Most of you are probably vaguely aware that Kant turned 300 on Monday, and a few of you have perhaps even heard that the German Federal Chancellor Olaf Scholz gave a speech commemorating the occasion. Now I’m not a great Scholz fan, for the most part (who is? — he and his coalition have reached lows in the polls that even Biden and Trump can only dream of), but I have to admit that this speech is really first-rate. If you read about it in the news anywhere, you probably assumed it was the usual politician thing — bumbling around trying to sound as if he had some idea what he was talking about and coming out sounding like a complete fake. Admittedly, that’s what usually happens. And you can’t imagine an American (or perhaps even a British) politician trying anything like this. But I have to admit that this speech really works. It lends weight and credibility to stuff that Scholz has been saying constantly (and too repetitively, too phlegmatically) since February 2022; the invocation of Kant is appropriate and measured and not at all weithergeholt (as they say here). So it’s not only unusually educated for a politician, it’s also rhetorically effective and politically smart. I urge you to read the whole thing.
Unfortunately I haven’t been able to find an English translation yet, and haven’t had time to provide one myself. If anyone finds a decent one, could they let me know? I’ll link that as well. (Or if enough people ask, and I still can’t find one, I might even translate it.)
Since we’re celebrating our old friend Immanuel, I’d like to call for a reevaluation (perhaps: ‘challenge’ is the more appropriate word) of the Carnapian program that you’re promoting here by referencing some history of German neo-Kantianism, broadly conceived, and try to relate it to development of Carnap’s metaphilosophical views.
It is well-known, especially thanks to Friedman’s scholarship (not because Friedman was the first to notice this or because his reading of Kant is correct, but because of the lack of any serious Anglophone Kant scholarship before people like Friedman), that Kant’s conception of synthetic a priori principles was supposed to reference not an epistemic, but a semantic (perhaps metasemantic), kind. It is through the Gültigkeit of a priori truths of intuitive geometry and of regulative principles coordinating our use of relational categories (causality) – corresponding to the Aesthetic and the Analytic of the first Critique – that perception and Erfahrung proper become possible (apologies for trying to imitate Kant’s own terminology). In other words, transcendental principles articulate the conditions under which contentful reference to reality is possible at all (even if these principles are limited to the empirically established truths of sense psychology regarding the structure of our internal perceptual spaces, like a popular misreading of Helmholtz holds).
A lot, however, can be established to follow deductively from such seemingly thin truths, as Kant demonstrates in his Critique of Pure Reason and later in his Metaphysical Foundations of Natural Science (even if, as is the case, his dynamic theory of matter is hopeless). This has motivated many thinkers, notably Helmholtz, to reevaluate the claim to universal validity of some cognitive norms that Kant asserts to be certainly valid, especially regarding intuitive geometry (because, as we know, Euclidean geometry isn’t really uniquely discerned by any physical fact among the class of possible geometries of constant curvature, even if it corresponds to our natural mathematical inclinations). Whether Kant is correct in assigning a particular role to the propositions that he does is thus very uncertain (or rather: was, before the development of modern physics, when Helmholtz and other neo-Kantians wrote about these issues) but it doesn’t change that certain truths must hold in order for our thoughts to acquire a meaningful relation to reality or, in Kant’s terminology, be able to constitute experience. In a slight paraphrase of Helmholtz’s words, ‘space might be transcendental, but axioms of geometry needn’t be’.
This motivates two distinctive responses: a conventionalist one (Poincare, Marburg school of neo-Kantianism, early Schlick) or a Bild-theoretic one (Helmholtz himself and his star-pupil Hertz, who eventually influenced Wittgenstein’s Tractatus, although the majority of physicists at the end of nineteenth century were in fact Bild-theorists). I reccommend reading Lydia Patton’s “Signs, toy models, and the a priori” for a brief overview of this topic. Either the normative constraints which guarantee that our representational systems bind with reality that Kant references under the name of synthetic a priori judgements can be stipulated arbitrarily as conventions or the a priori is relativized to a class of models that have only a restricted claim to validity (I am purposefully not considering ‘historicizing’ the a priori, which was also a part of the Marburg program, because this has deeper motivation than the problem of ‘coordination’ discussed by the first logical positivists, i.e. Reichenbach and Schlick).
But the first solution entirely avoids confronting Kant’s central point. It is not just the fact that mathematical concepts or symbolism of the physical theory is initially inhomogeneous with the appearances that is the issue (which is what Schlick seemed to think). The problem is with, to use Tractarian terms, the validity of our assumptions about geometry of the logical space (and the related geometry of physical space-time) itself. The internal manifolds of sense physiology that Helmholtz studied as a natural scientist are the very same manifolds of Kant’s Transcendental Aesthetic and thus conditions of possibility of experience. But they cannot be, contrary to Kant’s expectations, studied introspectively. And the stipulations present in Carnap’s Aufbau are completely indifferent to the problem. A reconstruction of a world based on 2D sense-data is not impossible, as Quine claimed, but pointless unless we map our stipulations back onto the manifold that has a objective, physical ground.
The essential founding myth, I think, of the syntactic program inaugurated by left-leaning members of the Vienna Circle, especially Carnap and Neurath, in the 1930s was that the possibility of metalinguistic analysis was recognized by them despite Wittgenstein’s constant assertions against it (they claimed: thanks to Tarski and other Polish logicians). But the truth is that the semantic account of symbolism that treats sentences as picture-models standing in an relation of representation (isomorphism) to an array of atomic facts, which is present in the Tractatus, is entirely self-sufficient, and intentionally so. What this account is designed to accomplish is to ensure that any symbolism that is coordinated with the logical space is guaranteed to be meaningful. Logic and mathematics – as Carnap and others have correctly noticed – is then only a byproduct of this coordination, free play of symbols. But no further account of syntax is necessary under the Tractarian picture (for a more detailed telling of this story, see David J. Hyder’s “Logicism and Atomism in Wittgenstein’s Tractatus”). Of course the early Wittgenstein himself makes the previously discussed Kantian error of thinking that he can intuit the geometry of (logical) space.
As you recognize in your post on Williamson, Frege’s (or early Wittgenstein’s) linguistic turn isn’t Carnap’s linguistic turn. The significant difference is that Carnap allows himself to utilise a formally constructed or otherwise specified metalanguage to provide an account of syntactic, semantic and pragmatic characteristics of a given language, and therefore to explicate conditions of proper usage of symbols (whereas for the logicisist universalist this must be specified ‘from the inside’, as Wittgenstein says in the Preface to the Tractatus) and the meaning which it conveys. The decision to use or abstain from using a language is an external one, a matter of convenience, broadly conceived. This constitutes a clear analogy with Poincare’s geometrical conventionalism (as you yourself point out, a radical version of this view was an inspiration for Carnap) – the usefulness of adopting a linguistic framework is akin to adopting a geometry, according to Poincare. …but we know that the geometry of space-time is NOT a matter of choice, it is a matter of fact (assuming a quite basic understanding of these terms). This can be extended to other domains.
In one of your post you’ve discussed Carnap’s late work on inductive logic, mentioning that he conceived attribute spaces as associated with a particular language – that is, as I understand it, a potentially interpreted formal syntactic system. I think this is a mistake, and one which is paradigmatic of logical empiricism and its response to Kant and early Wittgenstein as such. Choice of the geometry of an attribute space (which is analogous to Tractarian state-space in many ways) isn’t a matter of bridging, in an arbitrary way, a syntactic gulf between two kinds of sentences. There’s an issue of correctness and meaningfulness involved, just like with space-time geometry. And this perhaps even implies that the syntactic features of an inductive framework, just like in the Tractatus, should be here (at least partially) subordinated to semantics.
A natural Carnapian (perhaps not) response, along the lines of Poincare’s conventionalism, to this dillema would be to assert that it is a matter of convenience whether we utilise a certain framework for inductive reasoning. This response is possible, as has already been said, because we don’t have to worry about universal applicability of any concepts or frameworks – explication is pragmatic, Frege’s Begriffsschrift in contrast is metaphysical. But this would be in this case to make the Quinean move of rejecting the distinction between semantics and pragmatics. In Goodmanian terms, Carnap would have to say that attributing objective meaningfulness to one attribute space over another is due to the strong entrenchement of its associated perceptual predicates or something along these lines.
This, I believe, would ultimately threaten Carnapian non-cognitivism. I think we must be able to speak of correct applicability of a conceptual framework (a geometry or an inductive procedure) to the world in a way that doesn’t reduce to the question of whether it optimally accomodates our sensations according to some further unspecified criterion, per Quine’s “more thoroughgoing pragmatism” (Quine, “Two Dogmas…”), which ultimately boils down to the question of habit or custom. I think Carnap never really provided any resources to provide a general solution to classes which involve something like this (applicability) notion.
I think this case, even if I am somewhat misconstruing Carnap’s account of induction (which served just as an example of a certain continuity in his approach), might be a sign of a deeper problem with how the logical empiricists wished to accomodate Kant’s synthetic a priori judgements after 1920. Carnap, as far as I knew, never explicitly renunciated Poincare’s conventionalism in geometry (which he endorsed in Logical Syntax) and his account of science remained throughly syntactic through his career (in contrast with the model-theoretic approach which is motivated by concerns of answering the Kantian-Helmholtzian predicament). Is this possible while maintaining the same degree of (in Sellars’ words) conceptual voluntarism which is the core of Carnap’s metaphilosophy? What do you think about this?
Thanks in advance and Sapere Aude!
Apologies for leaving this un-approved (I haven’t looked at the site since my last post — sorry!), and taking so long to respond. Your comment raises too many issues for me to be able to respond quickly, but I will look at it more closely and respond properly as soon as I can.