What Dogs Know: Cancer at One Part Per Trillion, a Bridge in Scotland, and a Terrier Called Jaytee

This is the second of three connected pieces. The first, on what happens when science stops asking questions, sets up the argument. The third asks you to judge a piece of footage for yourself.

It started with a mole

The published scientific literature on dogs detecting cancer begins with a woman in the United Kingdom whose dog would not leave a mole on her leg alone. It sniffed at it persistently, and eventually tried to bite it off through her trousers. She had it looked at.

It was a malignant melanoma.

Nobody trained that dog. Nobody rewarded it. It simply knew something she did not know about her own body, and it would not let the matter drop. That single case is why the research field exists at all. As one investigator put it: if a dog will offer the behaviour spontaneously, then a dog can be trained to reproduce it.

Claire Guest and Daisy

The best-known version of that story belongs to Dr Claire Guest, a behavioural psychologist in Buckinghamshire.

One day her fox-red Labrador, Daisy, refused to jump out of the car with the other dogs. Instead she pressed her nose insistently into Guest's chest. That night, Guest checked. She found a lump. The lump turned out to be a harmless cyst — but while investigating it, doctors found something else nearby: early-stage breast cancer.

Guest went on to co-found the charity Medical Detection Dogs and to run the first trials of canine prostate-cancer detection approved by Britain's National Health Service. She was awarded an OBE in 2023.

Her own advice is worth repeating, because it is unglamorous and practical: we should not worry about dogs sniffing us, because sniffing is what dogs do. But if a dog pays particular and persistent interest to one part of your body, see a doctor.

What is actually documented

A dog's olfactory detection threshold has been measured at somewhere between forty parts per billion and 1.5 parts per trillion. To picture the lower end of that: take one tenth of a single drop of liquid and disperse it evenly through two Olympic-sized swimming pools. Some dogs can smell it.

Now the caveat, given properly rather than buried. A 2017 systematic review of seventeen prospective studies found significant methodological shortcomings in at least thirteen of them. Sample sizes are small. Dogs vary. Performance drops on a hot, humid day, because a dog is a living animal and not an instrument. This is a promising field, not a settled one.

But the direction of travel is not in doubt. Researchers are now trying to build machines that can do what the dog already does — and finding it extremely hard.

Why dogs, and not other animals

We say "man's best friend" so automatically that we have stopped hearing it. There turns out to be a measurable, physiological reason for the phrase.

In 2015, a team led by Miho Nagasawa published a paper in Science showing that when a dog and its owner look into each other's eyes, oxytocin rises in both of them — and the rise in one drives the rise in the other, in a feedback loop. The longer the mutual gaze, the larger the effect.

Oxytocin is the hormone that binds a mother to her newborn. Dogs and humans are running that machinery on each other.

The researchers ran the identical experiment with hand-reared wolves. The loop did not appear. Wolves rarely hold human eye contact at all.

(In fairness: other researchers have questioned whether captive wolves are a fair comparison group. The critique is reasonable, and the work continues.) Whatever else is going on with dogs, we are not imagining the bond. It has a chemistry.

Jaytee

Here is where it becomes contested, and where I am going to be careful with you.

In 1989, a woman named Pam Smart in Ramsbottom, near Manchester, got a mongrel terrier puppy called Jaytee. Her parents, who looked after him while she worked, began to notice that he would go and sit in the porch window before she came home — not at a fixed hour, but seemingly when she set off.

Rupert Sheldrake heard about it and ran more than a hundred videotaped experiments. Pam travelled at least seven kilometres away. The porch was filmed continuously with a running timecode. The tapes were scored blind.

In the trials where her return time was chosen at random, and she sometimes returned in unfamiliar vehicles such as taxis, Jaytee was at the window four per cent of the time during the main period of her absence — and fifty-five per cent of the time while she was on her way home.

Three honest things about this, which most write-ups leave out.

One: it was not every time. Television coverage claimed Jaytee did it without fail. He did not. Sheldrake's own published figure is eighty-six per cent. In one randomised trial out of twelve, the dog never went to the window at all. He reports this himself — and then spends years correcting the exaggeration that television put in his mouth.

Two: the famous debunking is weaker than its headlines. The psychologist Richard Wiseman ran four experiments and published in the British Journal of Psychology that he had found nothing. The press release went worldwide: psychic pets exposed as a myth. But Wiseman scored the dog against a criterion drawn from remarks by TV presenters, not from the behaviour Sheldrake had already documented across a hundred trials — and when Wiseman's own raw data is plotted the way Sheldrake plots his, the same pattern is there. Wiseman maintains his method was sound. He also wrote to both Sheldrake and Pam Smart to say he was appalled at how the newspapers had treated her.

Three: it has never been independently replicated. This is the real problem, and no amount of enthusiasm gets around it. Essentially all the positive data comes from one dog, one owner, one pair of investigators. That is not an accusation of dishonesty — it is the structural weakness of any single-lab result, in parapsychology or in cancer research. What turns an anomaly into knowledge is somebody with no stake in the outcome getting the same result with a different dog in a different town. Thirty years on, that has not happened.

Bobbie

In August 1923, a family called Brazier drove from Silverton, Oregon to Wolcott, Indiana, with their two-year-old Scotch collie mix. At a service station in Wolcott, local dogs chased him off. The family searched, left instructions, and eventually went home without him.

On 15 February 1924 — six months to the day — Bobbie scratched at their door in Silverton. He was emaciated, his eyes red with strain, his feet worn to the bone. He slept for three days.

He had crossed roughly 2,551 miles of plains, desert and mountain, over the Continental Divide, in winter.

This is not folklore. The Oregon Humane Society investigated the family's claim and confirmed it, reconstructing his route from letters written by people who had fed and sheltered him along the way. They could identify him by three scars. Parts of the reconstructed path retrace the family's outbound journey — he was coming home by way of places they had stopped.

Scent and memory, over 2,500 miles and six months? Perhaps. But notice that this is itself an extraordinary claim about what a dog's nose and mind can do.

The bridge

Overtoun Bridge, near Dumbarton in Scotland, was completed in June 1895 — three arches of rough-faced ashlar over a ravine called Spardie Linn, about fifty feet below. Since the 1950s, at least three hundred dogs have been recorded going over the parapet. At least fifty have died.

The pattern is extraordinarily specific. Nearly always the same few feet of stone, on the right-hand side walking away from the house. Nearly always on clear, dry days. Nearly always long-nosed breeds — collies, labradors, retrievers, spaniels. Precisely the dogs built to follow a thread of scent.

The best available explanation is physical, and it depends on a detail the viral videos never mention: a dog cannot see over that parapet. It is solid stone, waist-high on an adult. A dog crossing has no idea there is a drop. The coping stone on top is tapered, so it is not a flat ledge either. The long-term residents of Overtoun House, who have watched several of these incidents, describe it plainly — the dog catches a scent, jumps up on the wall, and because the top slopes away, it topples.

The scent is the contested part. An RSPB investigation found nests of mice, squirrels and mink on exactly the favoured side, and in a small test, seven of ten dogs offered mouse, squirrel and mink scent went straight for the mink — several of them dramatically. But a local hunter of fifty years insists there are no mink there at all. The SSPCA investigated and came away inconclusive.

Here is the piece of reasoning I would ask you to sit with, because it cuts against the direction most readers want to go:

Hachikō, and what I actually think

Every day, an Akita named Hachikō met his owner at Shibuya Station in Tokyo. In 1925 the owner died at work and never came back. The dog returned to that station, at the same hour, every day, for nine years.

Nothing about that is mysterious. And that, I think, is the point.

If you had told a physician in 1900 that a family pet could smell a tumour through a person's breath, you would have been laughed at — and you would have been right. We now know it happens; we still cannot fully name the molecule. If you had said that looking a dog in the eye changes the hormones in both your bloodstreams, you would have been mocked — and you would have been right.

So what I take from all of this is not that we should believe everything. It is that the boundary between impossible and documented has moved several times within living memory, in this exact area — and it moved because somebody took a dog owner's ridiculous story seriously enough to test it.

That is the argument of the first piece in this series. Not that the strange things are true. That the questions are worth asking properly, by people equipped to answer them, instead of being ruled out in advance. And if you would like to practise deciding for yourself, the third piece hands you a much harder piece of footage and asks you to judge it.

Ross Cline · rosscline.com

Sources: Nagasawa et al., "Oxytocin-gaze positive loop and the coevolution of human–dog bonds," Science, 17 April 2015. McCulloch et al. (2006), canine detection of lung and breast cancer from breath. Systematic review of olfactory detection of cancer by trained sniffer dogs (2017). Sheldrake & Smart, "A dog that seems to know when his owner is coming home: videotaped experiments and observations," and Sheldrake's commentaries. Wiseman, Smith & Milton, British Journal of Psychology (1998), and the subsequent exchange. Medical Detection Dogs; NPR, "For Cancer-Detecting Canines, The Nose Knows" (2015). Oregon Encyclopedia and Oregon Historical Society on Bobbie. New York Times, "Dog Suicide Bridge" (27 March 2019). Images via Wikimedia Commons, credited individually above.