STATHIS PSILLOS CARNAP, THE RAMSEY-SENTENCE AND REALISTIC

ABSTRACT. Based on archival material from the Carnap and Feigl Archives, this paper re-examines Carnap’s approach to the issue of scientific realism in the 1950s and the early 1960s. It focuses on Carnap’s re-invention of the Ramsey-sentence approach to scientific theories and argues that Carnap wanted to entertain a genuine neutral stance in the realism-instrumentalism debate. Following Grover Maxwell, it claims that Carnap’s position may be best understood as a version of ‘structural realism’. However, thus understood, Carnap’s position faces the challenge that Newman raised against Russell’s structuralism: the claim that the knowledge of the unobservable is limited to its purely structural characteristics is either uninformative or unsustainable. 1. LIBERALISATION By the early 1950s, empiricists have fully acknowledged that theoretical discourse has ‘excess ’ or ‘surplus ’ meaning. This idea is reflected in the standard empiricist way to represent a scientific theory: Carnap’s “The Methodological Character of Theoretical Concepts ” (1956) – henceforth MCTC. There, Carnap advances a general logico-linguistic framework L in which scientific theories can be developed. The total language of science is divided into two sub-languages: an observational language LO which is completely interpreted, and a theoretical language LT whose descriptive vocabulary VT consists of theoretical terms. The variables of LO range over concrete observable things and their domain is finite. The theoretical language LT is much richer: it contains a type-theoretic logic with an infinite sequence of domains D0, D1, D2,..., where Dn is the domain of the n-th level.1 Each variable and each constant belong to a definite level. D0 comprises the infinite sequence O, O 0, O 00,..., which can be thought of as the domain of natural numbers. Then the domain of each DnC1 is the domain of all subclasses of Dn. LT contains the whole of classical mathematics, i.e., expressions and variables for all objects of classical mathematics. Such strong language has a certain theoretical advantage: all phys-ical concepts occurring in theories can be shown to be represented b