Gottlob Frege

Gottlob Frege

Logician and philosopher of mathematics · 1848–1925 · Germany

Make the dependencies in a statement visible

Frege's Begriffsschrift (1879) introduced a powerful formal treatment of quantification, relations and inference. Its importance is not that earlier logic had been motionless since Aristotle. Boole, De Morgan, Peirce and others also transformed logic. Frege developed a distinctive system aimed at expressing and proving mathematical claims with their dependencies made explicit.

His original notation is two-dimensional, and his logical system includes higher-order resources. A present-day first-order formula is therefore a useful translation for teaching, not a facsimile or complete characterization of his system. Original edition record, technical account of the two systems.

Historical orientation

Frege was born in Wismar on 8 November 1848. After studying in Jena and Göttingen, he completed his doctorate in 1873 and returned to Jena, where he taught for most of his career. His principal landmarks include Begriffsschrift (1879), The Foundations of Arithmetic (1884), “Function and Concept” (1891), “On Sense and Reference” (1892), and the two volumes of Basic Laws of Arithmetic (1893 and 1903). Biography, chronological catalog.

Russell communicated the contradiction in Frege's foundational system in 1902, while the second volume was in production. Frege attempted a repair and continued to teach and publish afterward. It is inaccurate to say that the letter ended all his intellectual work. His 1924 diary contains antisemitic and fascist sympathies. Those documented views should be acknowledged without inventing a psychiatric explanation or presenting mathematical achievement as an excuse. Historical account.

Syntax, interpretation and inference

Consider Reviewer(x) and Reviews(x,y). The first is a one-place predicate; the second is a two-place relation. An interpretation supplies a domain of objects and specifies which objects or pairs make these predicates true. The formula is not made true merely by writing it in precise notation.

The sentence “Every paper has a reviewer” might be formalized as

∀p (Paper(p) → ∃r (Reviewer(r) ∧ Reviews(r,p))).

The witness r may depend on p because its existential quantifier lies inside the universal quantifier's scope. In contrast,

∃r (Reviewer(r) ∧ ∀p (Paper(p) → Reviews(r,p)))

requires one reviewer who reviews all papers. Moving the existential quantifier changes the claim, not just its typography. The symbols ∀ and ∃ here are modern conventions; Frege used different signs.

A formal proof also needs inference rules. From ∀x P(x) one can infer P(a) for an appropriate term a. The reverse inference is not justified just because P happens to hold for one named object. Generalization requires that the object be arbitrary and that the proof's undischarged assumptions do not improperly single it out.

Worked mechanism: a small countermodel

This is an original modern example. The domain contains two reviewers, Ada and Ben, and two papers, P and Q. Assume the relation contains exactly these pairs:

Reviewer	Paper reviewed
Ada	P
Ben	Q

Each paper has a reviewer. For P choose Ada; for Q choose Ben. Thus the first formula is true. Neither reviewer reviews both papers, so the second formula is false. This one interpretation is enough to disprove the proposed implication from the first formula to the second.

The implication in the other direction does hold. If a single reviewer reviews every paper, that same reviewer can be used as the witness for each individual paper. Notice the difference between constructing a countermodel and constructing a proof: the countermodel refutes a universal entailment, while the proof establishes it for all interpretations satisfying its premises.

Empty classes matter. If there are no papers, “Every paper has a reviewer” is vacuously true even if there are no reviewers. If the application requires papers or reviewers to exist, state those existence requirements. Formalization exposes a product requirement; it does not decide it for the designer.

Numbers and a failed abstraction principle

Frege's logicist project aimed to explain arithmetic through logic. One important idea relates number equality to a bijection: the number of F-objects equals the number of G-objects exactly when the F-objects and G-objects can be paired one-to-one. This is commonly called Hume's Principle.

Basic Law V is different. In simplified modern notation, it associates an object, an extension, with a concept and identifies two such extension objects exactly when their concepts agree on every argument. Combined with the expressive comprehension resources of Frege's system, it leads to contradiction. Calling Basic Law V simply “unrestricted set comprehension” loses that distinction. There are weakened systems in which related abstraction principles behave differently. Frege's Theorem and the faulty foundation.

A modern set-theoretic presentation displays the danger. Suppose every condition defines a set. Then take R = {x : x ∉ x}. If R ∈ R, its defining condition says R ∉ R. If R ∉ R, the same condition says R ∈ R. Both cases contradict themselves. This is a teaching translation of the paradox, not Frege's literal syntax.

The conclusion is about the inconsistent system. It does not refute all predicate logic, every analysis of number, or every result obtained using some of Frege's ideas. In particular, the later study of Frege's theorem separates what can be obtained from Hume's Principle from the inconsistent derivation of that principle using Basic Law V.

Sense and reference: a careful comparison

Two expressions may refer to the same object while presenting it differently. Frege used this distinction to explain why an identity can convey information. It is not simply the distinction between a pointer and its pointee: both a pointer value and an object address can be discussed without resolving Frege's philosophical account of sense.

For software, the safe comparison is narrower. Different expressions can produce the same value, but a compiler cannot therefore replace one with the other in every context. Evaluation may have effects, fail, diverge or depend on state. Those are operational issues requiring a language-specific equivalence relation. They are not automatically answered by citing “sense and reference.” Frege's semantic distinctions.

Exercises

1. Formalize “Every user can access some document” and “Some document can be accessed by every user.” Give a finite model making the first true and the second false, and state your nonempty-domain assumptions.
2. Explain why observing Approved(Ada) does not establish ∀x Approved(x). Contrast this with a proof in which x is introduced as an arbitrary user and the conclusion follows without assumptions specific to x.
3. A specification permits a set for every property, including self-membership tests. Reproduce the contradiction for R above. Then explain why restricting set formation to subsets of an already supplied set changes the argument. Do not claim that this single change proves the consistency of a complete set theory.

Compare your reasoning with the worked solutions

1. Quantifier order

The formulas are ∀u(User(u) → ∃d(Document(d) ∧ Access(u,d))) and ∃d(Document(d) ∧ ∀u(User(u) → Access(u,d))). Use two users U and V and two documents D and E, with only Access(U,D) and Access(V,E). Every user has a witness, but neither document works for both users. The domain is nonempty, and both relevant classes are nonempty. If User were empty, the first universal condition could hold without any accessible document, so those assumptions should remain explicit.

2. An instance is not a universal proof

A countermodel with Ada approved and Ben unapproved preserves the observed fact but refutes the universal conclusion. By contrast, if x was arbitrary and a derivation proves approval using only assumptions valid for every user, no special feature of x was used. Universal generalization can then be justified under the proof system's conditions. Calling a name “arbitrary” after exploiting a fact specific to that name does not repair the inference.

3. Restricted separation

Unrestricted formation yields R and therefore R ∈ R ↔ R ∉ R. With separation inside a supplied set A, one obtains only R_A = {x ∈ A : x ∉ x}. The definition gives R_A ∈ R_A ↔ (R_A ∈ A and R_A ∉ R_A). The argument shows that R_A cannot belong to A; it does not force R_A ∈ A. The missing universal set blocks this particular paradox. Consistency of an entire formal theory is a broader question requiring analysis of all its axioms.