Church's Basic Recursive Arithmetic

• Church's Basic Recursive Arithmetic (BRA) is a family of formal systems defined by recursive arithmetical constructions, restricted logical tools, and an internalised coding of syntax.
• It employs quantifier-free reasoning alongside primitive recursion and minimization to enable self-referential proofs and model partial functions effectively.
• The framework bridges pure arithmetic and metamathematical analysis by integrating categorical reconstructions, while-loop formulations, and rigorous internal consistency proofs.

Church’s Basic Recursive Arithmetic (BRA) denotes, in the literature considered here, a cluster of closely related formal systems centered on recursive arithmetical definition, restricted logical apparatus, and internal arithmetisation of syntax. In Church’s Princeton logic lectures of 1957–1959, and in Guard’s 1963 lecture notes, BRA is a quantifier-free arithmetic whose deductive machinery suffices for primitive recursion, coding, substitution, and Gödelian self-reference (Coquand, 1 Jun 2026). In Pfender’s categorical reconstruction, BRA is presented as the equational or quantifier-free theory of primitive recursive functions, extended by minimization as partial $\mu$-recursion, together with a universal coding and evaluation apparatus and the $\pi$-schema for non-infinite descent (Pfender, 2013). A further “BRA viewpoint” identifies Church’s characteristic restriction with the discipline that quantifier instantiations use only basic terms built from $0$, $S$, and variables; under full first-order induction, that discipline captures exactly the PA-provably recursive functions (Makarov, 2012). This suggests that “BRA” functions less as a single fixed formalism than as a historically connected family of recursion-theoretic arithmetics sharing a common concern with denoting terms, primitive recursive definability, and internalised metamathematics.

1. Historical and terminological setting

Church’s Basic Recursive Arithmetic is described in the 2026 Agda formalisation as a quantifier-free arithmetic “presented in Church’s Princeton logic lectures (1957–1959),” and Guard’s 1963 notes rename the same system “Binary Recursive Arithmetic,” emphasizing unary and binary arities together with binary pairing for coding finite sequences (Coquand, 1 Jun 2026). The same paper states that, in that usage, “BRA” denotes Church’s system as concretised by Guard, while “Binary Recursive Arithmetic” is Guard’s name for the same system.

Pfender’s reconstruction uses different but explicitly related categorical terminology. There, the core theory is PR or PRa, the partial-map extension is $\hat{PRa}$, the universe theories are PRX and PRXa, and $\pi R$ is PRXa plus the non-infinite descent axiom $\pi$ (Pfender, 2013). The reconstruction maps this framework “explicitly to the standard understanding of BRA as the equational/quantifier-free theory of (total) primitive recursive functions extended by the minimization operator yielding partial $\mu$-recursive functions, together with a universal coding/evaluation apparatus.”

A third strand appears in the characterization of provably total functions by basic-term quantifier instantiation. That paper states that Church’s BRA “may be understood as PA formulated with $0$ and $S$ as primitive constructors and with the basic-instantiation constraint on $\pi$0-elimination and $\pi$1-introduction,” while retaining full first-order induction (Makarov, 2012). The identification is not with Guard’s quantifier-free Hilbert calculus, but with the “basic eigenterms only” policy as the hallmark of the Church-style framework.

Source	Characterization
Guard/Church	Quantifier-free Hilbert system; unary and binary arities; binary pairing (Coquand, 1 Jun 2026)
Pfender reconstruction	PR/PRa core plus partial $\pi$2-recursion, universal evaluation, $\pi$3-schema (Pfender, 2013)
Basic-term viewpoint	PA with full induction and basic-term restriction on quantifier instantiation (Makarov, 2012)

A common misconception is that the name refers to a single presentation with a stable proof-theoretic profile. The sources instead document a terminological spread: one and the same historical label is used for a quantifier-free Hilbert system, for a categorical reconstruction with partial maps and a universal evaluator, and for a first-order system whose distinctive feature is restricted instantiation. The shared nucleus is not identical syntax, but a recursive arithmetic organized around denoting numerical terms and explicit function-building schemes.

2. Formal language and deductive core

In Guard’s presentation, the non-logical signature contains the constant $\pi$4, unary successor $\pi$5, unary function symbols $\pi$6, $\pi$7, and the compound-builder $\pi$8, binary function symbol $\pi$9, and primitive recursor $0$0, together with the binary identity relation $0$1 (Coquand, 1 Jun 2026). Variables are $0$2; terms are built from $0$3, variables, and applications $0$4 and $0$5; formulas are built from equations by negation and implication:

$0$6

Numerals are $0$7.

The Hilbert calculus has fourteen axiom schemes and three rules (Coquand, 1 Jun 2026). Among the defining equations are

$0$8

$0$9

$S$0

$S$1

Equality axioms provide transitivity-style reasoning and unary and binary congruence; propositional axioms are the standard implicational and classical negation schemes. The rules are modus ponens, substitution, and induction. Rule III performs single-variable instantiation on formulas, and Rule VI states induction. Guard formulates induction for $S$2, but induction on arbitrary variables is derivable (Coquand, 1 Jun 2026).

The same source stresses that there are no object quantifiers: theorems are open formulas implicitly universally quantified over their free variables, and Rule III supplies closed instances. That quantifier-free design is central for the Guard-style formalisation of incompleteness.

Pfender’s categorical reconstruction expresses the same recursive core in terms of objects $S$3, $S$4, and finite products, together with identities, composition, projections, pairing, and Surjective Pairing, and uses the Natural Numbers Object to obtain iteration (Pfender, 2013). Initialised iteration is governed by Freyd’s uniqueness scheme (FR!), and full primitive recursion is given by the schema

$S$5

with uniqueness. Peano induction is then derived, quantifier-free, from uniqueness of primitive recursion. Arithmetic operations are introduced by primitive recursion:

$S$6

$S$7

$S$8

This deductive core matters because it isolates a recursion-theoretic foundation in which equational reasoning, substitution, and induction are already sufficient for substantial metamathematics. In the Guard line, that suffices for arithmetisation and Gödel II; in the Pfender line, it supports universal evaluation and partial-map semantics.

3. Primitive recursion, minimization, and partiality

The primitive recursive content of BRA is explicit in both the Guard and Pfender presentations. In Guard’s system, Church’s combinators $S$9 and $\hat{PRa}$0, together with pairing, are said to align BRA “with PRA” in strength and intent, while addition and multiplication are definable rather than primitive (Coquand, 1 Jun 2026). In Pfender’s reconstruction, the PR or PRa core consists of composition, pairing and product, full primitive recursion, and the equational axioms for $\hat{PRa}$1, $\hat{PRa}$2, $\hat{PRa}$3, and $\hat{PRa}$4 (Pfender, 2013).

Pfender then extends this core with minimization as a theory of partial maps. A partial map $\hat{PRa}$5 is represented by a pair

$\hat{PRa}$6

satisfying right-uniqueness: if $\hat{PRa}$7, then $\hat{PRa}$8 (Pfender, 2013). Given a predicate $\hat{PRa}$9, the $\pi R$0-operator is defined as a partial map $\pi R$1 with domain

$\pi R$2

enumeration $\pi R$3, and value $\pi R$4. Its defining properties are existence and minimality:

$\pi R$5

$\pi R$6

Pointwise, $\pi R$7 is defined if there exists $\pi R$8 with $\pi R$9, and then yields the least such $\pi$0.

The important restriction is that $\pi$1 produces partial functions in general. The reconstruction states explicitly that BRA proves totality of all primitive recursive functions, but does not generally prove totality of arbitrary $\pi$2-definitions (Pfender, 2013). The standard example is

$\pi$3

If $\pi$4 is defined, then $\pi$5 and every other witness is at least $\pi$6; but totality requires additional termination arguments.

While-loops are treated as instances of $\pi$7-recursion and partial iteration. For a loop condition $\pi$8 and a step $\pi$9,

$\mu$0

Categorically,

$\mu$1

where $\mu$2 (Pfender, 2013). The Euclidean-division example is a while-loop subtracting the divisor until the remainder is smaller than the divisor, and is partial if the divisor is $\mu$3.

A recurring misconception is that adding $\mu$4 makes BRA a theory of total computable functions. The cited reconstruction says the opposite: $\mu$5 is accommodated as a partial operator, and totality must be established separately, for example by complexity-controlled iteration and the $\mu$6-schema.

4. Arithmetisation of syntax, coding, and universal evaluation

Internal coding is one of the defining metamathematical features of BRA. In Guard’s formalisation, the meta-level encoding $\mu$7 of a designator $\mu$8 is defined inductively using pair-coded sequences and fixed tags for variables, unary and binary applications, equality, negation, and implication (Coquand, 1 Jun 2026). Guard also introduces a unary internal functor $\mu$9 mapping terms to codes of their values. For numerals, BRA proves

$0$0

for every meta-natural $0$1 (Coquand, 1 Jun 2026). The paper emphasizes that $0$2 is not the identity on numerals: $0$3. This distinction is essential in the diagonal and substitution machinery.

Substitution is likewise internalised. There is a formula-level single-variable substitution functor $0$4 and a term-level wrapper $0$5 such that

$0$6

A key lemma is numeral-inertness:

$0$7

proved internally by induction (Coquand, 1 Jun 2026). The significance is technical: nested single-variable substitutions suffice because once a numeral code has been planted, outer re-scans leave it fixed.

The same source defines a verifier $0$8 by course-of-values recursion. If $0$9 is a Hilbert derivation of $S$0, then

$S$1

On malformed inputs,

$S$2

Provability is then represented without quantifiers: $S$3 is expressed by the open formula $S$4, and consistency by

$S$5

The paper stresses that a merely “complete” but unsound internal verifier yields incorrect incompleteness statements; the validating-decoder invariant and soundness-by-construction are therefore essential (Coquand, 1 Jun 2026).

Pfender’s reconstruction supplies a more general universal coding and evaluation framework. A universal object $S$6 is defined in PRa as a PRa-subobject of $S$7 containing numerals, nested pairs, and an undefined code $S$8; the universe theories PRX and PRXa internalise syntax in a single-object setting (Pfender, 2013). The set of map codes is denoted PRX, and code constructors exist for identity, zero, successor, projections, diagonal, composition, pairing, product, and iteration. The universal evaluator

$S$9

is defined by complexity-controlled iteration of an evaluation step together with a complexity $\pi$00 (Pfender, 2013). Its characteristic equations include

$\pi$01

$\pi$02

and the recursion equation for iteration codes $\pi$03$\pi$04f:A\to B$\pi$05$\pi$06$\pi$07num$\pi$08sbf$\pi$09sub$\pi$10thmT$\pi$11X$\pi$12ev$$internalise the syntax and execution of primitive recursive maps and their iterations. In both cases, BRA is not merely a theory about numbers; it is a theory that can encode and manipulate its own formal objects.</p> <h2 class='paper-heading' id='soundness-incompleteness-and-consistency'>5. Soundness, incompleteness, and consistency</h2> <p>The metamathematical reach of BRA is clearest in the incompleteness results. In Guard’s Agda formalisation, the Gödel sentence is built from the seed formula</p> <p>$$\pi$13

followed by the substitutions

$\pi$14

$\pi$15

$\pi$16

and BRA proves the diagonal identity

$\pi$17

(Coquand, 1 Jun 2026). The system then proves the fixed-point equivalence in the precise $\pi$18 form needed for the formal argument.

The main theorem is stated meta-level as follows: if BRA proves its own consistency schema $\pi$19, then BRA proves $\pi$20 (Coquand, 1 Jun 2026). Equivalently, if BRA is consistent, it does not prove $\pi$21. The proof is carried out entirely within the quantifier-free Hilbert calculus and uses the verifier’s closure properties, encoded modus ponens, encoded substitution, predicate-Leibniz, and the internal diagonal construction. The paper explicitly states that no $\pi$22-consistency or $\pi$23-consistency assumptions are needed beyond the Hilbert-derivability machinery.

Pfender’s reconstruction develops a related but distinct soundness-and-consistency theory. Internal PRX-equality soundness takes the form

$\pi$24

and objective soundness says

$\pi$25

(Pfender, 2013). Logical soundness is phrased as: if there exists a PRX-proof $\pi$26 of a PRa-predicate $\pi$27, then $\pi$28. Within PRXa, the key internal result is termination-conditioned soundness: if an internal equation $\pi$29 has a deduction tree whose evaluation terminates, then $\pi$30.

The constructive consistency claim is stronger than a bare external consistency assertion. Pfender states that $\pi$31 proves

$\pi$32

so there is no internal proof of false in PRX (Pfender, 2013). The proof uses the $\pi$33-schema to exclude infinite descent in deduction-tree evaluation. In the paper’s terminology, “constructive consistency” is achieved for the PRX-universe theory framed by $\pi$34 without recourse to external classical quantifiers.

The literature also records a cautionary episode. The initial autonomous attempt reported in the Agda paper used an insufficiently specified internal provability predicate based on a false theorem of Rose and produced a statement superficially resembling Gödel II but mathematically unrelated to it (Coquand, 1 Jun 2026). The point is substantive, not anecdotal: in BRA, the exact form of the internal verifier and its soundness clauses is not dispensable bookkeeping but part of the theorem’s content.

6. Proof-theoretic relationships and scope of the theory

Comparisons with neighboring systems depend on which BRA presentation is meant. In Pfender’s reconstruction, BRA is “strictly weaker than PA” because it lacks full first-order quantifiers and induction schemas over arbitrary formulas, but stronger than pure PR because it admits $\pi$35 and partial functions (Pfender, 2013). The same source notes that RCA$\pi$36 is not treated and is generally stronger than BRA in that setting, while proof-theoretic ordinals for PR, PRa, and $\pi$37 are not specified.

The Agda formalisation places Guard’s BRA above Robinson’s $\pi$38 and below $\pi$39 in a different sense. $\pi$40 is strictly weaker because it has quantifier-based axioms for addition and multiplication without induction; $\pi$41 is stronger because it has first-order quantifiers and $\pi$42-induction (Coquand, 1 Jun 2026). Nevertheless, Guard-style BRA suffices to arithmetise syntax, build a provability predicate via $\pi$43, establish the diagonal lemma, and prove Gödel’s Second Incompleteness Theorem.

The basic-term interpretation yields yet another comparison. The characterization theorem for “Provably Total Functions of Arithmetic with Basic Terms” states that a function is provably recursive in A(PR), equivalently PA, if and only if there exists a strongly coherent full program $\pi$44 such that

$\pi$45

(Makarov, 2012). Here $\pi$46 is the natural deduction system in which the eigenterms for $\pi$47-elimination and $\pi$48-introduction must be basic terms:

$\pi$49

The paper explicitly identifies this restriction as the hallmark of Church’s BRA and concludes that, for totality assertions of the form $\pi$50, BRA-style proofs are equivalent to PA proofs. It follows that the basic-instantiation discipline does not reduce the class of provably total computable functions when full induction is retained; the class remains exactly the PA-provably recursive functions, including non-primitive-recursive examples such as Ackermann (Makarov, 2012).

Taken together, these results show that BRA occupies several neighboring but non-identical proof-theoretic niches. In the quantifier-free Guard and Pfender line, it is a compact arithmetic of recursion, coding, and internal meta-reasoning. In the basic-term first-order line, it is a disciplined formulation of PA whose restriction falls on instantiation rather than induction. The stable conceptual theme is that recursive arithmetic can be made metamathematically expressive without granting unrestricted term formation free rein in proof rules.

7. Later formalisation and enduring significance

The 2026 Agda development provides a fully machine-checked formalisation of Gödel’s Second Incompleteness Theorem for BRA, with approximately 50,000 lines, no postulates, and trusted base exactly the 14 axioms and 3 rules of the object theory (Coquand, 1 Jun 2026). It reconstructs points that Guard left implicit: the internal numeral-encoding operation $\pi$51, the interaction between substitution and closedness, the predicate-Leibniz principle, the validating-decoder invariant for $\pi$52, and the use of nested single-variable substitutions in place of genuine simultaneous substitution. The implementation realizes course-of-values recursion inside Church’s non-standard recursor by a nested-$\pi$53 trick together with fuel-stability lemmas (Coquand, 1 Jun 2026).

Pfender’s reconstruction gives BRA a different kind of modern significance. By embedding primitive recursion, partial $\pi$54-recursion, universal coding, universal evaluation, and complexity-controlled iteration into a single categorical framework, it presents BRA as a self-contained foundation for recursion theory, internal syntax, and constructive consistency arguments (Pfender, 2013). The resulting picture connects traditional computability-theoretic notions—least-number search, partial functions, while-loops, s–m–n, and Kleene fixed points—to categorical formulations of maps, products, subobjects, and evaluation.

The basic-term characterization of provably recursive functions adds a further significance criterion. It shows that the Church-style requirement that quantifier instantiations use only numerals and variables is not a merely stylistic restriction but a semantically meaningful control on denotation: unrestricted instantiation can trivialize totality assertions, whereas basic instantiation forces quantification to track terms guaranteed to denote natural numbers (Makarov, 2012). In that precise sense, BRA isolates a boundary between genuine totality proofs and pseudo-proofs driven by non-denoting program terms.

Across these strands, Church’s Basic Recursive Arithmetic appears as a compact setting in which recursive definition, syntactic coding, and metatheoretic self-application can be studied with unusual explicitness. Its enduring role is not that it eliminates incompleteness or collapses distinctions between total and partial computation, but that it exhibits how much of arithmetic, computability, and internal proof theory can be carried by a deliberately austere recursive apparatus.