The Moth That Reads the Milky Way

An Insect With No One to Learn the Route From

Here is the part of this story that made our club stop and reread the sentence twice. The Bogong moth, Agrotis infusa, breeds across the lowlands of southeastern Australia. Each spring a fresh generation of adults, only weeks old, sets off after dark toward the Australian Alps and does not stop until it reaches a specific set of limestone caves, where it packs onto the rock walls at densities the paper's companion work puts near 16,000 animals per square metre and waits out the heat of summer in a torpor called aestivation [2]. Come autumn, the same individuals wake up, fly the entire distance back, breed once, and die. Nobody teaches them the route. There is no older generation making the trip alongside them, no return flight to correct a bad first attempt, and by the time they are capable of flying at all their parents are already dead. Every leg of an up-to-1,000-kilometre round trip is a genuinely novel journey for the animal making it [1].

Compare that with the two other famous long-distance insect migrants. Monarch butterflies split their journey across several generations, so no single monarch ever completes the whole loop, and the ones that do fly the long leg use a sun compass that reads the time of day off circadian clocks in their own antennae [5]. Migratory songbirds, the other classic star-navigators, get to practise: many species make the trip repeatedly over a multi-year lifespan, and some appear to calibrate their innate compass against what they actually see in the sky as fledglings [4]. A Bogong moth gets one shot, in each direction, and has to get it right the first time using a nervous system built in a matter of weeks. Something in that nervous system has to already contain a usable map of "which way is south" before the moth has ever looked at a star with any consequence riding on the answer.

The question this paper actually answers is not "can a moth see stars." It is "can a moth's genes, with no help from experience, wire up a working compass out of them."

Building a Planetarium for an Animal the Size of a Fingernail

You cannot follow a moth for 1,000 km and watch what it looks at. So the team brought the sky to the moth instead. Each animal was chilled for five to ten minutes to hold it still, had a small patch of scales removed from its thorax with a micro-vacuum, and had a fine tungsten stalk glued to the bare cuticle with contact cement. That stalk clips, through a short length of soft rubber tubing, onto the shaft of an optical encoder, which lets the moth rotate freely around its yaw axis — choose any compass heading it likes — while the encoder logs that heading 5 times a second at a resolution of 3° [1]. The moth is flying in place, a treadmill for direction rather than distance, and the encoder writes down every turn it tries to make.

Around that moth sits a transparent Perspex cylinder 50 cm across and 35 cm tall, and onto the inside of that cylinder the team projected a real night sky — generated in the planetarium software Stellarium, moonless, dated to actual spring and autumn nights (1 October 2018 for the spring trials; 21 and 27 March 2018–2019 for autumn), and filling 160° of the moth's field of view in the behavioural rig and 100° in the version built for recording from the brain, both centred on the zenith directly overhead [1]. As far as the moth's compound eyes can tell, it is flying under the actual stars of a specific Australian night.

Turning the Magnetic Field Off, on Purpose

Two candidate compasses were on the table before this experiment ever ran. Plenty of migratory animals steer by Earth's magnetic field, and if the moths were doing that, a fake sky projected overhead would be irrelevant scenery. So the team's first job was to make the magnetic explanation impossible, not just implausible.

The three pairs of Helmholtz coils shown above don't just weaken the ambient field, they actively cancel it, driven by constant-current supplies and checked with a magnetometer before each run, holding the field below 100 nanotesla in a small volume around the tethered moth [1]. Earth's surface field runs roughly 25,000 to 65,000 nanotesla depending on latitude, a range that is common background in geomagnetism rather than a number from this paper. Do the division on that ourselves and 100 nanotesla is somewhere between a 250-fold and a 650-fold reduction, comfortably below anything a magnetoreceptor could plausibly use as a directional signal. With the sky projected and the field nulled, a moth that still picks a consistent, seasonally correct bearing has only one candidate cue left to be using.

And that is exactly what happened, and rotating the projected sky 180° is what turned "the moth ignored the field" into "the moth was reading the stars."

Six Trials That Do the Real Work

The core experiment is almost embarrassingly simple to describe once the hard engineering is out of the way. Fly a moth under the field-nulled, projected night sky and see which way it points. Then fly a fresh moth under the identical sky flipped 180°. Then fly a fresh moth under the same stars with their positions scrambled into a random arrangement, keeping brightness and density the same but destroying every constellation. Run all three versions in spring, when free-flying moths head roughly south toward the caves, and again in autumn, when they head roughly north back to the breeding grounds.

Notice what the rotation trial actually rules out. A moth that was simply attracted to the brightest patch of projected light, or repelled by some artefact of the projector itself, would not flip its heading when the sky flips, because the physical hardware around it has not moved. A moth reading a magnetic cue would not respond to the sky at all, field or no field, since there is no field left to read. The only thing left standing after both of those explanations fail is the pattern of stars itself.

The Night the Stars Moved and the Moths Didn't

The planetarium trials prove the moths can use a fixed, projected sky. A separate set of trials, run outdoors under the real night sky with the coils switched off entirely, asks a harder question: can they keep their heading as the real sky drifts across the night, the way it always does because the Earth keeps turning underneath it?

Ninety-five moths were flown outdoors early in the evening, averaging 20:49 local time, and pointed a mean 5° — essentially due north, the correct autumn bearing. The same 95 animals were flown again roughly three hours later, averaging 23:42, and still pointed 355°, a shift of only 10° that a likelihood-ratio test could not distinguish from no change at all (χ²=2.663, p=0.103) [1]. We wanted to know how much the actual star field had moved in that gap, since the paper reports the headings but not the sky's own displacement, so we did the arithmetic ourselves from a fact the paper does not need to state because every astronomer already knows it: the sky appears to rotate around the celestial pole at essentially the Earth's own rotation rate, about 15° per hour.

That gap is the whole argument for something smarter than "orient toward one bright dot." Either the moths are tracking a broad pattern, such as the shape of the Milky Way, that rotates as a rigid whole and stays recognisable even as it turns, or they are doing something closer to what the paper's own discussion raises: a time-compensated star compass, of the kind already documented for the sun in monarch butterflies, which use a circadian clock in their antennae to shift their heading as the sun moves across the sky [5]. The paper does not settle which mechanism it is. It does rule out the naive one.

Put the indoor and outdoor results together and a division of labour appears. Stars alone, with the field nulled, are sufficient. Magnetism alone, with the sky hidden by cloud, is also sufficient. Scramble the stars and keep the field on, and the moths still orient by magnetism, largely unfazed. Take away both cues at once and, the paper reports, the moths are lost. Two independent compasses, either one enough on its own — which is a very different, and more robust, design than "one primary cue with the other as a vague assist."

A Compass Wired Into a Brain the Size of a Poppy Seed

Behaviour tells you a moth can do this. It says nothing about how. To find out, the team repeated the projected-sky experiment while recording electrical activity directly from neurons inside the moth's own brain, using the same slowly rotating sky as the visual stimulus. They sampled from 79 neurons in total. Of those, 28 — about 35% of the sample, an arithmetic we did ourselves on the paper's own raw counts — passed the criteria for reliable, repeatable tuning to the rotating sky and made it into the final dataset [1].

Most of those 28 cells, 18 of them, fired in a single broad peak: quiet most of the time, and most active when the moth's simulated heading pointed in one particular compass direction. A further 6 showed the mirror-image pattern, going quiet at one specific heading rather than firing there. A handful of others were sensitive to which way the sky was turning, and three, recorded in a brain region that generates steering commands, fired in two opposite peaks 180° apart rather than one [1]. Eleven of the 28 cells, roughly two in five, were filled with a tracer dye and anatomically traced back to their exact location, letting the team say with confidence which part of the moth's tiny brain was doing which job.

Why the Preferred Direction Is Always South

Here is the detail that we found genuinely strange, in a good way. Pool the tuned cells together and ask which direction each one prefers, and the answer clusters tightly around one heading: 171° ± 35°, which is to say, almost due south [1]. That number does not shift between seasons (a Mardia–Watson–Wheeler test found no statistical difference between the spring and autumn values, p>0.2), even though a real moth flying south in spring is heading toward the caves and a real moth flying south in autumn would be heading the wrong way entirely.

So this is not, on its own, a map of "which way to fly this season." It looks more like a fixed reference axis that the rest of the brain reads and reinterprets depending on the time of year, the way a compass needle always points to magnetic north regardless of which direction you actually intend to walk. A close analogy shows up in an entirely different animal: head-direction cells recorded from the brains of streaked shearwater chicks fire maximally when the chick's head points toward magnetic north, even though the very same chick will eventually fly south on its first migration [6]. A fixed internal "north," or in the moth's case "south," seems to be a cheap and general way to build a directional reference, leaving the actual seasonal decision (turn toward it, or turn away) to a separate piece of circuitry.

That last result is the one we keep coming back to. A point of light is a much simpler thing to encode than an extended, faint band stretching across a third of the sky. If the moth's compass worked by fixating on the single brightest star or star cluster it could find, the dot stimulus should have reproduced the natural tuning perfectly and the bar should have done worse. It went the other way. The bar — the shape, not the highlight — is the better model of what these cells are actually doing.

An Old Circuit, Doing a New Job

None of the three brain regions in the pipeline above were invented for this task. The central complex, sitting in the middle of that pipeline, shows up across essentially every insect order that has been examined for it, and a 2019 review co-authored by two of this paper's own scientists lays out its established repertoire: path integration, a polarised-light compass in desert locusts, and the transport-orientation system that lets a dung beetle roll a ball of dung away from a crowded pile in a dead-straight line [3]. Dung beetles, in fact, are the previous headline case of an insect reading the Milky Way — but only to hold a straight heading for a few metres away from competitors, not to pick out a specific compass direction toward a destination a thousand kilometres away [7]. The Bogong moth result does not introduce new hardware into the insect brain. It shows old, well-mapped hardware doing something nobody had caught it doing before: reading an extended star pattern and returning a genuine geographic bearing, one that can be held across an entire night despite the sky itself turning beneath it.

Why This Matters More Coming From a Moth Than From a Bird

Stellar compasses in animals are not new news on their own. Stephen Emlen's planetarium experiments on indigo buntings, run inside an actual planetarium projector rather than a 50 cm Perspex cylinder, established fifty years ago that a migratory songbird can derive a compass direction from the pattern of stars around the celestial pole [4]. The 2025 paper's own framing makes the comparison explicit: as far as its authors know, apart from humans and a handful of night-migrating songbirds, no other animal had been shown to use the stars to navigate toward a specific geographic direction [1]. A moth joining that list is not a modest addition. It is a different category of nervous system entirely.

A songbird's star compass sits inside a brain built along the same basic vertebrate plan as ours: a forebrain with dedicated regions for spatial memory and navigation, built from many millions of neurons, refined over a multi-year lifespan in which the bird gets to fly the route, or watch the sky rotate, more than once. The Bogong moth's entire tuned population in this study is 28 cells. Its whole compass has to be assembled, correctly, by a genetic program alone, inside a brain built in the handful of weeks between metamorphosis and the first flight that will decide whether the animal lives through summer at all — and it has to work the very first time it is used, in each direction, because there is no second migration to learn from. If a circuit this small, this fast to build, and this thoroughly disposable can extract a genuine compass bearing from a pattern as diffuse as the Milky Way, then the raw computation involved may be a great deal less demanding, in neural terms, than its scarcity in the animal kingdom had suggested. The interesting question stops being "how did a moth manage this" and becomes "why haven't we found it in more insects yet."

A brain that has to work perfectly on the first try, with no teacher and no do-over, is a strange place to discover one of the more sophisticated pieces of astronomy in the animal kingdom.

What the Paper Does Not Claim

The authors are careful about the limits of their own result, and it is worth repeating those limits rather than rounding them off. They write plainly that it remains unclear exactly how many of the sky's brightest stars a moth's small compound eyes can actually resolve as distinct points [1], which means the "extended shape" interpretation of the tuning curves in Figure 4 is an inference from the neural data, not a direct measurement of what the moth's eye optically registers. Whether the mechanism is genuinely time-compensated, tracking the sky's motion the way a monarch's antennal clock tracks the sun, or relies on some rotation-invariant feature of the Milky Way's shape that never needs correcting in the first place, is explicitly left open rather than resolved. And every number in this note describing "the moth's compass" comes from animals tethered in a lab, flying in place under a projected sky; nobody has yet recorded from a free-flying moth's brain 1,000 km from home to confirm the same 28 cells are doing the same job outdoors.

What We Would Still Want to See

Three follow-ups strike us as the obvious next moves, and none of them require new technology, only more of the same. First, resolve the compound eye's actual resolving power for faint, extended sky features, to settle whether "reading the shape of the Milky Way" is optically plausible or whether the neurons are doing more computational heavy lifting than the raw image supports. Second, run the artificial-stimulus test from Figure 4 with a continuous range of stimulus shapes between a point and a full arc, rather than just the two endpoints tested here, to map out exactly how much extent a stimulus needs before the tuning locks on properly. Third, and hardest: find a way to record from, or at least track, a moth's heading decisions partway through an actual free flight, to check that the laboratory's 171°-south reference axis is the same one steering the animal across real Australian terrain.

What We Would Tell Somebody Who Read Only the Headline

Four things. The moth is not "using the stars" the way a sailor with a sextant does; nobody has shown it can identify individual constellations by name, only that it responds to the rotating pattern as a whole and to an elongated shape resembling the Milky Way in particular. The magnetic sense has not been demoted: it is a fully functional backup compass in this species, not a vestige, and the two systems appear to operate independently of each other. The 171° "south" tuning found in the brain is a fixed reference axis, not a seasonal instruction, and something else downstream decides whether to fly toward it or away from it depending on the time of year. And the headline claim, "first invertebrate star compass," is not our club's summary or a journalist's shorthand; it is closer to the authors' own words, and every DOI and figure in this note traces back to a paper that, as far as we can verify, says exactly that.

References

1. 1.

   Dreyer, D., Adden, A., Chen, H., Frost, B., Mouritsen, H., Xu, J., Green, K., Whitehouse, M., Chahl, J., Wallace, J., Hu, G., Foster, J., Heinze, S. & Warrant, E. (2025). Bogong moths use a stellar compass for long-distance navigation at night. Nature 643, 994–1000. doi:10.1038/s41586-025-09135-3

2. 2.

   Warrant, E., Frost, B., Green, K., Mouritsen, H., Dreyer, D., Adden, A., Brauburger, K. & Heinze, S. (2016). The Australian Bogong moth Agrotis infusa: a long-distance nocturnal navigator. Frontiers in Behavioral Neuroscience 10, 77. doi:10.3389/fnbeh.2016.00077

3. 3.

   Honkanen, A., Adden, A., da Silva Freitas, J. & Heinze, S. (2019). The insect central complex and the neural basis of navigational strategies. Journal of Experimental Biology 222(Suppl_1), jeb188854. doi:10.1242/jeb.188854

4. 4.

   Emlen, S. T. (1975). The stellar-orientation system of a migratory bird. Scientific American 233(2), 102–111. doi:10.1038/scientificamerican0875-102

5. 5.

   Merlin, C., Gegear, R. J. & Reppert, S. M. (2009). Antennal circadian clocks coordinate sun compass orientation in migratory monarch butterflies. Science 325(5948), 1700–1704. doi:10.1126/science.1176221

6. 6.

   Takahashi, S., Hombe, T., Matsumoto, S., Ide, K. & Yoda, K. (2022). Head direction cells in a migratory bird prefer north. Science Advances 8(5), eabl6848. doi:10.1126/sciadv.abl6848

7. 7.

   Dacke, M., Baird, E., Byrne, M., Scholtz, C. H. & Warrant, E. J. (2013). Dung beetles use the Milky Way for orientation. Current Biology 23(4), 298–300. doi:10.1016/j.cub.2012.12.034