There is a rock in the Baltic with sheer sides and a cave near the shore.
Nothing lives on it that walks. No deer, no hare, no fox, no land mammal at all.
Seabirds nest on the cliffs and seals haul out below, and that is the whole menu.
Then somebody put two old canid bones from that cave through a mass spectrometer.
The chemistry came back reading of the sea.
So how does a bone remember a diet?
Why the sea leaves a mark that meat cannot fake
Bone is not inert. The collagen inside it is rebuilt continuously from the protein an animal eats.
Carbon and nitrogen both come in more than one weight, and the ratios differ between a forest and an ocean.
Marine food chains are longer and start from a different carbon pool, so anything living off them carries heavier values in both elements.
A wolf eating elk in a Swedish forest signs itself one way. A wolf eating seal signs itself another, and the two cannot be confused.
The reading is also slow.
Collagen turns over across years, so what comes out of the machine is an average of a life rather than a last meal.
Which makes the awkward part unavoidable. On a rock with nothing to hunt, a wolf with a marine signature was either catching seals or being handed food.
The rock and the hole in it
Stora Karlsö is roughly a mile long and half a mile wide, just under a square mile in total.
It sits about 4 miles off the west coast of Gotland, and Gotland itself is some 47 miles from the Swedish mainland.
The cave is called Stora Förvar, and it opens close to the water.
Inside it, people left a deposit more than 13 feet deep, built up from the Mesolithic through to the Iron Age.
This was a working station rather than a village: a place seal hunters and fishers came back to, season after season, for thousands of years.
Everything in that deposit had to arrive by boat.
The stone, the tools, the people and, it turns out, the canids.
What the two bones actually said
DNA identified both animals as gray wolves, with no dog ancestry anywhere in the genome.
Radiocarbon put them between roughly 5,000 and 3,000 years old, one of them Bronze Age.
Their carbon and nitrogen values sat in marine territory and lined up with the humans from the same site.
Set against the 72 ancient wolf genomes published in 2022, the better covered of the two had the lowest genetic diversity of any ancient wolf yet sequenced.
One animal carried something else: severe damage to a limb bone that it survived, an injury that should have ended a wild hunter’s ability to run down prey.
The work was published in a national academy journal in November of 2025 by teams in Sweden, England and Scotland.
The authors describe wolves living alongside people, eating what those people ate, in a place reachable only by boat.
Four things two bones cannot settle
It is two animals, and two animals are an observation rather than a population.
The cave was dug out in the eighteen nineties in flat 12 inch spits across a sloping floor, so depth in that deposit is not a reliable clock. Every date here comes from the bone itself, not from the layer it sat in.
Low genetic diversity has two readings and the data cannot separate them: a small isolated island population inbreeds, and so does a kept lineage.
A marine diet is not proof of a feeding hand either, because dead seals wash ashore and a wolf will take one.
Transport by people is the best available explanation and not a demonstrated fact, since the Baltic freezes and canids swim, as other wild canids living at the edge of human space keep showing.
What is not in doubt is that the crossing happened somehow, on a sea people had been working for millennia.
The experiment that did not take
The instinct is to read this as an early stage of dog domestication, and the genome forecloses it.
These were wolves at the end, not dogs at the beginning, and they left no descendants in any dog alive.
That makes the find more interesting rather than less.
Dogs were already common across Europe by the time these two were living on that island, so this was not a first attempt at anything.
It was people who already had dogs, keeping wolves anyway, and getting nowhere with it, as the published paper lays out.
Domestication reads in textbooks like a single event with a single outcome, and it looks more like a thing tried repeatedly in different places, as the research institute puts it.
Most of those attempts went nowhere.
These two wolf bones are one of them, kept by accident in a cave.
Read the whole thing?
Get the week's signal, not the noise
Our sharpest reporting on energy, climate and nature — free, once a week.
