During the past few years a lot of rubbish has been circulating about Carnapian frameworks. I have been watching this infestation with dismay, but so far addressed it only occasionally, e.g. here with respect to Chalmers, or here with respect to Eklund (and that was a while ago). I’m very glad to see that someone has now decided that enough is enough, and sprayed some serious ant killer on this irruption of philosophical insect life. Bravo to Gabriel Broughton for having taken on this unpleasant task in his new paper “Carnapian Frameworks” (Synthese)!
The paper is narrow in scope — a little too narrow, some might argue — but perhaps that is the cost of addressing one particular misconception about frameworks (Eklund’s) with the thoroughness required to show just how threadbare these new clothes actually are. And the paper does, in fact, reveal this particular little emperor to be completely naked. After trying charitably to spell out Eklund’s conception as precisely as possible, it shows first that Carnap can’t have meant frameworks to be natural languages, since of course he didn’t think anything could be clearly defined as “internal” or “external” to a natural language. Then, the paper shows that Carnap thought of frameworks as formal semantical systems; finally, it deals briefly with Carnap’s supposed “ontological pluralism” (which I have also addressed, here and here, from a slightly different angle).
Of course I have a few minor quibbles. Broughton says, for instance, that “Carnap clearly considered all normative questions to be external questions. Semantical systems, after all, were explicitly designed to deal exclusively with ‘cognitive’ claims.” So how are normative languages to be characterized, such as those Carnap lays the groundwork for in his reply to Kaplan and urges us to go on with (pp. 1012-13), and in fact made a start on himself in his fragmentary continuation of that reply published in Synthese a few years ago? In both of those texts, Carnap introduces the normative concepts model-theoretically — i.e. as equipped with a semantics — so why on earth would he disqualify the system he sketches there from being “semantical”? The normative component of a language presumably belongs to pragmatics, in Carnap’s classification, but that doesn’t mean that a sentence with a pragmatic component doesn’t also have a syntax and a semantics — just as a sentence with a semantic component doesn’t lack a syntax.
Broughton also claims that “Carnap agreed with Quine that [the internal/external distinction] was tied up with [the analytic/synthetic distinction] (e.g., Carnap1950a, p. 32 fn.2). While it’s always possible that Carnap somehow misunderstood his own views, surely, all else being equal, we should prefer an interpretation that avoids this result.” The reference is to a footnote in ESO where Carnap says that “Quine does not acknowledge the [internal/external distinction], because according to his general conception there are no sharp boundary lines between logical and factual truth, between questions of meaning and questions of fact, between the acceptance of a language structure and the acceptance of an assertion formulated in the language.” Carnap acknowledges that Quine thinks this, in other words, i.e. that in Quine’s own view (to which he takes exception), the denial of a distinction between logical and factual truth (etc.) has the consequence that Quine also rejects the distinction between internal and external questions — but he doesn’t endorse this view of Quine’s. Also, Carnap allowed that a language without analytic sentences could be constructed (though of course it would be pretty useless for just about any purpose we construct languages for), and a framework based on such a language would still have “internal” and “external” sentences (and one could ask the corresponding questions). So Broughton’s presumption that Carnap agreed with Quine that “the” analytic-synthetic distinction underlies the distinction between internal and external questions is mistaken.
(Others seem also to regard it as at least debatable; Gary Ebbs recently devoted a paper to the question whether the analytic-synthetic distinction depends on the internal-external distinction, and if so, in what way.)
Finally, the paper cuts some corners in its discussion of Carnap’s empiricism. Referring to the ESO passage where Carnap says that those who take external questions at face value have been unable to articulate them (or their answers) in “the common scientific language,” Broughton asks how this should be interpreted. His answer is to remind his readers that ESO is addressed to empiricists, specifically empiricists skeptical of abstract objects; so Carnap “effectively equates scientific language with empiricist language.” Not quite. First of all, the language of science — as this phrase was ordinarily employed among logical empiricists — is an object language; empiricism is a possible (constraint on the) common metalanguage to be employed when talking about scientific knowledge, how its components fit together, etc. (see “Testability and Meaning,” also Ricketts’s paper in Reading Putnam). Carnap takes for granted that the language of science itself does not contain synthetic a priori statements (though he’d probably have been willing to concede that, as a matter of descriptive pragmatics, this is ultimately an empirical question), so this wouldn’t have been a requirement for the language of science, even in the 20s.
Of course back then, the Vienna Circle was still fixated on the Tractatus, and took very seriously the idea (which they thought they saw in it) that the language of science had to be finitist and (as Carnap would later put it) “molecular,” despite obvious appearances. Carnap spent years trying to figure out how to reduce axiomatic systems to this kind of molecular form. And it was this effort that ultimately led him to his “sleepless night” of the syntax breakthrough and, a year later, to the principle of tolerance. At that point, the whole idea of “the” language of science became moot, and Carnap engaged in serious dialogue with Neurath about what the common language of (and metalanguage for) science should be, not just ideally, but in actual practical life. Unfortunately they got hung up on arguing over whether semantics should be added to the metalanguage (Neurath wanted to stick with just syntax), and Neurath died before they had the chance to sort this out. But both I (from the Carnap side) and Thomas Uebel (from the Neurath side) think a rapprochement between their language-of-science conceptions would have been (and remains) possible.
So what did Carnap mean by “the common language of science” in ESO? It would seem odd for him to have meant the object language of science, in the old Vienna-Circle sense, since all meta-scientific, meta-knowledge questions are in a metalanguage. It seems more plausible, rather, that he meant something like the (meta-scientific) “scientific vernacular” Neurath had been programmatically advocating (“universal slang,” he called it), as the basis for the “encyclopedic” unity of science — or perhaps some imagined fusion between that and Carnap’s own preferred deductive model of the unity of science (a fusion both Thomas and I think ultimately feasible). Would that metalanguage be empiricist? Of course! So Broughton isn’t exactly wrong; but empiricism is only one constraint among many — simply equating “the language of science” with “an empiricist language” sweeps too much under the rug.
All that said, Broughton’s discussion of Carnap’s empiricism isn’t really too far off the track; it’s just a bit oversimplified, which might even, in the context of this paper, be perfectly justified. A signpost about the hidden depths would have been helpful, though — not just in this case but also in the two others I’ve mentioned. The absence of such signposts hardly detracts, though, from an otherwise very thorough and conscientious refutation of an insidious (since superficially innocuous and thus hard to flush out) Carnap misinterpretation. It’s good to see younger people getting their elders out of the ruts they’ve so perversely managed to get stuck in. It makes me feel much better about the state of the field.
PS: Eklund has (somewhat belatedly) replied to Broughton, and I’ve commented briefly here.