Modal Revisionism and Moderate Externalism
The ultimate goal of Part III is to further ground and develop the constitutive a priori as a comprehensive, satisfactory stance on Plato’s problem. chapter 5 has gotten the ball rolling, when it comes to how to absorb revisability and externalism into a constitutive a priori framework. The aim of this chapter is to further develop the contours of my modal revisionist response to the challenge of revisability and of my moderate externalist response to the externalist’ challenge. Further, the significant confluences between these responses will be highlighted and drawn out.
I begin in §6.1 with a more detailed sketch of some (connected, overlapping) versions of the constitutive a priori orientation. Next, in §6.2, I give a thorough treatment of a serious challenge to that orientation. Then §6.3 integrates the revisionist response to the challenge of revisability, §6.4 covers some terrain which overlaps both revisability and externalism, and §6.5 develops the moderate response to the externalist challenge. Finally, §6.6 concludes Part III by mapping my own favored version of the constitutive a priori view.
§6.1: some versions of the constitutive a priori
I will distinguish three (overlapping, related) versions of the constitutive a priori orientation as (i) the Wittgensteinian, (ii) the Carnapian, and (iii) the Kantian. (That is a bit awkward, as (i) and (ii) contain heavy doses of Kantianism; but the third is distinguished by its more explicit commitment to transcendental philosophizing.) Among the core ties that bind them all together are, first, the idea that a priority is at least in part a matter of status, function, or role, as opposed to marking off some categorically different kind of content or faculty of mind, and, second, that it is a necessary condition for any inquiry that some rules or principles have some such status.
[§]
Our story begins with Wittgenstein’s struggles within the philosophy of logic, prior to the World War I, and traces the continued development of the core idea that there is a crucial difference in status between ‘rules’ and ‘propositions.’ The latter bookend will be provided by excerpts from Wittgenstein (1969), his most sustained treatment of epistemological questions, written in the last 18 months of his life. The present aim is to document Wittgenstein’ seminal and enduring commitment, stretching between those bookends, to a constitutive a priori approach to Plato’s problem.1
It may strike a discordant note to portray this most anti-historical and non-traditional of philosophers as struggling with Plato’s problem. However, Wittgenstein certainly engaged—throughout his entire career—with the notion of immunity to counterexample as it crops up in the philosophy of logic. In the Notebooks 1914–1916 (1/6/15), Wittgenstein puts this struggle in familiar terms: ‘The great problem around which everything that I write turns is: Is there an order in the world a priori, and if so what does it consist in?’ Like Kant (even if not explicitly inspired by Kant), he develops an original answer to that question, which self-consciously differs from the more familiar varieties of rationalism, empiricism, and skepticism.
I quote next an excerpt from Coffa (1991: 163), which describes some themes which are germane to this ongoing discussion:
What justifies the logical laws and rules? … In Wittgenstein’s view, an accurate diagnosis of the situation has been prevented by the propositionalist prejudice, by the insistence on looking at logic and the other a priori disciplines as being expressed in statements that convey facts, just like any a posteriori statement—except that the facts in question are somehow otherworldly. If logical laws were essentially like all other statements, they would require some sort of justification, as every statement does. … Wittgenstein’s solution is to say that logic … has no justification [in the sense that] it cannot be conveyed by means of claims. There isn’t a ‘hard’ domain of a priori truth and a ‘soft’ domain of a posteriori truth. … This does not eliminate the a priori but rather locates it as a ‘hardness in the soft.’
The allusion toward the end is to another quote from the Notebooks (1/5/15): ‘My method is not to sunder the hard from the soft, but to see the hardness in the soft.’ Plato’s dualistic metaphysics is the classic example of sundering the hard from the soft, and Russell (1912) supplies some fairly pristine examples of its continuing prevalence (one instance of which was cited in §4.2).
Let’s take ‘the propositionalist prejudice’, as it pertains to the philosophy of logic, to be the view that all claims are to be treated equally, when it comes to justification—that is, they are justified iff they correspond to some mind- and language-independent fact, which is their linguistically determined truth-condition. With a few pinches of salt, propositionalist justification can work for many varieties of statement (‘The cat is on the mat,’ ‘2 + 2 = 4,’ ‘F = MA,’ ‘Murder is wrong’). However, in the philosophy of logic, it forces a (familiar enough) choice between an obscure Platonism and an unsatisfactory relativism. For cases like &-elimination (Φ & Ψ, ∴Φ), for example, it seems evident that we are justified in believing it to be completely immune to counterexample. Is that because of some eternal Platonic super-fact? Or is that semblance mere illusion?
Wittgenstein charts a third option. What if &-elimination is a rule, not a proposition? The reason why that pattern of inference will never lead you astray is not that there is anything magical about the proposition expressed, but rather that it encapsulates a rule which constitutes what it means to employ ‘&.’ (Justification in this case is more like exceptionless generalizations regarding bishops moving diagonally in chess, than like staking a claim which may or may not turn out to be subject to counterexample.) Immunity to counterexample in logic is more a matter of consequences of implicit definitions, than the discovery of a mind- and language-independent law.2 Consider a few other quotes from the Notebooks, articulating this idea:
(2/9/14): It must in a certain sense be impossible for us to go wrong in logic.
(8/9/14): The ‘self-evidence’ of which Russell has talked so much can only be dispensed with in logic if language itself prevents every logical mistake.
On into the Tractatus we get the following ideas, already discussed in §5.2 above:
5.473: In a certain sense, we cannot make mistakes in logic.
5.4731: What makes logic a priori is the impossibility of illogical thought.
Logical truth has the status of the cement holding together the foundations of a language game, of thereby defining what counts as an intelligible move within the game.
This early Wittgensteinian philosophy of logic is a clear and seminal case of the constitutive a priori. What is distinctive about logic is its status. Immunity to counterexample in logic is a matter of the role or function which certain axioms play in the framework; and this orientation stiff-arms many traditional metaphysical questions about the (otherwordly) nature of (certain kinds of) truth-makers. (If there is a truth-maker, it is not a hunk of being [i.e., some or other sort of fact, or state of affairs] but rather resides in the rule’s playing a certain role or having a certain status.)
Further, this thread runs throughout Wittgenstein’s entire career, as the a priori gets embodied in the guise of ‘grammar’ in the middle works, and on into its most extensive development in On Certainty. One of the things for which that work is known is Wittgenstein’s ‘Moorean turn’—Wittgenstein reportedly (1969: vi) loved Moore’s (1925) ‘Defense of Common Sense,’ and engages with that essay in various ways at various points. In terms of the a priori, the Moorean turn is amenable to Wittgenstein for its seeing the hardnesss within the soft—that is, its exploration of various senses in which various sorts of seemingly empirical matters of fact should also be seen as immune to counterexample. One of the directions in which Wittgenstein takes this line is toward the conclusion that a priori and a posteriori are not discrete monoliths, but rather come blended together in various degrees, among language games and world pictures:
§52: There isn’t a sharp boundary line between [rules and propositions].
§308: Not everything that has the form of an empirical proposition is one.
§309: Rule and empirical proposition merge into one another.
§318: There is no sharp boundary between methodological propositions and propositions within a method.
§319: But wouldn’t one have to say, then, that there is no sharp boundary between propositions of logic and empirical propositions? The lack of sharpness is that of the boundary between rule and empirical proposition.
§454: There are cases where doubt is unreasonable, but others where it seems logically impossible. And there seems to be no clear boundary between them.
On Certainty also instances, related to this Moorean turn, the second aspect of the contingent a priori view, in addition to this first point about status—namely, the idea that holding something to that status is a necessary condition for any inquiry. (This second aspect is especially stressed in version (iii) below, which I am calling the ‘Kantian’ constitutive a priori). Namely: it is a necessary condition for any intelligible, systematic inquiry that some things get afforded this special status, function, or role:
§337: One cannot make experiments if there are not some things that one does not doubt. But that does not mean that one takes certain presuppositions on trust.
§341: That is to say, the questions that we raise and our doubts depend on the fact that some propositions are exempt from doubt, and are as it were like hinges on which those turn.
§342: That is to say, it belongs to the logic of our scientific investigations that certain things are in deed not doubted.
§343: We just can’t investigate everything, and for that reason we are forced to rest content with assumption. If I want the door to turn, the hinges must stay put.
§415: Certain propositions underlie all questions and all thinking.
§450: A doubt that doubted everything would not be a doubt.
§509: I really want to say that a language game is only possible if one trusts something (I did not say ‘can trust something’).
This is a core pillar of the constitutive a priori orientation—that is, that treating some things as having the status of immunity to empirical counterexample is a necessary condition for any rational inquiry. This is perhaps most stark (though as always, under-explained) here:
§425: Although ‘I cannot be making a mistake about it’ that still does not entail that ‘I am infallible about it’.
(This is similar to what Poincare says about the status of a definition in geometry, as we will again see below.)
Relatedly, the [UJ] connections, which form the spine of understanding-based approaches to a priority, are also evident in On Certainty:
§80: The truth of my statements is the test of my understanding of these statements.
§83: The truth of certain empirical propositions belongs to our frame of reference.
§§96–99 blends all of these diverse strands together, in the course of developing some of the most widely cited metaphors from On Certainty:
It might be imagined that some propositions, of the form of empirical propositions, were hardened and functioned as channels for such empirical propositions as were not hardened but fluid; and that this relation altered with time, in that the fluid propositions hardened, and hard ones became fluid. The … river-bed of thoughts may shift. But I distinguish between the movement of the waters on the river-bed and the shift of the bed itself; though there is not a sharp division of the one from the other.
But if someone were to say ‘So logic too is an empirical science’ he would be wrong. Yet this is right: the same proposition may be treated at one time as something to test by experience, at another as a rule of testing. And the bank of the river consists partly of hard rock, subject to no alteration or only to an imperceptible one, partly of sand, which now in one place now in another get washed away, or deposited.
To summarize: one thread which runs through the entirety of Wittgenstein’s career is a seminal commitment to a constitutive a priori orientation, when it comes to justification in the philosophy of logic. Wittgenstein is clearly a proponent of both, first, the idea that a priority is (at least in part) a matter of status, function, or role, as opposed to marking off some categorically different kind of content or faculty of mind, and, second, that it is a necessary condition for any inquiry that some rules or principles have some such status. There is a strong dose of context-sensitive framework-relativity evident in Wittgenstein’s distinctive orientation towards Plato’s problem.
One yawning avenue into which I have not yet turned is, namely, that the specter of relativism haunts—that is, the worry that Wittgenstein’s later philosophy is inhospitable to, or even incompatible with, the very idea of mind- and language-independent objective truth. I should not wade into that corner of Wittgenstein scholarship here—it is perhaps enough to say that while there is ample textual evidence of a variety of relativism in Wittgenstein (1969, as in some other places), there are also fairly categorical disavowals of any strong form of relativism (e.g., 1969: §§108, 317, 336).3 In any case, the specter of relativism haunts all non-absolutists. Our primary interest in this visit with Wittgenstein is his relevance to the constitutive a priori position on the questions of metaphysical, semantic, and epistemic immunity to counterexample. Elements of Wittgenstein’s life-long engagement with the philosophy of logic will be seen to permeate and influence lots of subsequent work on those matters.
[§]
The two other versions of the constitutive a priori picture are also closely related in various ways, but are rooted more firmly in the philosophy of science. (Wittgenstein’s anti-scientism was notorious, and another constant throughout his complex career.4) In particular, the drastic episodes in the history of ideas known as scientific revolutions provide some clear examples of the more mundane phenomenon which I am calling conceptual evolution.5 Reichenbach (1920) and Pap (1946) are two early proponents of a certain neo-Kantian line of response to scientific revolutions which bear the key hallmarks of the constitutive a priori orientation. This line develops into Carnap’s (1937, 1950) well-known account of frameworks and its attendant distinction between internal and external questions.
This terrain was already charted, in a preliminary way, above in §5.2; it will be more extensively developed throughout the rest of this chapter. The primary present point is to document the allegiance to this variety of constitutive a priori orientation on the part of Reichenbach (1920) and Pap (1946), as well as fellow-travelers Poincare and Carnap.
Chronologically, this strand begins with Poincare’s (1899, 1900) conventionalism (which, as remarked above, overlaps with Wittgenstein’s philosophy of logic). Famously, Poincare rejects the idea that geometric axioms express propositions; they are rather on his view disguised definitions which are meaning-constituting. In this respect, Poincare was a proto-framework-relative constitutivist. Coffa (1991: 140) and Friedman (2000: 376; 2007: 99), for example, both cast Poincare’s view as very much akin to Carnap’s (1950)—that is, what appears to be an arbitrary convention from the outside at the same time appears unshakably necessary from the inside:
Thus, in the case of a priori claims, one and the same linguistic form may be seen as playing two radically different … roles: When regarded from outside a linguistic framework, it must be seen as … a definition in disguise; when regarded from within the defined framework, that very sentence now expresses a claim, one true in virtue of the constituted meanings and therefore necessary. (Coffa 1991: 140)
Like more or less any understanding-based approach to a priority, the constitutive a priori orientation certainly contains an element of this kind of conventionalism—that is, Poincare’s ‘convention,’ akin to Wittgenstein’s ‘rule,’ is precisely a special status or role.
Our next step, after Poincare, is Reichenbach’s (1920) effort to absorb the theory of relativity into a broadly Kantian epistemology. Consider the following morals:
The doctrine of the a priori has been transformed into the theory that the logical construction of knowledge is determined by a special class of principles, and that this logical function singles out this class, the significance of which has nothing to do with the manner of its discovery or the duration of its validity. (p. 94)
‘A priori’ means ‘before knowledge,’ but not ‘for all time’ and not ‘independent of experience’. (p. 113)
Here we see an even more explicit statement than in Poincare of the move to treating a priority as a matter of status, not just of content, or of faculty of mind.
Moving on into Pap (1946), who explicitly builds on both Poincare and Reichenbach, we get a kind of Wittgen-Moorian context-relativity, as the soft hardening over time:
If … our point of view is dynamic or developmental, we shall find that what were experimental laws at one stage come to function, in virtue of extensive confirmation by experience, as analytical rules or ‘conventions’. (vii)
A priori is characterized in terms of functions which propositions may perform. … A proposition which is a priori in one context of inquiry may be a posteriori in another context. (viii)
A priority is a matter status, and status is a highly context-sensitive notion. (The importance of this temporal dimension is further explored below at §6.3, and will significantly influence the final maps to be drawn in Part IV.)
Pap segues smoothly into Carnap’s (1937, 1950) better known work on framework-relativity. The conventional-from-the-outside dimension, with its attendant specter of relativism, is there loud and clear: ‘In logic there are no morals’ (Carnap 1937: 52), and the scientific spirit in philosophy demands tolerance as to the proliferation of linguistic frameworks (1950: 40). As Friedman (2000: 371) puts it, for Carnap at this stage: ‘All standards of “correctness,” “validity,” and “truth” … are relative to … linguistic framework. … Such rules are constitutive of the concepts of “validity” and “correctness.”’
The work of explaining how and why this conventional-from-the-outside is also, seen from the inside, a robust and significant kind of immunity to counterexample a primary aim of the rest of this chapter. How and why is this not just a nihilistic relativism? How does it not fall prey to some of the daunting challenges to a pure conventionalism in the philosophy of logic (e.g., Quine [1936], Prior [1960])? Answering that will necessitate delving into the complexities of conceptual evolution, and into the shifting sands around the border between change of language and change of theory.
[§]
Version three is the most explicitly Kantian take on the idea; its most widely read proponent is Friedman (1992, 2000, 2001, 2011), but other proponents include DiPierris (1992), DiSalle (2002), Richardson (2002), and Franco (2011). While elements of Wittgenstein, Reichenbach and others discussed above are evident here as well, the explicit and self-conscious Kantian talk of transcendental philosophy as a required a priori supplement to scientific theories provides something of a third variation on this theme, in addition to Carnap’s (1950) frameworks and Wittgenstein’s (1953) language games or (1969) world pictures.
The guiding analogy is to space and time in Kant (1781), enjoying a status undreamt of by previous rationalists and empiricists—that is, transcendental preconditions for experience. Space and time, for Kant, are more like rules of the game than like Platonic otherworldly superfacts or Humean inductions from experience. To experience the happening of an event in space and time is, quite simply, constitutive of experiencing it at all. In the constitutive a priori tradition, this distinctively Kantian status is then transposed from preconditions for experience to preconditions for intelligible, tractable inquiry. Here we see the key respect in which the constitutive a priori is a neo-Kantian orientation—Kant’s innovative move in epistemology is precisely about a new status, previously undreamed of in the long history of rationalism-empiricism debates.
In Friedman’s (2000, 2001) version of this general program, he distinguishes in status three different levels or components of scientific inquiry: (1) at the base level, we have concepts and principles of empirical natural science; (2) at the second level, there are the constitutively a priori principles, within which alone empirical testing at the base level is possible; (3) at the third level, there are philosophical meta-paradigms or meta-frameworks.6
The relations between (1) and (2) are familiar enough by now—as this point has been explicitly with us since at least §5.2. Other examples of the myriad ways in which some things must be held fast in order to generate and intelligibly isolate specific questions are discussed by Friedman (2000), who develops at some length the examples of the relations between the mathematics and the physics in both Newton’s and Einstein’s epochal scientific innovations. It is the (2)–(3) relations which are distinctive here, and which proponents take to be where this third Kantian sort of constitutivist moves beyond conventionalism:
It is precisely here that we need to move beyond both Carnap and Kuhn, by describing a fundamentally new function for what Kant called transcendental philosophy. (379)
The enterprise that Kant called transcendental philosophy—the project of investigating and philosophically contextualizing the most basic constitutive principles defining the fundamental framework of empirical natural science—plays an indispensable orienting role with respect to conceptual revolutions within the sciences … thereby makes available prospective notions of inter-framework rationality. (382)
Friedman sees his Kantian constitutivism as an advance over previous versions, on which it is a further development, precisely for its stance on progressive relations between frameworks.
Relatedly, and significantly, these most recent, most Kantian constitutivists understand a large part of their job to be to explain why a seamless Quinean holism cannot afford an adequate epistemology in general, or account of the history of science in particular. Friedman (2000: 376) says: ‘It will not do … to view … the constitutively a priori parts of our scientific theories as simply relatively fixed or entrenched … relatively difficult to revise.’ In Stump’s (2011: 188) terms: ‘The constitutive principles are not merely more entrenched—they have never been directly tested.’
Akin to their Wittgensteinian and Carnapian allies, then, these Kantian constitutivists in the philosophy of science also clearly adhere to these core tenets of: (i) a priority is a matter of status, role, or standing, not merely a special kind of content or faculty of mind, and (ii) a priority thus understood is a necessary condition for any intelligible inquiry.
§6.2: a case study in understanding and justification
It is clear from chapter 5 that the [UJ] connections and principles are absolutely essential, the spine of any understanding-based approach to a priority. The next issue then is a long focused case study into what is perhaps the most significant prima facie counterexample, threatening to undermine these connections. Working through this challenge is not only important in itself, but will also help to hone some distantly related contours of the constitutive a priori orientation.
I will use Conjunction Elimination—the inference from a conjunction to one of its conjuncts—as my stock example of a pattern of inference which is safely known to be valid:
Hence, any specific instance of [&E] counts as a safely known logical truth. Some of the perennial questions within the epistemology of logic are due to the sense that such knowledge exhibits a remarkable immunity to counterexample: that is, it is not just that I have yet to encounter a situation in which a conjunction failed to entail one of its conjuncts (which would be remarkable enough, to be sure), but, further, there is the atavistic intuition that such a scenario would be both epistemically inconceivable and metaphysically impossible. And hence, questions about the epistemology of logic are entangled with some rather large philosophical issues, such as a priori knowledge and necessary truth.
The [UJ] connection is perhaps most strongly evident in the case of logical truths. (Surely the claim that one who understands ‘&’ is thereby justified in believing an instance of [&E] to be a logical truth is safer than Anselm’s claims about what is entailed by understanding the concept of God!) According to this approach to the epistemology of logic, one’s justification for holding that [&E] is a valid pattern of inference is grounded in one’s grasp of what ‘&’ means. (Alternatively, holding that [&E] is immune to counterexample is a necessary condition for competence with the concept of conjunction.) Versions of this understanding-based epistemology of logic are rather ubiquitous—for example, variants can be found in the work of both Leibniz and Hume—well prior to Poincare or Wittgenstein—and instances of it are recently developed as a version of ‘rationalism’ by Peacocke (2000, 2004) and as a version of ‘empiricism’ by Boghossian (1997, 2000, 2011).
Given both the prevalence of this approach to the justification of logical truth, and the sense that logical truth is perhaps the most viable case of this core [UJ] connection, the deviant logician objection threatens to wreak considerable havoc in the house of philosophy. For the claim pressed by the proponents of the DLO is precisely that even here in the pristine confines of pure logic, understanding falls decidedly short of affording justification. (And see Williamson [2008] for a sustained attempt to draw out sweeping, revisionary meta-philosophical conclusions from the DLO.)
As preliminary, I should sketch (at least a little bit) what makes for ‘deviance’ in logic. There is general consensus as to what constitutes ‘standard’ or ‘classical’ logic, fundamental tenets cementing the foundation of the enterprise, which hold constant from Aristotelian categorical logic through (and beyond) modern propositional and predicate logic. Core here are the Law of Excluded Middle and the Law of Non-Contradiction:
Deviant logics are those which transgress such standard, classical tenets. Thus understood, deviant logics are hardly a novel phenomenon: the idea that LEM is subject to counterexample (for future contingents, say, or conditionals with a false antecedent) was fairly prevalent throughout Ancient and Medieval philosophy. However, the monolithic status of standard, classical logic is more drastically under siege in the current era than at any previous time. Intuitionist logics, many-valued logics, and fuzzy logics are some fairly well-known, fairly recently well-developed logics which categorically reject LEM, as do many contemporary theories of vagueness. There are paraconsistent logics which develop the idea that rejecting LNC is the best way to handle the semantic paradoxes (e.g., ‘This sentence is false’), among other phenomena, and quantum logics also reject LNC. In these liberal times, it is even fairly common and plausible to work with different logics for different purposes, in different contexts.7
Of course, in logic (as in life) deviance has its price. Perhaps most notably, one cannot have proof by contradiction (a.k.a. indirect derivation, reductio ad absurdum) without LEM, and many pillars of both logic and mathematics have as yet only been proved in this way. Before getting back to our main themes, I will quote Quine’s statement of the (open-minded but just barely so) orthodox party line on this question:
[L]et us not underestimate the price of deviant logic. There is a serious loss of simplicity, especially when the new logic is not even … truth-functional. … [T]he price is perhaps not prohibitive, but the returns had better be good. (1970: 86)
[§]
The primary target of the DLO, then, is this general category of views which ground justification for logical truths in understanding their components:
[UJLT]: our justification for logical truths is grounded in our understanding of their constituent concepts
(This is a relatively bulletproof, bedrock instance of the [UJ]s.) The strategy is to undermine [UJLT] by offering counterexamples to the following putative corollary:
[UJLT corollary]: competent agents who share the same understandings of logical primitives could not coherently disagree as to whether something constructed out of commonly-shared primitives counts as a logical truth.8
To the extent that one can motivate the notion of disagreement about what ought to be counted as logical truths, among those with a shared understanding of the meanings of the constituent primitives, one thereby motivates skepticism that understanding could suffice for justification.
Now, one possible line of response to the DLO would be to question whether [UJLT corollary] really is entailed by [UJLT]. However, for present purposes, I will concede the corollary. Prima facie, it seems that proponents of [UJLT] must classify any disagreement as to whether something ought to count as a logical truth as ultimately stemming from one of the following two sources: (i) at least one party falls short of a competent, comprehensive grasp of one of the relevant concepts, or (ii) at least one of the relevant concepts is ambiguous, understood in different senses by the different parties. In referring back to these, I will call (i) ‘the incompetence option,’ and (ii) ‘the ambiguity option.’
All versions of the DLO, then, argue that there can be disagreements about logical truth which involve neither incompetence nor equivocation. One version of the objection can be found in Horwich (2000: 158–59, 2006: Ch. 6), focused on disputes between an intuitionist and a classical logician about whether instances of LEM should be counted as logical truths. I take it that it would not be remotely satisfying for a defender of [UJLT] to avail of the incompetence option—that is, to just insist that the intuitionist ipso facto lacks a competent grasp of the classical concepts of negation or disjunction. To the contrary, the intuitionist challenge presupposes a grasp of those concepts, and rejects some of their relatively unpalatable consequences.
A defender of [UJLT] might attempt to take refuge in the ambiguity option—that is, the idea that, as a result of their disagreements about what ought to count as valid, the intuitionist ends up with distinct concepts of negation, disjunction, etc. However, this option too encounters some complications. For example, the ambiguity objection does not seem to be strong enough to quell the worry. What resources does the classical logician have to handle an intuitionist who obstinately insists, in the face of the ambiguity objection, ‘NO! I mean exactly what you do by the terms “negation” and “disjunction”?’ Nothing, it seems, but the fallback to the incompetence option, which we already found to be wanting.9
So, what does Horwich’s deviant logician show? Does reflection on the intuitionist challenge to classical logic show up something deeply suspicious about [UJLT], and, more generally, about the alleged core [UJ] connection?
Versions of the DLO are developed more thoroughly by Williamson (2006: §2, 2008: Ch.4). Williamson argues that logically competent agents can even have unequivocal, informed, engaged disagreements about whether something of the form ‘All As are As’ is an instance of a logical truth. For example, there are (sophisticated, considered) reasons to worry about the relations between existential import and truth-conditions—that is, to hold that any statement that purports to refer to ‘A’s can only be true if there exist As (in the relevant context). To the extent skepticism about the existence of As can be motivated, then we can imagine someone with (sophisticated, considered) reservations about whether a particular instance of ‘All As are As’ should be counted as a logical truth. (E.g., Is ‘All unicorns are unicorns’ a logical truth?) Again, the incompetence and the ambiguity options do not have much promise to handle all possible dissenters (as Williamson argues).10
Another example developed by Williamson concerns the logic of vagueness. Many theories of vagueness posit truth-value gaps. To the extent that one can motivate the claim that ‘A’ is vague, to proponents of such a theory, then, again, we can imagine someone whose considered judgment is to balk at whether a particular instance of ‘All As are As’ (e.g., ‘All tall people are tall’) should be counted as a logical truth. Again, Williamson argues that neither the incompetence nor the ambiguity option can save [UJLT] on this front.
At the same time, Williamson holds (as do I) that any statement of the form ‘All As are As’ is a logical truth. Hence, Williamson believes he has provided counterexamples to [UJLT corollary]—that is, logically competent agents who understand instances of logical truth, but yet do not assent to them. So, then: Does the DLO show that there can be no deep constitutive connection between understanding and justification, even in the relatively straightforward case of logical truth? Does the DLO prove, a fortiori, that nothing can have the status such that assenting to it is a necessary condition for understanding it?11
[§]
What then should a proponent of the constitutive a priori orientation say about this challenge? Well, they hold that it is an important lesson of mid-twentieth-century modal epistemology—a legacy of Wittgenstein, Carnap, and others—that such notions as a priority are comprised of two distinct, separable factors. A priority depends on both intrinsic content and place in a framework. If we appreciate this point, then the fact that two competent agents could unequivocally agree about something’s intrinsic content, while attaching different statuses to it, is no knock-down challenge to core [UJ] connections. Horwich’s and Williamson’s challenges pose no more of a problem for [UJLT], for constitutitivists, than does the fact that two agents might agree that every event has a cause, and yet for one of them this is an a priori regulative rule while for the other it is an a posteriori inductive generalization (cf. §5.2).
In other words, for proponents of constitutive a priori orientation, there are at least two separable factors which constitute ‘shared understanding’—that is, sameness of content, and sameness of status. Hence, this orientation offers proponents of the [UJ] connections a principled defense from the DLO. For it offers a way to articulate and develop the intuition (which is no doubt motivating the authors alluded to in note 8) that deviant logicians do not in fact instance a ‘shared understanding’ of the basic tenets of standard logic (as is demanded by any notion that the DLO undermines [UJLT]). So, even if their conceptions of the content are identical, their attaching different (and deviant) status to that content undermines their promise to afford counterexamples to [UJLT].
Consider again the intuitionist challenge to LEM. LEM is a fruitful simplifying element of many branches of logic; but there are deep and ancient reasons to countenance counter-instances. This situation is completely amenable to an understanding-based, constitutivist-style explanation. There are external (instrumental, conventional, pragmatic, etc.) questions about what we want from a logic. Again, these days it is even fairly plausible and common to work with different logics for different purposes and contexts, where the appropriate external questions vary from case to case. For some of these external questions, simplicity and fecundity will receive a high ranking, and LEM has proven to be expedient toward those ends. For other external questions, comprehensiveness and integrity may trump simplicity and fecundity, and, accordingly, the putative counter-instances to LEM may be judged to be decisive.
The important point for present purposes is that intuitionism is a distinct framework from classical logic, as befits its different answers to external questions about logic. Hence, for proponents of the constitutive a priori orientation, that the LEM is a standard, classical logical truth (i.e., derivable as a theorem in frameworks which adhere to the traditional answers to external questions about logic) while being invalid within intuitionist logics is no serious challenge to the core, constitutive [UJ] connections. Common understanding demands the sharing of both status and content, not just of content. Since, accordingly, the understanding of LEM varies between intuitionists and classical logicians, it is unsurprising and inconsequential that its justifiability (or lack thereof) does too.
Given certain aims and interests, it is reasonable to take LEM as a priori (i.e., like the principle of sufficient reason in some frameworks, simply not subject to empirical counterexample).12 However, when it comes to [UJ] connections, there is little to be said in favor of the claim that assenting to LEM is a necessary condition for understanding it. Contrast this with [&E], where this tight [UJ] (or meaning-constituting) connection is unshakably evident (despite Williamson’s audacious effort mentioned in note 11). Espousers of [UJLT] can and should treat LEM and [&E] differently, because of this clear difference.
What about Williamson’s deviant logicians? Again, proponents of the constitutive a priori orientation should take the moral to be that foundational questions about the relevant framework are conceptually prior to what turns out to be constitutive a priori within any specific framework. Williamson’s deviant logicians are, like the intuitionist, working within a non-standard framework, which rejects certain canonical answers to external questions about logic. One exceedingly complex challenge stemming from the development of deviant logics is that it is difficult to judiciously and comprehensively settle external questions about logic; nonetheless, this defense of the core [UJ] connections relies only on the crucial distinction, in the epistemology of logic, between internal and external questions, not on any specific answers to the external questions.
Distinct logics of vagueness (say) will countenance disjoint sets of logical truths. It is not a trivial matter to decide which logic of vagueness one ought to prefer, all things considered; but constituent [UJ] connections are a separate matter. Once we have settled, however tentatively, what we want from a logic of vagueness, we will accordingly settle on answers to the appropriate external questions. In due course, there will issue framework-relative logical truths, such that justification for them is grounded in understanding them. One is, of course, free to deviate; but one thereby changes the framework.
Hence, proponents of the constitutive a priori need not take the DLO to have undermined the core [UJ] connections, because they can explain how it is that deviant logicians do not share an understanding of the relevant tenets with their classical opponents. For example, there is a clear sense in which an intuitionist and a classical logician do not instance a ‘shared understanding’ of LEM—even if (contra note 9) they semantically associate exactly the same content with the formula ‘Φ v ~Φ.’ Likewise, while I concede that the prospects are dismal for dismissing Williamson’s deviant logicians (pertaining to ‘All As are As’) on grounds of incompetence or ambiguity, still these cases instance such different orientations with respect to external questions about logic, and attendant differences in status as to these instances of ‘All As are As,’ that constitutivists should hold that these are not instances of ‘shared understanding’—again, as the DLO needs them to be.
[§]
What deviant logicians show about the epistemology of logic, then, is that proponents of understanding-based accounts of justification for logical truths are well-advised to endorse the constitutive a priori. There is a distinctive notion of a priority (of which logical truth is a distinctive case in point) which is well-equipped to meet the deviant logician’s challenge.
The key step in any version of the DLO is to argue that there can be disagreements about logical truth between competent agents who share the same understandings of logical primitives. However, if we concede that a priority depends on both intrinsic content and place in a framework, then the fact that two competent agents could unequivocally agree about something’s intrinsic content, while attaching different statuses to it, is neither surprising nor disturbing. If logical truths are constitutive a priori truths, two competent agents could unequivocally agree about something’s intrinsic content, while attaching different statuses to it. Hence, then, the DLO does not afford counterexamples to [UJLT].
To treat something as a priori, as simply not possibly subject to empirical disconfirmation, is to mark off a certain content as having a certain privileged status. I have been using [&E] as a case where the intrinsic content all but guarantees the special modal status; but such meaning-constituting cases are relatively rare. The case of LEM helps to illustrate what a daunting job it is to get from [&E] to a comprehensive epistemology of logic, let alone to a comprehensive account of a priori justification. For now, though, the moral is that for proponents of the constitutive a priori, the possibility of a deep constitutive connection between understanding and justification in (and beyond) the epistemology of logic survives the challenge of the deviant logician.
What the prevalence of deviant logics shows is not that the core [UJ] connections are completely untenable, but rather that it is untenable to approach the notion of ‘logical truth’ as if it designates a monolithic block of eternal superfacts.
Hence understanding-based accounts of a priority in general, and constitutive a priori orientations in particular, have not been besmirched on this core front. The [UJ] connections still stand. Furthermore, and to the contrary, working through these details has just added to ongoing amassing of considerations in favor of this orientation. On then with the task of showing how this orientation provides responses to the challenges of revisability and externalism.
§6.3: revisability and conceptual evolution
So: that defense of the [UJ] connections on a constitutive a priori picture is a main plank in the main business of this present chapter. Again, my own constitutive a priori orientation is made up of a modal revisionist response to the challenge of revisability, and a moderate externalist response to the externalist challenge. This present section will continue this development and defense work, in ways focused most specifically on modal revisionism; §6.4 will segue into issues that pertain to both revisability and externalism; §6.5 will move on to issues which pertain more squarely to moderate externalism; and §6.6 will come back to the two challenges’ confluence and collective upshot.
The complexities of change over time need to be taken into account, when it comes to the modal revisionist’s framework-relative answer to Plato’s problem, and the ways in which this consideration differs for the cases of metaphysical, semantic, and epistemic immunity to counterexample. Not only are multiple frameworks generally always applicable, when it comes to more or less any isolated question or embedded inquiry, but more or less all frameworks are more or less constantly under revision. Scientific revolutions may be rare and drastic, but conceptual evolution is more on the order of death and taxes. (Any scientific revolution would involve lots of conceptual evolution, but, for the most part, not vice versa.) This is a crucial aspect to be reckoned with and rendered, when it comes to articulating how it is that immunity to counterexample is retained, on this orientation.
Paying heed to this historical dimension brings into clearer relief some of the complexities within a satisfactory constitutive a priori picture. It is easy enough to distinguish paradigm cases of internal questions from external ones, or as Carnap later puts it, change of language from change of theory:
I should make a distinction between two kinds of readjustment in case of conflict with experience, namely, between a change in the language, and a mere change in, or addition of, a truth-value ascribed to an indeterminate statement. … A change of the first kind constitutes a radical alteration, sometimes a revolution, and it occurs only at certain historically decisive points in the development of science. On the other hand, changes of the second sort occur every minute. (1963: 921)
However, in the course of ongoing inquiry there is clearly a continuum between these extremes. The re-categorization of whales, or the spitting of the atom, involved both change of language and change of theory, and neither can be seen as an easily discretely separable sub-component.13
Both theories and languages change over time, which involves, and prompts, change of meanings, and of the frameworks they compose. This is a glaring and significant upshot of the challenge of revisability, which it is crucial to sort out in this present project. Since languages are organic entities which change over time, that will have palpable effects on precisely what it is to understand the meaning of a given word (e.g., ‘whale’ or ‘atom’). In turn, any effect on what it takes to count as understanding a word or concept is going to have knock-on effects to any [UJ] connections or principles.
The limit of revisability is precisely the sense in which immunity to counterexample is retained. Changes as to what counts as analytic or a priori can only occur where there exists conceptual evolution, which in turn suffices for a change of framework. Within a framework, though, on the constitutivists’ view, immunity is not only always possible but, further, required. (As broached above in §6.1, Poincare engaged with this Janus-faced quality, whereby what appears to be malleable from one perspective can seem mandatory and unshakeable when approached from another vantage point.) One task of Part IV will be to chart the metaphysical, semantic, and epistemic consequences, when it comes to immunity to counterexample, of this dimension of change over time.
While conceptual evolution is most obviously and directly pertinent to the challenge of revisability, it will also have some deep bearing on the externalist challenge. Following up on §5.4, for starters, conceptual evolution is bound to be a very different thing for the case of natural versus conventional frameworks. In general, anything which depends on concepts will be affected by conceptual evolution; any categorical or principled difference between sorts of concepts (i.e., natural vs. conventional) will track or mark differences in these sorts of effects. Conceptual evolution in cases like ‘whale,’ ‘planet,’ or ‘metal’ is in myriad ways different from cases like ‘marriage’ or ‘money’ (i.e., our views about this mind-independent phenomenon are changing in light of new discoveries vs. if there is consensus to change this institution, then so be it).
[§]
I will next return to one of our ongoing themes of sorting through various senses in which something could be said to be immune to counterexample, and further unpack some differences between metaphysical modality, on the one hand, and semantic and epistemic modality, on the other, when it comes to the challenge of revisability.
As detailed above in Parts I–II, that which is necessary is not revisable, for it is precisely the things which are the most firmly bolted down that that concept is used to single out. (Of course, any particular agent’s guesses as to what is necessary may be revised over time, as a function of evidence, insight, etc.; but the truth-conditions of ‘It is necessary that P’ do not change over time.) However, for finite, fallible agents like us, it is only prudent good sense to hold that what counts as justified a priori will vary as do our beliefs and the concepts which they involve. So, for a post-Kripkean framework-revisionist, a major difference between attributions of metaphysical modality (e.g., ‘It is necessary that P’) and attributions of epistemic modality (e.g., ‘My belief that P is justified a priori’) is that the truth-value can change over time only in the epistemic case. Metaphysical necessity is framework-independent, whereas a priori justification is framework-relative. (Semantic modality is also framework-relative, for similar reasons; but here let us focus on the contrast between metaphysical and epistemic modality.)
For example, if it is necessary that heat is the motion of molecules, then it did not just become necessary once it occurred to someone, or once it became sufficiently verified and accepted by experts. Metaphysical necessities could not have been otherwise, and are categorically indifferent to whether anyone knows anything about them. In contrast, my favored revisionism holds that truth-conditions could change over time, for cases like ‘It is justified a priori that whales are fish’ and ‘It is justified a priori that atoms are indivisible.’ Note though that change in the truth-conditions of an attribution of epistemic modality can occur only given an updating of the relevant framework—that is, such a change depends on a significant change to the content of at least one of the relevant constituent concepts. This explains the qualified but still significant sense in which immunity to counterexample is retained on this sort of revisionism. My ancestors’ belief that whales are fish has not been falsified, because whales still do satisfy the relevant superficial, unscientific criteria—that is, their concept of ‘fish’ was something like ‘anything that lives in water, swims, and is more-or-less shaped like a tuna.’ But, in my dialect, it is not true, let alone justified a priori, that whales are fish. My concept of ‘fish’ is a rather distinct ancestor of theirs, a distinct biological kind term, which differs in conditions of satisfaction.
So, Kripke’s distinction between necessity and a priority lends credence to the revisionists’ tenet that epistemic modalities are revisable, framework-relative, subject to re-evaluation in cases of conceptual change. Thus, absolutism about epistemic modalities is not just immodest; further, Carnap (1950), Quine (1951), Wittgenstein (1969), and Kripke (1972) converge in showing absolutism to betray a deep and arrogant mistake. Our beliefs get updated, our concepts evolve; it is to be expected that any non-empirical (semantic, conceptual, rational, etc.) sources of justification are directly and thoroughly effected by such epistemic developments and conceptual evolution.
[§]
In addition to this dynamic source of complexity, which is the main present focus (i.e., frameworks are more or less constantly under revision) another source of complexity for this constitutive a priori picture is the purely static consideration that multiple frameworks might simultaneously apply in the same inquiry (whether we are talking about an isolated individual at a given time, or across a community of agents). Generally, out here in the real world, both the static and the dynamic sources of complexity (among other things) are always at play. The upshot is a deep and through context-sensitivity, which pervades and motivates the framework-relative approach to a priority. Railton (2000) is instructive on this front, going beyond Wittgenstein’s own complex position on the context-sensitivity of the rule/proposition boundary. Railton’s is a multi-dimensional dynamism, in which ‘norms’ are propositions given the status of rules for the purposes at hands, and ‘supernorms’ are created whenever there is systematic rational deliberation about norms. Neither norms or supernorms are ways of classifying content, per se, and neither is context-independent. (It is a question of how it is used, not of what it says or how you came to believe it.) There are multiple shades of grey and directions of fit.14
This static kind of factor is most familiar perhaps from the literature on scientific revolutions, in cases of severe Kuhnian crises. Which theoretical lens is brought to bear on the data will significantly impact how we conceive of its nature and upshot. Precisely what the question at issue is is framework-relative, let alone how it should be answered. Consider some of the cases already discussed, to motivate the two separable dimensions of a priority (i.e., content vs. status), such as ‘all events are caused’ from §5.2 (between an agent who takes that to be a regulative rule, and an agent who takes it to be an inductive generalization), or LEM from §6.2 (between an intuitionist and a classical logician). As we saw, theories in which the notion of status plays a key role complicate what counts as ‘shared understanding’; this will also drastically affect the scope and range of the core [UJ] principles.
Another familiar way in which this static complexity factor plays out in the literature is in the guise of Quine/Duhem holism (cf. Quine [1951], Quine & Ullian [1970]), which is particularly influential among naturalists. A guiding idea here is that there are many ways to iron in recalcitrant data (e.g., an unexpected and problematic experimental result), and hence that the unit of experiential confirmation not an individual statement. Down this avenue, shades of grey stretch between updating an isolated, individual belief versus rejecting an entire theory. The web of belief is a compelling metaphor, for such cases. We’re all repairing our ship while out at sea, after all—to borrow the metaphor which Quine borrows from Otto Neurath—at least those of us who are not absolutists. Here again we steer into the inevitability of jettisoning any pretense of a firm distinction between change of language and change of theory. Conceptual evolution essentially involves the former, and is also an inevitable component of the latter.
The static source of complexity flows confluent with the temporal, dynamic source. With conceptual evolution (and constitutive thereof) is alternation of framework; attendant upon that too is change with respect to what gets treated as immune to counterexample, what counts as analytic and hence a priori. Only in hindsight can conceptual evolution be neatly sorted—and, at that, that must be to some degree stipulative.
[§]
I close this section by honing my own constitutive a priori position on conceptual evolution via quick comparison with three assorted authors, from outside this tradition—Peacocke’s (2000, 2004) program of moderate rationalism, Field’s (2000, 2005) evaluationism, and Maddy’s (2000, 2007) naturalism.
Peacocke’s self-styled ‘moderate rationalism’ is an understanding-based account of a priority which aims to fashion an adequate account of the a priori without taking on any anti-naturalistic commitments associated with the acquaintance-based tradition. The notion of ‘meaning-constitution’ plays a key role in his program. Compare again:
[1] Squares have four sides.
[2] Neptune has four moons.
[1] is an example of a meaning-constituting belief. Hence, Peacocke’s ‘meaning-constitution’ is very close to my [UJ], an understanding-based approach to immunity to counterexample . It bears many similar influences, from Poincare through Carnap.
However, Peacocke’s a priori is more of a special kind of content, as opposed to the constitutivists’ status. This makes Peacocke more explicitly a kind of rationalist than those of us in the constitutive a priori camp, who privilege the importance of status, in addition to content, when it comes to a priority.
Next, Field’s (2000, 2005) evaluationism is a novel, status-conscious, stance on a priority. The core idea is that, while most of our beliefs are ‘non-basic’ in the sense that they can be justified by appeal to other beliefs (tenets rules principles etc.), not all of our beliefs can have this status, or else the process of justification would be circular, or would go on forever. Some of our beliefs must be basic, and to them we are entitled ‘by default,’ not because we can justify them by argument. There is nothing mysterious or obscure about default entitlement; it is an ‘attitude of approval’ that is accorded to certain core constituents of our ‘logic and methodology’ (2005: 86).
Now, this is very similar to the constitutive a priori orientation. It too is a modal revisionist response to the challenge of revisability; further, it also endorses the two prongs of the constitutive a priori view—that is, the status-orientation, and the point about necessary preconditions for inquiry. It may not be fair to force Field into the constitutivists camp, whether he likes it or not; but, given the above, it is tempting to take the arguments he gives for evaluationism as also further bolstering the case for the constitutivist status-conscious orientation on a priority.
Finally, Maddy (2000, 2007) styles herself as a naturalist heir of Quine. However, as I have set things up here, the particular substance of her explicit departure from Quine is quite significant (cf. Maddy [2000: 108ff, especially pp. 112–14]). Crucially, there is a seam in her web of belief, a ‘brute methodological distinction’ (113), of the exactly sort that has been familiar to us since §5.2 (e.g., rule/proposition, or status/content). Though not by that name, the constitutivist status of a priority—a very familiar kind of immunity to counterexample, for certain special aspects of theory—is alive and well within Maddy’s self-styled naturalism. Because of this, rather than countering the anti-naturalist arguments of the Kantian constitutivists, Maddy ends up espousing a terminological variant of the view.15
§6.4: semantic deference
Next on the agenda is to delve into the notion of semantic deference, which plays a role in most recent conception of language, which appreciates its communal dimensions. An individual language user is a part of a system, which enhances the conceptual resources available to the individual. Consider our evident abilities to think and talk about Ancient Greek politics or the geography of Mars, or all manner of things with which no one I have ever met has had any first-hand acquaintance—let alone non-actual or impossible objects, which are rather thoroughly unacquaintable. Semantic deference is the appropriate attitude of the individual speaker toward the community-wide communicative powers of the terms in their thought and talk. (Natural kind terms provide a nice clear paradigm case, to which we turn below—for example, I believe that aluminum is a metal, would assent to that and bet on it and etc., and yet have no real capacity to divide either aluminum or metals from any close competitors. My attitude toward the criteria for the correct application of both terms ‘aluminum’ and ‘metal’ is deferential.)
Properly excavated, semantic deference promises to have some deep relevance to both revisability and externalism—let alone to many other ongoing discussions, about whales and atoms and the broad distinctions between natural and social-conventional reality.
As a point of contrast with some of the sorts of cases to be considered below, let us next delve a little further into the notion of semantic deference as it applies to ordinary speakers’ usage of natural kind terms. For ordinary (i.e., non-expert) speakers, to use a term as a natural kind term involves a Lockean, ‘I know not what,’ deferential intention.16 That is, on such uses, terms like ‘tiger’ and ‘water’ are used to refer to a mind- and language-independent kind of thing or stuff, the precise criteria of identity for which is typically unknown to speakers who nonetheless count as competent with the term. (As came up in §5.4 [cf. especially note 11] and will be further explored below, there are of course lots of non-typical uses of these terms [as of others].) Typically, and tellingly, such non-expert speakers are relatively open to correction by experts, when it comes to the exact criteria for the correct application of such terms. Clearly, in these cases of deferential uses of natural kind terms there exists an externalists’ gap [EG] between the conditions which typical individual speakers associate with the terms, and their actual precise technical conditions for correct application.
(Herein lies another way in to why natural kind terms play such an important role in recent work in the theory of reference, alongside the perennial model case of proper names. Using a term as a natural kind term is the very paradigm of semantic deference, and semantic deference is essential to semantic externalism.)
There are, however, stretches of the lexicon concerning which this [EG] does not seem to exist. To take an extreme example, consider the word ‘and,’ whose meaning I will take to be constituted by the standard introduction and elimination rules. The traditional motivations for semantic internalism still seem to be completely applicable here—that is, competence with ‘and’ is a matter of grasp of meaning, and that meaning determines the term’s extension. So, here, we do not get the [EG] between what is constitutive of competence and what determines the extension. When I encounter the conjunction of two propositions P, Q into ‘P and Q,’ the semantic contribution of ‘and’ is completely transparent to me. As opposed to the case of natural kinds, I am autonomous, not the least bit deferential, about the semantics of ‘and.’17
Though ‘and’ stands at the extreme, in this respect, it is not hard to find kinds of expression which are more like ‘and’ than like ‘tiger.’ Terms of elementary arithmetic, such as ‘equals’ and ‘even number’ provide strong candidates, as do simple geometric terms like ‘square.’ Even further, though, many common social kind terms, which target categorizations which humans construct, as opposed to discover, also provide plausible instances. So, for example, the content of my ‘grandmother,’ ‘bachelor,’ or ‘widow’ thoughts and utterances is more autonomous and transparent than deferential. Here again the [EG] between what typical speakers typically associate with the term and what determines the extension does not get much purchase. That which is constitutive of competence is also that which determines the extension, and so content is transparent to competent speakers.
Tying back in to some points first mentioned in §5.4, degree of resistance to correction is a primary indicator of where a particular use of a particular term belongs on this deference-transparency spectrum. And so compare what it would take to convince you to reject the following:
1. Aluminum is a metal.
2. Widows are formerly married women whose spouse has died.
3. If P and Q, then P.
The move from a paradigm natural kind case to a paradigm social kind case is pretty drastic, along this index, as is the subsequent the move from there to a meaning-constituting axiom of elementary logic. These are steps further and further into the un-black-swannable terrain, where the [UJ] connections and principles hold fast. Instead of an [EG], in these latter cases, we have a rather fast and tight link between meaning and extension. Its relevance to immunity to counterexample is glaring.
This line of thought requires careful handling and extensive refinement; I will begin this job immediately below. Still, though, it runs clearly counter to the metasemantic thesis about the [EG] (i.e., the idea that there is an [EG] for all types of usage of all types of term). Statements [2] and [3] provide examples of mundane counterexamples to such extreme or unqualified claims about competence, transparency, and externalism (which were discussed above in §5.3, and to which we will return below in §6.5). Since the [EG] is a source of potential threat to the [UJ] principles, questioning the tenacity of the foundational links between meanings and extensions, investigating its proper range is of utmost import for this ongoing project.
[§]
I now turn to some of the most important refinements, further developing this idea of an inverse relation between deference and transparency, before turning back to the question of the limits of externalism.
First, it is evident that there are many different kinds of use of any particular term; and so it is important to point out that the above remarks are premised on certain conceptions about typical uses by typical speakers. For example, notoriously, many uses of the term ‘water’ are not uses as a natural kind term, as above characterized. To delve further into this issue, consider Putnam’s (1962) thought experiment which concerns the surprising discovery that cats are actually Martian robots. Putnam (1962: 661) points out that intuitions may be divided between two different reactions to this surprising discovery:
i. Wow! It’s turned out that cats are not animals after all!
ii. Wow! It’s turned out that there aren’t and never were any cats!
One fundamental difference between [i] and [ii] concerns exactly how the term ‘cat’ is used. Statement [i] involves a deferential, whatever-it-is-exactly-that-constitutes-the-real-essence-of-this, use of the term; whereas [ii] involves a more autonomous, my-meaning-determines-my-extension, use of the term. (The externalists’ gap is open in [i], but not in [ii].) In my idiolect, cat is a [i]-type, natural kind term; and so [i] would be the correct response, while [ii] is confused.
Of course, someone else might insist on using ‘cat’ more autonomously, and insist that [ii] is correct of their use of the term. (Cf. Loar [1991: 120]: ‘Social meanings do not deprive me of autonomy when I insist on it.’ Deference cannot be forced, or legislated, onto a stubborn speaker.) This will afford more transparency and autonomy to their ‘cat’ thoughts and utterances. Fair enough; but, given that such stubborn speakers are not using ‘cat’ as a natural kind term, they pose no counterexample to idea that there is an inverse proportion of transparency and deference. The gains in transparency here come proportional to the lack of deference. However, such stubborn speakers do complicate any simple articulation of the inverse proportion relation, which does not take into account that any term can be used in multiple sorts of ways, according to the speaker’s intentions.18
Again it is important to ward off oversimplistic misimpressions, caused by focusing exclusively on paradigm cases at the extreme poles. These paradigm cases are dialectically important; but clearly the lexicon embodies many complex and heterogeneous divisions, which map onto each other in complex ways, between deferential uses of ‘aluminum’ and transparent uses of ‘and.’ There may even be lots of cases which, upon reflection, we would want to classify in purgatory, between transparency and deference. For example, consider Putnam’s (1962) ‘Pencils are artifacts.’ How would I respond to the discovery that pencils are actually intelligent organic Martian spies? Is this a discovery that there aren’t and never were any pencils, or a big surprise about pencils? I for one have no strong intuitions either way on this one; which is to say that I use ‘pencil’ as neither a paradigmatically deferential natural kind term nor as a paradigmatically autonomous and transparent sort of term. (Cf. Putnams’s [1975: 248] remarks about differences between ‘cat’ and ‘pencil’; they come up again in the next section.)
Both of these refinements (i.e., that pertaining to different types of usage, and that pertaining to different types of term) show that we are clearly dealing with a complex variety of differences of degree along a continuum, as opposed to any straightforward, categorical difference in kind, from deferential uses of ‘aluminum’ to autonomous uses of ‘widow’ or ‘and’ (or: between natural kind terms on to social kind terms and logical particles). Further, there are experts who are not generally deferential about a given range of natural kind terms; and there are plenty of non-natural-kind technical terms about which most speakers are rather deferential (e.g., ‘hedge fund,’ ‘spandrel,’ ‘junta’). The picture is messy, as any picture of the semantic aspects of our linguistic behavior must be, if it aims at comprehensiveness. Nonetheless, if the broad contours of the picture traced herein are accurate, this has considerable significance for questions about the range of semantic externalism.19
To summarize and consolidate some of the points made about transparency and deference, then, consider the following five statements:
1. Aluminum is a metal.
2. Cats are animals.
3. Pencils are artifacts.
4. Widows are formerly married women whose spouse has died.
5. If P and Q, then P.
They are ranked in increasing order of resistance to correction. (The difference between [1] and [2], I take it, is that the aforementioned stubborn, non-deferential, non-natural kind uses of ‘cat,’ like of ‘water’ [cf. note 18], are more prevalent than such uses of ‘aluminum.’ No one would say that there isn’t and never was any aluminum!) They illustrate the inverse proportion of semantic transparency and semantic deference, in that each successive step brings with it less deference and more transparency. Furthermore, with each successive step, there is less and less purchase for the [EG] between what is constitutive of competence and what determines the extension.
§6.5: the limits of externalism
There are multiple different ways in which semantic externalism might be limited. For starters, we might distinguish two among them: limiting the RANGE of externalism versus limiting the DEPTH of externalism. For example, to hold that the seminal externalist arguments apply only to proper names and natural kind terms, and have no obvious relevance beyond those types of referring expression, is to limit the RANGE of externalism. In contrast, to hold that, even in the case of incompletely mastered referring expressions, there are still significant limits to the amount of ignorance or mistakes which are compatible with competence, is to limit the DEPTH of externalism. (Surely even the most radical externalist has to agree that someone who thinks that ketchup is a kind of marsupial cannot be counted as competent with the term!) However, not much work has been done, when it comes to serious investigation of the DEPTH of semantic externalism—that is, defining minimum thresholds for incomplete-mastery-yet-competent.20 Perhaps the only possible answer to that is stipulative. In any case, that is one pertinent sense in which semantic externalism is limited—no one is an externalist all the way down, holding that there are no substantive requirements for competence whatsoever, even in the case of deferentially used (Millian-tag-like) referring expressions.
I am presently concerned with the RANGE question: that is, How far, beyond the bounds of deferential uses of proper names and natural kind terms, can or should the seminal externalist arguments be extended? I will not touch on the DEPTH questions except in passing—not to mention on some other significant questions in the neighborhood, such as how externalism relates to the rule-following considerations, or to self-knowledge, or whether externalism entails a priori access to things one otherwise would have thought to be a posteriori (or vice versa), etc.
Recall from §5.3 the following three theses about the [EG]: the Pragmatic thesis takes the [EG] to be a property of certain limited type of linguistic usage; the Semantic thesis takes the [EG] to be a property of certain limited type of linguistic expression; and the Metasemantic thesis takes the [EG] to be a general discovery about the nature of language. I turn next to adjudicating this issue, and connecting it to our ongoing discussion.
For starters, then, why might one conclude, with the Pragmatic thesis, that the seminal externalist arguments tap into something distinctive about certain sorts of uses of a limited class of terms, as opposed to something about semantic competence in general? Well, the dialectic in Kripke’s seminal work goes: (i) recent developments in modal semantics strongly suggest that proper names are much more like Mill thought them to be than like Frege or Russell thought them to be, (ii) and lo! They also show that natural kind terms are in many respects (i.e., those described above in §6.4) similar to names. That is, Kripke is quite plausibly read as arguing that proper names and natural kind terms (in typical sorts of usage) are distinctively non-descriptive. (Bach’s [1987] ‘relational determination’ and Recanati’s [1993] ‘psychological neutrality’ are some influential ways to further articulate what is distinctive about the mechanisms of reference involved in this circumscribed class of non-descriptive terms, ways that do not [at least not immediately, or obviously] extend across to linguistic meaning generally.)21
So, there are some positive reasons to think that the seminal externalists’ arguments have tapped into something unique about a specific, circumscribed kind of referring expression, and hence to think that minimal conditions for competence apply distinctively to proper names and natural kind terms. (Again, this is most starkly evident in their deferential uses by non-expert speakers.) Insofar as ‘Einstein’ or ‘tiger’ is merely a tag or label—in a way that some if not most sorts of linguistic expression are not—then (i) the minimal threshold for competence, in such cases, is appropriately thought of as relatively insubstantial and undemanding, and (ii) competent grasp of such expressions will not bring in its train transparent access to much of anything. But this consideration does not yet apply to expressions which are not so plausibly thought of as mere tags or labels.
Note that Kripke himself enters some remarks contrary to Metasemantic thesis’ range-extension, in the course of discussing the famous case of Pierre and ‘London’/‘Londres’:
Not that the puzzle extends to all translations from English to French. At the moment, at least, it seems to me that Pierre, if he learns French and English separately, without learning any translation manual between them, must conclude, if he reflects enough, that ‘doctor’ and ‘medecin,’ and ‘heureux’ and ‘happy’ are synonymous, or at any rate, coextensive; and potential paradox of the present kind for these word pairs is blocked. (1979: 256)
(Not to suggest that Kripke is here explicitly concerned, let alone solely concerned, with the question of the range of the EG. However, there is surely a relevant distinction articulated here—albeit tentatively—about differential criteria for competence, which relates to transparency and externalism.) The thought here is that apart from the externalist-friendly (distinctively non-descriptive) cases of proper names (e.g., ‘London’) and natural kind terms (e.g., ‘furze’/ ‘gorse’), more substantive requirements are appropriate for competence.
Also, in the middle of the third Naming and Necessity lecture—Kripke’s most extensive treatment of the [EG] for the case of natural kind terms—he expresses a similar tentative qualification to its range, citing ‘foolish,’ ‘fat,’ and ‘yellow’ (1972: 127) as examples of sorts of expression to which the ongoing discussion may not apply. So Kripke himself holds that it would not be easy, or obviously appropriate, to motivate an externalists’ gap for the likes of ‘foolish,’ ‘fat,’ ‘yellow,’ ‘doctor’ or ‘happy.’ Hence, it is plausible to read Kripke as holding that many common expressions are descriptive. Further, similar remarks apply to Putnam too. He holds (1975: 233) that ‘some words do not exhibit any division of linguistic labour: “chair,” for example.’ Further, he also concedes (1975: 248) that the case for an [EG] with respect to ‘cat’ ‘has more plausibility’ than the analogous case with respect to ‘pencil’.
So, there do exist grounds for reading some of the great seminal externalists as moderate externalists, when it comes to the crucial question of differential criteria for competence, and for the relation between what constitutes competence and what determines reference. These big-picture considerations and historical precedents are confluent with the above cases considered, which run contrary to the Metasemantic thesis that the [EG] applies to all corners of the seas of language.
[§]
The externalist challenge, recall, is the consideration that factors external to individual speakers play various roles in determining the extensions of the terms which they employ in their thought and talk. It is most deeply relevant to charting a priority, because of its potential to undermine, or at least drastically alter, traditional conceptions of the relations between meaning and extension. The knock-on effects on the [UJ] connections would be immediate and drastic. Hence any radical form of semantic externalism would obviously affect our considered conception of (at least) semantic and epistemic immunity to counterexample (cf., e.g., Williamson [2008], Russell [2012] for concrete integuments of this line of thought).
Reasons are emerging, though, to hold that the Metasemantic thesis about the [EG] looks to be an unsupported RANGE-extension, when it comes to understanding the upshot of the seminal externalist arguments. There are some good reasons to hold that those arguments tap into something distinctive about a limited class of expressions; and there are some fairly strong counterexamples to radical externalist claims—as they pertain to, say, ‘grandmother,’ ‘square,’ and ‘and.’ In such cases, meaning limns extension, and so U can afford J.
What the counterexamples to semantic internalism (i.e., semantic content supervenes on the intrinsic states of individual speakers) clearly show is that otherwise competent speakers are often deferential as to the criteria for the correct application of the terms tokened in their thought and talk. In such cases, content is generally not transparent to individual speakers, and the externalist gap is evident between what is constitutive of competence and what determines the extension. However, there are vast stretches of the lexicon regarding which typical speakers are not typically (nor ought they to be) deferential about the criteria for correct application. These cases are beyond the scope of the seminal externalist arguments, per se, and in these cases there is no [EG] and content is transparent to competent speakers. As Burge (1979) points out, and we have further explored, degree of resistance to correction is a primary determinant for where a particular use of a particular term belongs on this deference-transparency spectrum.
In general, there is an inverse proportion of semantic transparency and semantic deference—that is, transparency wanes to the extent or proportion that deference waxes. To the extent that the speaker is deferential in their use of a term, and is open to correction, then they lack transparent access to the conditions for the correct applicability of their terms. In contrast, to the extent that a speaker is not deferential in their use of a term, and is not open to correction, then the speaker does have transparent access to the conditions which determine the term’s extension. (These non-deferential speakers are liable to make mistakes—incompetent moves—in a way that their deferential counterparts are not.) Clearly, a wide and complex range of cases stretches in between the extremes (illustrated herein by deferentially used natural kind terms and transparently used logical constants). Still, this might be seen to go some way toward undermining the Metasemantic thesis, which would be hostile to the [UJ] connections. The seminal externalist arguments have not shown the transparency of semantic content to be obsolete, beyond the bounds of deferentially used proper names and natural kind terms. It has not been shown that, in general and unequivocally, there is a lot less to semantic competence than the traditional internalists presumed. Incomplete mastery can amount to semantic incompetence, in lots of mundane cases, wherein deference is absent.
I next consider two questions which have come up:
[1] How far, beyond the range of referring expressions such as proper names and natural kind terms, does this externalist challenge extend?
The answer here depends upon the speaker’s intentions—any term (even ‘and’) could be used deferentially. However, though, lots of terms are not in fact typically so-used. Furthermore, in any case, there are reasons to think that incomplete mastery is uniquely compatible with competence for the circumscribed cases of deferentially used proper names and natural kind terms, in a way that does not extrapolate across to ‘pencil,’ ‘widow,’ ‘and,’ etc.—because of certain distinctive properties of that circumscribed category of referring expressions. Insofar as ‘Einstein’ or ‘tiger’ is merely a tag or label—in a way that ‘pencil,’ ‘widow,’ ‘and,’ etc., are not—then the minimal threshold for competence, in such circumscribed cases, is appropriately thought of as relatively insubstantial and undemanding.
[2] Has it been proven that there is, quite generally, a lot less to semantic competence than traditional internalists would have us believe?
No, given that competent speakers are not always, and perhaps not even typically, deferential. (After all, I’ll bet that in your house, like mine, there are a lot more chairs than tigers) Incomplete mastery is compatible with competence just to the extent that the speaker is deferential; this will differ significantly (in degree) for distinct types of term, as well as for distinct types of usage for any given term. In any case, neither the seminal externalist arguments nor subsequent externalist work suffice to ground the need for any drastic, general re-conceptions of semantic competence. In lots of mundane cases, transparency holds, and (if the speaker in question is not appropriately deferential) incomplete mastery can still amount to semantic incompetence.
To sum this up, then: To endorse the Pragmatic thesis about the [EG] is one way to be a moderate externalist, conceding a main thrust of the exernalist arguments but being discriminating about the range of the terrain they impact. All things considered, it is my preferred response to the externalist challenge—that is, there is an [EG], but it is a function of how certain limited kinds of terms are typically used by ordinary speakers. (Semantic externalism may nail shut the coffin lid on the very idea of a priori access to the essence of aluminum, but [&E], widows, and fortnights remain steadfastly unaffected.) Furthermore, a moderate externalist response to the externalist challenge dovetails with a modal revisionist response to challenge of revisability, constituting a stable ground between absolutism and skepticism. Hence, the [UJ] connections can thereby survive these two formidable and important challenges, within a constitutive a priori orientation.
§6.6: Summary of Part III
Next, I summarize what the constitutive a priori orientation is, in general, and how it absorbs the challenges of revisability and externalism, in particular. Thus continues the ongoing process of charting my considered stance on the different senses of immunity to counterexample—mindful as always of the framework-relativity of immunity to counterexample, the context-relativity of frameworks themselves, and the important differences between different kinds of frameworks.
The most general question, over here in the corner of epistemology which overlaps with the philosophy of language and the metaphysics of modality is: Is there or is there not a non-empirical source of justification? Can an adequate epistemology be developed without appeal to one? A priorists are those who answer YES THERE IS to the first and NO IT CAN’T to the second. Some variants of classical skepticism and contemporary naturalism are two varieties of anti-a priorist; but probably most and in any case very many important philosophers are in the a priorist camp.
Historically, over the almost twenty-five centuries between Plato’s ‘light of the mind in her own clearness’ and BonJour’s (1998) ’defense of pure reason,’ most a priorists have espoused a variant of what was first distinguished in §1.4 above as an acquaintance-based approach, according to which a priority is taken to be a quasi-perceptual relation between minds and the objects of knowledge, often described as a kind of ‘mental seeing’. The alternative understanding-based approach was perhaps first clearly articulated by Hume (1748) under the guise of ‘relations of ideas’, though one can clearly discern seeds of this approach in Hobbes and Locke, and perhaps going all the way back to Aristotle’s opposition to Plato’s extravagant, otherworldly epistemology. The guiding idea here is that the a priori is not an extra, elaborate cognitive faculty, but can rather be fashioned out of ingredients to which we are all committed (i.e., a comprehensive account of what it takes to grasp or understand a concept). The signature objection to acquaintance-based approaches is that they amount to little more than obscure, unhelpful metaphors; and a key challenge to understanding-based accounts is to explain the possibility of ampliative, substantive a priori knowledge. Critics of the latter allege that what it calls justification does not really amount to any such thing.
There are many significantly distinct sub-varieties of both acquaintance-based and understanding-based approaches to a priority. For example, even after narrowing our focus onto the understanding-based orientation, there are adherents to this view who take themselves to be rationalists, and others whose approach is styled in the empiricist tradition.
Following the lead of some major twentieth-century thinkers and developments, I am defending a constitutive a priori approach to this nexus at which epistemology overlaps with the philosophy of language and the metaphysics of modality. This is a broadly neo-Kantian status-refinement within the understanding-based tradition—that is, not only does a priority not apply to some special faculty of mind, as on the understanding-based tradition, it also does not solely apply to some special category of content, as the moderate rationalist would have it. It essentially also has to do with status, role, or function. To categorize something as a priori is, in part, to say something about its place in a given framework (or language game, world picture, theory, etc.).
I take the challenge of revisability, associated with Quine circa 1950, and the externalist challenge, associated with Kripke circa 1970, to be huge recent jolts to the a priorists’ world-order. They render if not obsolete, then at the very least in need of extensive refinement, many traditional ideas about epistemology, semantics, and metaphysical modality. I have argued that these challenges are well met by a constitutive a priori picture which espouses a modal revisionist response to revisability (i.e., the a priori is real but framework-relative) and a moderate externalist response to the externalists’ challenge (i.e., the [EG] is very real and deeply significant but limited in scope to certain types of uses of certain types of terms). This leaves a fairly broad and significant terrain open for business for acquaintance-based a priori ratiocination, amenable to [UJ]-underwritten immunity to counterexample.
Recall from §5.1 the following [UJ] principles, which I characterize as the spine of an understanding-based approach to a priority, and which articulate the direct contact which all of these finer points have to Plato’s problem:
UJP1 tells us that if there are [UJ] connections, then analyticity can ground (at least some cases of) a priori justification. UJP2 tells us that the realm of the un-black-swannable is the range of cases in which semantic intuition can underwrite the [UJ] connections. Basically, to the extent that the [UJ] connections hold up, semantic intuition can be relied on to accomplish what has been traditionally demanded of rational intuition. In a vast range of cases, semantic intuition can underwrite immunity to counterexample.
Natural reality is the range of the necessary a posteriori; social-conventional reality is the range over which sense determines reference. It is here that the [UJ] connections are resistant to the [EG]. Again, though, bear in mind that these categories are not discrete monoliths. There are many shades of grey, and directions of fit, and multiplicities of overlapping frameworks, between deferential uses of ‘Aluminum is a metal’ and transparent and autonomous uses of ‘A fortnight is a period of 14 days,’ or ‘If P and Q then P.’
Ultimately that is why my constitutive a priori view is a variety of empiricism, not rationalism—that is, the range of the a priori is the framework, and the concepts they compose; so there is no a priori knowledge of the world on this conception of this orientation. This I see as a natural, though perhaps not inevitable, consequence of the move to an understanding-based view, which denies that a priority is a distinct faculty of mind. It is starkly put by the externalist challenge, but the writing has been on the wall at least since the development of non-Euclidean geometries. Traditional conceptions of rational intuition are not the potent weapons they were once conceived to be.
In any case, the main thesis of Part III is that the constitutive a priori orientation on Plato’s problem, and the framework-relative [UJ] connections which form its spine, can survive the challenge of revisability by endorsing modal revisionism, and the externalist challenge by endorsing a moderate variant of semantic externalism.
NOTES
1. Wittgenstein scholarship is a notoriously tangley, contentious, dangerous business—his work is subtle, nuanced, complex, and not to be read as supporting any simple, straightforward philosophical thesis—so I enter the disclaimer that I have no pretense of having discovered the one true reading of Wittgenstein on the a priori. Rather, the following is a plausible reading of a certain thread running through Wittgenstein’s thought, backed up with some textual evidence, chosen because I take it to well-serve my ongoing project. Not all experts will agree with all of it.
2. This may strike an echo to Poincare’s conventionalism: ‘Although the influences on Wittgenstein’s work are notoriously difficult to interpret, there are clearly evident similarities to Poincare’s fundamental views’ (Heinzmann & Stump, 2013, §5). Relatedly, cf. Friedman (2007: 96–99) for conceptual ties between Poincare and Reichenbach. There is a bit more discussion of Poincare below.
3. For a sample of the literature on this issue cf. O’Grady (2004).
4. Wittgenstein wrote in a letter to Schlick in 1932 that ‘neither Poincare nor Reichenbach could have the same conception [of hypotheses as I do] because they do not share my conceptions of propositions and grammar’. Cf. Stern (2007: 325) for discussion.
5. As mentioned in the Preface, the constitutive a priori orientation has deep roots in the philosophy of science literature, from canonical European figures like Poincare and Cassirer, to classical American pragmatists Pierce, Dewey, and Lewis. My aim here is not so much to exhaustively cover that historical terrain as to argue for its more broad applicability to the ways in which twentieth-century philosophy of language has altered traditional epistemology.
6. Friedman (2000) is extensively cited here as a representative illustration of the constitutive a priori, not because I think it is the final truth on the matter. Friedman’s view continue to change and evolve (e.g., [2007, 2008, 2011], and so on), and there has been much critical discussion of Friedman’s views which I will not turn to here. (For a variety of critical engagements with Friedman’s work, cf. Korkut [2011], Angeloni [2012], Uebel [2012], Everett [2015], Stump [2015].)
7. Cf. Haack (1974) for a canonical discussion of deviant logics; cf. Beall & Restall (2006) for a case in favor of logical pluralism.
8. Williamson’s version of what I am calling [UJLT corollary] is: if something is a logical truth, then assenting to it is a necessary condition for understanding it.
9. Actually, the ambiguity option might have some real purchase in this case. One could argue that ‘~Φ’ or ‘Φ v Ψ’ literally means something different for an intuitionist, as opposed to a classical logician. (Thanks to Wayne Myrvold for pressing this case in the discussion period after I presented the argument of this section.) To the extent that this is so, then this points to a rather clear difference between the DLOs of Horwich and of Williamson. Note well though that taking this path would undermine the promise of Horwich’s DLO to support any drastic conclusions about [UJ] connections (which could only be derived from a SHARED understanding, not from diverse understandings).
10. As Flanagan (2013: 346–47) documents, multiple authors have tried to answer Williamson’s challenge in one (or both) of these ways, but in Flanagan’s (and my) assessment, that will not do. As Williamson (2009: 135; 2011: 499) insists, this is a case of ‘theoretical disagreement,’ not equivocation or incompetence.
11. Williamson (2008: 95) concedes that [&E] may ‘have the best chance’ as far as candidates for [UJ] connections go, but argues that even it is subject to competent, unequivocal dissent. However, his putative counterexamples (p. 96) are relatively weak and problematic. (Cf. Boghossian [2011], Peacocke [2011] for discussion.) In my opinion he does a much better job of motivating counter-instances for ‘All As are As’, and they are enough to force a challenge to [UJLT].
12. Indeed, within propositional logic, one might take the LEM and LNC to implicitly define the term ‘proposition’. (The price of that move is that ‘There will be a sea battle tomorrow’ or ‘Erin is tall’ might fail to express a proposition.)
13. Maddy (2000: 113) reads Carnap in a way that is incompatible with this kind of complexity; I will not get into a battle of Carnapian exegesis but will explicitly disavow any such static, categorical reading of the internal/external or language/theory distinctions as applied to frameworks.
14. Railton’s messy Wittgensteinian picture stands in stark contrast with Friedman’s much more regimented three-level picture. This is due in no small part to differences of range—that is, Friedman is focused on the history of science, while Railton is evidently talking about inquiry in general.
15. And this is not even to mention her huge qualification about the case of logic in her note 43. If we compare the quote from Quine (1951: §6) cited at the opening of §5.1, about revisability within logic, there are some serious doctrinal differences between Maddy’s naturalism and Quine’s.
16. As alluded to above in §5.4, Locke’s (1690: Bk III) distinction between terms which target a real essence and terms which target a nominal essence is an important early analysis of this core distinction. That widows are formerly married women whose spouse has died articulates a nominal essence, while that aluminum is a metal targets a real essence. Deference, openness to correction by experts, and the possibility of community-wide error are distinctively evident in cases of intending to target real essence.
17. Here we brush up against the issue of tacit knowledge, a general issue in the philosophy of psychology concerning the different senses in which (say) competent speakers should be said to ‘know’ the rules which govern their linguistic behavior. For example, anyone who has taught Introductory Logic will have encountered students who unfailingly use ‘and’ in accordance with the standard introduction and elimination rules, but who nonetheless have great difficulty in attaining a theoretical grasp of those rules. Such agents tacitly know these rules, as is attested by their (deliberate, rule-governed, intentional) ground-level thought and talk involving ‘and’, even despite their theoretical difficulties.
I take it that such agents are not counterexamples to the idea that competence with ‘and’ is constitutively tied to the standard introduction and elimination rules—only someone who we would want to call competent with ‘and’ but yet reflectively conceded counterexamples to those rules would. (Cf. §6.2 for extensive discussion of such deviants.) However, they do show that finessing is required as to exactly what is meant by ‘transparency’ in this context. Such agents show that my claim has to be that grasp of intension can be transparent while being merely tacit, as opposed to the stronger claim (which I do not want to make) that transparency demands the kind of theoretical grasp which is necessary for success in a Logic course. What matters for the claims about ‘transparency’ made herein is (deliberate, rule-governed, intentional) action in accord with the rules, and that such rules are in principle accessible to reflective agents—despite the evident non-trivial differences between competent speakers, when it comes to the ability to consciously grasp the rules.
The general issue of tacit knowledge has some relevance to this ongoing strand of debate between the different theses about the [EG] introduced in §5.3 and returned to immediately below, insofar as they are essentially (if only partly) concerned with what counts as semantic competence. However, this relevance is relatively tangential—it is not as if the extreme differences between Locke (1690) and Soames (2002), when it comes to competence, have anything to do with tacitness per se—and so I will go no further into it here.
There is a bit more on Williamson’s challenge to the common presumption that competence with ‘and’ is constitutively tied to the standard introduction and elimination rules below at note 19.
18. We need not even go to such far-fetched thought experiments to illustrate this point. For example, reactions [i] and [ii] might surely have applied to the scientific recognition that whales are mammals and not fish, or, more recently, to the de-classification of Pluto as a planet. (This is a nice recent illustrative example: media reports I encountered about the decision to de-classify Pluto from the ranks of the planets tended to include a stubborn person-on-the-street insisting ‘As far as I am concerned, Pluto will always be a planet!’ This is a transparent but not deferential response.)
This first refinement is also applicable to the Chomskyan (1993, among other places) sort of pessimism about the externalist tenet that H2O is the essence of water, on the grounds that what counts as water in lots of places is a lot less purely H2O than Sprite or tea is. (Cf. note 11 from chapter 5.) Many such uses of ‘water’ are not uses as deferential, essence-targeting natural kind terms, but rather uses as relatively crude practical kind terms (i.e., ‘whatever it is that flows out of this tap’).
19. This point is relevant to debates between Boghossian and Williamson, over the [UJ] sort of epistemology of logic I am appealing to for ‘and’ (i.e., the presumption that competence with ‘and’ is constitutively tied to the standard introduction and elimination rules). One of Williamson’s challenges to Boghossian’s defense of this epistemology of logic is to say: if all your epistemology can guarantee is such pristine and sterile instances like ‘and’-elimination, then your epistemology is not terribly worthwhile:
Strategically, Boghossian’s response is not very promising. If he can rely on understanding-assent links only for ‘and’-elimination and a few other equally banal rules … then he is in no position to base either a general epistemology of logic or a general account of the understanding of logical constants on understanding-assent links. (2011: 498)
I see in this inverse proportion relation between transparency and deference the promise for a general reply to Williamson on this front, which applies to varying degrees, as appropriate, beyond the bounds of ‘and’. (Of course, I am not sure that Boghossian would want to endorse anything along these lines.)
20. There is of course lots of relevant literature, on similar and related questions—such as, for example, the rule-following considerations (e.g., What exactly constitutes meaning addition by ‘+’?) My claim is just that direct attempts to explicitly investigate the depth of competence required in these externalists’ incomplete-mastery cases are relatively rare. (Cf., Sullivan [2010] for discussion and references.)
21. This line of thought is extensively developed in Sullivan (2012).