Zeno's Dichotomy Paradox

There are three quite different things people mean when they say the Dichotomy is solved, and they are worth separating.

One: the arithmetic is not in doubt. That an infinite series of positive terms can have a finite sum is a theorem, given rigorous form by Cauchy in the 1820s and Weierstrass later in the century. Nobody disputes it. If Zeno's argument requires the sum of the sub-distances to be infinite, the argument is simply unsound.

Two: whether that is what Zeno's argument requires is contested. Read the Dichotomy as a claim about distance and the series settles it. Read it as a claim about doing things — each sub-journey is an act, and infinitely many acts cannot be got through — and the series is beside the point. It shows the durations sum to ten seconds. It does not, on its own, explain how a task with no last member gets completed, or how a task with no first member gets begun.

Three: Aristotle's answer was not the modern one. He gives two, roughly forty pages apart, and the second is a retraction of the first.

• In Physics VI.2 he says the length and the time to cross it are divisible in exactly the same way. If the distance is infinitely divisible then so is the hour, and an infinitely divisible distance can be crossed in an infinitely divisible time. This is essentially the modern reply, minus the arithmetic.
• In Physics VIII.8 he calls that answer inadequate — "a solution sufficient as against the questioner, but not adequate to the facts" — and replaces it. A line, he says, is not actually divided into infinitely many parts. It is potentially so divided: the halfway points exist only when someone actually makes the division, and you cannot make infinitely many divisions. The runner crosses one undivided distance. The infinite stages were never there.

That distinction between actual and potential infinity governed the subject until Cantor made completed infinities respectable in the 1870s. It is also, awkwardly for the tidy version of the story, an answer that modern mathematics rejects rather than vindicates. Contemporary analysis treats the interval as actually containing its infinitely many points. So the standard modern reply does not extend Aristotle; it takes the position he abandoned.

Who holds what today. Adolf Grünbaum (1967) argued that the classical theory of the continuum, properly applied, disposes of the paradoxes completely, including the "act" reading. Max Black, writing in Analysis in the early 1950s, took the opposite view: the idea of an "infinity machine" that performs endless operations is not merely unrealised but incoherent, so summing a series answers the wrong question. James Thomson (1954) invented a device to make Black's worry bite, and Paul Benacerraf (1962) showed that Thomson's device does not do what Thomson claimed. Wesley Salmon's 1970 anthology collects that whole exchange, and the fact that it was worth collecting is the point: this was a live argument among careful philosophers three hundred years after Newton.

The honest summary is narrow but firm. The mathematics is settled and the metaphysics is not. Anyone who tells you calculus killed Zeno in the seventeenth century is telling you something that is true about the sums and false about the history.

What is documented. Zeno of Elea lived in the Greek colony of Elea, in what is now Campania in southern Italy, roughly 490 to 430 BCE. He wrote one book. It is lost. Every word we have of him comes through other people's quotation of it.

Plato's dialogue Parmenides (127a–128e) has Zeno describe his own project: the book was written in his youth as a defence of Parmenides, showing that anyone who believes in plurality is committed to consequences even more absurd than Parmenides' monism. That is the best evidence we have for what Zeno was doing, and it comes from a philosophical dialogue written perhaps eighty years after his death by an author who was not writing history. The dramatic scene — Zeno at forty, Parmenides at sixty-five, Socrates very young — may be an invention.

Proclus, writing in the fifth century CE, says the book contained forty arguments. Nobody knows whether that number is reliable.

The four arguments against motion — the Dichotomy, the Achilles, the Arrow and the Stadium — are known only from Physics VI.9, where Aristotle summarises each in a sentence or two before refuting it. He is a hostile witness working from a text he has read and we have not. His name for the first one, "the dichotomy", is the name it still carries.

Around a dozen verbatim lines of Zeno survive, all quoted by the Neoplatonist Simplicius in his commentary on the Physics in about 540 CE — roughly a thousand years after Zeno wrote. Every one of those quoted lines belongs to the arguments against plurality, not against motion. There is no surviving sentence of Zeno's own on the Dichotomy. What we have is Aristotle's paraphrase of it and the commentators' paraphrases of Aristotle.

What is tradition. Diogenes Laërtius (IX.26–27) reports that Zeno joined a conspiracy against a tyrant, was tortured, and bit off his own tongue and spat it at the man. He gives several mutually inconsistent versions, with different tyrants. It is a good story and there is no reason to believe any particular version of it.

What happened afterwards. The paradoxes were discussed continuously in the Greek commentary tradition — Themistius, Philoponus, Simplicius — and again by medieval writers on the continuum. Pierre Bayle's Dictionnaire historique et critique (1697) put them back in front of a modern audience, and their pressure is visible in Hume and in Kant's antinomies of infinite divisibility.

The mathematics arrived in the nineteenth century, not the seventeenth. Newton and Leibniz computed with infinitesimals that they could not justify; Cauchy in the 1820s and Weierstrass afterwards replaced them with limits defined by inequalities, and Cantor from the 1870s made completed infinite collections a legitimate object of study. Only then did "an infinite sum with a finite value" become a precise statement rather than a suggestive one.

Bertrand Russell, in The Principles of Mathematics (1903), called Zeno's arguments "immeasurably subtle and profound" and complained that posterity had treated him unjustly. The twentieth-century argument then reopened, with Max Black, James Thomson, Paul Benacerraf and Adolf Grünbaum disagreeing sharply about whether the new mathematics answered the old question — a dispute collected in Wesley Salmon's 1970 anthology and not concluded there.

It forced infinity to be defined. For two thousand years, mathematics followed Aristotle in refusing to treat an infinite collection as a completed object. As late as 1831 Gauss wrote to Schumacher that he protested "against the use of an infinite magnitude as something completed, which is never permissible in mathematics". Zeno is a large part of why that caution existed, and the machinery eventually built to answer him — limits, convergence, the rigorous continuum — is the foundation of calculus as it is actually taught and used today. The epsilon-delta definition of a limit is, historically, an answer to Zeno.

It is the cleanest example of a bad premise hiding in plain sight. Every step of the Dichotomy is obviously true except one, and the false one — that an infinite sequence of stages cannot be got through — is the step nobody notices because it never gets stated. Arguments fail this way constantly. Zeno's is the canonical case because the conclusion is so plainly false that you know in advance a premise must go, which turns the exercise into a pure hunt for the unstated assumption.

It made the structure of space and time an open question rather than an assumption. Whether space is infinitely divisible is not something that can be settled by looking. Zeno turned it into a question with consequences, and it has stayed one: the same question is live in quantum gravity, where several programmes posit a shortest meaningful length.

It gave philosophy a genuinely unresolved technical problem. The supertask literature, running from Thomson in 1954 to work on relativistic computation today, exists because the Dichotomy would not stay solved. That is unusual. Most ancient paradoxes are studied for what they reveal about their authors; this one is still generating papers about physics.

Zeno behaviour in hybrid systems. In control theory and formal verification, a hybrid automaton exhibits Zeno behaviour when it takes infinitely many discrete transitions in a finite time interval. The standard example is a bouncing ball whose bounces converge on a finite instant. This is not a metaphor or a naming joke: a simulator hitting a Zeno point will loop forever making infinitesimal progress, and model checkers for timed automata must explicitly exclude Zeno runs before they can certify anything. Zeno's argument became an engineering failure mode.

Numerical computation. Every iterative algorithm is a Dichotomy: infinitely many steps that must be stopped somewhere. Whether the sequence converges, how fast, and whether the limit is the answer you wanted are the daily questions of numerical analysis, and the difference between a geometric series and a harmonic one is the difference between a routine that terminates usefully and one that does not.

The quantum Zeno effect. Misra and Sudarshan (1977) showed that a quantum system measured sufficiently often can be prevented from decaying — repeated observation freezes the state. They named it after Zeno's Arrow, and the effect has since been observed experimentally in trapped ions and other systems. The connection is an analogy rather than a resolution of anything, but it is a real physical effect that carries his name.

Rendering and level-of-detail. Any system that keeps subdividing until the subdivisions fall below a threshold — adaptive mesh refinement, recursive ray tracing, progressive image loading — is running the Dichotomy deliberately and relying on convergence to make it terminate. The reason such algorithms work is exactly the reason the runner reaches the door.

The argument in standard form

Let a runner traverse the interval $[0, 1]$ at constant speed $v$. The Dichotomy observes that the traversal decomposes into the sequence of sub-intervals

$$\left[0, \tfrac12\right],\ \left[\tfrac12, \tfrac34\right],\ \left[\tfrac34, \tfrac78\right],\ \ldots,\ \left[1 - 2^{-n}, 1 - 2^{-(n+1)}\right],\ \ldots$$

whose lengths are $2^{-(n+1)}$ for $n = 0, 1, 2, \ldots$. Reconstructed as a valid argument:

1. To traverse $[0,1]$ the runner must traverse every member of this sequence.
2. The sequence has infinitely many members.
3. A process consisting of infinitely many successive members cannot be completed.
4. Therefore $[0,1]$ cannot be traversed.

Premises 1 and 2 are true. Premise 3 is what fails, and everything turns on how it fails.

The series

$$\sum_{n=0}^{\infty} 2^{-(n+1)} = \lim_{N \to \infty} \sum_{n=0}^{N} 2^{-(n+1)} = \lim_{N \to \infty}\left(1 - 2^{-(N+1)}\right) = 1$$

At constant speed the corresponding durations are $t_n = 2^{-(n+1)}/v$, and

$$\sum_{n=0}^{\infty} t_n = \frac{1}{v}$$

which is the elementary answer. The partial sums form a strictly increasing sequence bounded above by $1/v$; by the monotone convergence property of the reals it has a supremum, and that supremum is the value of the sum. Note that the limit is not attained by any partial sum. "The runner completes the journey" corresponds to the limit, not to a final term, and there is no final term.

The regressive form of the argument is the sequence $\left[2^{-(n+1)}, 2^{-n}\right]$ for $n = 0, 1, 2, \ldots$, ordered by containment towards $0$. This sequence has order type $\omega^{*}$: it has a last member but no first. Its lengths sum to $1$ just as before, so convergence is not the issue. The issue is that the runner's tasks are not well-ordered, and "perform them in order" has no first instruction. Any resolution of the progressive form by summation leaves this untouched.

Supertasks

Thomson (1954) named the completion of an $\omega$-sequence of actions in finite time a supertask, and argued against its possibility with the lamp.

A lamp is off at $t = 0$. It is switched at $t = 1 - 2^{-n}$ for $n = 1, 2, 3, \ldots$, so the switch is thrown infinitely often before $t = 1$. What is its state at $t = 1$?

It cannot be on, since every "on" is followed by an "off". It cannot be off, for the mirror reason. Thomson concluded that the supposition of a completed supertask is contradictory.

Benacerraf (1962) identified the gap. The specification fixes the lamp's state at every $t < 1$ and says nothing whatever about $t = 1$. There is no principle requiring the state at $1$ to be a limit of the states before it — that function has no limit — so no contradiction is derivable. Thomson's argument establishes only that the description is incomplete. Thomson later accepted the point.

This exchange is the crux of the modern situation. Benacerraf's reply defends the consistency of supertasks; it does not exhibit a mechanism for them, and the lamp remains a good demonstration that a convergent series does not by itself license every conclusion about what happens at the limit point. Later work has gone in both directions: Pérez Laraudogoitia (1996) constructed Newtonian particle systems that violate energy conservation through supertask-like interactions, and Malament–Hogarth spacetimes in general relativity allow an observer to receive the output of an infinite computation in finite proper time. The literature is surveyed in the Stanford Encyclopedia entry on supertasks.

The at-at theory

Russell (1903) proposed, and Salmon developed, the view that removes the "task" framing entirely. On the at-at theory, motion is nothing more than occupying different positions at different times: a continuous function $x(t)$ from an interval of times to an interval of positions. There is no additional activity of "getting from one place to the next", no smallest completed step, and hence no infinite sequence of achievements to be got through. The runner's being at $1/2$ at one moment and at $3/4$ at another is a fact about the function, not a job the runner performed.

If motion is at-at, premise 3 is not false so much as ill-formed, since there is no sequence of acts for it to be about. The at-at theory buys the resolution at a price: it makes motion a matter of the pattern of positions across time rather than anything happening at an instant, which many find as counterintuitive as Zeno's conclusion. That is the trade, and the trade is why this is still a subject rather than a footnote.

If space were not continuous

The whole argument presupposes that the interval can be halved indefinitely. Deny that — posit a smallest indivisible length — and the Dichotomy never begins, because the sequence of sub-intervals terminates.

The cure is worse than the disease. Hermann Weyl's tile argument (1949) points out that in a square grid of indivisible cells, the diagonal of an $n \times n$ square is $n$ tiles long, exactly like each side, so the Pythagorean relation fails at the fundamental level and no continuous rotation is available. Ancient atomism ran into a version of the same wall. Zeno's remaining arguments, particularly the Stadium, appear to have been aimed precisely at the discrete-space escape route: he seems to have set out a dilemma in which both horns — continuous space and atomic space — lead to absurdity. Reading the four motion paradoxes as a coordinated attack on a disjunction rather than as four unconnected puzzles is a reconstruction, plausible but not documented; Aristotle does not present them that way.