The Ross-Littlewood Paradox

Does the answer depend on the definition of "limit"? The paradox exposes a tension between different ways of extending finite intuitions to infinite processes. The cardinality argument (9 balls added per step, so infinitely many remain) uses one notion. The set-theoretic limit (every specific ball is removed) uses another. These two notions agree for finite processes and diverge for supertasks.

Is the answer zero or indeterminate? Some philosophers argue that the question has no well-defined answer: a supertask is not a real physical process, and asking what is in the vase "at noon" requires specifying what happens at the limit, which the problem does not do. On this view, the set-theoretic limit is a choice about how to extend the problem to the limit, not a discovery. The contrasting version shows that different choices give different answers.

Physical realisability. No physical vase can contain infinitely many balls. No physical process can perform infinitely many operations in finite time. The paradox is purely mathematical: it asks what we should say about a mathematical model, not about the physical world. Its interest lies in what it reveals about the mathematics of infinity, not in any physical implications.

The puzzle in approximately this form was discussed by John Edensor Littlewood in A Mathematician's Miscellany (1953) as an illustration of how infinity can surprise. Sheldon Ross included a version in his probability textbook and the puzzle became attached to both names.

The puzzle belongs to a family of supertask paradoxes that became a topic of philosophical analysis in the 1950s and 1960s, alongside Thomson's Lamp (James Thomson, 1954) and Zeno's paradoxes. The philosophical literature on supertasks was substantially developed by Adolf Grünbaum, Paul Benacerraf, and José Bernadete.

It separates cardinality from set identity. The paradox shows that counting (cardinality) and identity (which specific elements) are not interchangeable for infinite sets. This is a fundamental feature of infinite set theory that has no finite analogue.

It shows that supertask outcomes depend on implementation details. The two versions of the problem — remove the lowest-numbered ball vs. the highest — differ only in which ball is removed each step, yet produce opposite limiting results. This is a strong demonstration that the abstract description of a process (add 10, remove 1) does not determine its limit without further specification.

It is a vivid illustration of set-theoretic limit vs. cardinality limit. These notions agree for finite sets and diverge for infinite ones. The paradox makes this divergence concrete and memorable.

Database transactions. In systems that perform infinite sequences of insertions and deletions, the "final state" depends on the order and identity of operations, not just on the net count. The Ross-Littlewood paradox illustrates why order of operations matters in infinite-state systems.

Limit analysis in algorithm design. Algorithms that iteratively update sets — adding and removing elements — can exhibit paradoxical behaviour when the iteration count is unbounded. The paradox is a reminder that the cardinality of a set at each step does not determine the set's identity at the limit.

The formal setup

At step $n$ ($n \geq 1$), define:

• Added: $A_n = {10n-9, 10n-8, \ldots, 10n}$ (ten balls)
• Removed: $r_n = n$ (one ball)

The contents of the vase after step $n$:

$$V_n = \bigcup_{k=1}^{n} A_k \setminus {1, 2, \ldots, n} = {n+1, n+2, \ldots, 10n}$$

(Easily verified: $|V_n| = 9n$.)

The set-theoretic lim inf is:

$$L = \liminf_{n\to\infty} V_n = \bigcup_{n=1}^{\infty} \bigcap_{k=n}^{\infty} V_k$$

For any fixed $m \in \mathbb{N}$: ball $m \in V_k$ iff $k < m \leq 10k$, i.e. $m/10 \leq k < m$. For $k \geq m$, ball $m$ has been removed ($r_m = m$) and $m \notin V_k$. So $m \notin \bigcap_{k=n}^{\infty} V_k$ for any $n \geq m$. Hence $m \notin L$.

Since $m$ was arbitrary, $L = \emptyset$.

The lim sup, by the same argument, is also $\emptyset$ (since every ball eventually leaves $V_n$ forever). So $\lim_{n\to\infty} V_n = \emptyset$ in the set-theoretic sense.

The conditional convergence analogy

The paradox is analogous to the conditional convergence of infinite series. The series $\sum (-1)^{n+1}/n = 1 - 1/2 + 1/3 - 1/4 + \cdots$ converges to $\ln 2$. But by the Riemann rearrangement theorem, rearranging the terms can make it converge to any real number or diverge. The sum depends on the order of the terms, not just their values.

The Ross-Littlewood paradox is similar: the "limit" of the vase's contents depends on which balls are removed at each step (the labelling), not just on the net count added per step. The relabelling version makes this explicit.