Hume's Principle: Foundations of Arithmetic

• Hume’s Principle is an abstraction principle that defines numbers by equating concepts that are in one-to-one correspondence.
• It bridges the existence of bijections between sets with the identity of numerical objects, thereby recovering arithmetic in a second-order logical framework.
• The principle extends its impact to cognitive studies and set theory by explaining how symbolic counting and urelement models realize exact cardinality.

Searching arXiv for recent and foundational papers on Hume's Principle. Hume’s Principle is the abstraction principle that identifies numbers through equinumerosity. In its standard Fregean formulation, it states that the number of the $F$s is identical with the number of the $G$s if and only if there is a bijection between the $F$s and the $G$s. Within contemporary philosophy of mathematics it is the paradigmatic abstraction principle, central to neo-Fregean logicism, to the modern reconstruction of Frege’s theorem, and to debates about categoricity, predicativity, and the ontology of arithmetic (Walsh, 2014, Walsh et al., 2014).

1. Formal statement and abstraction structure

In the standard second-order setting, Hume’s Principle is written

$$\forall F \forall G\,\bigl(F \approx G \leftrightarrow \#F=\#G\bigr).$$

Here $F$ and $G$ are concepts, $F \approx G$ says that $F$ and $G$ are equinumerous, and $G$0 is a type-lowering cardinality operator from concepts to objects. Equinumerosity is standardly defined by the existence of a bijection between the extension of $G$1 and the extension of $G$2. In the set-theoretic-looking notation used in some contemporary presentations, the principle takes the form

$G$3

The principle is an abstraction principle in the precise sense that it introduces a new object $G$4 whose identity conditions are fixed by an independently given equivalence relation, namely equinumerosity. In the generic abstraction schema

$G$5

Hume’s Principle is the instance in which $G$6 is equinumerosity and $G$7 is $G$8 (Walsh, 2014).

Its logical architecture has two sides. On the right-hand side lies a relation between concepts, $G$9. On the left-hand side lies identity between objects, $F$0. The operator $F$1 is therefore what bridges one-to-one correspondence and number as object. In many model-theoretic treatments, $F$2 is represented by an equivalence class of equinumerous concepts, but the principle itself is not merely a set-theoretic recipe; it is an abstraction principle that fixes the identity conditions of cardinal objects through an equivalence relation (Pal, 7 Jul 2026).

2. Fregean arithmetic and the modern form of Frege’s theorem

Hume’s Principle acquired its contemporary prominence because second-order logic plus Hume’s Principle yields arithmetic. This result is the modern form of Frege’s theorem: second-order Peano arithmetic is interpretable in second-order logic with Hume’s Principle and full impredicative comprehension (Walsh, 2014).

The usual Fregean reconstruction defines zero as the number of a concept with no instances, for example

$F$3

and defines successor by passing from a concept $F$4 to a concept with exactly one more instance. With second-order comprehension one proves the characteristic Peano clauses: zero is not a successor, successor is injective, and second-order induction holds. In this way arithmetic is recovered not from primitive numerical objects but from the values of an abstraction operator applied to concepts.

In the strong setting, the relation to arithmetic is exact at the level of interpretability strength. One result gives

$F$5

and another gives the converse, so that $F$6 and $F$7 are mutually interpretable (Walsh, 2014). This is one reason Hume’s Principle is treated as the central abstractionist basis for arithmetic. By contrast, Basic Law V, Frege’s original principle for extensions,

$F$8

is inconsistent with full impredicative comprehension, whereas Hume’s Principle is consistent in that setting and strong enough to recover arithmetic (Walsh, 2014).

3. Predicative, restricted, and modal settings

The logical strength of Hume’s Principle depends sharply on the surrounding comprehension principles. In predicative settings, Hume’s Principle alone no longer reproduces the full Fregean result. Earlier work showed that HP plus predicative comprehension does not interpret $F$9; more specifically, the hyperarithmetic fragment of Hume’s Principle does not interpret the hyperarithmetic fragment of second-order Peano arithmetic, so that in that sense there is no predicative version of Frege’s theorem (Walsh, 2014).

A different predicative result is obtained by enlarging the abstractionist environment. The predicative Fregean theory $G$0, which contains all abstraction principles whose underlying equivalence relations are provably equivalence relations in a weak background second-order logic, interprets second-order Peano arithmetic. In that framework, HP is included but is not by itself sufficient; the interpretation of arithmetic leans heavily on the presence of Basic Law V and other abstraction principles, together with the theory’s ability to recover full comprehension for $G$1-formulas (Walsh, 2014).

A separate weakening concerns ontology rather than predicativity. In a modal setting of potential infinity, Hume’s Principle is reformulated as necessary across worlds,

$G$2

with $G$3 rigid across worlds. In potentially infinite models, one can define $G$4, successor, addition, multiplication, and a Fregean ancestral relation, and interpret first-order Peano arithmetic. What one cannot recover is full second-order Peano arithmetic. The result is a trade-off: weakening commitment to actual infinity weakens the mathematics recovered from Hume’s Principle (Stafford, 2021).

4. Relative categoricity and comparison with other abstraction principles

A major later development concerns the question whether Hume’s Principle determines a unique pure number structure up to isomorphism. The relevant notion is natural relative categoricity. One studies a double-abstraction theory with two operators $G$5 and $G$6 satisfying the same abstraction principle, and then asks whether the natural bijection

$G$7

induces an isomorphism between the two pure abstract structures (Walsh et al., 2014).

For Hume’s Principle, the answer is affirmative. Hume’s Principle is naturally relatively categorical, and this fact is characterized in the general theory by two equivalent conditions: injection invariance on abstracts and cardinality coarsening on abstracts. For HP the proof is immediate, because its underlying equivalence relation is itself equinumerosity; whenever $G$8, the required $G$9 already holds (Walsh et al., 2014).

The philosophical significance is that agents who each possess a number-of operator satisfying Hume’s Principle agree on pure numerical truths, even if they differ on how numbers are embedded among other objects. The paper formulates this by considering the induced structures on the ranges of the two operators, and the natural bijection preserves truth in pure numerical discourse.

This behavior is exceptional. Most other abstraction principles are not naturally relatively categorical. The paper gives counterexamples for New V, the Bicardinality Principle, and the Complementation Principle. The Nuisance Principle satisfies natural relative categoricity only in a degenerate finite-domain regime, and is incompatible with Hume’s Principle, which forces infinitude. The general moral is that Hume’s Principle occupies a special position within the abstractionist landscape: it combines logicality, strong cardinality-based invariance, stability, and determinacy of pure arithmetical truth in a way most abstraction principles do not (Walsh et al., 2014).

5. The operator $$\forall F \forall G\,\bigl(F \approx G \leftrightarrow \#F=\#G\bigr).$$0 as symbolic and cognitive technology

Recent work has extended the analysis of Hume’s Principle beyond formal logic to the cognitive realization of exact cardinality. On this view, the crucial distinction is between equinumerosity as a relation and the cardinality operator $$\forall F \forall G\,\bigl(F \approx G \leftrightarrow \#F=\#G\bigr).$$1 as a term-former. The right-hand side of HP, $$\forall F \forall G\,\bigl(F \approx G \leftrightarrow \#F=\#G\bigr).$$2, can be behaviorally manifested by one-to-one matching. The left-hand side, $$\forall F \forall G\,\bigl(F \approx G \leftrightarrow \#F=\#G\bigr).$$3, presupposes a practice that yields stable, reusable cardinal objects (Pal, 7 Jul 2026).

The Pirahã case is treated as a precise localization of this distinction. The relevant claim is not that the Pirahã lack equinumerosity. They can execute one-to-one matching where pairing can be enacted, and the reach of that matching can be extended under local training. What they lack is the operator side of Hume’s Principle: they have no counting or arithmetic, and no stable symbolic practice corresponding to $$\forall F \forall G\,\bigl(F \approx G \leftrightarrow \#F=\#G\bigr).$$4. The proposed identification is that number-word practice is the cognitive realization of $$\forall F \forall G\,\bigl(F \approx G \leftrightarrow \#F=\#G\bigr).$$5, or more generally that any stable, reusable marker preserving exact cardinal identity across absence, rearrangement, delay, or modality shift can realize $$\forall F \forall G\,\bigl(F \approx G \leftrightarrow \#F=\#G\bigr).$$6: a spoken numeral, a scratch on a stick, or a knot in a cord (Pal, 7 Jul 2026).

This recasts the number-as-cognitive-technology thesis in formal terms. The cognitive boundary is located at symbolization, not recursion. The claim is explicitly constitutive, not causal: the point is not that number words cause the existence of cardinal objects, nor that the thesis is a neural or acquisition hypothesis, but that in cognitive terms having $$\forall F \forall G\,\bigl(F \approx G \leftrightarrow \#F=\#G\bigr).$$7 consists in having a symbolic token-practice with the right identity conditions. The same framework is used to interpret converging evidence from Nicaraguan homesigners, numerate adults under verbal interference, and cross-linguistic numeral gradients. A plausible implication is that the reach of $$\forall F \forall G\,\bigl(F \approx G \leftrightarrow \#F=\#G\bigr).$$8 is graded rather than all-or-nothing, because it tracks the reach of the token practice that carries it (Pal, 7 Jul 2026).

6. Set-theoretic realizability and other uses of the label

In urelement set theory and class theory, Hume’s Principle becomes a question about whether there exists a definable map from sets or classes to objects satisfying the Humean biconditional. For sets, this is

$$\forall F \forall G\,\bigl(F \approx G \leftrightarrow \#F=\#G\bigr).$$9

where $F$0 says that there is a bijection between $F$1 and $F$2. If the urelements form a set, Scott’s trick yields a realization of HP for sets; if the Axiom of Choice for sets holds, one can define $F$3 as the least ordinal equinumerous with $F$4. But HP is not automatic in universes with urelements. There are models of $F$5 in which HP does not hold for sets, and in fact no parametrically definable map fulfills HP. There are also models of KMCU with a global well-ordering but no definable map fulfilling HP for classes. These results connect the realizability of Hume’s Principle to “the size of reality,” especially to the number and distribution of urelements and to definability constraints (Yao, 28 Aug 2025).

The label “Hume’s Principle” is also used in other ways, and this is a recurring source of confusion. In one recent historical-philosophical reconstruction, “Hume’s Principle” is not the Fregean abstraction axiom at all but a meta-level conception of the laws of reasoning: preserve them to the furthest extent possible, while allowing exceptions when reasons for violating a law outweigh reasons for preserving it. That reconstructed Humean principle is used to interpret Peacock’s principle of permanence and Hamilton’s development of quaternions (Toader, 8 Mar 2026). In another context, the phrase is used for Hume’s problem of induction, namely the challenge that there is no logical guarantee that the future will resemble the past; there the relevant topic is probabilistic confirmation rather than abstraction by equinumerosity (Yang, 2023).

In contemporary logic and the philosophy of mathematics, however, the dominant meaning remains the Fregean one. Hume’s Principle is the abstraction principle

$F$6

and its enduring significance lies in the fact that it simultaneously serves as a foundation for arithmetic in strong second-order settings, a test case for predicative and modal reductions, a benchmark for relative categoricity, and a framework for analyzing the symbolic realization of exact cardinality across formal, cultural, and cognitive domains (Walsh, 2014).