Is that just the placebo effect?

No. Gate control and descending inhibition are specific, measurable neurological mechanisms, not an absence of a mechanism. Contextual effects are also real effects: they change how much pain a person is actually in.