Inductive logic, run as arithmetic
Carnap's Logical Foundations of Probability (1950) fixes a language by its atomic sentences; a state-description is one complete assignment of truth-values to those atoms, and a proposition's range is the set of state-descriptions in which it holds. Every notion below is a fact about ranges: confirmation c(h, e) = m(h & e) / m(e), analyticity is range-equals-all, and the verifiability criterion sorts a sentence by whether it has confirmation conditions at all. Distilled from the Chapter 9 book pass and the c-function test suite.
Confirmation is a number you can compute
Propositions are written as func(sd) predicates over a declared vocabulary. A hypothesis confirms itself completely (c(h, h) = 1); given p, the confirmation of p & q is exactly 0.5 — half the p-worlds also have q. That is a Bayesian update, run to the digit.
The lambda-continuum: does the method learn?
The deepest move in The Continuum of Inductive Methods (1952): one parameter tunes how much evidence moves belief. The logical measure m-dagger weighs every state-description equally, so evidence is inert — two positives leave the third at the prior 0.5. Carnap's preferred c* (lambda = 2) instead learns, raising it to 0.75; as lambda climbs it slides back toward m-dagger.
Relevance: evidence that confirms, disconfirms, or does neither
Carnap's relevance is the shift c(h, e) − c(h): positive means e supports h, negative means it undercuts it, zero means it is idle. Because p and q are independent atoms, learning q tells you nothing about p — relevance is exactly 0. This is the graded successor to a yes/no verification.
Analyticity is a fact about range
In Meaning and Necessity analyticity is not a mystery — it is range-equals-all. A tautology holds in every state-description (l_true), a contradiction in none (l_false), and deductive entailment is range-inclusion (l_implies). Confirmation theory subsumes deduction as its limit case: when e entails h, c(h, e) = 1.
The Scheinsatz test: meaningless, not false
The verifiability criterion sorts every sentence into three bins: analytic (L_true, confirmation 1.0), synthetic (meaningful, confirmation 0.5), or — with no verification conditions in either bin — a Scheinsatz, Carnap's pseudo-statement. "The Absolute is incapable of evolution" is not false; it makes no claim that could be true or false. diagnose runs the classifier.
Take confirmation further
This is the machinery Ayer's half of the Mises study runs on. See it adjudicate a live philosophy-of-economics dispute — or fork the c-function suite, swap the language, and watch the digits move.