Rejecting the Analytic-Synthetic Distinction
If we cannot make sense of the idea of a synthetic truth, it looks as if we have to reject the analytic-synthetic distinction (the reverse of post-Quine orthodoxy). There is nothing coherent for the concept of analytic truth to contrast with (a genuine distinction requires meaningful things to be distinct). Yet there is some sort of distinction because not all statements are analytic (unless we could somehow make good on such a thesis—no easy task). It just needs to be re-conceptualized, re-thought. We must drop all epistemic attempts to capture the distinction and abjure such locutions as “true in virtue of the world”. Perhaps we need to dig deeper and more boldly; perhaps we need to upend some prominent assumptions. I am going to do just that, taking my cue from Kant’s characterization in terms of “explicative” and “augmentative” propositions. But I am not aiming at Kant exegesis, just philosophical truth (much easier than Kant exegesis). I will aim for maximum clarity and minimal defensiveness—simple not subtle (we can add the subtlety later). So, here goes: analytic propositions (so-called) are really identity propositions concerning concepts (also properties); synthetic propositions (badly so-called) are propositions asserting the co-instantiation of distinct concepts (also properties). Analytic statements say, “The concept F is identical to the concept G”, while synthetic statements (read non-analytic statements) say, “The concept F is co-instantiated with the concept G”. For brevity I will say that the former are identity statements and the latter are co-instantiation statements. The basic picture is that our conceptual scheme (system of concepts) has two sorts of structure, constituent structure and associative structure, and these structures concern conceptual relations in a broad sense. Thus, a given concept can be related to its conceptual constituents via constituent structure (what it is “made of”), and a concept can also be related to other concepts (not its constituents) by associative structure (what it can combine with). For example, the concept bachelor is related to the concept unmarried via the relation of constituency, while the same concept can be related to the concept happy via the relation of association (joining, juxtaposition). Distinct concepts can be combined in the same proposition, and concepts can contain other concepts. I take it this idea will be familiar; it is intended to be. Then the thesis is that the latter relations give rise to analytic truth whereas the former relations give rise to synthetic truth (as it is misleadingly labeled). We might instead call these “truths of containment” and “truths of association” so as to rid ourselves of unhelpful connotations (they are as bloodless as I can manage). Kant’s terminology is tacitly epistemic: “explicative” means “articulates antecedently possessed knowledge”, while “augmentative” means “adds to our pre-existing knowledge”. I am just talking about abstract relations between mental elements that may map onto such epistemic notions but are not defined by them. In fact, I want to take a further step away from the cognitive and say that the official theory is to be stated in terms of properties (references of concepts): properties can contain properties and they can also be co-instantiated with properties. These are relations that obtain in the non-mental world—in the world of external objects. We may say (if we like) that analytic statements correspond to identity facts involving properties, while synthetic statements correspond to facts of co-instantiation involving properties. We may also say that analytic statements are about concept parts while synthetic statements are about concept partners (or property parts and property partners). What is crucial to the distinction is that partners are not parts; co-instantiation is not identity. It is this that distinguishes the synthetic from the analytic. It has nothing intrinsically to do with knowledge or experience or justification. The distinction is emphatically not the same as the a priori-a posteriori distinction. It is about truth-makers, i.e., about two categories of truths. It says nothing about what is trivial or non-trivial, informative or uninformative, known by experience or known independently of experience, having cognitive value or not having cognitive value, being intuitive or being perceptual. It is a distinction drawn at the level of metaphysics not epistemology. It has nothing intrinsically to do with rationalism and empiricism, or revisability, or infallibility. The claim I am making is simply that this is the best way to re-conceptualize the traditional distinction labeled “the analytic-synthetic distinction”; and it may not correspond to the intentions of people casually employing that phrase. I would be quite happy to allow that it is not a version of that distinction (whatever it is exactly) but a new distinction altogether—a better, more intelligible, distinction. I am not doing intellectual interpretation; I am doing philosophical excavation—of metaphysical reality not intellectual history. The distinction, as I understand it, is basically about the ontology of properties—how they stand in relation to each other. Does this suggestion serve to unify the class of non-analytic truths? Yes and no. Yes, in that it says what is in common to all non-analytic truths, viz. concept association (without concept identity); but no, in that there is a huge variety of what we call properties, ranging from physics to ethics, psychology to arithmetic, with nothing to unite them except the very general notion of a property (attribute, characteristic, predicate). There is nothing like the traditional idea that synthetic truths are all known by sense experience, or add to our knowledge, or produce a sense of cognitive augmentation. The class of non-analytic truths is not unified in these ways; it is simply the class consisting of truths that concern distinct properties that are said to be co-instantiated. This may have consequences for epistemology and psychology, but it is not defined that way. We can put the point as follows: propositions have the power to contain concepts that have other concepts as their constituents, and they also have the power to contain concepts that are instantiated together—the former gives rise to analytic truths, the latter gives rise to synthetic truths. We can thus preserve the substance of the analytic-synthetic distinction without running afoul of the problems surrounding the usual ways of formulating it. It is really much more of a truism than has been supposed, though a truism with serious metaphysical commitments. The analytic-synthetic distinction is rooted in deep facts of nature, ultimately the nature of properties.
 This paper follows on from my “Are There Synthetic Truths?”
 I think the fundamental problem with the usual discussions of the analytic-synthetic distinction is that people decline to recognize that the question is metaphysical, not linguistic or epistemological. Even Kant gave it an epistemological slant that influenced all later discussions, though it is au fond not an epistemological distinction. It is about the nature of truth (or truths) as such not about our knowledge of truth (truths). We should not psychologize the distinction.