One of the chief playgrounds of the new, supposedly Carnapian, metaontology has been Carnap’s distinction between internal and external questions in his “Empiricism, Semantics, and Ontology” (ESO). David Chalmers (in the 2009 Metametaphysics volume, pp. 80-85), for instance, acknowledges that there is “something natural” about the distinction as it “seems to reflect a distinction in our practice of raising questions about existence.” However, he thinks Carnap’s terminology is “suboptimal” as it is “too closely tied to Carnap’s theoretical apparatus involving frameworks to serve as a neutral starting point.” So he proposes a “relatively pretheoretical” replacement for the internal-external distinction “that almost anyone can accept, regardless of their theoretical inclinations” (p. 80). His proposal distinguishes not between internal and external, then, but between ordinary and ontological existence assertions. What Chalmers has in mind when considering this pair to be less “theoretical” and more “neutral” (or perhaps less burdened by philosophical baggage) than internal and external, doesn’t emerge here; I may come back to this in a later post.
Whatever aspects of Carnap’s distinction this proposed replacement reproduces, others are left out. It was one of Carnap’s priorities, for instance, to distinguish two fundamentally different kinds of languages (which I’ve sometimes called “constructed” and “evolved” languages; Carnap’s own terminology varied with time and context). Internal questions were those that could be answered deductively or procedurally in explicitly constructed languages. External questions arise in natural (or evolved) languages, specifically in a natural metalanguage used in the employment of a constructed object language. These aren’t language-relative, i.e. internal, questions to which answers are deducible or procedurally attainable within the object language but wider-ranging, intuitively plausible questions about e.g. existence or truth in general, without reference to any particular constructed object language. These become, of course, the metaphysical questions that have preoccupied philosophers since Plato. The later Carnap did not exclude them altogether, but suggested that we re-interpret them, and treat them, in practice, as questions about which language to use (or which explication to prefer). This put them in the appropriate light, he thought, and turned what had been intractable doctrinal questions into practical ones about the costs and benefits of different languages or alternative clarifications of a concept. Like Poincaré, he thought questions of language choice extremely important, indeed fundamental (Schilpp volume, p. 862).
Carnap’s objective, then, was precisely to rule out the specifically “ontological” use (in Chalmers’s terms) of external questions and external discussion, and replace it by discussion of the costs and benefits of different languages, or more usually, of different proposed explications for a vague explicandum (“pragmatics,” he called this realm of discussion). So the mature Carnap admitted both descriptive and normative discourse. Descriptive sentences could be true or false, and must be internal; normative ones could not be true or false, though they could contain true or false descriptive components, and were external. Chalmers’s proposed replacement distinction leaves all this aside and makes nonsense of Carnap’s original terminology. Chalmers’s internal sentences have nothing to be “internal” to since they aren’t conceived as framed in a constructed language; it’s only with respect to a framework of such languages that the “internal” status of a sentence is well-defined. For Chalmers, though, “ordinary” ontological questions and claims need not be “internal” in any such well-defined sense. They are held to be “correct” (which needn’t mean “true,” p. 82) relatively to “commonsense ontology” which is committed to ordinary objects and ordinary (including mathematical) reasoning but not “distinctively ontological truths and distincively ontological reasoning.” Thus “commonsense ontology appears to be committed to particles, ordinary macroscopic objects such as tables, and at least some abstract objects such as numbers, but it does not appear to be committed to more recherché objects such as arbitrary mereological sums.” (p. 86) But where is the boundary between the ordinary and the recherché? Doesn’t this differ, even from person to person? At very least, then, Chalmers’s distinction between ordinary and ontological questions is extremely vague, and certainly much vaguer than the Carnapian distinction it is intended to reconstruct in more neutral terms.
Actually, though, it’s less neutral than Carnap’s distinction, in that in invokes standards of “correctness” for ontological claims (ibid., p. 82). Ordinary existence statements are insensitive to ontological standards of correctness, Chalmers says, while ontological existence statements are sensitive to them – they are correct (or not) relative to ontological doctrines, such as realism or nihilism. And while Chalmers does not equate “correct” with “true,” the ontological doctrines with respect to which claims are judged correct or incorrect are considered true or false. The point of Carnap’s distinction, however, had been precisely to undercut the traditional interpretation of such doctrines as true or false. He regarded them as ill-posed, underspecified, the epitome of the kinds of things ordinary language doesn’t have the resources to deal with (Schilpp volume, pp. 938-9). Chalmers’s proposed replacement for Carnap’s distinction leaves all this out.
Carnap was neither an anti-realist nor a meta-(ontological) anti-realist. Far from denying that “there are objective answers to the basic questions of ontology” (Chalmers p. 77), he thought such questions unintelligible. Chalmers, though insisting that what he is exploring is a “Carnapian variety of ontological anti-realism,” is “not certain that it is true” (p. 79). Carnap didn’t think that his view could be true (or false).
UPDATE (13 April): This post now has a sequel.
UPDATE (9 May): And another one.