Yet Another Victim of Kripkenstein’s Monster: Dispositions, Meaning, and Privilege

1. Semantic Dispositionalism and the Standard Arguments against It

In the sense in which I use the term, there is more to dispositionalism than just the idea that the meaning of a linguistic expression is in some way a matter of how speakers are disposed to use it. Such a use of “semantic dispositionalism” is not especially idiosyncratic, but it is nonetheless worth explaining. “Semantic dispositionalism”, in this sense, refers to a family of views Kripke (1981/1982: 22–37 and 110–12) criticizes in the context of his discussion of (what he takes to be) Wittgenstein’s rule-following paradox. And such views do not share just the commitment to the very general idea that meaning is in some way a matter of dispositions; they also share two other important features: they argue for an a priori reduction of meaning to speakers’ dispositions and they assume a certain view about the relation between the notion of meaning and that of reference. Let us start with the second point.

Semantic dispositionalism, in the Kripke-inspired sense in which I am using the term, relies on a certain picture of the concept of meaning (for which see, e.g., Chalmers 1996 and Kaplan 1989), what is sometimes called the “referential picture” of meaning. According to this picture, the meaning of a linguistic expression is something like a rule explaining how the referent, the denotation, of that expression depends on certain parameters—in more formal terms, a function which takes possible referents as values.³ The meaning of “I”, for example, would tell us that its referent depends (roughly) on the identity of the speaker, and the meaning of “the president of the United States” would tell us how its referent depends on features of the conversation’s topic (are we talking about the latest news or about the Cuban missile crisis? About this world or about some merely possible one?). Now, against the background of the referential picture, asking what makes it the case that a certain expression has the meaning it has is the same as asking what makes it the case that, given the value of the relevant parameters, a certain expression refers to what it refers. What determines the reference of, say, “red”? What makes it the case that “+” refers to addition, and not to some other function?⁴ It is to such questions that the dispositionalist answers that it is all a matter of the speakers’ dispositions.⁵

However, talk of dispositions determining reference, or making it the case that a certain sign denotes a certain entity, is somewhat vague (in the everyday, not the technical, sense of the word). Is the idea that facts about reference supervene on facts about the speakers’ dispositions? Or are such claims better interpreted, say, in terms of grounding? Or maybe in terms of identity? And is dispositionalism a reductionist view or not? And if it is, what kind of reduction are we talking about? An a posteriori reduction, such as that of water to H₂O? Or is the dispositionalist making a conceptual analysis claim, thereby championing an a priori reduction of meaning and reference to speakers’ dispositions?

Kripke is not very explicit about such issues, but it seems clear to me that the dispositionalists he has in mind must be arguing for some form of a priori reduction. In reading Kripke this way I agree with Boghossian (2015: 336); however, unlike Boghossian, I also believe that Kripke had excellent reasons to focus on conceptual analysis varieties of the dispositional view. Kripke begins his exposition of (what he takes to be) Wittgenstein’s argument by casting doubt on the very intelligibility of the notion of meaning, referentially conceived, and so the challenge his book revolves around is primarily to make sense of that concept. And in order for it to be addressing this problem, dispositionalism must, of course, be put forward as an exercise in conceptual analysis.

Here, then, is the kind of view I have in mind when I speak of “semantic dispositionalism”. You start with the background assumption that the concept of meaning must be made sense of in terms of reference, roughly along the lines explained above. Then you are confronted with the usual battery of Kripkensteinian arguments aimed at showing that the referential picture of meaning is of dubious coherence. And this leads you to try to make sense of the results of your referential analysis of meaning in terms of dispositions. Semantic dispositionalism, in the sense in which I am using the term, is therefore the general view that meaning, referentially conceived, can be a priori reduced to speakers’ dispositions.

That the dispositionalists’ goal is to provide a reductive account of the concept of meaning is especially worth stressing, for it entails that—on pain of their analysis turning out to be viciously circular—dispositionalists cannot make, in their account, any use of that concept. This is a key requirement that any attempt to answer Privilege—and, more in general, any attempt to defend dispositionalism—must meet, and which Kripke himself (1981/1982: 27–32) explicitly calls attention to. We will encounter it again in Section 3, during our discussion of some of the ways in which a dispositionalist may try to meet the challenge raised by Privilege.

Having clarified what the dispositionalists’ thesis amounts to, it is useful to say something about its prima facie plausibility. This comes, I believe, mainly from the fact that the dispositional strategy is suggested quite naturally by the standard story about past usage’s inability to determine reference.⁶

Let us imagine that we have been using “karatakai” to talk about the color a certain fruit is when perfectly ripe—where the description should be read de re. We now bump into a hue (“H”, for short) we had never seen before. Should H things be labeled as “karatakai”, or not? Since H is a hue we had never seen before, it is clear that past usage, by itself, is not enough to answer such a question.

One could, however, try to tell a more complex story, of which past usage is but one component. One could, for example, try to argue that every time we used “karatakai” we also grasped a certain universal, and that that is enough to determine whether H things should be labeled as “karatakai” or not—if H things instantiate the universal in question, it is correct to call them “karatakai”; otherwise, not.⁷

But the most natural way to muscle up the intuition that the way a word should be used must have something to do with the way it has been used in the past is, of course, to go dispositional. Every time we used “karatakai”, our utterance was the manifestation of just one of our countless dispositions concerning that word. Most of those dispositions have never been manifested, usually because we have never bumped into the relevant stimuli, but they were there nonetheless; and among them there were also dispositions concerning H, which determine whether H things should be labeled as “karatakai” or not—if these dispositions were dispositions to categorize H things as karatakai, then it is correct to call H things “karatakai”; otherwise, not.

As natural as it is, the idea that the reference of words and sentences is determined by the speakers’ dispositions has, as I have already noted, many enemies. Following Kripke (1981/1982), skepticism about the dispositional strategy is usually motivated by an appeal to some subset, proper or not, of a group of three arguments.

First, since we are all disposed to make (what we would usually regard as) mistakes, for example, categorizing white objects as red under certain illumination conditions, a dispositional analysis assigns the wrong meanings to our words and can therefore be discarded out of hand as empirically inadequate (the Mistake Argument, or just Mistake).

Second, we do not have enough dispositions, and so—once again—a dispositional analysis does not assign the correct meanings to our words (the Cardinality Argument, or just Cardinality). Take, for example, the case of “+”. The dispositional analysis tells us that “+” means plus if and only if our dispositions concerning that sign track addition, but our dispositions cover only a finite segment of addition—since we lack any dispositions about additions involving huge numbers—and therefore the dispositional analysis tells us that “+” does not mean plus. But, of course, “+” does mean plus and, therefore, the dispositional analysis is empirically inadequate.⁸, ⁹

Third, meanings entail categorical oughts. If, for example, “+” means plus, then you ought to answer “125” if asked for “68 + 57”—and this ought does not depend on your desire to give the correct answer, nor on any other particular desire of yours.¹⁰ But having certain dispositions does not entail any such ought. Therefore meaning cannot be analyzed in terms of dispositions (the Ought Argument, or just Ought).¹¹

These are, of course, also the arguments on which the friends of dispositionalism focus when they have to defend the strategy. There is, therefore, something like a consensus that the destiny of dispositional analyses of meaning hinges on the status of these arguments (see, e.g., Båve 2020; Blackburn 1984; Cheng 2009 and 2010; Ginsborg 2011; Heil & Martin 1998; Jones & Podlaskowski 2012; Kowalenko 2009; Kusch 2005; Podlaskowski 2012; and Schlosser 2011). This is unfortunate, since there are other arguments which provide, I think, stronger reasons to be wary of this approach. The Privilege Argument is one of them.

2. Cardinality, Mistake, and Privilege

The best way to introduce the Privilege Argument is against the background of Cardinality and Mistake. This is because the Privilege Argument can be viewed as arising from a weakness in the standard strategy for dealing with Mistake, which is in turn related to Cardinality.

Let us start, then, by asking what a successful reply to Cardinality must look like. The answer is, I think, obvious. In order to successfully deal with it, a dispositionalist must show that even though it may seem that I have no disposition to give the arithmetically correct response “Z” if asked for the sum of two huge numbers X and Y, in fact I do have such dispositions.

Building on previous work by Blackburn (1984), Shogenji (1993), and others, Warren (2020: 262–68) has recently offered a nice implementation of this strategy, and having taken a look at it here will come in handy later on. The basic idea behind Warren’s approach is simple. For some pairs of (smallish) numbers, there is no doubt that I have the disposition to give the arithmetically correct response if queried about their sum. For some others, even though I do not have any simple disposition to give the correct answer, I have suitable complex dispositions: if asked for “1,091 + 2,906” I would answer “3,997” as a result of the execution of some intermediate steps (i.e., adding 1 to 6, 9 to 0, etc)¹²—even though I have no disposition to give that answer directly, without going through the various steps of the algorithm. When the addenda are really huge I have neither a simple nor a complex disposition to give the arithmetically correct response—even if I were to start going through the relevant intermediate steps, I would die of old age before getting halfway through;¹³ however, in such cases I still have an appropriate composite disposition, where a subject S has a composite disposition to perform action A if and only if A would be the output of the manifestation of certain linked dispositions of S’s.

The key claim is, of course, the last one, so let us take a closer look at it. Let d₁ … d_n and g₁ … g_m be, respectively, an n-digit and an m-digit number—where n and m are assumed to be huge and, for simplicity’s sake, it is also assumed that n > m. One way of having a complex disposition to give the correct response to “d₁ … d_n + g₁ … g_m” would be to have the disposition to go through the following steps: first, I add d_n to g_m, I get the result t₁t₂ (where t₁ is either 1 or 0), and I write down “t₂”; second, I add d_n-1 to g_m-1, I get the result t₃t₄, to which I add t₁, thereby getting t₃t₅ (of course, sometimes t₄ = t₅), and I write down “t₅” to the left of “t₂”; and so on—modifying the procedure in the obvious way once I get to d_n-m. Now, let us say that this is a disposition I do not have—because by the time I get to d_n-p (where p is assumed to be smaller than n but still huge) I will be long dead. My lacking this disposition is perfectly consistent with the notion that, for each digit d_i of d₁ … d_n, I do have the disposition to implement the n-i+1th step of the algorithm above. And having these dispositions is just having a composite disposition to give the correct response to “d₁ … d_n + g₁ … g_m”. Let us imagine, for example, that I am really slow at adding and that, because of this, I lack the complex disposition to correctly compute “714,231,723,055,439 + 231,135,230,349,825”—by the time I get to the fourteenth step of the algorithm above I will be six feet under. This does not mean that I do not have, say, the disposition to compute “1 + 3” and write down the result!

Warren’s strategy to deal with Cardinality looks, I think, promising enough. Let us now turn to Mistake and ask what a successful reply to this problem would look like. I think it is clear that a solution to Mistake will have to be somewhat more complex than one to Cardinality. More precisely, I believe that a little reflection is enough to see that this time the advocates of the dispositional strategy must both (1) convince us that there is a kind of dispositions k such that, for any pair of numbers, the answer I am disposed_k to give if asked for their sum is the arithmetically correct one and (2) modify the original, unqualified, dispositional analysis so that the meaning of a word depends only on our dispositions_k.

Warren (2020: 270–73) offers, once again, quite a neat concrete case to focus on. First, let us define the notion of a general disposition by saying that S has a general disposition to perform A in a class of situations γ if and only if in the overwhelming majority of the situations in γ S performs A (this is what Warren calls an “M-general” disposition). Second, let an answer to a given question be stable if and only if checking and rechecking tends to produce the very same answer over and over again. Third, let us define the normal conditions as those in which neither external nor internal factors are interfering with the functioning of our cognitive apparatus.¹⁴ Finally, let us put these three notions together and define a disposition of kind gsn (a “disposition_gsn”, for short) as follows: S has the disposition_gsn to give a certain response to a certain addition problem if and only if S has the general disposition to stably give that response when queried about that problem in situations in which the conditions are normal. Now, Warren maintains that, for any pair of numbers, the answer I am disposed_gsn to give if asked for their sum is the arithmetically correct one (see point (1) above) and, accordingly, he suggests that we should take “+” to mean plus if and only if our dispositions_gsn concerning that sign track addition (see (2) above).

Keeping all this in mind, we can now introduce the Privilege Argument. As we have just seen, dispositionalists are committed to the idea that there is some kind of disposition k such that, for any pair of numbers X and Y, the answer we are disposed_k to give if queried about “X + Y” is “Z”—which, here and in what follows, will always denote the sum of X and Y; however, dispositionalists do not deny that for some other kind of disposition j and some pair of numbers X and Y, the answer we are disposed_j to give if queried about “X + Y” is not “Z”—we are disposed_j to make a mistake.¹⁵ In other words, even though they believe that, for some k, our dispositions_k track addition, dispositionalists are willing to grant that there are js such that our dispositions_j do not track that function—if one is drunk, or overly tired, they will make some mistakes. Dispositionalists, however, are keen to assure us that this does not really matter, since the dispositions on which the meaning of “+” depends are only our dispositions_k; dispositions_j are entirely irrelevant. But why so? Why are our dispositions_k privileged over all our other dispositions? Why does a word’s referring to a certain entity reduce to our being disposed_k, rather than disposed_j, to use it in a certain way?¹⁶ Dispositionalists do not tell us—and, moreover, it is not clear what they could say. And if dispositionalists cannot answer such questions, the dispositional strategy does not really explain what it was supposed to explain, and it should therefore be abandoned. This is, in a nutshell, the Privilege Argument.¹⁷

That Privilege raises quite a serious problem for any dispositional view of reference becomes especially clear if we pay due attention to the fact that facts about reference are normative in nature: if “rosso” refers to the class of all the red things, then it is correct to call “rosse” the red things but not the non-red ones. If we keep this well in mind, the questions dispositionalists find themselves facing take a different form: why should the answer we are disposed_k to give also be the answer we ought to give? And where does the alleged normative oomph of these dispositions come from? And as soon as we turn our attention to such questions, the dispositionalist’s inability to provide satisfactory answers becomes, I believe, glaringly apparent.

It is important, though, not to misunderstand the nature of the relevant normativity. Unlike the oughts around which the Ought Argument revolves, the oughts (and the normative oomph) I have in mind here are unremarkably hypothetical. The relevant notion of the answer I ought to give is just the notion of the answer I ought to give if I care about mathematical truth—and, more in general, the relevant notion of the way one ought to use a word is just the notion of the way one ought to use that word if they want to tell the truth, or if they want to communicate. This means that what is usually referred to as “the debate on the normativity of meaning” (see, e.g., Boghossian 2003 and 2005; Glüer and Wikforss 2009; Hattiangadi 2006 and 2007; Whiting 2007 and 2009; and Wikforss 2001) is utterly orthogonal to my argument. The notion of normativity at issue in that debate is a very strong one: the oughts which people like Daniel Whiting have argued follow from semantic facts are categorical (they do not depend on any desire of ours). None of the participants to the debate on the normativity of meaning regard the idea that there are merely hypothetical semantic oughts as particularly problematic (see, e.g., Hattiangadi 2006: esp. §1 and 2007: esp. what she says about normativity vs. norm-relativity). Even more importantly, that there are such oughts is something that semantic dispositionalists themselves accept. Indeed, that there are such oughts is a straightforward consequence of one of their key theses, namely of the thesis that meaning can be made sense of in terms of reference (see again §1 of Hattiangadi 2006, but also Boghossian 1989: 513 and 530). If “rosso” refers to the class of all the red things, then it is correct to call “rosse” the red things but not the non-red ones, which in turn straightforwardly entails that if one wants to use that word correctly they ought to call something “rosso” only if it is red.

It is also important not to misunderstand the role that considerations about normativity play in the challenge raised by Privilege. The challenge is to explain why the fact that a certain word has the meaning it has reduces to facts about a certain kind of disposition—what is so special about those dispositions? Hence, the challenge can be formulated without any mention of oughts and the like, and in principle it could be issued to any view which identifies meaning facts with facts about a certain kind of disposition—even to views which reject the notion that meaning is in any sense normative. Now, since, as a matter of fact, dispositionalism (in the sense of the term defined in Section 1) assumes meaning facts to entail certain (hypothetical) oughts, the challenge can also be phrased in normative terms. And phrasing the challenge this way no doubt makes it especially vivid, which is why I think not only that it is useful to give this formulation too, but also that sometimes it is more effective to focus on this dimension of the argument. But that should not obscure the fact that the challenge is primarily that of explaining why meaning has to do with that particular kind of disposition—no mention of oughts needed. So there is definitely a sense in which the Privilege Argument can be described as a normativity argument (provided, of course, that one is clear about the kind of normativity involved),¹⁸ but by so doing one would be focusing on a somewhat superficial feature of the problem.

The challenge raised by Privilege—this, too, is definitely worth stressing—is also quite different from the one raised by Cardinality and Mistake.¹⁹ As a matter of fact, dispositionalists do not want their analysis to entail any revisionist corollaries about the referents of the words of our language; they do not want to find themselves committed to the notion that, say, “+” refers to something other than the addition function. The point of both Cardinality and Mistake is that such corollaries do follow from the dispositional view, either because we do not have enough dispositions or because we are all disposed to make (what we would usually regard as) mistakes. This is not the point of Privilege. The point of Privilege is not that dispositionalism sometimes assigns the intuitively wrong referents to the words of our language but, rather, that it cannot justify the assumption—which it has to make—that certain dispositions have a normative oomph that all the others lack and, more in general, that only those dispositions are relevant to the determination of a word’s meaning and reference.

One way to appreciate the difference between, on the one hand, Privilege and, on the other, Cardinality and Mistake is to focus on the fact that answering Cardinality and Mistake does not, in and by itself, give us an answer to Privilege. Let us assume, for example, that Warren’s answers to Cardinality and Mistake are successful. That would mean that if, first, we take care to analyze the meaning of “+” in terms of the relevant dispositions_gsn (i.e., our general dispositions to stably give certain responses when queried about “+”-problems in situations in which the conditions are normal) and, second, we keep in mind the need to quantify over composite dispositions too, our analysis will deliver the desired result about the reference of “+” (i.e., that “+” refers to the addition function). That, however, is not yet having shown that the correct way to use a word is the way we are disposed_gsn to use it.

The point is that Warren has proved, at most, that if the correct way to use a word is the way we are disposed_gsn to use it, then it is not the case that we are hopelessly wrong about the correct way to use most (maybe all) of the words of our language—let me stress it: a conditional statement. If we agree that any analysis that entailed that we are hopelessly wrong about the correct way to use most words is a non-starter, that is a fine result; Warren has shown that the hypothesis that the correct way to use a word is the way we are disposed_gsn to use it is not, unlike other dispositional analyses of meaning and reference, clearly absurd. But Warren has not proved the antecedent of the conditional, that is, that the correct way to use a word is indeed the way we are disposed_gsn to use it.²⁰

Here is another way to see that the point I am calling attention to has little to do with the general line of argument instantiated by Cardinality and Mistake. As I have already noticed, both Cardinality and Mistake rely on the assumption that any analysis that entailed that we are hopelessly wrong about the correct way to use most words (i.e., any analysis that assigns the intuitively wrong referents to most words) is a non-starter. Let us now try to drop this assumption and see what happens. Of course, neither Cardinality nor Mistake has any force anymore: if revisionism is not a flaw, entailing that “+” does not refer to addition cannot be a mortal sin. Not so with Privilege, though. Whatever dispositions the dispositionalist ends up identifying as the meaning-determining ones, one can still ask why those and not others.²¹ Why is the answer we are disposed to give in these conditions the right one? Why not the answer we are disposed to give in those other conditions? And so on.

One can think of Privilege as having three steps. The first step consists in realizing that dispositionalists like Warren are making the above assumption, that is, that the only meaning-determining dispositions are our dispositions_k. Second, one has to recognize that that assumption is in need of justification. And, finally, we end up realizing that no such justification is forthcoming. Since dispositionalists usually make the assumption in question tacitly, the first step is not trivial, but by now it should be clear that, indeed, dispositionalists do make this assumption. In the remainder of this section I want to say something more on the argument’s second step, before finally turning to the third one in the next section.

The reason why I believe it is useful to say something more on Privilege’s second step is that one may feel that when we ask the dispositionalist to justify the assumption that the only meaning-determining dispositions are our dispositions_k we are making a demand that we would never issue to the proponents of other reductive theories—which, in absence of a convincing explanation of the asymmetry, would of course give us very strong reasons to suspect that here there is really no genuine explanatory burden for the dispositionalist.

My answer to this, very natural, worry is that, contrary appearances notwithstanding, there is no asymmetry here. The explanation we are asking of the dispositionalist is no different in kind from the ones that we routinely expect from a reductive theory. Hence, the point of Privilege is just that dispositionalists do not seem to be able to meet the explanatory standard that successful reductive theories of other phenomena effortlessly meet.²²

Take, for example, the case of memory. The common wisdom in neural science is that facts concerning the storage of general knowledge about the world, say that elephants have trunks, should be identified with certain neocortical facts (see, e.g., Schacter & Wagner 2013: 1449–51).²³ Now, when one makes such a claim, they must be able to justify it: what is so special about (the relevant areas of) the neocortex? And that is indeed what any good neural science textbook tries to do. Of course, ideally, what we would want is some kind of causal (see Lewis 1986) and/or mechanistic (see Machamer, Darden, & Craver 2000) story showing exactly how the manifestation of the knowledge in question is linked to the relevant neocortical facts—which is not really what we find in a neural science textbook. However, we are at least given, first, a general idea of how such a story might go and, second, some reasons (having to do with the observation of patients with brain lesions and/or with the use of neuroimaging techniques) to believe that here there indeed is a story to be told.

In a parallel fashion, if dispositionalists want to identify, say, “+” ’s referring to addition with certain dispositional facts, they have to show us how our having the dispositions in question explains the properties of the phenomenon they are trying to reduce/the facts that constitute “+” ’s referring to addition. Of course, there are also important differences between the two cases. For one, facts about what refers to what are normative in nature, in the sense explained above (if “rosso” refers to the class of all the red things, then it is correct to call “rosse” the red things but not the non-red ones), while mnestic facts are not²⁴—which suggests that the required explanation would have to be somewhat sui generis. Still, the general dialectic is the same.

3. Three Strategies to Try to Handle Privilege

One may agree that Privilege calls attention to a perfectly legitimate explanatory demand and still be skeptical that the argument raises any serious problem for dispositionalism—simply because they are confident that the dispositional view has the resources to meet the challenge. In this section I will argue that there are strong reasons to believe that such trust in the dispositionalist program is misplaced.

I shall discuss three strategies a dispositionalist can try to develop to meet the explanatory demands introduced in Section 2, starting with the least and concluding with what I take to be the most promising: the answer I am disposed_k to give is privileged because it is the right answer, the answer I am disposed_k to give is privileged because it is the answer I would give in ideal conditions, and the answer I am disposed_k to give is privileged because it is the answer I would give in standard conditions.

The lessons I will draw from my discussion are quite easy to generalize to other cases. That being said, the primary goal of this section (and, indeed, of the paper as a whole) is just to show that the challenge raised by Privilege should be taken seriously and that, by ignoring it, dispositionalists are dodging an issue that, on the contrary, they should regard as quite pressing. Hence, here I am not committing myself to the thesis that my arguments are generalizable to any attempt to answer Privilege dispositionalists may come up with—even though I suspect they are.²⁵