The next two chapters are dedicated to further analyses of the substance of the concepts necessary, analytic, and a priori. I will also consider some arguments for and against the intelligibility and worth of each concept. Looking ahead, one aim of these two chapters is to pare down these notions, to isolate their core out from amongst many closely nested questions and issues. The general goal is to arrive at concepts that are refined enough so that specific theses about their substance and interrelations can be tested, but yet not so over-refined as to be out of touch with their broad historical roots.
§3.1: more on necessity
‘Necessary’ is opposed to ‘contingent’ or ‘accidental.’ Something is necessarily so if and only if it could not possibly not be. Necessary truths are absolutely firm, unalterable; they are fashioned from the hardest of steel. They just simply are the case, regardless of time or place. They could not be otherwise, irrespective of how contingent matters of fact might be altered. We have no choice or influence in the matter.
Two concepts that are closely tied up with necessity, both historically and conceptually, are eternal truth and universal truth. A link between necessary and eternal is evident in Plato’s works, for instance—he talks of necessary truths as timeless, as not subject to temporal change. To cite another example, Leibniz explicitly uses ‘eternal truth’ and ‘necessary truth’ as synonyms (cf. Pap [1958: Ch. 1]). ‘Eternal truth’ contrasts most specifically with ‘historical truth’; a historical truth is one whose truth-value is contingent upon specific historical developments, and so could vary across the ages. So there is reason to think that a truth must be eternal, not subject to change over time, to be necessary.
It is also commonly claimed that all necessary truths are universal. The idea here is that, if a truth just applies to some particular matters of fact, at a certain time or place, or in virtue of certain specific and local contingencies, then it is not necessarily the case. To illustrate, suppose it is claimed that it is necessary that the conjunction of three conditions—let us call them A, B, and C—precipitate the occurrence of some further condition D. (D could be, for instance, a specific disease, an economic recession, or a forest fire.) The idea that all necessary truths are universal comes down to this: the claim of necessity is that, absolutely anywhere that A plus B plus C occurs, D occurs. If the claim is qualified, so that A plus B plus C lead to the occurrence of D only at a certain place or time, or only given certain other specific local circumstances, then, given that universality is a criterion of necessity, such qualifications undermine the claim to necessity.
To be sure, though, claims of necessity do not depend on there actually being any specific minimal number of instances. (Analogously, the claim ‘Trespassers will be prosecuted’ is not rendered false or meaningless if no one, in fact, ever happens to trespass.) A plus B plus C might only actually co-occur once, or might not co-occur at all, without undermining the claim that the asserted connection to D necessarily holds. So, for instance, the claim that the presence of oxygen is necessary for fire is not a claim about how many fires there are; and, if now true, it did not just become true the first or tenth or millionth time there was a fire. Rather, it was true all along, determined by the very nature of the phenomena.
One factor that complicates this link between necessity and universality we might call Heraclitus’ worry. Heraclitus is said to have remarked that one can never step into the same river twice. One thing he is taken to have meant by this remark is that, strictly speaking, no situation ever occurs more than once; if this is so then all phenomena are specific and local, and the very idea of universal truth might seem to be useless artificial abstraction. For any event in the real world—diseases, recessions, fires, and so on—there are an awful lot more than three causally relevant factors to consider. It may well be the case that, for any particular phenomena, the causally relevant factors are never exactly repeated. If so, all connections between phenomena would be, in a sense, contingent on local circumstances, and universality would seem idle and insignificant.1
Note, though, that although these considerations serve to underline the complexity of the study of metaphysical modality, and to guard against oversimplification, they do not undermine the interest or worth of the core notion of necessity. These complexities are real and important, but they do not entail that there is no such thing as factors that necessitate diseases, recessions, or fires, or that it is folly to keep looking for them.
The notion of generality lies at the base of these links between necessary, eternal, and universal. It is widely held to be part of the content of the concept of necessity that, in principle, necessary truths apply to more than one particular situation or individual. The scope of a claim of necessity includes all places and times; following Plato, this is sometimes glossed as the thought that necessary truths are outside the bounds of space and time. And so, the more instances we observe of a connection between two things—oxygen and fire, say, or cessation of heartbeat and death—the stronger is the case for the claim that the connection between these phenomena is necessary, not just accidental.
Necessity, per se, has nothing to do with anyone’s knowledge of anything, or with what any linguistic expressions mean. It has to do with the nature of things, and while we certainly try to fashion our beliefs to track the mind-independent nature of things, and to craft our terms so that we can accurately discuss the phenomena which matter to us, that something is necessary is not a claim about knowledge or meaning, and does not immediately entail any such epistemological or semantic claim. It is a matter of metaphysics, and is something about which all humans could be mistaken or ignorant, or lack the conceptual means to engage with in their thought and talk.
[§]
The distinction between essence and accident is one of the central distinctions in modal metaphysics. Take any individual, or kind, or phenomena. What changes can it persist through, and what changes would be so drastic as to preclude the sustenance of its identity? Sure, Gillian might have been a lawyer instead of a teacher, but presumably she couldn’t have been a rock; sure, the Olympic Games would still be the Olympic Games if cross-country wrestling were admitted as a sport, but probably not if it no longer contained any sports and did not involve international competition. Similarly for diseases and sub-atomic particles and political parties: an integral part of understanding any kind of phenomena is to sort what is essential to it from what is accidental.
There is much controversy concerning the precise content and intelligibility of these kinds of modal attributions, and the relations between them. These and related questions form some ancient threads in metaphysics, still vibrant today. For instance, consider the question: Do individuals have an essence? Aristotle held that essences are species-wide properties. (Indeed, ‘species’ and ‘essence’ come from the same root word, the former stemming from a Medieval Latin variant on the older Greek term.) All tigers, and all eyes, share an essence, down this avenue; the essence of something is what it does; and all tigers, all eyes—all members of any one species—should in this sense be seen as doing the same thing. Therefore individuals do not really have distinct essences, on Aristotle’s view. Kripke (1972), in contrast, defends a very different view, according to which individual humans (and some other kinds of objects) do have unique individual essences. There is something unique about being me, or being a particular tiger, on Kripke’s view. He makes some substantial and controversial conjectures about individual essence.2
Those who believe in distinct individual essences are called ‘haecceitists.’ (The word comes from the Medieval Latin translation for the Ancient Greek word for ‘this.’) Haecceitists believe that there is something that constitutes the essence of a particular thing, something it necessarily has, without which it would not be itself, and that nothing else could have. There is something unique and essential about an entity, which endures through the change of all accidental properties. Anti-haecceitists reject this claim. Anti-haecceitists are skeptical of the coherence or usefulness of the distinction between individual essence and accidental properties.3
[§]
I will next discuss the way in which the exact strength or force of modal terms like ‘necessary’ can vary with context. In most cases, modal terms are not used in a completely unrestricted manner, but are rather implicitly restricted to some (more or less vaguely defined) contextually salient set of possibilities. For example, suppose I sincerely utter:
1. I would really like to make this $50,000 investment, but I am afraid that it is just not possible right now.
Clearly, not the whole of the vast expanse of metaphysical possibilities is relevant to the content of such a statement. Rather, only those possibilities in which the constraints on my resources are (more or less, roughly) held constant are relevant. It would be obtuse for one to respond that [1] is false because there is a perfectly possible situation in which I could easily make this investment (e.g., if I had won the lottery yesterday). Such remote possibilities are irrelevant to what statements like [1] express, which is to say that statements like [1] involve implicit restriction to a contextually salient proper subset of the set of all possibilities.
Another way to illustrate this phenomenon is with reference to counterfactual conditionals—that is, hypothetical ‘if’/‘then’ statements whose ‘if’ clause is false. Counterfactuals, like probabilistic reasoning, involve reasoning about non-actual possibilities; and, from scientists to Monday morning quarterbacks, we use them all the time in everyday reasoning:
2. If Hitler had not been born, then World War II would not have happened.
3. If the Seahawks had punted instead of gambled on the last play of the third quarter, then they would have won the game.
Counterfactuals also clearly instance this phenomenon of implicit restrictions on possibilities. [2], for example, asserts that Hitler was the, or at least a central, cause of World War II. Again, it would be obtuse to argue against [2] on the grounds that even if there were no Hitler, still it is possible that World War II was started by, say, a one-armed Australian sheep farmer. [2] is about the loosely defined set of possibilities that are most like the actual world except that Hitler does not exist; it is not the claim that it is metaphysically impossible for there to be World War II but no Hitler. Similarly for [3], which does not assert that it would contravene the ultimate laws of the universe for the Seahawks to both punt and lose. Only a selected, target set of possibilities are involved, in the context of such discussions.
This is an important point to beware of, generally. In ordinary discourse, there is much vagueness, ambiguity, and shiftiness inherent in metaphysical modal claims. When one encounters such a claim, one needs to be keen to various contextual clues, in order to determine precisely how restricted, or unrestricted, the scope of the claim is intended to be. The cases of analyticity and a priority differ fairly drastically, in this respect. While, as we will see, they do instance a distinctive and pervasive sort of framework-relativity, they are not subject to exactly the same sort of context-dependent shiftiness, illustrated by [1]–[3] above.
[§]
A related refinement is that there are different kinds, or strengths, of necessity. One distinction which is drawn in this terrain is that between physical and metaphysical necessity. ‘X is physically necessary’ means that X is necessitated by our actual laws of nature (such as E = MC2, or G = M1 × M2/D2). ‘Y is metaphysically necessary’ means that Y is necessitated by the laws of metaphysics (such as that everything is self-identical, or that no individual simultaneously both has and lacks a given property). Physical necessities hold throughout the set of physically possible worlds, that is, the worlds in which the physical laws of the actual world hold. Metaphysical necessities hold throughout all metaphysically possible worlds. Reichenbach (1953) gives a nice way to illustrate this distinction. Consider:
4. There is a solid one-ton sphere of gold.
5. There is a solid one-ton sphere of uranium 238.
Both are actually false. However, [5] is physically possible, while [6] is not. That is, while the sphere of gold could exist—maybe Oprah could build one, if she wanted to—the sphere of uranium could not, as a matter of physical law, because it is radioactive, and would explode. However, if one thinks that the uranium sphere could exist, in possible worlds in which the laws of nature are different, then one thinks that [6] is metaphysically possible, even though physically impossible.
Intuitively, the idea behind this distinction between physical and metaphysical possibility is that there is some degree of contingency to the actual laws of nature. It is in some sense possible that, say, gravity or magnetism might have worked differently, or might not have been at all, and that would not spell the end of the world. If so, we are describing physically impossible but metaphysically possible worlds. Or, again, take Planck’s constant, which gives the ratio of the frequency of radiation to its quanta of energy. Planck’s constant has an approximate value of 6.625 × 10–27. Now surely, one might think, things did not have to be this way. Planck’s constant might have had the approximate value of 6.624 × 10–27, or 6.626 × 10–27, and the universe would still exist (in more or less the same fashion). If that is so, then there are metaphysically possible worlds in which Planck’s constant has a slightly different value; but these would be physically impossible worlds, because the actual laws of physics would not hold. If pigs can fly, and cows can jump over the moon, then this happens in physically impossible worlds. (In the actual state of things, pigs have not shown any aeronautical capability, and cows cannot jump over squat.)
[§]
Metaphysical modality was an area of vibrant debates in the Ancient and Medieval periods in Western philosophy, but was to subsequently endure various degrees of neglect or hostility, until well into the twentieth century. For example, during the Modern period, epistemological questions were given foundational priority over metaphysical ones; and, during the first half of the twentieth century there was active antipathy toward many such metaphysical notions—some of the reasons for which will be discussed below in §3.3.
The notion of mind- and language-independent necessity became much more widely accepted in the latter decades of the twentieth century, after enduring this period of relative neglect. Developments pertaining to the sophistication of our understanding of semantic matters makes it possible to detect fallacies and confusions in several varieties of argument against the coherence of metaphysical necessity (cf. especially §3.3 below), and rigorous semantics for modal discourse were developed.4 We post-Kripkeans are more at home with many varieties of modal attributions; even if there is still much controversy as to understanding the exact content and import of such attributions. That is, metaphysicians of many different stripes and orientations can agree that it is not possible for something to have the property ‘square’ but not be four-sided, or to have the property ‘water’ but not be H2O. It is at least commonplace, if not orthodox, in these liberal times, to accept mind- and language-independent necessary connections among properties.
§3.2: analyticity revisited
A truth is analytic if and only if it is true by virtue of the meanings of the terms involved; alternatively, the denial of an analytic truth is contradictory. Analyticity is a kind of semantic guarantee of immunity to counterexample. Candidates for analyticity include:
1. Squares have four equal sides.
2. No grandmother is childless.
3. Bachelors are unmarried men.
Recall Plato’s problem: Our experience is particular and limited, but yet we still manage to attain knowledge of some universal, necessary truths. As mentioned in chapter 1, many philosophers—including especially empiricists who countenance universal, necessary knowledge—have hoped that the notion of analytic truth can ground a non-obscure but adequate solution to Plato’s problem. The thought is that, in certain privileged cases, just grasp of meaning is sufficient to justify the belief that the statement is universally and necessarily true. (In other words, semantic intuition is not an obscure, non-empirical faculty of mind, and it might be capable of delivering all that was wanted of rational intuition.)
The term ‘analytic’ comes from the root ‘analysis,’ another word with Ancient Greek roots. An analytic approach to a problem proceeds via separating it into component parts or constituent elements. ‘Analytic’ contrasts with ‘synthetic.’ In the case of a synthetic statement (such as that Neptune has four moons or that Alice is a grandmother), more than just analysis of the constituent terms is required in order to judge whether it is true. ‘Synthetic,’ from ‘synthesis,’ means having been put together. Analysis is a process of taking something apart; synthesis is a process of putting something together.
What exactly, is analyticity a property of? Sentential characters are not the best candidates for analyticity, because sentences per se are not true or false (a fortiori not true or false in virtue of anything). Rather, a use of a sentence in a context can express something, and the question of truth or falsity arises with respect to what is expressed. So, is analyticity then a property of what is expressed—that is, a proposition, or sentential content? First and foremost, my reason for avoiding that option is the point from §2.4 about vehicle-indifference (i.e., relevantly different sentences can be used [in context] to express the same proposition). For a variety of sorts of cases, some of which will be important below, it is important to leave open the possibility that different ways of expressing the same proposition might differ with respect to either or both semantic or epistemic status. Again, examples include:
‘Cordates are cordates’ versus ‘Cordates are renates’.
‘I am here now’ versus ‘Arthur is in St. John’s on October 18, 2017’.
‘H2O contains hydrogen’ versus ‘Water contains hydrogen’.
Even if such pairs are taken to express the same proposition, still there might be good reasons to count only the first member of each pair as analytic (and/or a priori).
So, the question of truth brings propositions into the picture; for this question does not arise at the level of sentences. However, given that there can be relevant differences between distinct sentences which express the same proposition, then propositions are not the only thing in the picture. Thus, the bearer of analyticity will have to encompass both a proposition (in order to accommodate the question of truth) and the means of expressing that proposition (in order to accommodate the non-truth-conditional element in analyticity). That is, analyticity is not exclusively a question about sentential characters or about sentential contents, but essentially includes both dimensions.
Hence, I take the primary bearers of analyticity to be statements, where a statement is a particular, dated use of a sentence to express a proposition. Analyticity, then, is a property not solely of symbols, or of contents, but of specific uses of symbols to express contents.5
[§]
At this juncture, I should explicitly flag two fundamental distinctive properties of analyticity, as distinct from necessity or a priority. Both will be revisited at several junctures below. The first pertains squarely to Plato’s problem. I will call it the ‘from coherence to worth’ worry: that is, for the case of analyticity, more so than for the cases of necessity or a priority, there prevails the worry that even if the notion is coherent, and uncontroversial examples could be found, still it is not yet clear that analytic truth is of any philosophical interest, as opposed to being merely verbal trifle. One would hardly hear either ‘Sure, that is a necessary truth; but necessity itself is of no philosophical interest,’ or ‘OK, fine, that is a priori knowledge; but that category of knowledge is insignificant.’6 However, this kind of worry is commonly pressed for the case of analyticity: ‘Sure, OK, I grant that “All bachelors are married” is immune to counterexample in virtue of meaning. But how does that get us even one iota toward solving Plato’s problem?’ Again, this ‘from coherence to worth’ worry threads throughout several strands of the ensuing discussion. One reason why I flag it here is that it is crucial—for both proponents and opponents—to distinguish objections to the coherence (or intelligibility) of analytic truth from objections to the worth (or usefulness) of the notion.
A second distinctive complication is that analytic truths are essentially about meanings, and meanings are notoriously difficult things to talk about in a non-contentious, non-tendentious way. Not only have there been considerable and influential skeptics about the notion of meaning, there have also been lots of non-trivial differences between the non-skeptics, as to how to understand this complex notion. So, it seems that lots of honing and framing of the notion of meaning is a precondition for a worthwhile discussion of analyticity—again, as opposed to the cases of necessity and a priority. Of course my point is not that there are no substantive philosophical questions about the notions of metaphysical reality or of epistemic status, but rather that the notion of meaning is relatively worse off, when it comes to getting a pre-theoretical handle on the very idea, in order to get a philosophical discussion rolling.
So, one of the reasons why there was relatively little discussion of the notion of analytic truth in the latter half of the twentieth century is that there is such a wide variety of distinct theoretical approaches to meaning. There are philosophers who take the basis of meaning to be sense, reference, use, intentions, truth-conditions, causation, teleology, etc., and each of these orientations has distinct sub-varieties. As a result, it is difficult to be precise about exactly what ‘truth in virtue of meaning’ comes to without assuming a specific and controversial take on the nature of meaning. To proceed in terms of any of sense, reference, use, intention, truth-conditions, causation, teleology, etc., would be controversial; while to proceed in terms of all of them would be untidy and vague.
For present purposes, though, it is neither necessary nor desirable to pin down ‘meaning’ more precisely. My view is that if meaning is determinate, then there will inevitably be a class of statements that is true in virtue of meaning.7 (There may be borderline cases, of course, but the present point is that there will still be paradigm cases. From the fact that there is a broad spectrum of shades of grey, it hardly follows that nothing is either black or white.) That is, whichever approach to meaning to which one subscribes, as long as it is possible, if not common, for some specific content to be semantically expressed with a statement, then the brute datum described above (of immunity to counterexample in virtue of meaning) is bound to arise. Hence, it is appropriate for this discussion to stay as ecumenical as possible about the correct theoretical account of meaning.
Of course, some of the most influential critics of the notion of analyticity have also (and not coincidentally) been skeptics about the determinacy of meaning. (Quine is a seminal case in point.) I will not say much that is directly addressed to refuting such meaning-skepticism here, but will rather confine myself to two pertinent points. One minimal point is that it is important to distinguish the following two questions, with respect to any use of a sentence (in context):
1. Can we know for sure exactly what a certain speaker is intending to communicate, with a certain utterance?
2. Is there a determinate meaning semantically expressed?
The first (interpretive) question is a complex and difficult knot, which it would take considerable hard work to even begin to untangle. However, even if the interpretive question should ultimately be answered in the negative, that does not mean that the second (semantic) question should also be answered in the negative. There are lots of familiar, intuitive reasons to think that the second question should be answered in the affirmative,8 and that is enough to get the problematic of truth in virtue of meaning up and running. Again, if meaning is determinate, then there will inevitably be a class of statements that is true in virtue of meaning. (With a nod back to §1.3, this is to distinguish between epistemic from metaphysical questions, in this particular domain.)
Second, also note that while meaning-skepticism was an option for a naturalistically inclined philosopher in the mid-twentieth century, when behaviorists and structuralists still ruled the human sciences, this is no longer so at this juncture, since the cognitive revolution. The reason is that what are currently the most successful research programs in the study of cognition are up to their minds/brains in intentional semantic notions. To dismiss these research programs as wrong-headed is to presume a perspective outside of science from which science can be evaluated. So, a naturalist must respect these research programs, and to do that is to reject meaning-skepticism.9
Of course all that still leaves us with the hard work of explaining how something which might seem as trivial as ‘truth in virtue of meaning’ has any promise to afford a substantive and weighty answer to Plato’s problem. A successful defense of the coherence of analytic truth would not yet amount to a positive case in favor of its significance or worth.
[§]
While the explicit employment of the notion of analyticity is a relatively modern phenomenon, the core idea underlying the notion has been around for a long time. For one thing, many propositions that have been of perennial interest in philosophy are candidates for analyticity, such as:
1. No proposition is simultaneously both true and not true.
2. Every even number is divisible by two.
3. Murder is wrong.
The strategy of using something like analytic truth to fashion an answer to Plato’s problem is almost explicitly formed in the work of some Modern empiricists, such as Hobbes and Hume. Hobbes seems to have espoused the view that the root of all necessity lay in language, that what we think of as necessity is properly traced back to linguistic conventions.10 Variations on this conventionalist theme have been players on the scene ever since.
With Hume, we come very close to an explicit statement of this style of answer to Plato’s problem:
All of the objects of human reason or enquiry may be naturally divided into two kinds, to wit, Relations of Ideas, and Matters of Fact. Of the first kind are the sciences of Geometry, Algebra, and Arithmetic: and, in short, every affirmation which is either intuitively or demonstratively certain. … Propositions of this kind are discoverable by the mere operation of thought, without dependence on what is anywhere existent in the universe. … Matters of fact … are not ascertained in the same manner; nor is our evidence of their truth, however great, of a like nature with the foregoing. The contrary of every matter of fact is still possible. (1748: Sect. IV, Part 1)
This is a seminal statement of what many empiricists want to say about our universal, necessary knowledge: We have a priori knowledge of certain things because they are analytic—that is, because of intrinsic relations among the relevant concepts.
To be sure, though, what Hume calls ‘relations of ideas’ involves an undifferentiated mush of what I want to distinguish as necessity, analyticity, and a priority. Further, as Kant stresses, it is far from clear that such Relations of Ideas can afford the basis for substantive, informative, judgments (as opposed to merely a priori trivialities).
[§]
As with the cases of a priority, a defensible notion of analytic truth must steer clear of some of the features that have been historically associated with the concept. It is crucial to sharpen the concept of analyticity, because, even though the notion has a respectable and important place in the philosophers’ toolkit, there are serious problems with many of its traditional associations. Thus, I must disavow some of the prevalent associations of ‘analyticity,’ and develop some refinements. While many would hold that it is precisely those associations which constitute the main philosophical interest in analyticity, I will argue that the conception which remains is not only defensible but quite substantive and worthwhile.
To begin this pruning process, I will start from a quote from Coffa (1991: 9–10):
One of the many ways philosophers have tried to understand meaning might be called the ‘chemical theory of [propositions],’ using an analogy occasionally found in the writings of Locke … and Kant. According to this theory, [propositions], like chemical compounds, are usually complexes of [meanings], which may themselves be complex. … Analysis is the process through which we identify the constituent [meanings which make up a proposition, and then, in turn, break down these meanings into their constituents]. It is a process that must come to an end … in the identification of simple constituents. … To know a concept fully, for example, is to define it; and definition is no more and no less than exhaustive and complete analysis.11
I will not attempt to document the historical prevalence of this chemical theory. Rather, first I will briefly explain why, given the chemical theory of propositions, the analytic/synthetic distinction is both sharp and important. Second, I will point out that there are lots of reasons to reject the chemical theory. Finally, while this does suggest that the analytic/synthetic distinction is less sharp than many had thought, it does not entail that the distinction lacks importance for philosophy. Rather, the shortcomings of the chemical theory point to some valuable refinements to the notion of analyticity.
So, first: if the chemical theory of propositions were an accurate picture of semantic content, then the analytic/synthetic distinction would be crystal clear, and its philosophical importance would be absolutely paramount. The guiding idea here is that meanings, and the propositions which they compose, are objective, mind- and language-independent entities (with objectively identifiable compositional structures). Whether one is more inclined toward a dialogical sort of view that getting in touch with these objective entities is essentially an intersubjective affair, or toward a more subjectivist view that one isolated mind is capable of getting acquainted with them, either way there is an objective fact of the matter as to whether a statement is true in virtue of meaning (or, alternatively, of whether its denial is contradictory). The question of whether a certain judgment is analytic is, like the question of whether a certain sample of ore contains any iron, a matter for objective discovery.
Further, given this chemical theory of propositions, the analytic/synthetic distinction is rather vital to the enterprise of philosophical inquiry as a whole. Indeed, there is a strand within philosophy which takes the analytic/synthetic distinction to be definitive of the very subject matter of philosophical inquiry, as distinct from scientific inquiry. While both sorts of inquiry might be objectively factual, synthetic matters of fact are appropriately left to scientists, while the distinctive subject matter of philosophy is conceptual analysis. The unique expertise of the philosopher, on this orientation, lies in mining, amplifying, and tracing inferential patterns among analytic judgments, and in cultivating this kind of specific knowledge into general wisdom.
However, throughout the twentieth century, in various ways, it has become generally acknowledged that concepts and propositions are not so simple as the chemical theory would have it. To the contrary, languages are organic entities, and the conventions pairing expressions with meanings evolve over time. Languages, meanings, and conventions are non static, objective entities inhabiting some realm out there, waiting for us to get acquainted to them, but rather, are constantly being revised, expanding while here and decaying there. New meanings and conventions are ever suddenly being coined, and gradually dying off. These considerations greatly complicate any links between linguistic conventions and necessary truth.
Relatedly, the idea that there are objective mind- and language-independent facts about meanings and propositions is no longer generally conceded as unproblematic. Many philosophers hold that not just linguistic conventions, but meanings and propositions themselves, are, in some significant senses, human-dependent. To the extent to which the very content of the concepts at issue is, at least in part, up to us—and so can vary from time to time and place to place—the very idea of objective facts about analyticity is deeply problematic. The idea that there are timeless objective facts about analyticity now seems rather archaic, given the general recognition that languages and conceptual schemes are organic entities which evolve over time. (There will be much more discussion of this kind of conceptual evolution in Part III; especially §6.3.)
I will close this particular thread of the discussion by entering two critical remarks, pertaining to aspects of Kant’s views which must be either rejected or substantially refined. The first point is that Kant was an adherent of what was above called ‘the chemical theory of concepts,’ as is evidenced by another definition of analyticity he endorses which appeals to the metaphor of containment—that is, a judgment is analytic iff the predicate concept is contained in the subject concept. Not only does this criterion just apply to a limited class of statements,12 it also seems to clearly presuppose the picture of concepts as static entities out there waiting for someone to get acquainted with them, of there existing timeless objective facts about what is or is not analytic, which is now generally rejected.
So, there is a real definite lack of substantive criteria for analyticity here. Notoriously, controversies as to whether a contested candidate should be considered to be analytic or synthetic are not resolvable on the basis of anything in the Kantian corpus. For this reason, debates between Kantians and their opponents had, by the mid-twentieth century, ‘settled, or bogged, down around a handful of particular cases: for example, “nothing can be red and green all over” and the transitivity of “earlier than”’ (Quinton 1963: 31). Contrasting positions on the synthetic a priori status of such cases deteriorated into articles of faith, because of the ineffectuality of the extant ways of drawing the analytic/synthetic distinction.
For these and other reasons, Kant’s conception of the distinction has been subject to an awful lot of criticism.13 So, even though the view developed in Parts III and IV owes a lot to Kant, below I will depart from the specifics of his views at several junctures.
For now, one important moral is that the analytic/synthetic distinction is no longer taken to be as crisp as Hume or Kant, say, thought it was. Rather, it is perhaps more plausible to view it as a continuum with paradigm cases at the extremes and a range of shades of grey stretching in between. Even so, my view is that, suitably refined, it is still of philosophical importance.
For instance, despite the demise of some of these oversimplified ideas about meanings and propositions, there is still the brute datum—that is, we are justified in believing some statements that are universal in scope, and not subject to refutation by contingent happenstance, such as:
1. All squares have four sides.
2. No grandmother is childless.
3. Two is a factor of every even number.
These brute data provide strong prima facie paradigm cases of analytic truths. Among other things, this may yet hold the key to a satisfactory solution to Plato’s problem. The idea is that grasp of the meanings of the constituent terms is sufficient to justify the belief in the truth of such statements. In such cases, understanding of meaning grounds the recognition of truth.
[§]
A couple of final refinements to analyticity, before turning to some claims staked by skeptics. One stone which has remained as yet unturned concerns how ‘ampliative’ an analytic truth can be, with Kant (1781) and Frege (1884) at the opposed poles. For Kant, analytic judgments cannot afford new knowledge, but rather just merely unpack the content of what is already known. In contrast, Frege derides ‘the widespread contempt for analytic judgements’ and its attendant ‘legend of the sterility of pure logic’ (1884: 24). Frege’s metaphor that the fruits of analysis ‘are contained in the definitions, but as plants are contained in their seed, not as beams are contained in a house’ (1884: 101) illustrates this nicely. Logical and semantic analysis can afford new knowledge, as opposed to telling us what we already knew—as, indeed, anyone who has had to work hard on finding a derivation of a self-evident theorem of logic, or found a simple, elegant proof of something entirely non-trivial, can appreciate. So, Frege clearly rejects conceptions of analyticity which would tie it to lack of substance, to emptiness of content. Analytic judgments can be ampliative for Frege (i.e., substantive increases in knowledge), unlike for Kant.
This is especially relevant to our ongoing ‘from coherence to worth’ struggles, for Frege (and many of those influenced by him) rejects the idea that analytic truths are trivial and uninteresting. Relatedly, it is also pertinent to ongoing questions about the ampliative power of semantic intuition itself, and ultimately to upon which side of the rationalism/empiricism divide a variety of the constitutive a priori orientation is situated.
Another barely yet turned stone: In §2.3 I distinguished the notion of ‘reference determiner’ from both meaning and extension; this notion has become important in the wake of the externalist challenge. In penetrating recent work, Russell (2008, 2012) has argued that three more precise notions should now replace the old ‘truth in virtue of meaning’—namely, (i) truth in virtue of character, (ii) truth in virtue of content, and (iii) truth in virtue of reference determiner. On her view, analytic truth is worth saving, and (iii) is the best way to do so, in the wake of reviseability, externalism, and other considerations.
In the context of the present inquiry, a main part of which is tracing the overlaps and distinctions between analyticity and the closely related concepts of necessity and a priority, I have opted for keeping meaning in the picture. Continuity with that aspect of the traditional discussion is both intrinsically desirable and extrinsically useful. Furthermore, the response to the challenge of revisability which I develop in Part III is rather different from Russell’s, one result of which is that the notion of a reference determiner (as distinct from meaning and extension) will play a more limited role in my work than in hers. This is hardly a criticism of Russell’s work,14 but rather just a note that, while I find her work very interesting and worthwhile, I will not be following it very closely, in this respect.
§3.3: skepticism about necessity and analyticity
Skepticism about metaphysical necessity is not hard to motivate, or to sympathize with. Claims about essences are relatively obscure and mysterious; and they are awfully difficult to conclusively support. They are rather remote from everyday experience and practical concerns. For these and other reasons, there have been skeptics about necessity for as long as there has been modal metaphysics. Next I will discuss two variants or aspects of this skepticism: first, the epistemological charge that claims of necessity could never be justified; and second, the stronger charge, partly semantic and partly metaphysical, that such distinctions as necessary/contingent or essence/accident are incoherent. The first is easier to establish but less consequential, and the second would be quite serious, but is rather hard to establish.
[§]
There is some reason to think that we could never conclusively verify the claim that a proposition is necessary. By definition, a claim of necessity extends beyond the range of not only any particular individual’s observation, but, even further, beyond the range of the observations of all actual individuals, past and future. Consider, for instance:
1. It is necessary that the amount of force an object will exert on another will rise in proportion to an increase in its velocity (i.e., the faster it’s moving, the harder it’ll hit).
[1] is not just a claim about all cases that have been, or will be, observed. It is the claim that it is a fact determined by the very nature of things and unconstrained by specific spatio-temporal contingencies, that it is not possible to increase an object’s velocity without increasing its potential energy. This claim extends indefinitely through time, and throughout all possible ways the world could have been—that is, the scope of [1] is so broad that it applies regardless how the world might have been contingently different, wherever and whenever we choose to consider. Given that we can never conclusively verify claims of necessity, why should we believe in them, and what use are they?
Even if, strictly speaking, we can never conclusively evaluate such claims to necessity, a reasonable response is to admit that our judgments of necessity are fallible, and subject to change in the face of counter-evidence, as opposed to outlawing the term ‘necessity’ from rigorous and systematic thought. The more instances that are observed of a connection between two things—such as velocity and force, or fire and heat, or cessation of heartbeat and death, and so on—the less plausible it is that the connection is accidental, and the greater the reason one has to believe that the connection between them is necessary, is rooted in the very essence of the phenomena. Further, the positing of necessary connections, where we have good reason to, has great explanatory value. In general, the more well-confirmed hypotheses of necessary connections we have, the more comprehensive is our understanding of how things work, and the greater is our ability to predict future conditions, and to manipulate conditions to our advantage. Indeed, this kind of non-accidental, mind-independent connection between phenomena is precisely the target at which scientific inquiry (as well as much non-scientific inquiry) aims.
So, we can have very good reason to believe in claims of necessity, asymptotically approaching but never reaching certainty, even though they extend beyond the range of our experience. The usefulness of the notion can be demonstrated by its conceptual role at the foundations of the scientific enterprise, which has led to countless useful improvements in our ability to predict future occurrences, and to manipulate natural conditions to our own advantage.
Regardless, it is important to underline that claims of necessity are not claims about knowability, verifiability, or usefulness. Those are important and difficult issues, but reasons to be pessimistic about them do not undermine the intelligibility of the notion of necessity. Further, it is crucial to separate them out from necessity, in order to isolate the relevant notion, before we can productively address these questions about how claims to necessity are established or what use they might be.
[§]
Hume is a seminal source of skepticism about metaphysical necessity. He is often read as arguing that such concepts are things that minds impose on, rather than discover in, nature. Nowhere in our experience do we observe causation, law, or necessity; rather, these are just parts of our conceptual repertoire that are useful in helping to organize our beliefs and theories about nature. Necessity is no more out there in the world than are centers of gravity or the equator, according to Humeans. These are one and all human constructions that have no real, objective mind- or language-independent correlatives.15
It is not possible to decisively counter Hume’s objections. We are not in the realm of conclusive proof, when it comes to the hypothesis that notions like law, causation, and necessity do in fact have real, objective mind- and language-independent correlatives. But who should have the burden of proof here: the skeptic or the realist?
Well, there is no shortage of reasons to think that modal concepts are rather important, useful, significant. We will soon turn to showing that certain influential arguments to the contrary can be countered, given due attention to fine distinctions within the philosophy of language, as well as to distinctions between the epistemological and the metaphysical aspects of an issue. For another thing, people do engage in modal speculation all the time, and it seems reasonably clear what they are talking about—that is, how familiar objects and individuals would have fared under different circumstances, or would have endured alterations to their accidental properties. Hume’s point is a valuable one, exposing the lack of conclusive support for many of our most fundamental beliefs in epistemology and metaphysics; but it is a long way from this point to categorical skepticism about any and all claims concerning the metaphysical modalities. (It is a long way from a ‘lack of conclusive support’ to a ‘conclusive lack of support’.)
[§]
Throughout I have been careful to distinguish between the issues of: (i) the coherence of analytic truth, and (ii) the worth of analytic truth. I will likewise arrange this section along these two dimensions, when it comes to engaging with skepticism about analyticity. Further, as for the charge that the notion of analytic truth is incoherent, I distinguish between what I will call ‘the weak coherence charge’ and ‘the strong coherence charge.’ The weak coherence charge is the idea that the analytic/synthetic distinction has never been drawn in a comprehensive and satisfactory way. The strong coherence charge is the much bolder—and, consequentially, more difficult to justify—notion that the very idea of the analytic/synthetic distinction is irredeemably confused.
It would be difficult to deny the weak coherence charge. While one can clearly see progress in the sophistication of discussions of the analytic/synthetic distinction—from Hobbes to Kant to Frege, and on into recent work by, say, Boghossian (1997) and Russell (2008)—still I cannot imagine any proponent of analyticity claiming that the matter has been conclusively laid to rest. Indeed, it is hard to see how there could be consensus concerning the analytic/synthetic distinction until there was consensus concerning the proper theoretical treatment of the notion of meaning, and it is safe to say that neither philosophy nor any of the cognitive sciences is anywhere near such a state. Among other complications, the analytic/synthetic distinction can only be as firm as meaning is determinate (here compare the discussion of the chemical theory of concepts in §3.2). While I see little force in sweeping, global skepticism as to the determinacy of meaning, given the reasonable view that meaning is thoroughly context-sensitive (i.e., that meanings are malleable, relative to, and as mandated by, the context of utterance), any formulation of the analytic/synthetic distinction will have to be programmatic, hedged, shifty, and at least somewhat imprecise.
[§]
However, the strong coherence charge (i.e., that the very idea of the analytic/synthetic distinction is irredeemably confused) is another matter entirely. Could it be conclusively established? When we factor out Quine’s dated behavioristic skepticism about meaning, and the aspects of his arguments which apply not intrinsically to analyticity per se but extrinsically to some of the things to which certain empiricists attempted to apply this notion, what arguments are there for the strong coherence charge?
It must be noted that the degree of even Quine’s commitment to the strong coherence argument is questionable. His allegiance to at least the weak coherence point is firmly established in his (1951: 32), and is nicely stated thus:
I do not know whether the statement ‘Everything green is extended’ is analytic. Now does my indecision over this example really betray an incomplete understanding, an incomplete grasp of the ‘meanings’, of ‘green’ and ‘extended’? I think not. The trouble is not with ‘green’ or ‘extended,’ but with ‘analytic.’
However, there are passages in which it sounds like Quine (1951: 41) endorses something much bolder:
That there is such a distinction to be drawn at all is an unempirical dogma of empiricists, a metaphysical article of faith.
As Grice & Strawson (1956) quickly pointed out, though, the argument in Quine (1951: Parts I–IV) cannot suffice to support the strong coherence charge, without the aid of the virtually unsupportable premise that all possibilities have been exhausted. (That is, it is an elimination argument, that neither A nor B nor C nor D provides satisfactory ways of drawing the analytic/synthetic distinction.) Furthermore, Quine (1991) explicitly stops short of the strong coherence charge, stating there that the problem is not that it is incoherent to claim that (say) ‘All bachelors are unmarried’ is analytic, but rather that such trifles are of no philosophical interest or worth.
In any case, there is an argument for the strong coherence charge in Quine’s corpus, and it has been influential. (Cf., e.g., Boghossian & Peacocke (2000: 4): ‘Our own view is that Quine decisively refuted the idea that anything could be true purely in virtue of meaning.’) The core idea is that ‘no [statement] is true but reality makes it so’ (Quine, 1970: 10).16 Quine’s (1970: 10–12) argument in favor of this dictum runs as follows: The truth-conditions of any statement can be specified ‘as Tarski taught us’ via the schema: ‘S’ is T iff S. The disquoted S on the right hand side of the biconditional stands for a fact, an element of reality; and this is so whether the left-hand quoted ‘S’ expresses a contingency (such as ‘Quine speaks Portuguese’) or a triviality (such as ‘All Portuguese speakers speak Portuguese’). Therefore, no statement is true but reality makes it so.17 (Cf. Boghossian [1997: 334–37] for a very similar argument, inspired by Quine.)
While this argument has an unassailable air, I will argue that it commits a fallacy of equivocation. For ease of exposition, I will abbreviate Quine’s dictum (i.e., ‘no statement is true but that the facts make it so’) as ‘TVF’ (or ‘true in virtue of facts’), and the very idea that a statement might be true in virtue of meaning as ‘TVM’. The structure of the argument, then, is as follows:
P#1: TVF is unassailable.
P#2: TVF is incompatible with TVM.
Therefore, TVM is untenable.
There is a sense of ‘fact’ according to which TVF is unassailable, and a sense of ‘fact’ according to which TVF is incompatible with TVM. However, these are distinct senses of the term. There is no one reading that makes both premises true. Thus, this case against TVM is flawed by a fallacy of equivocation.
Consider first P#1, which Boghossian motivates with the following sorts of consideration:
After all, if a statement is known a priori, then it must be true. And if it is true then it must be factual, capable of being true or false. (1997: 334)
In this sense of ‘factual,’ to be factual is to be truth-evaluable—it is to make a claim, or to have truth-conditions. Note how heterogeneous the correlative conception of a fact must be, metaphysically speaking. For example, that no grandmothers are childless, that unicorns have one horn, that there is an even prime, that Germany is not in Asia, that hydrogen is less dense than gold, that justice is a virtue, that all humans are mortal, that there is no rhinoceros in this room, etc., are one and all facts, in this sense. The facts, in this sense, are as diverse in nature and status as are the indefinite range of things about which humans can think and talk.
This is the sense of ‘factual’ involved in Quine’s (1970: 10–12) Tarski-inspired argument described above. If the criterion for counting as factual is to issue in a truth when plugged into the schema: ‘S’ is T iff S, then it is not a very discriminating property. Since this notion of ‘fact’ does not correspond to any precise or homogeneous metaphysical category, I will call it the ‘Tarski-semantic notion of fact’.
Given this Tarski-semantic sense of fact, P#1 is uncontroversial, but P#2 is far from obvious. Consider, for example:
[1] No grandmother is childless.
It is, to say the least, not clear that there is any tension whatsoever between, on the one hand, the claim that [1] is T iff no grandmother is childless, and, on the other hand, the claim that one might be justified in believing [1] based solely on semantic intuition—that is, on understanding what it takes to be a grandmother—as opposed to based on empirical investigation or mystical Platonist intuition. That a sentence issues in a truth when plugged into the Tarskian schema (‘S’ is T iff S) is entirely irrelevant to questions about what it takes to understand or justify the proposition expressed. (Otherwise, that would spell trouble for a Tarskian approach to truth, one of whose key virtues is the clear distinction between the semantic concept of truth and, say, epistemic concepts like justification.) Thus, if we read ‘fact’ in this Tarski-semantic way, then, TVF is compatible with TVM. To say that a statement is truth-evaluable does not rule out any possibilities as to what counts as understanding it, or how one might be justified in believing that it is true. On this Tarski-semantic reading of ‘fact,’ then, P#2 should be rejected.
Burge (2000: 16) affords another way to articulate this problem with P#2. When Leibniz or Frege contrast a priori truths of reason with a posteriori truths of fact, according to Burge, the point of the contrast is not that a priori truths are not factual, but rather that they are not merely factual. The claim is that such truths are not subject to refutation by contingent happenstance, not that they are entirely and categorically unrelated to contingent happenstance. So, [1], for instance, could be not merely factual, but not thereby non-natural, otherworldly. Grasp of its meaning might suffice to justify belief in its truth, but nonetheless the statement is still about our own flesh-and-blood grandmothers. So, again, it appears that P#2 is seriously flawed. A statement could be at once both factual, in the Tarski-semantic sense, while also being not merely factual, in the Leibniz-Frege sense. Thus, one and the same statement could be Tarski-semantic-factual and yet still be TVM.
There is a distinct, more metaphysically robust, conception of fact, given which P#2 fares much better. It is a long way from Tarskian schemas, though, closer to what Armstrong (1996) calls ‘truth-makers,’ or to Russell’s (1918) and Wittgenstein’s (1921) logical atoms. Facts in this sense are discrete mind-independent entities to which (typically: contingent, empirical) statements stand in some specific semantic relation (such as representation). Famously, Russell and Wittgenstein quarreled over whether it is a fact, in this sense, that there is no rhinoceros in the room. With respect to metaphysical worries about this variety of fact, negative existentials are just the tip of the iceberg. Russell (1918) seems to have never been able to convince himself of the existence of such general facts as that all humans are mortal, though he recognizes that his current views of meaningfulness seem to commit him to such entities. (Cf. Lewis [1998] for some related objections to Armstrong [1996].18) Russell (1918: Lecture 3) reports having ‘nearly produced a riot’ at Harvard in 1914 by arguing for the existence of negative facts. Surely, the claim that ‘“There is no rhinoceros in this room” is T iff there is no rhinoceros in this room’ would not have provoked such a reaction—even if the pragmatists would have disputed it.
In this metaphysical sense of ‘fact,’ there is definite tension between TVF and TVM.19 To be factual, in this metaphysical sense, is to represent a (typically: contingent, empirical) state of affairs; for any statement that represents a (contingent, empirical) state of affairs, there is reason to think that the ‘TVM’ label is probably inappropriate.20 The price of this strategy of saving P#2 from imposing a false dilemma, though, is P#1. If we read ‘factual’ in this metaphysical sense, then TVF is eminently assailable, as is evidenced by Hume’s (1748) reasons for positing relations of ideas in the first place, by Russell’s (1918) struggles with negative and general facts, by Lewis’ (1998) criticisms of Armstrong (1996), etc. In short, in order to make P#2 true, you have to think of TVF and TVM as mutually exclusive answers to an Armstrong-style demand for truth-makers. This issues an understanding of TVF that may well be interesting and even defensible, at least for a broad class of statements, but is certainly assailable. So, given this second, metaphysical sense of ‘fact,’ P#1 is controversial, not the sort of thing to which one can help oneself on the strength of vague slogans, or disquotational schemas.
To sum up: Quine (1970: 10–12) and Boghossian (1997: 334–37) provide clear illustrations of a prevalent line of thought that is widely but mistakenly thought to spell the end for TVM. The semblance of a compelling case against TVM depends on taking ‘factual’ in the Tarski-semantic sense in P#1 while taking it in the metaphysical-truth-maker sense in P#2. Once we recognize and guard against this slide, either P#1 is extremely contentious (if we adhere to the metaphysical-truth-maker sense of ‘factual’) or else P#2 is false (if we adhere to the Tarski-semantic sense of ‘factual’). So, there is no cogent case against TVM forthcoming down this avenue.
Given that these arguments against the coherence of TVM are lame, and provided that we have an account of analyticity that unequivocally rejects the notion that meaning has supernatural truth-making powers, analyticity may yet hold some promise to provide a compelling account of justification, for at least some of our universal, necessary knowledge.
[§]
One important and pertinent point which Quine (1951) is commonly taken to have established is that ‘analyticity’ admits of no satisfactory non-circular, reductive conceptual analysis. That is, analyticity cannot be reduced to, or defined purely in terms of, some other less obscure notion. Hence Quine is often taken to have demonstrated that there is something deeply suspect about appeal to this notion. It might even be taken to be another sort of coherence argument: analyticity can play no role in a rigorous philosophical theory, because there is no acceptable, non-circular definition of the notion.
For example, if we had a satisfactory criterion for synonymy, then we could get from there to a satisfactory definition of analyticity. (That is, if it were possible to conclusively establish that ‘bachelor’ is synonymous with ‘unmarried man,’ then it could be demonstrated that it is analytic that all bachelors are unmarried men.) Vice versa, if analyticity were unproblematic, then a crisp, clear definition of synonymy would be forthcoming. But one cannot help oneself to synonymy in definiendum analyticity, so the objection goes, since the defiendium is equally as obscure and suspect as the definiens.
First: Note the role that meaning-skepticism plays here. This line of argument draws force from the sentiment that meaning is indeterminate, inscrutable, and hence that synonymy is deeply suspect. Again, as pointed out above, this line of argument is considerably less weighty in the wake of the cognitive revolution.
Second: Of course, lots of (probably, most?) concepts cannot be reduced to others without remainder.21 Why think that they ought to be? Why hold ‘meaning’ to a higher standard than the rest of the lexicon? This elementary status should not be all that surprising, and is hardly a defect. In general, there is no good reason to hold that that renders them suspect, or second-rate. For analyticity, as for any other concept, showing that it cannot be reduced to any other concepts is far from sufficient for showing it to be incoherent (without the aid of a premise, or presumption, that there is something deeply suspect or second-rate about meaning).
[§]
On these grounds, I take coherence of analyticity to be a reasonable tenet. Assuming that we all concede the weak coherence charge, the live coherence dispute is then over the question of whether the analytic/synthetic distinction is worth further investigation. There is weighty philosophical-historical precedent, as well as the brute data, in favor of further pursuit of a comprehensive, satisfactory conception of analyticity. There are no compelling arguments to the contrary of which I know—provided that we are careful about boundaries between metaphysics, semantics, and epistemology, and pay clear-headed heed to fine semantic distinctions between questions that pertain to the conventional links between S and p and questions that pertain to the truth-conditions of p.
Some prevalent arguments against analyticity do not in fact bear upon its coherence or worth. The notion has not been proven incoherent, and it still may well be our best hope for an adequate, non-obscure solution to the problem of a priori knowledge—among other things.
A reason to retain analyticity is that the brute data is not going away. Given the lack of other viable accounts of universal, necessary knowledge, this avenue is still eminently worth further exploration.
There remains of course the massive and daunting challenge of revisability. This is a good segue to chapter 4 as that issue is also deeply pertinent to the notion of a priori knowledge. On, then, to chart the a priori.
NOTES
1. Cf. Cartwright (1983) for thorough recent development of this line of thought.
2. Cf. especially Kripke (1972: Lecture II). Roca-Royes (2011) is a good general discussion of these issues.
3. Cf. Lewis (1986) for development and defense of an anti-haecceitist approach.
4. Cf. Kripke (1972, 41–53; 1980, 15–20).
5. See Boghossian (1997) and Sullivan (2008) for further justification for taking the statement as the primary bearer of analyticity.
6. Williamson (2007, 2014) and Hawthorne (2012) have recently argued for some such conclusion about a priority; but even still, there remains a vast gap between it and analyticity (let alone necessity) in this respect.
Note that I do not mean to imply that all or even most philosophers dismiss analyticity as a worthless notion. To the contrary, consider for example Grice’s (1987: 344) claim that the analytic/synthetic distinction is ‘one of the most important topics in philosophy, required in determining, not merely answers to particular philosophical questions, but the nature of philosophy itself’.
7. I argue this at length in Sullivan (2008). Grice & Strawson (1956) made an early, forceful case in favor of this claim, and for a varied sample of subsequent supporters of this claim cf. Fine (1994), Katz (1997), and Gertler (2002).
8. For some recent statements cf. Sullivan (2003a) and Cappelen & Lepore (2005). Here is a classic statement of the core idea from Frege (1892b: 46):
Nowadays people seem inclined to exaggerate the scope of the statement that different linguistic expressions are never completely equivalent, that a word can never be exactly translated into another language. One might perhaps go even further, and say that the same word is never taken in quite the same way even by men who share a language. I will not enquire as to the measure of truth in these statements; I would only emphasize that nevertheless different expressions quite often have something in common, which I call the [meaning], or in the special case of sentences, the [proposition]. In other words, we must not fail to recognize that the same [meaning], the same [proposition], may be variously expressed. … It is possible for one sentence to give no more and no less information than another; and, for all the multiplicity of languages, mankind has a common stock of [propositions]. If all transformations of the expression were forbidden on the plea that this would alter the content as well, logic would simply be crippled; for the task of logic can hardly be performed without trying to recognize the [proposition] in its manifold guises. Moreover, all definitions would then have to be rejected as false.
As with some other excerpts, I have freely substituted some of Frege’s terms (e.g., ‘sense’, ‘thought’) with similar terms (e.g., ‘meaning’, ‘proposition’), in order to better fit this excerpt with the broader discussion.
9. Cf. Sober & Hylton (2000) for development of a similar theme. For an introduction to the cognitive revolution, cf. Pinker (2011).
10. For discussion see Munsat, ed. (1971: 19–20).
11. Again, as with some other excerpts, I have freely substituted some of Coffa’s terms (e.g., ‘concept’, ‘representation’) with similar terms (e.g., ‘meaning’, ‘proposition’), in order to better fit this excerpt with the broader discussion.
12. See Sullivan, ed. (2003) for extensive treatment of this allegation. The core idea is that, whereas traditional logic viewed the content of any significant proposition along the lines of:
[All/some/no] S [is/is not] P
Various nineteenth-century (and hence post-Kantian) logicians discovered that this does not capture the correct logic of various sorts of propositions.
13. See especially Frege (1884) and Quine (1951) for seminal statements, and Coffa (1991: Part 1) for discussion.
14. For critical discussion of Russell (2008) cf., for example, Wilkfors (2008), Boghossian (2010).
15. Quine is commonly thought to be a leading proponent of the stronger charge against the coherence of metaphysical necessity. However, Quine’s target is not the intelligibility of metaphysical necessity per se, but rather the prevalent early-twentieth-century empiricist idea that all necessity reduces to analyticity, and that that affords a decisive solution to Plato’s problem. For good discussions cf. Kaplan (1986), Neale (1990: Ch.4), Marcus (1991).
16. Here and throughout, to fit the terms of the present discussion, I use ‘statement’ in place of Quine’s nominalist adherence to the term ‘sentence’.
17. In additions to Boghossian (1997), Cassam (2000) is another recent writer who is clearly influenced by this argument. The case of Boghossian is complicated, though. He distinguishes between metaphysical and epistemological conceptions of analyticity, and argues that while Quine does present a conclusive case against the metaphysical conception, it does not affect the worthwhile and valuable epistemological conception. While I agree with much of what Boghossian has to say about what he calls ‘the epistemological conception of analyticity’, there is much that I disagree with here. First of all, analyticity is a semantic notion, and so should be firmly distinguished from both metaphysical and epistemological concepts. Second, as mentioned above (in §2.1) , I think that what Boghossian calls ‘the metaphysical conception of analyticity’ is completely a straw target anyway—no one that I can think of believes that our meaning p by S makes it the case that p.
18. Note that something like Quine’s (1970: 10) dictum that no statement is true but reality makes it so explicitly plays a role in Armstrong’s (1996) case for positing truth-makers.
19. It is in this second, metaphysical sense of ‘fact’ that the approach to a priori knowledge associated with Hume (1748) and Ayer (1936) is not unfairly glossed as the view that a priori knowledge is devoid of factual content. (I take it that Hume and Ayer are saying something different from the above-discussed view that a priori knowledge is not merely factual.) That is, what Hume and Ayer claim is that the likes of:
[2] All squares have four sides.
do not rule out any contingent empirical possibilities, and this explains why they are not subject to refutation by contingent happenstance. Obviously, but nonetheless crucial for present concerns, the Hume-Ayer claim is certainly not in the slightest tension with anything along the lines of: [2] is T iff all squares have four sides. Hume and Ayer have no reason or inclination to deny that such statements are Tarski-semantic-factual.
20. Though this does accord with Hume (1748), and the wealth of tradition, it might run afoul of Kripke (1972) on the contingent a priori. In any case, as I explain immediately below, my defense of TVM in no way depends on siding with Hume against Kripke on the contingent a priori, because P#1 is false on this second understanding of ‘fact’ anyway. (There is more on the contingent a priori in Part IV.)
21. Compare the remarks about ‘truth’ (or ‘good’, or ‘art’) in §2.1.