Rozeboom Rozeboom

You are, I am sure, just as aware as I am that the operational nodes of a complex problem, the points at which it can be split open to yield nuggets of new insight or achieve lasting advances, often lie in tediously technical details perhaps incomprehensible to all but specialists in the matter and anyways totally lacking in the romance and easy excitement which attract the topic's dilettantes. I would like you to hold fast to this appreciation, for the concerns I shall raise here are technical indeed. They are, however,seminaltechnicalities which, as we seek to fathom them, should fundamentally reshape our comprehension of the logic of confirmation, causal connectedness, and the foundations of statistical inference. In brief, it will be seen that certain powerful everyday intuitions concerning which propositions are confirmationally irrelevant to which others are exceedingly difficult to justify.

As anyone who has ever seriously attempted to analyze the semantico-epistemological status of scientific theories has soon discovered, it is not easy to reconcile the belief that theoretical terms have genuine cognitive properties with the empiricist tenet that all knowledge derives from experience. Even if it be granted that knowledge can originate in experience without being about experience, it still remains to develop a coherent metalinguistic account of the truth-conditions of theoretical propositions and the designata of denotative expressions in the theory language. In a recent article, “On the use of Hilbert's e-operator in scientific theories” [1], Carnap has broadened his analysis of theories to include provision for the referential properties of individual theoretical terms. But while his account focuses more closely than ever upon the fundamental semantical problems of a theory language, his specific proposals amount to a repudiation of the very logical empiricism which he has so strongly championed these many years. In what follows, I use Carnap's notation and terminology throughout.

In a recent note, Horwich challenges the foundations of Hempel's classic paradox of confirmation by a clever example purporting to show that under Nicod's Criterion, data can be made to confirm a hypothesis with which they are logically incompatible. Specifically, Horwich observes that 'Pb' is formally equivalent to ''. The latter has form '' with '∼P___ · ∼Pb' for 'Ψ' and '___ ≠ b' for 'ϕ', while the observation that distinct objects a and b both lack P, i.e. that —Pa · ∼Pb · a ≠ b, can be expressed as 'Ψa · ϕa' for these same instantiations of the predicate markers. Accordingly, if an uncertain generality '' were always to be confirmed by an observation of form 'Ψa · ϕa', as Nicod's Criterion has long been presumed to say, then we could confirm that b has P by observing that b and some other object both lack P—a flagrant absurdity.

Part I is concerned with the tenet of modern Emperical Realism that while the theoretical concepts employed in science obtain their meanings entirely from the connections their usage establishes with the data language, the referents of such terms may be "unobservables," that is, entities which cannot be discussed within the data language alone. Such a view avoids both the restrictive excesses of logical positivism and the epistemic laxity of transcendentalism; however, it also necessitates a break with classical semantics, for it follows from the empirical realistic position that a theoretical term must in principle be capable of simultaneously designating not just one entity, but indefinitely many. Drawing upon the Carnapian explication of "analytic truth," Part II examines a possible axiomatic basis for the empiricist theory of scientific meaningfulness to demonstrate that even if theoretical terms are able to designate entities inaccessible to the observation language, as held by Empirical Realism, so long as the meanings of theoretical terms derive from their connections with the observation language, the meaning content of a theory is exhausted by its observational consequences