Hair colour is almost entirely a lightness

Take the whole human range — Scandinavian blonde to the black common across Asia, Africa and southern Europe, and every brown and red between — and it occupies a startlingly small piece of colour space. Morel and Christie, reviewing the chemistry of permanent hair dyeing for Chemical Reviews, put numbers on it: natural hair colours have dominant absorption wavelengths between 586 and 606 nanometres, while lightness runs from 1.8 to 90 on the L* scale. Twenty nanometres of hue. Nearly the whole scale of value. Hair colour is a lightness with a narrow warm bias, and every argument about tone happens inside that band.

The proportions behind it are less intuitive than the names. Borges and colleagues degraded the two melanins to their characteristic markers and measured the yields: black hair came out at roughly 99% eumelanin to 1% pheomelanin, red at 67% to 33%. Brown and blonde came out the same as each other in that assay — 95% and 5%. Blonde is not a different mixture from brown. It is less of the same mixture. And even the ratio does not settle the colour: Morel and Christie note that granule size, granule shape, how the granules are distributed and how they are packed all move it too, and say outright that the interrelationships remain incompletely understood.

So what is a colour service doing? Permanent colour is not painted onto hair. The bottle holds small molecules that are not the colour yet — a primary intermediate such as p-phenylenediamine or p-aminophenol, and a coupler such as resorcinol or an m-phenylenediamine, sorted in the trade by whether they push the result yellow-green, red or blue. They are small enough to get into the fibre; a size limit somewhere around 9.5 ångströms has been proposed for getting in at all. Mixed with hydrogen peroxide and ammonia at about pH 9.5 and left on for twenty to forty minutes, they oxidise and couple to one another inside the cortex into molecules considerably larger than the ones that went in. The colour is manufactured in place, and then it is too big to leave.

The peroxide is doing two jobs at once, and the entire product ladder falls out of that. It oxidises the dye precursors, and together with the alkali it lightens the pigment already there. Swap ammonia for monoethanolamine and it lightens less; use sodium carbonate or aminomethylpropanol and colour still develops while the lightening stops more or less altogether. Permanent, demi-permanent, and the no-lift systems the trade calls deposit-only are not three kinds of dye. They are one reaction with the second job turned down.

Lightening is not removal. It is destruction, and it has been watched. Kojima and colleagues cut bleached and virgin black hair in cross-section and imaged them by nanoscale secondary ion mass spectrometry: after bleaching, the oxygen content of the melanin granules was higher. The pigment is oxidised where it stands. The same chemistry does not stop at the pigment — Morel and Christie record that oxidative bleaching significantly raises the number of anionic groups in the fibre through cysteic acid formation, which is a mild way of saying the sulphur bonds holding keratin together are being converted into something else. Inoue and colleagues mapped where, by X-ray absorption spectromicroscopy: in bleached hair the cysteic acid was concentrated in the cuticle, while in permed hair it was spread evenly through the section. Two treatments, two different patterns of damage.

Brassiness has a cause and it is not carelessness. Wolfram and Albrecht compared how brown and red hair behave under hydrogen peroxide and under sunlight and found pheomelanin more resistant than eumelanin to both. Lighten a dark head and the brown-black component goes first; what stands longest is the warm one. The orange stage is not a failure of the process, it is the process, and a toner is a correction for a difference in rate.

"Permanent" is a claim about the dye, not the appointment. Hair grows about a third of a millimetre a day, so a permanent colour needs redoing every four to six weeks for new growth — a qualification the same review makes of its own word.

What has actually been measured about hair and how old you look is narrower than the salon conversation, and blunter. Gunn and colleagues photographed 102 pairs of female Danish twins aged 59 to 81 and 162 British women aged 45 to 75 and asked what predicted looking old for one's age. Three things came out independently: skin wrinkling, hair greying, and lip height. The heritability analysis put greying and hairline recession mostly down to genes and hair thinning mostly down to environment. Colouring grey out is the one intervention in this piece aimed squarely at a variable with evidence behind it — and the study was written by Unilever researchers, who say so in the paper.

What has not been measured is the rule the consultation runs on. Whether your skin decides which hair tone suits you — ash if you are cool, golden if you are warm — has no test behind it that the indexes can find. Search hair colour against skin colour with any word for matching, and what comes back is melanoma risk-factor studies, because that is the only context in which the two have been recorded together. That is not a contested finding. It is an absent one, and it belongs in the same drawer as the wrist test: taught everywhere, checked nowhere.

Which leaves a service that is mostly a change of level, argued over inside twenty nanometres, priced on the argument, and paid for in cystine. Worth doing: covering grey is the one appearance decision here with a measured target under it. Worth knowing that the part carrying the strongest opinions is the part with nothing measured under it at all.