No. 024 October 11, 2026 8 min read

Ruled Out in 2012, Required in 2026

Iron-rich cells in a pigeon's beak were demoted to immune-system debris fourteen years ago, and the search for the real magnetoreceptor moved on. A paper in Science puts the same cell type back at the centre of the compass — in the liver, and only when the sun is hidden.

The 2012 result I keep coming back to is a negative one. David Keays’s lab in Vienna went looking for the pigeon’s magnetic compass in the place the field had agreed it should be: six loci in the upper beak, where earlier work had described magnetite-bearing trigeminal nerve endings. They built a three-dimensional MRI and CT map of the beak and they found the iron. The iron was in the wrong kind of cell. It sat in macrophages — immune cells packed with ferritin-like granules, siderosomes and haemosiderin, carrying MHC class II on their surface — and their number and position varied from bird to bird in a way no sense organ should.

“We had hoped to find nerves with magnetic crystals,” Keays said then, “but unexpectedly we found thousands of macrophages, each filled with tiny balls of iron.” The paper’s closing line was that the work “necessitates a renewed search for the true magnetite-dependent magnetoreceptor in birds.”

For fourteen years, it’s just macrophages has been how you dismiss an iron-rich magnetoreceptor claim. Then in May, Science published a paper arguing that macrophages are the magnetoreceptor — not in the beak, in the liver.

pigeons in the flight test

half depleted about 24 h before

19 km

release distance

a single route, west to east

~70 min

the controls' homing time

none of the depleted birds made it that day

field used to image the cells

reported only as far stronger than Earth's

What they did

Clivia Lisowski, Martin Wikelski, Christian Kurts and sixteen colleagues surveyed pigeon tissue for magnetic response — beak, eye, brain, spleen, liver — and the liver won, by a lot. Liver macrophages were the cells that stuck to a magnetic column; Ulf Wiedwald, who did the magnetometry, put it as finding “by far the strongest magnetic response in liver tissue.” Electron microscopy placed those iron-loaded cells against nerve fibres, which is where the proposed mechanism comes from: the macrophages sense, and hand something to the nervous system, possibly by way of the vagus nerve.

Then the behavioural half. Thirty-four homing pigeons trained on a 19-kilometre west-to-east route. Half of them received a macrophage-depleting treatment — clodronate, per Chemistry World’s account — roughly a day before release. All the birds were driven out and let go under heavy overcast, carrying GPS loggers. The controls were home in about seventy minutes. The depleted birds scattered in no particular direction and none of them got home that day; they came back the next morning, after the sun was out. Depleted birds released on sunny days flew straight home.

That last point is the design’s load-bearing beam, and Kurts said as much: “It’s very important that the birds don’t have a clue where the sun is.” If the treatment had simply made the birds ill, they should have failed in sunshine too. They didn’t. So the claim is narrow and, on its face, well aimed — these cells are not the navigation system, they are the backup you need when the sky is closed.

Why the cell type is both plausible and suspicious

Macrophages in the liver and spleen have a day job: dismantling worn-out red blood cells. Haemoglobin iron comes out of that process and gets parked in ferritin, whose shell holds up to about 4,500 iron ions, and some of it ends up as iron-oxide nanoparticles. That makes the cells magnetic as a side effect of being good at their actual work.

We know this, in fact, from Kurts’s own lab. In 2015, Franken and colleagues — Wiedwald and Kurts among them — reported that splenic red pulp macrophages are intrinsically superparamagnetic, with a room-temperature response much like commercial magnetic nanoparticles and, at 10 K, hysteresis with a coercive field of 780 oersted. Kupffer cells, the liver’s resident macrophages, behaved the same way.

The reason that paper existed is the part I find delicious. These cells were a problem. Being magnetic on their own, they stick to magnetic separation columns without being labelled, and so they contaminate the cell isolates immunologists live on: up to about a third of a CD11c+ isolate, five to ten per cent of B and T cell preparations, quietly spiking the cytokine readouts and the RT-PCR with macrophage signal that gets attributed to dendritic cells. The 2015 paper’s contribution was a protocol for getting rid of them. The same magnetism that was a laboratory nuisance in 2015 is the sensory mechanism in 2026.

That is a good story and it is also exactly the shape of a story that should make you careful. Every cell that recycles blood is full of iron. That is a reason to expect magnetism in the liver, and not, by itself, a reason to expect a sense of direction there.

Where it is thinner than the headlines

The headlines said pigeons navigate with their livers. The paper says something much more conditional, and the people who have spent careers on this are not persuaded.

Joe Kirschvink, who has argued the magnetite case at Caltech for decades and has no stake in this paper, told Scientific American flatly, “I am not convinced,” adding that he was surprised it cleared review and that superparamagnetic receptor models have hit a dead end after many failed attempts. Carl Meyer, in Hawaii, put the physics objection directly: there is no evidence that superparamagnetic iron in a macrophage can register a field as weak as Earth’s and pass that to a neuron — and he is wary “not least because of all the previous declarations of victory.” The field strength the team used to image the cells was, by the same report, far stronger than the geomagnetic field; the actual number does not appear anywhere in the public coverage.

This is not a vague worry. There is a quantitative literature on exactly this gap — Winklhofer and Kirschvink’s 2010 assessment of torque-transducer models for magnetoreception is the obvious reference — because a superparamagnetic particle has no permanent moment to twist on, which is why single-domain magnetite chains have been the preferred candidate for so long. A model that routes the signal through clustered superparamagnetic particles has to say how the cell beats thermal noise. Lisowski is straightforward that they can’t yet: “We have our theory, but right now it would just be speculation.”

Pascal Malkemper raised the confound that bothers me most. Clodronate does not politely remove one hepatic cell population and leave the bird otherwise unchanged; it kills phagocytes wherever they take it up, and in the mouse literature a single dose clears better than 90 per cent of splenic macrophages by 48 hours, with repopulation taking one to two weeks. A bird that has just lost its phagocytes is a bird whose immune system is down. Malkemper’s version is that the drug might simply have agitated the animals, and that the evidence is largely correlational.

The sunny-day flights are the reply to that, and they are a real reply — the next-morning return happened well inside the depletion window, so those birds were still depleted when they flew home competently. What I cannot check from outside the paywall is the thing that decides how strong the reply is: whether depletion was verified in each bird at the time of each flight, and how many days separated the cloudy release from the sunny one. Overcast and sunny are different days, not different treatments, so wind, time of release and motivation ride along with the cue manipulation. Kirschvink’s objection is precisely that sunny-day behaviour can differ for motivational reasons.

Knocking out a cell type shows that the bird needs it. It does not show that the bird reads it.

That asymmetry is the whole thing. A depletion experiment establishes necessity. Magnetoreception is a claim about transduction — that a field direction becomes a neural signal — and no amount of necessity gets you there. Susanne Åkesson’s questions in Science News are the right ones: how would the information reach the nervous system, which brain regions receive it, and do the songbirds, bats and sharks this is supposed to generalise to even have such cells? John Phillips, also outside the work, split the difference nicely — “there are certainly going to be nonbelievers,” but the work is good enough that they can’t ignore it.

Why I care about this one

Magnetoreception is the best natural experiment I know in how a field behaves when its central object keeps dissolving. The beak magnetite went away. The iron-rich trigeminal neurons went away. Every few years a receptor is announced and then turns out to be a cell doing something else, and the reason that keeps happening is structural: iron is common, the signal is weak, and the behaviour is noisy because animals treat the magnetic sense as a backup and only consult it when everything better has failed. The condition in which the sense is detectable is the condition in which the behaviour is least reliable.

So I notice that the most defensible thing in this paper may not be the compass at all. It is the relocation. The search has been in the head for forty years — beak, eye, inner ear — and this says look at the liver, where the iron actually is, and gives a specific, falsifiable target: deplete those cells, verify it, fly the birds in a cue-controlled field rather than under whatever the weather is doing, and show the vagal signal. All of that is doable, and none of it is done.

Wikelski says, “What we think we found here really fits all the evidence that’s out there.” The trouble with a result that fits all the evidence is that a result which fit none of it would be easier to test.