Abstract
“Almost every” and “almost none” carry two different meanings — smallness measured by measure, and smallness measured by category in Baire's sense. This paper calls them by name and shows in numbers that the two return opposite verdicts. NO NEW MATHEMATICAL THEOREM IS CLAIMED — all of this is standard.
First, on countable sets the two agree — which is why one word has served for both (Section 2).
Second, the fat Cantor set is nowhere dense and of measure $1/2$. After $14$ steps the measure remaining is $0.50003051757813$ and the longest surviving piece is $3.05\times10^{-5}$ — nothing at all to category, half the line to measure (Section 3).
Third, the reverse also happens. Covering the rationals with intervals of length $\varepsilon2^{-n}$ gives an open dense set of measure at most $\varepsilon$ (at $\varepsilon=0.001$, measure $0.0010000000$ against a complement of $0.9990000000$) (Section 4).
Fourth, and this is the core. The real line splits into a null set and a meagre set. One of them is nothing to measure and everything to category; the other is exactly the reverse (Section 5).
Fifth, there is a named witness: the Liouville numbers. The bound on the measure of the covering diverges at $n=2$ and falls to $1.693$ at $n=3$ and $0.109$ at $n=6$ — null to measure and almost everything to category (Section 5).
Sixth, there is a third sense of small. The middle-thirds Cantor set has measure zero and Hausdorff dimension $0.630929753571$, which is not zero (Section 6).
Scope (what is not claimed)
NO NEW MATHEMATICAL THEOREM IS CLAIMED — Baire's category theorem, the fat Cantor set, the decomposition of the line into a null set and a meagre set (Oxtoby), the fact that the Liouville numbers form a dense $G_\delta$ of measure zero, and Hausdorff dimension are all standard. Nor does this paper rule on which sense of “small” is correct — what can be written is what each of them measures. What is added is setting the two on one computation and making them point opposite ways, carrying the fat Cantor set over from Paper 168 and measuring it from both sides, and actually summing the bound on the measure of the Liouville numbers.
1. Notation and position
A set is null when it can be covered by intervals of arbitrarily small total length. It is nowhere dense when its closure has empty interior, and meagre when it is a countable union of nowhere dense sets. A set whose complement is meagre is comeagre. The measure of a union of intervals was computed here by merging overlaps by hand.
Position: Paper 168 wrote that Riemann integrability is decided by the measure of the discontinuities and not by their density. This paper stands beside it — “small” itself is two words, and the two disagree.
2. Where they agree
Observation 2.1 A countable set is null. It is also meagre, a single point being nowhere dense and there being only countably many of them.
| set | measure | category |
|---|---|---|
| the integers | $0$ | meagre |
| the rationals in $[0,1]$ | $0$ | meagre |
| any finite set | $0$ | meagre |
Observation 2.2 So far the two words point the same way. That is why they have been used interchangeably. They are not interchangeable.
3. Nowhere dense, and of measure $1/2$
Verification 3.1: Remove from $[0,1]$ the open middle of length $4^{-n}$ at step $n$ (the same set built in Paper 168).
| step | pieces | measure remaining | longest piece |
|---|---|---|---|
| $1$ | $2$ | $0.75$ | $0.375$ |
| $3$ | $8$ | $0.5625$ | $0.0703125$ |
| $6$ | $64$ | $0.5078125$ | $0.00793457$ |
| $10$ | $1024$ | $0.50048828125$ | $0.000488758$ |
| $14$ | $16384$ | $\mathbf{0.50003051757813}$ | $\mathbf{3.05194\times10^{-5}}$ |
The measure tends to $1/2$ and the longest piece to $0$. (QED)
Observation 3.2 That the longest piece falls to zero means the set contains no interval at all — it is nowhere dense, and therefore meagre. And it occupies half the line.
Observation 3.3 To the eye of category it is nothing; to the eye of measure it is half.
4. Dense, and of arbitrarily small measure
Verification 4.1: Enumerate the rationals of $[0,1]$ and place an interval of length $\varepsilon2^{-n}$ about the $n$-th. The union is open and dense, and its measure is at most $\varepsilon$.
| $\varepsilon$ | measure of the cover | measure of the complement |
|---|---|---|
| $0.5$ | $0.4999997313$ | $0.5000002687$ |
| $0.1$ | $0.1000000000$ | $0.9000000000$ |
| $0.01$ | $0.0100000000$ | $0.9900000000$ |
| $0.001$ | $\mathbf{0.0010000000}$ | $\mathbf{0.9990000000}$ |
(The rationals were enumerated to denominator $60$, giving $1103$ of them. QED)
Observation 4.2 The cover surrounds every rational, so it is dense, and its measure can be made as small as one likes. Its complement holds almost all of the length while containing no open interval.
Observation 4.3 The experiment truncates the enumeration, so apparent gaps remain in the complement. They are artefacts of the truncation and vanish in the limit.
5. The core: the line splits into two, each calling the other almost nothing
Observation 5.1 Let $G_k$ be the cover of Section 4 with $\varepsilon=1/k$, and put $H=\bigcap_k G_k$.
| $H$ lies inside every $G_k$ | so its measure is $\le 1/k$, that is, null |
|---|---|
| $H$ is a countable intersection of dense open sets | so it is comeagre |
Observation 5.2 The line therefore splits as follows.
| to measure | to category | |
|---|---|---|
| $H$ | $\mathbf{0}$ (nothing) | almost everything |
| the complement of $H$ | almost everything | meagre (nothing) |
Observation 5.3 The two words do not merely differ. On this pair they return opposite verdicts. What one eye sees as almost everything, the other sees as almost nothing.
Verification 5.4: There is a named witness. The Liouville numbers form exactly such a dense $G_\delta$. The bound on the measure of the covering, summed to $q\le2000$:
| $n$ | $\sum_{q\le2000}2(q+1)q^{-n}$ | verdict |
|---|---|---|
| $2$ | $15.64560459$ | diverges |
| $3$ | $1.69298194$ | finite |
| $4$ | $0.5687600238$ | finite |
| $6$ | $\mathbf{0.1085416343}$ | falls with $n$ |
(QED)
Observation 5.5 From $n\ge3$ the bound is finite and falls with $n$, so the intersection over all $n$ is null. Yet every stage is open and dense, so the intersection is comeagre. To measure the Liouville numbers are not there; to category they are almost everything.
Observation 5.6 The same thing happens in a more famous place — the normal numbers have measure $1$ and are meagre (Borel $1909$, together with the standard result on category). “Almost every number is normal” and “a typical number is not normal” are both true. The category side is not computed here — it is cited as a standard fact.
6. A third sense of small
Verification 6.1: The middle-thirds Cantor set.
| measure | $0$ |
|---|---|
| Hausdorff dimension | $\mathbf{0.630929753571}$ |
| is that zero? | no |
For contrast, the fat Cantor set has measure $1/2$ and dimension $1$. (QED)
Observation 6.2 The chain “countable $\Rightarrow$ null $\Rightarrow$ of dimension zero” does not hold. The middle-thirds set is uncountable, null, and of positive dimension (the same place at which Paper 94 wrote that dimension is not one thing).
7. The census
Observation 7.1
| sense of small | meaning | relation to the others | witness |
|---|---|---|---|
| countable | the points can be counted | null and meagre | where the two agree |
| null | length $0$ | need not be meagre | the Liouville numbers |
| meagre | countable union of nowhere dense | need not be null | the fat Cantor set |
| dimension zero | Hausdorff dimension | stronger than null | the middle thirds, $0.6309$ |
Observation 7.2 Which to use is settled by what is being asked. Probability and analysis take measure; arguments about what is typical or generic take category. One word, “almost”, pointing at two different ledgers.
8. Conclusion
- On countable sets the two agree (Obs. 2.1).
- Nowhere dense and of measure $1/2$ (Ver. 3.1): longest piece $3.05\times10^{-5}$.
- Dense and of measure $0.001$ (Ver. 4.1).
- The line splits into a null set and a meagre one (Obs. 5.2): each calls the other almost nothing.
- The Liouville numbers are the witness (Ver. 5.4): bound $0.1085$ at $n=6$.
- Null is not dimension zero (Ver. 6.1): $0.630929753571$.
References
- Baire, R. (1899). Sur les fonctions de variables réelles. Annali di Matematica Pura ed Applicata, 3.
- Smith, H. J. S. (1875). On the integration of discontinuous functions. Proceedings of the London Mathematical Society, 6.
- Volterra, V. (1881). Alcune osservazioni sulle funzioni punteggiate discontinue. Giornale di Matematiche, 19.
- Liouville, J. (1844). Sur des classes très étendues de quantités dont la valeur n'est ni algébrique ni même réductible à des irrationnelles algébriques. Comptes rendus de l'Académie des Sciences, 18.
- Borel, É. (1909). Les probabilités dénombrables et leurs applications arithmétiques. Rendiconti del Circolo Matematico di Palermo, 27.
- Hausdorff, F. (1919). Dimension und äusseres Mass. Mathematische Annalen, 79.
- Oxtoby, J. C. (1971). Measure and Category. New York: Springer.
- Yamagishi, Y. (2026). “Dimension” Is Not One Thing [Paper 94]. Zenodo. 10.5281/zenodo.21758581
- Yamagishi, Y. (2026). There Are Spaces With No Points at All [Paper 155]. Zenodo. 10.5281/zenodo.22084848
- Yamagishi, Y. (2026). “A Good Code Exists” and “Here Is a Good Code” Are Different Sentences [Paper 163]. Zenodo. 10.5281/zenodo.22092677
- Yamagishi, Y. (2026). Probability Has Four Addresses [Paper 164]. Zenodo. 10.5281/zenodo.22092815
- Yamagishi, Y. (2026). What Lebesgue Bought Was Not a Bigger Class but Closure under Limits [Paper 168]. Zenodo. 10.5281/zenodo.22093320
- Yamagishi, Y. (2026). Kodaira's Fence Is Integrality, Not Positivity [Paper 173]. Zenodo. 10.5281/zenodo.22094507
Reference
Yamagishi, Y. (2026). Small Is Two Words, and They Do Not Agree [Paper 176]. YOI: https://www.yamagishiyuuki.com/yoi/176
DOI: 10.5281/zenodo.22097994
ORCID: 0009-0006-8587-3450
PDF (English) · PDF (Japanese) · Date proof of the English PDF (OpenTimestamps .ots) · SHA256SUMS
The record page for this paper (in Japanese, with the PDFs in each language and the date proof): YOI 176. This paper is a preprint and has not been peer reviewed.