The twin who comes back younger

One twin stays at home. The other travels out at a large fraction of the speed of light, turns round and comes back. When they meet, the traveller has aged less, and the discrepancy can be made as large as desired by going faster or further.

The apparent paradox is that motion is relative. From the traveller’s point of view it is the Earth that recedes and returns, so by the same argument the stay-at-home should be the younger one. Both cannot be right, and the reunion is a single event about which there is no room to disagree.

What is actually being compared #

The quantity each twin accumulates is proper time: the time measured by a clock carried along a particular path. It is a property of the path, not of a frame, in the same way that the length of a road is a property of the road.

That analogy is exact and worth taking seriously. Two roads between the same pair of towns can have different lengths, and nobody finds it paradoxical; asking which length is “really right” is a confusion, because both are correct measurements of different things. Proper time works the same way, with one inversion: in spacetime the straight path has the longest proper time, not the shortest.

So the resolution can be stated in one line. The twins take two different paths between the same two events, the paths have different lengths in proper time, and the straight one is longer.

Where the slowing comes from is a light clock carried sideways: its pulse takes a longer, slanted path between the same two mirrors, and since the speed of light is the same for everybody, the tick takes longer. At 0.8 of the speed of light the factor is 1.67, so a traveller’s clock accumulates three fifths of what a stationary one does over the same coasting stretch — and nothing about the clock’s construction enters, which is why every clock the traveller carries agrees.

Why the straight path is the long one #

The inversion deserves its own explanation, because it is the one feature of the geometry that has no everyday analogue and it is where the sign of the answer comes from.

In ordinary geometry the distance between two nearby points is $\mathrm{d}s^2 = \mathrm{d}x^2 + \mathrm{d}y^2$, with both terms positive, so any detour adds. In spacetime the corresponding quantity for a clock is

$$c^2\mathrm{d}\tau^2 = c^2\mathrm{d}t^2 - \mathrm{d}x^2,$$

with a minus sign. Moving through space subtracts from the accumulated proper time, so any detour in space costs. A path that goes out and comes back spends time moving through space, and every second of that motion is a second not spent accumulating $\tau$.

That single minus sign is the whole content of the paradox. It also gives the general principle underneath: among all paths between two events, the one an unaccelerated body actually takes is the one of greatest proper time. Written that way it becomes a variational principle — a body free of forces moves so as to age as much as possible — and it survives into general relativity unchanged, where free fall maximises proper time and gravity stops being a force.

The triangle inequality is therefore reversed rather than absent. Two sides of a spacetime triangle are shorter than the third, always, provided both are paths a clock could take.

The numbers #

Take a destination four light-years away and a speed of 0.8$c$, with $\gamma = 5/3$.

The stay-at-home measures the round trip as $2\times4/0.8 = 10$ years. The traveller’s clock accumulates $10/\gamma = 6$ years. They meet four years apart in age, and the four years are not an illusion, an appearance, or a matter of interpretation: both twins are present at the reunion and both read their own clocks.

The factor is 1.25 at 0.6 of light speed, 1.67 at 0.8 and 3.20 at 0.95, and its shape is why the effect is invisible at everyday speeds and dominant near the speed of light. Every practical journey a person has taken sits at a $\gamma$ that differs from one in the twelfth decimal place — which is why the paradox is an argument rather than an observation, and why settling it took a spacetime diagram rather than an experiment.

Where the symmetry actually breaks #

The argument above is complete, and it does not explain what goes wrong with the traveller’s own reasoning — which is the part that makes the paradox feel like one.

The traveller’s error is in stitching two frames together. During the outbound leg the traveller is in one inertial frame, and in it the stay-at-home’s clock does run slow. During the inbound leg the traveller is in a different inertial frame, and in that one the stay-at-home’s clock also runs slow. But the two frames disagree about what “now on Earth” means, and switching between them at the turnaround skips a large interval of the stay-at-home’s life.

The same slicing at 0.8 of light speed tilts further and the gap grows accordingly. What is worth holding on to is that the traveller’s “now” at the far end sweeps across years of the stay-at-home’s worldline during the turnaround — not because anything happens to the stay-at-home, who notices nothing, but because the traveller’s definition of simultaneity swings round as their velocity does. The years are not skipped; they were never on the outbound slicing to begin with.

The skipped interval is not something that happens to the stay-at-home. Nothing happens to the stay-at-home at all: the turnaround is an event four light-years away, and the Earth’s clock ticks steadily throughout. What changes is which set of distant events the traveller would call simultaneous, and that is a convention rather than a fact.

The version with no diagram at all #

There is a way of getting the answer that involves no simultaneity, no diagram and no frame-switching, and it is the version to reach for when the argument gets tangled: count signals.

Let each twin send a pulse once a year by their own clock, and let each count what arrives. Counting is unambiguous — a pulse either arrives or it does not, and both twins agree on the total.

The relativistic Doppler factor is what decides that received rate. At 0.8 of the speed of light an approaching source’s signals arrive three times as often as they were sent and a receding source’s a third as often — the two being exact reciprocals, which is the property the whole count depends on.

The traveller sees the Earth’s pulses arriving at a third of a per year for the outbound leg (three years by the traveller’s clock, so one pulse) and at three per year for the inbound leg (three years, nine pulses): ten in total. So the traveller knows the stay-at-home aged ten years.

The stay-at-home’s count is asymmetric in a different way. The turnaround is seen late, because the light announcing it takes four years to arrive: for nine of the ten years the Earth receives pulses at a third of a per year (three pulses), and for the last one at three per year (three pulses): six in total. So the stay-at-home knows the traveller aged six years.

Both counts are objective, both twins agree with both, and there is no paradox to resolve. The asymmetry is in when each sees the turnaround, and that in turn is because one of them is at the place where the journey starts and ends.

Acceleration is not the cause #

The turnaround is often given as the resolution — the traveller accelerates, the stay-at-home does not, and that is that. It marks the asymmetry, and it is not the cause, and the difference matters.

The test is to vary things independently. Keep the turnaround identical and double the outbound distance: the age gap doubles. Keep the distance and make the turnaround gentler by taking a wide, slow arc: the gap is essentially unchanged. The difference accumulates during the coasting legs, in proportion to their length, and the turn contributes almost nothing.

The cleanest demonstration removes the acceleration entirely. Use three clocks, all moving inertially and none ever accelerating: one on Earth, one flying outward past the Earth and setting its clock to the Earth’s as it passes, and a third flying inward, which sets its clock to the outbound clock’s reading as they pass at the far point. When the third clock passes the Earth, its reading is compared with the Earth’s. The result is the same six against ten, with no acceleration anywhere in the experiment.

What that shows is that the asymmetry is between one inertial worldline and two joined together, rather than between accelerated and unaccelerated motion. A bent path is shorter in proper time whether the bend is sharp or smooth, and whether it is travelled by one clock or by two in relay.

The version that has been measured #

The effect is not hypothetical, and the versions that have been checked are not all about speed.

Height and speed pull in opposite directions in any real flown-clock experiment. At an aircraft’s altitude the height term dominates for a westward flight and the speed term for an eastward one, because the Earth’s own rotation adds to one and subtracts from the other — so the two directions give differences of opposite sign. That is a much sharper test than a single flight, because a systematic error in the clocks cannot change sign with the direction of travel.

The muon case is the one with the largest margin, and it is worth a paragraph because it is a twin experiment that happens continuously overhead. Muons created by cosmic rays some fifteen kilometres up have a half-life of 1.5 microseconds at rest, in which light travels 450 metres — so essentially none should reach the ground, and a great many do. At the typical $\gamma$ of around twenty, the surviving fraction rises by a factor of tens of thousands. The muon’s own account is the same arithmetic seen the other way: in its frame the atmosphere is contracted to under a kilometre, which it crosses easily in its ordinary lifetime. Both accounts give the same number of arrivals, which is what they have to do.

Hafele and Keating flew caesium clocks round the world in both directions in 1971 and found differences of −59 and +273 nanoseconds against predictions of −40 and +275. Muons produced in the upper atmosphere reach the ground in numbers that require a factor of thirty in their decay time, which is the same arithmetic with a much larger $\gamma$. And storage-ring experiments have carried unstable particles round a circle at $\gamma$ of nearly thirty and measured their lifetimes to a fraction of a per cent — a genuine twin experiment with the traveller returning repeatedly to the starting point.

The assumption hidden in “acceleration is not the cause” #

Saying that the turn contributes nothing is a stronger claim than it looks, and it rests on something that is not a theorem of relativity but an extra postulate about clocks.

The postulate is the clock hypothesis: an ideal clock’s rate depends on its instantaneous speed and on nothing else — not on its acceleration, not on its jerk, not on how long it has been accelerating. Grant that, and proper time along any worldline is the integral of $\mathrm{d}t/\gamma$ over the path, so the turnaround’s contribution is whatever tiny interval it occupies and no more. Deny it, and the whole integral has to be rewritten with acceleration-dependent terms in it.

It cannot be derived, because it is partly a definition. A clock is a physical object with internal structure, and structure can be broken: a pendulum clock at ten $g$ reads wrong, and at a thousand it reads nothing. What the hypothesis asserts is that this is a failure of the instrument rather than of time, and that a sufficiently robust clock is unaffected.

So it has to be measured, and it has been, at accelerations no engineering could otherwise reach. The muon storage ring at CERN held muons at $\gamma = 29.3$ on a circle seven metres across, which is a proper acceleration of about $10^{18}$ times the acceleration of gravity. Their measured lifetime matched $\gamma$ times the rest lifetime to two parts in a thousand, with no residual term depending on the acceleration — a bound tight enough that any correction would have to set in eighteen orders of magnitude beyond anything a traveller would experience.

That is the experiment which licenses the three-clock version and the “gentler turnaround” argument alike. Without it, “the difference accumulates during the coasting” would be an assertion rather than a result.

The journey with no coasting at all #

The other way to remove the sharp turn is to remove the coasting instead: accelerate continuously at one $g$ for the first half of the trip and decelerate at one $g$ for the second. That is the comfortable journey, since everyone aboard has a floor, and the proper time is now an integral along a smoothly curved worldline rather than a sum of two straight pieces.

The integral is elementary and the numbers are startling. The natural unit is $c/g$, which is almost exactly one year, so the arithmetic is unusually clean.

To Alpha Centauri, 4.3 light-years: 5.9 years pass on Earth and 3.6 aboard. To the centre of the galaxy, 27,000 light-years: 27,000 years pass on Earth and about 20 aboard. To the Andromeda galaxy, two and a half million light-years: 28 years aboard.

The reason the ship’s figure grows so slowly is that proper time accumulates logarithmically in the distance once the speed is close to $c$ — each doubling of the range costs the crew about eight months. There is no limit to it in principle, and a crew that kept the engine running would cross the observable universe within a working lifetime of their own.

None of which makes it a proposal. The energy required to reach even a modest $\gamma$ is the problem the final section names, and it is not an engineering difficulty but an arithmetic one. What the calculation establishes is that the barrier to travelling a long way is the fuel and never the time, which is the opposite of what the light-year makes it sound like.

An argument that ran for fifty years #

The puzzle is Langevin’s, from 1911, and he presented it as a demonstration rather than a paradox: his traveller goes out and back at a speed close to $c$ and returns to find two centuries have passed. Einstein had noted the effect in the 1905 paper, for a clock carried round a closed curve, without making anything of it.

What followed is a genuinely odd episode in the history of physics. Herbert Dingle, a respected physicist and historian of science, spent the last two decades of his life arguing in print that the effect was a logical contradiction and that relativity was therefore inconsistent. The exchanges ran through Nature and elsewhere from the mid-1950s into the 1970s, generated a large literature of rebuttals, and ended with the objection unsustained.

The interesting part is not that he was wrong but where he was wrong: he insisted on the symmetry of the two twins’ descriptions and would not accept that the traveller’s frame is not a single inertial frame. That is exactly the step the naive argument gets wrong, and it survived so long in the hands of a professional because the two coasting legs really are symmetric — every objection is correct up to the turnaround.

The episode is also the reason the counting argument is worth knowing. It contains no frames at all, and there is nothing in it to disagree about.

What it costs, and where the model stops #

Nothing here is an appearance. The difference is not about what each twin sees; it is the readings on two clocks brought together at one event. Any account that resolves the paradox by talking about light travel times has answered a different question.

Proper time is not the only quantity that behaves this way. Everything carried along the worldline ages by the proper time: a biological clock, a decaying nucleus, a wristwatch. That is not an extra assumption but a consequence of the theory applying to everything.

The three-clock version is not the same experiment. It reproduces the arithmetic without a single object making the round trip, and something is genuinely lost: no one clock’s reading is being compared with anything, only a relay’s total. Whether that is a legitimate version of the paradox is a question about what the paradox is about, and the honest answer is that it isolates the geometry and discards the biology.

The effect is not reversible. Nothing on this page provides a way to travel into the past or to make a clock run fast; the traveller arrives in the stay-at-home’s future having spent less of their own life getting there, which is a one-way trade. The energy cost is what makes it academic: reaching $\gamma = 5/3$ means carrying enough energy to give a body two thirds of its rest energy again, which for a person is comparable with the yield of a large weapon.

Gravity has been left out. In a gravitational field the stay-at-home is not on a straight worldline either, and the comparison is between two curved paths. The general rule that survives is that free fall maximises proper time locally, which is the statement the whole of general relativity’s dynamics is built on.

The one-sentence version, and why it took so long to find #

The resolution can be compressed to a sentence: proper time is a path length, the paths differ, and the straight one is longer.

That it took decades of argument to settle on that is not a failure of anybody’s intelligence. It is what happens when a subject is taught through its equations rather than its geometry. The Lorentz transformation makes the effect look like a fact about clocks observed from frames, and once the problem is posed that way the frames multiply and the contradictions follow. Minkowski’s reformulation in 1908 — spacetime as a geometry with an indefinite metric, worldlines as curves in it, proper time as their length — makes the paradox evaporate before it can be stated, and Minkowski’s version was available the whole time the argument was running.

The general lesson is one this collection keeps meeting: a change of representation is not a presentational convenience. The same physics written two ways produces two different sets of questions, and one of the sets can be unanswerable.

The ladder from here #

Later rungs on this anchor: proper time as a path length, with the reversed triangle inequality stated properly; the Doppler-count method generalised, which is the cleanest calculational route for any such problem; the twin problem in a gravitational field, where the free-falling twin ages most; the closed-universe version, in which a traveller returns without ever turning round and the resolution has to come from somewhere else entirely; and the experimental record, which is by now enormous.

The neighbouring ladders are simultaneity, whose tilting slices are what the traveller’s frame-switch skips over, and the relativistic Doppler effect, which supplies the counting argument that makes the whole thing elementary.