How to distinguish the semantic and metaphysical components of scientific realism? Is there a distinction at all?

Here are some short reflections that are only indirectly related to the main theme of this blog: about the traditional way of framing the thesis of scientific realism, that always confused me a bit.

The components of scientific realism

It is custom to distinguish the components of scientific realism very roughly as follows:

Metaphysical thesis:

there is a mind-independent structure of natural kinds

Semantic thesis:

scientific theories are "about" this mind-independent structure

Epistemic thesis:

we are in a position to know that they are true

Although this three-fold distinction seems to make sense, it isn't so easy to classify various philosophical positions in this scheme, notably when it comes to pragmatist views, Kantian views or varieties of constructivism. Should they be interpreted metaphysically or semantically? Are they thesis about representation or about the nature of reality? I have always struggled a bit to respond (usually interpreting kantian thesis metaphysically, and pragmatist ones semantically, but then doubting it). On reflection, I think that the problem comes from the fact that mind-independence is ultimately a semantic notion, and so, metaphysical realism and semantic realism are almost indistinct. This is what I will argue in this post.

The metaphysical thesis is semantic

Why do I think so? Well, firstly because few scientific realists would deny that there are such things as minds or mental entities (desires, beliefs, etc), even if they are not fundamental. Or even when they are eliminativists about the mind, they don’t necessarily deny something weaker: that it makes sense to be a realist about minds or mental entities. It’s a meaningful position. The point of metaphysical realism was never to deny that minds are real. However, there is a problem here: minds and mental entities are ontologically mind-dependent, and so it would seem that they cannot count as real for the metaphysical realist. A mind cannot be mind-independently real. So, what is going on?

This kind of issue is acknowledged in various places, and the traditional response amounts to qualifying the mind-independence thesis by reference to agents holding representations or concepts of the object we are concerned with: what matters is not whether something is ontologically independent of the mind in general, but rather whether it exists and belongs to its class independently of being represented as such. And of course, someone’s mind can exist mind-independently in this sense, at least in principle, for minds could exist as minds without being represented as such: claiming that this is so is a meaningful position.

As an illustration, see how Psillos defines mind-independence:

To say of a non-mental entity which features in one’s ontology that it is mind-independent is to say that assertions about this entity are true because and insofar as their truth-conditions obtain, and not because and insofar as such assertions can be verified, rationally accepted, believed and cognate epistemic notions. (Psillos. 1999. Scientific Realism: How Science Tracks Truth. p. 13)

Now, recall the metaphysical thesis “there is a mind-independent structure of natural kinds”. Coinjoined with Psillos's definition, this implies that assertions about a structure of natural kinds are true in a way that doesn’t depend on epistemic notions. But here one could wonder: what has all this talk of truth and truth-conditions to do with a metaphysical thesis? Isn't truth a semantic notion? It seems that according to Psillos's definition, the metaphysical and semantic thesis collapse, or almost so, since the metaphysical thesis already mentions assertions about the structure of reality. At most the semantic thesis locates scientific assertions among these assertions, which is to say that scientific discourse doesn’t deserve a sui generi interpretation. The metaphysical thesis of scientific realism would say "there is some discourse that is not epistemically constrained" and the semantic thesis would say "scientific discourse is one such discourse". Kind of misnomers...

(another implication of this definition that I won't discuss here is that the only way of characterising both of these theses is in negative terms: epistemic notions are not relevant when interpreting assertions of existence; arguably, this makes realism a bit mysterious since contrarily to appearances, it doesn't have a purely positive characterisation of existence, but to be fair, any position needs some primitive notions).

I don't think that this aspect is specific to Psillos's particular proposal. The broad point that can be made is that in order to formulate the metaphysical thesis of scientific realism, we need to assume some very general form of semantic realism. If we don’t, even the thesis of metaphysical realism cannot be stated meaningfully.

Imagine you were to say “There is a mind-independent structure of natural kinds”, and then you add “by natural kind, I mean the reference of a projectible predicate, and by mind-independent, I mean that they don’t need to be the reference of a predicate at all to exist”. This is plainly contradictory. So, the metaphysical thesis already assumes the semantic thesis of realism, at least with regards to philosophical discourse (here we could quibble about possible and actual reference, but I don't think that this will lead us very far).

In sum, denying the semantic thesis in full generality (not only when it comes to scientific theories) implies denying the metaphysical one as well by making it unintelligible. Furthermore, affirming the semantic thesis in full generality implies affirming the metaphysical one as well as a requirement, since it is cast in terms of some discourse being "about" a real structure. If both theses are implied by the other, then the two are equivalent. I don’t think that there is a way to escape this conclusion without falling prey to our initial problem (denying that mental entities are real). At best, as already suggested, the semantic thesis of scientific realism could be understood as a particular application of the more general metaphysical thesis (that realist discourse is possible at all) to a specific sub-domain of discourse (scientific discourse should be interpreted realistically). But if it is understood as a general thesis about meaning and language, then it cannot be distinguished from the metaphysical thesis.

Consequence for the classification of non-realist positions

There are consequences of these remarks for the way philosohical positions are classified in contrast with realism.

If this is the correct view, then the metaphysical thesis opposes various versions of idealism that concern all representations whatsoever, not only scientific ones, and indeed this claim is often made. It also makes sense of the idea that the semantic thesis opposes various versions of instrumentalism, to the extent that they would be concerned with scientific theories in particular, not all of language.

But then, what about the pragmatist theory of truth? It should be classified as metaphysical non-realism, since it applies to all forms of discourse. And what about the reductive empiricism, logical empiricist’s version? It seems to be committed to a general semantic theory too: that all meaningful statements can be analysed in observational terms, not only scientific language. So, they too should count as opposing the metaphysical thesis, contrarily to what is often claimed: the metaphysical thesis becomes impossible to state meaningfully with this kind of view (its apparent meaningfulness would disappear if we analyse it in observational terms). This interpretation meshes well with the pretension of Carnap to go past the (related) physicalism/idealism debate by taking a neutral position according to which the debate is meaningless, or amounts to a choice of language. But the positions that really count as denial of the semantic component of scientific realism (specific to scientific discourse) are few in the end.

Having said that, maybe we could distinguish the early phenomenist versions of logical empiricism that viewed observational language as mind-dependent (when Carnap was inspired by Russell’s sense-data theory) and the versions that view observations as publicly assessable. The former are idealist theories that are not very far from Berkeley-style empiricism. Perhaps (perhaps!) the latter could count as a semantic thesis that only affects theoretical discourse but preserves a realist understanding of ordinary discourse. In this case, logical empiricism could be interpreted as a semantic rather than as a metaphysical thesis. But this doesn’t go without saying: even if observations are public and inter-subjective, they can still be epistemically constrained somehow, and if this is so, then we are back to a version of pragmatism that is better classified as a denial of (the intelligibility of) the metaphysical thesis.

What is a bit surprising, if my analysis is correct, is that most versions of non-realism count as metaphysical positions in so far as they apply across the board (including to philosophical discourse and representation). Kantianism, constructivism and pragmatism aren't very far from each other, and cannot really be differentiated by the way in which they oppose realism only (Peirce was notoriously imspired by Kant's critique, so this might not be very surprising).

Perhaps a more surprising consequence, and something that brings us again to the main theme of this blog, is that the best candidate for being a genuine semantic anti-realism that still subscribes to metaphysical realism is not logical empiricism, but one of its main opponents, the semantic conception of theories. According to this conception, theories are not statements, not linguistic entities, but collections of models, and theoretical statements do not describe the world, but identify relevant structures for scientific purposes. These guys do treat scientific discourse (at least theoretical discourse) differently from the rest of language. In so far as they are semantic realists about the rest of language, we have here the clearest version of a denial of the semantic thesis (for the sub-domain of science) and an acceptance of the metaphysical thesis (for the rest of language).