In 1897, G. Stanley Hall and Arthur Allin filled the opening forty-one pages of volume 9 of the American Journal of Psychology with tickling. Their argument, and Louis Robinson’s a decade later, was that the ticklish places are the places where a blow would do real damage; the evolutionary gloss the new paper draws from it is that tickling may train us to guard them. It was a serious idea, and by that paper’s account nobody had ever tested it.
That is roughly the condition of the entire subject. Aristotle attributed ticklishness to the delicacy of human skin, which could imply it should be stronger where the skin is thinner. Darwin, in 1872, guessed at something close to the opposite: that the rarely touched areas are the ticklish ones. Two later accounts tie ticklishness to ordinary touch sensitivity, and to the erogenous zones. Those four plus Hall and Allin’s make five explanations, and, by the authors’ own account of the literature, no empirical test of any of them against a map of where people actually feel ticklish.
A paper published in Nature Human Behaviour on 10 August supplies the map. Ziliang Xiong and Konstantina Kilteni surveyed 448 people across three language groups, asked the 441 who said they had ever been tickled to draw one, and measured the resulting topography against all five theories. One held up: Darwin’s. In a combined model across eleven body areas, two of the others pointed the opposite way.
What the volunteers actually did
No one was touched. The phrase “bodily map” invites a picture of a laboratory and a feather, and there was neither.
Participants aged 18 to 30 were recruited through Prolific and Radboud University’s participant pool, 150 per language group, prescreened so that each answered in their first language. Two were dropped for inattentive performance on the colouring task, doodling in the authors’ example, leaving 149 Chinese-speaking, 150 Dutch-speaking and 149 Greek-speaking respondents. Of those, the 441 who said they had ever been tickled (145, 149 and 147 by group) were asked to draw the ticklishness map. The authors then excluded maps with fewer than five pixels drawn, along with any ticklishness map from someone who reported no tickling in their lifetime: 6.7 per cent of the 448 family-and-friends maps and 16.5 per cent of the partner ones, which leaves roughly 418 and 374 maps in the analysis. Those percentages are the paper’s, and they run against all 448 — the denominator it uses for every map type, including ones no ticklishness rule could touch. It ran in under an hour in a browser, for about £6.50 or €7.50, or course credit for some Dutch participants. The authors state plainly that the study was not preregistered.
So: everyone under 31, answering online, with the childhood-and-adulthood comparisons resting on what the paper calls retrospective reports, and three languages is three. The paper’s own list of things to do next includes more cultures, wider age ranges, and designs that do not lean on recall.
The map came from a colouring task adapted from the method behind the bodily maps of emotions that circulated widely twelve years ago. Two silhouettes, front and back, enlarged so that the armpits and the soles were actually visible, giving 91,617 colourable pixels. Participants dragged a ten-pixel brush over the places where they feel ticklish when family or friends do it, and again for an intimate partner if they had one. Colouring a spot twice deepened the colour, and deeper colour was the instruction’s stand-in for more frequent or more intense sensation.
The colouring items ran in a pseudo-randomised order, so the ticklishness maps did not necessarily come first. Six more asked where they are sensitive to even the lightest touch, where they are sensitive to even the lightest pain apart from the visceral and muscular kinds, where touch feels pleasurable, where they touch themselves during the day, where family or friends touch them, and where a partner does. Those six are what the ticklishness map was later weighed against.
The evidence is unusually inspectable. The body maps and the theory datasets are posted publicly on figshare, the analysis code in a second record; the survey responses sit in the Radboud Data Repository behind a data-use agreement the Donders Institute requires to protect participants’ sensitive demographic information.
The map came out nearly the same in three languages
Aggregated and tested pixel by pixel, the family-and-friends maps produced significant clusters on the neck, the armpits, the belly and the soles of the feet. The partner maps produced the same picture plus the groin and inner thighs, which is where partners usually touch and family and friends do not. Two of those regions, the armpits and the soles, are among the three the authors went on to mask and test as the “hotspots”; the third mask covers the sides of the abdomen rather than the belly, and the neck is not masked at all, so the hotspot trio and this cluster list are two different objects.
The cross-cultural agreement is the part that is hard to argue with. Overlap between any two groups’ maps ran from 0.78 to 0.86 on a scale where 1 is complete overlap, for the family-and-friends version, and from 0.72 to 0.81 for the partner version. A classifier trained on two language groups and tested on the third, which had contributed no pixels to its training, told ticklishness maps from everything else — the two ticklishness maps merged into one composite, the two social-touch maps merged the same way, and the five resulting non-ticklishness types pooled into a single class — with an area under the curve of 0.87 for Chinese speakers, 0.91 for Dutch and 0.94 for Greek. Trained on women and tested on men it scored 0.85, and 0.86 the other way round.
The hotspots carry the signal. When the authors blanked out the pixels covering the armpits, the sides of the trunk and the soles — 8,575 of them, about 9 per cent of the colourable area — the classifier’s performance fell reliably, by two to four points of area under the curve, the 95 per cent interval the paper prints for that drop without ever saying what the interval measures. That is a modest drop but a reliable one, and the model still beat chance afterwards, which the authors read as a sign that the non-ticklish regions carry information of their own. Blanking the same number of pixels at random locations made no significant difference (q = 0.784). Repeating the random version a thousand times never produced a drop as large as the hotspot one.
The ticklishness map is also its own object. A clustering analysis put the two ticklishness maps together first, before anything else joined them; touch sensitivity came closest, as a spatial overlap, which is a different question from the theory tests below; self-touch, partner touch, pleasure and pain sat further out; and the map of where family and friends actually touch you sat furthest of all. Asked to tell all six map types apart, the classifier did best on ticklishness, at 0.90 to 0.93 across folds, the top of a 0.71-to-0.93 range that spans all six classes, ticklishness included.
Only Darwin’s answer survived a test on its own, and the best model did better with two of the others in it
To test three of the theories, the authors had to leave their own dataset. Touch-fibre density, epidermal thickness and arterial diameter came from previously published measurements in other people rather than from anything the survey asked, compared region by region across the body: 13 anatomical areas for innervation, 23 for skin thickness, 15 for arteries. The authors rejected the self-report alternatives one by one — for touch, because participants had already answered items about ticklishness, so their own tactility maps could not be treated as independent. The remaining two were tested against the participants’ own drawings, each parcelled into 35 areas: the pleasure map for the erogenous-zones account, and an inverted composite of the self-touch and social-touch maps for Darwin’s.
Tested one at a time, four came back empty. Touch-fibre density: no significant relationship. Epidermal thickness, Aristotle’s: none. Arterial diameter, Hall and Allin’s, standing in for how dangerous a wound there would be: none. Pleasure: none, and the flattest of the four, at an R² of 0.02.
Touch exposure was different. The authors built an inverse map, the places least touched by yourself, your family, your friends and your partner, and it predicted ticklishness at a clearly significant level, holding up whether self-touch or social touch was used on its own.
The head-to-head comparison, though, comes from somewhere narrower than that. Run all five theories together through a variance-decomposition model — possible only across the eleven body areas for which all five datasets exist — and the combined model explains more of the pattern than touch exposure alone. Its single significant component is dominated by rarely-touched-ness, and two variables load strongly on it in the wrong direction for their own theories: the denser a region’s touch fibres, and the more pleasurable it is to have touched, the less ticklish it tends to be. Skin thickness and arterial diameter both fell below the threshold the authors set for a meaningful contribution, so the others here means two. The improvement is real but bounded: an adjusted R-squared of 0.57 against 0.47 for touch exposure alone, and the authors read the two models’ near-identical information criteria, 30.05 against 30.29, as evidence that the gain is not simply the reward for a more complicated model.
That is the sharpest result in the paper and the easiest to overstate. The palms are densely innervated, on the separate whole-body dataset the authors imported for that test, and they land among the least ticklish regions here. The soles of the feet are among the areas participants themselves marked as most sensitive to touch, and they are one of the three hotspots. Those are two different measures, which is part of why the pair settles nothing. So this is a tendency across regions rather than a rule that holds patch by patch, and it rests on eleven of them.
Four null results are not four refutations either. With as few as 13 anatomical regions as the unit of analysis, and three of the five comparison datasets drawn from different people measured for different reasons, a real but modest relationship could fail to show up here at all. The surviving result is correlational, and its arrow is unresolved: rarely touched areas may be ticklish because contact there is unexpected, or they may be rarely touched precisely because they are ticklish. The authors raise that loop themselves and cannot close it. Their own summary is that the sensation is “at least tuned by prior sensory experience that is further informed by physiological and hedonic inputs,” which is some distance short of a solved puzzle.
Two narrower limits matter, because headlines tend to swallow them. The study is about the intense, laughter-producing kind of tickling, not the light feather-on-the-arm kind, which the authors note might have a different topography. And it never asked about the situation: intent, consent, whether the touch was wanted. The paper allows that such factors probably decide whether tickling is experienced as playful, affiliative, intrusive “or, in some cases, abusive,” and it measures none of them. Nobody on this desk examines patients; what is on offer here is one paper, read closely. Where unwanted touch is a live problem in someone’s life, a counsellor is the right first call.
Universal, and thinning out
Being tickled is close to universal in this sample. Of the 448 surveyed, 98.4 per cent had been tickled at least once, and 95.8 per cent had done the tickling. The body regions involved were the same in all three languages. Tickling runs downhill in age, happens mostly between people who are close, and is reported less often after childhood: 95.5 per cent reported being tickled as children, against 86.8 per cent as adults, on a base the paper does not state.
Which leaves one thing sitting on the table. The paper says the map is tuned by prior sensory experience, and separately that being tickled thins out once childhood ends. It does not say the second is what does the tuning, and it does not answer the obvious follow-up: is the map a body carries at thirty the same one it carried at eight? The authors ask for exactly that study, and could not run it here.
