Cappelen's Four Arguments Against N-Theories of Assertion

In his paper "Against Assertion" (to appear in Assertion, eds. Brown & Cappelen (2010), OUP), Herman Cappelen defends a bold thesis which he calls the No Assertion View:

No Assertion View: Sayings are governed by variable norms, come with variable commitments and have variable causes and affects. What philosophers have tried to capture by the term 'assertion' is largely a philosophers' invention. It fails to pick out an act-type that we engage in and it is not a category we need in order to explain any significant component of our linguistic practice. (Cappelen 2010: 1)

His strategy for defending the No Assertion View (hereafter NAV) is to argue straightforwardly against what he calls Normative Theories of Assertion, which he calls N-Theories of Assertion:

Normative Theories of Assertion (N-Theories of Assertion): An assertion is a saying essentially governed by one or more norms.

In a previous post on this blog, I considered Peter Pagin's arguments against Normative Accounts of Assertion; Cappelen's own arguments are quite distinct and quite interesting. In what follows, I will examine these arguments, after which I'll offer some critical remarks.

Argument 1: Modal Judgments about norm variability for 'assertions'


The following is my own formalisation of Cappelen's first argument against N-Theories of Assertion:

1. According to N-Theorists, there is a norm, N, such that it is impossible for there to be an assertion that is not governed by N.
2. If conceivability is a guide to possibility and we can conceive of paradigtmatic assertions as governed by norms other than N, N-Theories are false.
   
3. We can conceive of paradgmatic assertions as governed by norms other than N.
4. Therefore, N-Theories are false.
   

The crucial premise here is (3). In support of (3), Cappelen basically asks us to imagine first that someone (here, Mia) says 'that Mandy forgot to pay her cell phone bill last week.'

Call this act E. What we're interested in is whether E could have been performed under a variety of different default norms. In other words, we can ignore the question of just what norm governs E in this world, and just focus on whether we can conceive of it as governed by a variety of norms. If we can, we have direct evidence against all N-Theories... (Cappelen 2010: 10).

Cappelen then asks whether we could have conceived of Mia performing E if the default assumption was that:

• she only assert p if she believes that p
• she assert p only if she is committed to defending p in response to objections
• she only assert p if p is true (10)

As Cappelen see is, we can duly conceive of any of these possibilities being the case, though the N-Theorist must say no in 'all cases except those cases that happen to coincide with the theorists's favoured norm.' So (3) is true, and from (1-3) we get the conclusion that the No-Assertion view is false.


Argument 2: Actual Variability in Norms

1. There are context-sensitive norms governing sayings.
2. If the N-Theorist is correct, there cannot be context sensitive norms.
3. Therefore, the N-Theorist is not correct (2010: pp. 12-13)

Cappelen defends (1) somewhat dogmatically, by appeal to Janet Levin. The more interesting premise, though, is (2). Cappelen's sums up his reasoning behind (2) as follows:

It's not as if we can easily go from e.g. the Knowledge Norm to contextually variable norms, and leave the rest of the framework intact. Remember that according to N-Theorists, the norms that govern assertion are constitutive of the act--the norm tells us what assertion is. Different token acts in different contexts are acts of the same type because they are governed by the same norm (or set of norms) that govern all these acts, so it seems the theory provides no account of what makes it the case that all the token acts are of the same type. If the norms vary between contexts, it is hard to see how the appeal to norms can tell us what makes an assertion an assertion. (2010: 13)

Argument 3: We never accuse speakers of having 'broken the rule of assertion' or 'cheated in assertion'

According to Cappelen, 'You will never find speakers being described in any of these two ways (at least outside a philosophy text):

(a) That assertion was cheating.
(b) That assertion broke the rules

or, with no reference to 'assertion':

(a) That was cheating
(b) That broke the rules

Argument 4: Infrequency of Assertion Attributions

The main idea here is that, according to the N-Theorist, the 'default assumption when someone utters a declarative is that she has performed an N-Assertion.

If that were so, we should expect that there be a word for that speech act and that that word be the default description of the acts in question. It should be like chess and tennis--when people play those games, there are expressions that denote the activity of playin those games and those are the default descriptions of the players. But, it is not like that for the game of assertion we allegedly play. This, I take it, is some evidence that we don't play assertion games. (Cappelen 2010: 16)

Some critical remarks

First, a dialectical point. Cappelen's four arguments against N-Theorists are supposed to generalise; they are meant to undermine not only the most substantial rival to the No-Assertion View, the N-Theorist's View, but also three other candidate positions according to which some account of assertion can be given. Those other rival candidate positions he taxonomises as claiming that:

(ii) Assertions are those sayings that have certain effects
(iii) Assertions are those sayings that have certain causes
(iv) Assertions are those sayins that are accompanied by certain commitments

As a purely dialectical point, it's not entirely clear how Cappelen's arguments, as he says 'provide...a fairly straightforward model for how to argue against views of type (ii), (iii) and (iv).' (2010: 2). I'll set this matter aside, though. What's more interesting are his arguments against the N-Theorist's view.

As I see it, the N-Theorist will have an easier time rebutting against Arguments 3 and 4 than 1 and 2. With respect to Argument 3: the N-Theorist can simply refute the inference Cappelen makes from the fact that cheating accusations are not leveled as such in the 'assertion game', that therefore, there is no assertion game for which alleged-norm-violating assertions would play the role of a cheating move. Even if we grant Cappelen his observation that cheating accusations are uncommonly made, the N-theorist can still maintain their position and offer some explanation for why this is so: perhaps, for one thing, we use a variety of expressions to capture the idea that some rule is broken., when it is. To say 'someone is cheating' is a particularly germane way to capture this idea when the game at issue is competitive (like chess or tennis). It is plausible though to think that competitive games, where the word 'cheating' is used, are not the only activities of which rules are constitutive--and in non-competitive activities, we just capture the observation of rule-breaking with different declarative descriptions. For instance, Lynch (2009) and Shah (2005) have suggested that inquiry is a goal-directed activity and that the notion of belief is a normative one such that the truth norm (i.e. believe only what is true) is constitutive of (and individuating of) the very concept of belief. Normativism about belief is a plausible view, and consider that we don't use the word 'cheating' to criticise a belief in violation of the truth norm. We don't call false believers cheaters, we just tell them they got it wrong. To the extent that normativism about belief is a live option, and one whereby no cheating accusations are made as such, it seems that normativism about assertion (N-Theorists) needn't have their backs to the wall simply on the observation that cheating fouls aren't called, as such. Extending this defensive analogy we can see that Cappelen's observation that we don't refer to the game of assertion (as we do to the games of tennis and chess) loses traction as an argument that we don't play the assertion game. After all, we play the inquiry game, and we don't refer to it as such. This latter observation I think undercuts much of the force of Cappelen's Argument 4--an argument that relies on pointing out our infrequently reference to any assertion game, as such.

That being said, I think that Arguments 1 and 2 pose difficult problems for the N-Theorist. Argument 2 can be sidestepped so long as the N-Theorist gives up what Pagin (2008) calls the constitutivity thesis--the claim that it is in virtue of being governed by some rule that something counts as an assertion, and endorses instead the slightly watered down position according to which some norm uniquely characterises assertion. Such a position claims, according to Pagin (and assuming something like the knowledge norm to play the role of the relevant norm):

UC: An utterance u by speaker S is an assertion that p iff it is necessary that (u is correct iff S knows that p. (Pagin 2008: 4)

As Pagin says:

Without the constitutivity idea it may be held that there is some factor F that makes u into an assertion, and that it is in virtue of being an assertion that the rule applies to u.

Problematically, though, it seems that even if the N-theorist gives up the constitutivity claim and holds instead only UC, that Argument 1 still poses a problem. Specifically, the necessity claim upheld in UC looks threatened by Cappelen's conceivability argument.

What moves are available to the N-Theorist (now slightly watered down to UC)? One move is to deny Cappelen's contention that conceivability is a mark of possibility. But that route doesn't look promising. Plenty of people take conceivabilty as a mark of possibility, so the N-Theorist should address the criticism while taking on board this assumption of the argument.

That said: I think there are two routes of defense. The first is more concessionary, the second less so. The first strategy would be to defend the N-Theory by giving up the necessity claim as part of UC and claim instead that:

UC*: An utterance u by speaker S is an assertion that p iff (u is correct iff S knows that p. (Pagin 2008: 4)

UC* allows the possibility that in some worlds, assertion is governed by a norm other than the one (say the knowledge norm) that governs assertion in this world.

A second, less concessionary line would be to try to link assertion to inquiry. One could argue by the following very broad strategy: that the value of truth is unrestricted, and so that belief aims at truth in all possible worlds. Something is a belief iff necessarily it it aims at truth (or some such claim as the normativists about belief hold). From this, one might argue that the norms that govern inquiry fix what norms govern moves in the game of inquiry--questions, assertions, etc. This is of course a very loose sketch of how such an argument would go. I'm not sure if it's plausible. As I see it, though, something like this argumentative strategy would have to be employed to save N-theories of assertion that don't wish to water themselves down by dropping the 'necessity' claim that framed part of UC.

Pagin Against Normative Accounts of Assertion

In "Against Normative Accounts of Assertion", Peter Pagin offers a challenge to what he calls normative accounts of assertion--accounts of assertion according to which something counts as an assertion in virtue of being governed by some norm. Put another way: Pagin is challenging the claim that some norm (i.e. for example, the knowledge norm) is constitutive of assertion. His argument takes the form of presenting a dilemma for those, such as Williamson, who endorse the constitutivity claim. Before engaging with Pagin's argument, I want to first relay some distinctions that helpfully inform it. First, consider that the knowledge rule is, taken by itself, simply a rule that we would say characterises assertion. An endorsement of the rule is not a claim about what assertion is, but rather, a claim that articulates a rule that applies to whatever counts as an assertion. The knowledge rule states:

KR: One ought: Assert that p only if one knows that p

Again, all this tells us is that, if something (for whatever reason) counts as an assertion, then a rule applies to it: assert only if you know. As Pagin notes, one 'adds an extra claim by saying of a rule that it not only characterises assertion but also that it uniquely characterises assertion. Concerning some rule R', that would amount to a claim that:

An utterance is an assertion iff it is governed by R'. (Pagin 2008: 4)

Speaking to the knowledge rule (KR), Pagin expresses the claim that KR uniquely characterises assertion as follows:

UC: An utterance u by speaker S is an assertion that p iff it is necessary that (u is correct iff S knows that p. (Pagin 2008: 4)

The claim that KR is is constitutive of assertion is however stronger than either KR or UC. This is important. Here's Pagin:

Finally, we can add the claim that a particular rule not only uniquely characterizes assertion but also is constitutive of assertion. Exactly what constitutivity amounts to is debatable (cf. Gluer and Pagin 1999) but it at least involves the idea that it is in virtue of being governed by a particular rule that an utterance u is an assertion. Without the constitutivity idea it may be held that there is some factor F that makes u into an assertion, and that it is in virtue of being an assertion that the rule applies to u.

With the constitutivity idea, the order is reversed (Pagin 2008: 5)

Pagin goes on to note that:

It is clear that Williamson favors the idea that the knowledge rule...is constitutive of assertion in some mainstream sense of constitutitivity (Williamson 2000, 238-43), and further that he thinks that the knowledge rule uniquely characterizes assertion. (Pagin 2008: 5)

Call those accounts that endorse the view that the KR (or some other rule) is constitutive of assertion normativity accounts of assertion.

Pagin's dilemma against normativity accounts takes the form of two forks:
Either:

(1) Norms are not involved in constituting an act as as assertion, and in that case we still need a non-normative account of assertion; or else
(2) Norms are involved, but then we need an empirical basis for the application
of norms, and such a story is hard to find. (Pagin 2008: 6)

Pagin's strategy, more specifically, is to argue by reductio--if we claim via (2) that 'norms are involved', we will be faced with a task we cannot satisfy: the task of providing an empirical basis for the application of norms. And so, (1): norms are not involved in constituting an act as an assertion (and we need to give a non-normative account).

His reasoning underpinning his reductio of (2) begins with the premise that 'the idea of constitutivity itself is normally taken to involve some awareness of rules as being in force' (Pagin 2008: 8). One isn't playing chess unless one is aware of the rules of chess as applying to one's movement of the pieces (8). Is assertion like this? Pagin suspects not.

When rules of games are put in force, this happens by way of explicit or at least conscious agreement. There is nothing corresponding to this in the case of language use. People speak and people listen and do not consciously agree on letting norms of assertion be in force. (Pagin 2008: 7)

Two considerations that support this suggestion: first, children seem to be making assertions without any recognition that rules are applying to their assertions. Secondly: Pagin claims that he and his wife do not take a normative attitude to the use of language and are nonetheless making assertions (as opposed to, say, 'quassertions') (8).

As Pagin claims, there seems to be no clear empirical test for determining when norms governing assertion are in play, when they are. We thus lack an empirical basis for the application of norms (i.e. that would govern assertion), something we would need, given that norms governing language aren't explictly or consciously agreed upon by language users, if norms were to be constitutive of assertion. Thus: Norms are not involved in constituting an act as as assertion, and in that case we still need a non-normative account of assertion.

Some thoughts on Pagin's Argument

(i) First, I think it's important to be clear about what Pagin is not claiming, in virtue of having claimed that normative accounts of assertion (such as the knowledge account) are indefensible: he is not claiming that when people claim that "knowledge is the norm of assertion" (or, say, justification is the norm of assertion), that they are incorrect. For all Pagin has said, any of the three main rival accounts (justification, knowledge, truth) which puport to answer the question "What is the norm of assertion" could be correct. Importantly, the varying claims about what is the norm of assertion--what rule governs correct assertion--are compatible with a rejection of either the claim that some norm R uniquely characterises assertion or the claim that some norm R is constitutive of assertion.

(ii) It's not clear that one of the key premises of Pagin's argument is correct. Consider again the analogy between chess and language. In chess, an activity where constitutivity holds (i.e. something counts as a bishop in virtue of being governed by the norms that govern correct moves for a bishop), we explicitly endorse and are aware of the rules of chess before playing. (If we are moving pieces of wood whilst oblivious to the rules of chess, we aren't playing chess, but schmess). In language discourse, Pagin thinks we don't explicitly endorse rules that govern language use before playing. We can 'play' the language game (i.e. engage in the practice of asserting) without any prior awareness of or endorsement of any norms. And so, as Pagin suggests, the analogy breaks down--we have a factual difference relevant to the notion of constitutivity that stands between chess (where norms are constitutive of what counts as chess moves) and language (where norms are says Pagin not constitutive of what counts as assertions). A problem I see is that Pagin hasn't done enough to convince that we don't implicitly endorse such norms before engaging what we'd recognise as assertoric discourse. To take an extreme case: if, suppose, someone uttered English sentences arbitrarily, I think our verdict should be that they're not 'playing the game' i.e. they're not making assertions. They're doing something else. It seems then quite plausible to think of language as much more akin to chess than Pagin does: you're making moves in neither of them if you would recognise something as a 'move' in the absense of recognising any rule that governs the move.

(iii) I can imagine one citing facts about disagreement, such as in the contemporary literature, about what rule (i.e. knowledge, justification or truth) governs correct assertion, and trying to argue from this to the Pagin's conclusion: that norms aren't constitutive of the practice of assertion because these norms (obviously, with reference to disagreements about rules of correctness) aren't reflectively assessible to all who make legitimate assertions.

I think such an argument could be blocked; even though we disagree sometimes about how to formulate the rules that govern epistemic (or, as Pagin says, 'semantic') correctness of assertion, it doesn't follow from that that the thought that the very nature of what counts as an assertion must be expressible in terms of such rules is problematic. This is because of a rather simple observation about rule-following: One isn't always necessarily able to articulate the rules one follows. This is true for rules that might be followed arbitrarily as well as for rules that are constitituve of a practice. The observation that we sometimes follow rules without reflecting on how to formulate them funds space for dispelling the tempting thought that--because we disagree about what rule governs correct assertion--we therefore aren't asserting in accord with some rule in virtue of which our assertions count as such.

Is Assertion Relative to a Question? Some Thoughts on Schaffer's "Knowledge in the Image of Assertion"

In "Knowledge in the Image of Assertion," Jonathan Schaffer (2008) defends a view of assertion according to which:

KQ: S ought: assert that p in context c only if s knows the answer (p) to the question under discussion in C. (Schaffer 2008: 10).

He gives several (four in total) interesting cases which he takes to support his question-relative account of assertion (which he takes in turn as counting in favour of his question-relative account of knowledge.) Here I want to examine specifically the plausibility of his question-relative account of assertion and will do so by considering two of these example-cases which he cites in support of the question-relative view of assertion: 'Stop Thief' and 'Millard Fillmore.'

A quick note on my dialectical aim. Although I am not here presupposing any particular (normative) account of assertion to be correct, I want to argue by the strategy of showing that, were I out to defend something like the 'justification account of assertion,' (i.e. Douven 2008, Kvanvig 2010, Lackey 2008) Schaffer's cases would support it (the justification account) no less than they appear to support his question-relative account of assertion. First, here's "Stop Thief", a case originally from Dretske (1970):

Stop Thief: Black has stolen the opals from the locked safe. The detective arrives on the crime scene, find Black's fingerprints all over the safe, and on that basis reports that Black has stolen the opals.

Schaffer's idea is that whether the detective's assertion is proper depends on the question: if the question concerned who stole the opals, then the detective has asserted correctly. But if the question concerned what Black stole then, according to Schaffer, the detective cannot assert appropriately that Black has stolen the opals.

A second case: 'Millard Filmore':

Consider the following easy question:

Who was the thirteenth president of the United States?

A. Millard Filmore

B. Hillary Clinton

Says Schaffer:

"Your average student, faced with this question, can appropriately assert that Millard Fillmore was the thirteenth president of the United States. But consider the following hard question:

Who was the thirteenth president of the United States?

A. Millard Fillmore

B. Zachary Taylor

According to Schaffer: "Your average student, faced with the hard question, cannot appropriately assert that Millard Filmore was the thirteenth president of the United States," (7) and this is so, Schaffer thinks, even though the student can appropriately assert "Millard Filmore is the thirteenth president of the United States" when it is the answer to the easy question.

I think that the apparent force of these examples in favour of a question-relative account of assertion can be explained away once we consider a more fundamental difference in the positions of the asserters in Schaffer's cases: a difference that can be accounted for independently of any appeal to what question is asked, but which an appeal to the differences in questions asked must be explained in terms of.

Put simply: it is a difference in epistemic position in each of the two cases which explains why the assertions were proper, when they were, and not proper, when they were not. To support this notion, let's now employ the dialectical strategy I alluded to. Suppose I'm trying to support the 'justification account of assertion' according to which:

JA: One ought: Assert p only if one is appropriately (epistemically) justified in one's belief that p.

In the thief case, consider that, in the instance in which the question was "Who stole the opals", one's evidence for it being Black who stole the opals is:

(i) Black's fingerprints are on the safe

(ii) Someone stole opals

Together, (i) and (ii) count as rather strong justification for the claim that Black stole the opals. When the question however is what it is black stole, then the evidence set eliminates (ii), and so the epistemic support is weaker:

(i) Black's fingerprints are on the safe

(ii) Something was stolen

Together, (i) and (ii) are not enough to justify one's assertion that "It was the opals that Black stole."These considerations would support a justification account of assertion--one which itself would need not make any further appeal to facts about what questions are asked in order to account for the impropriety of the assertion at issue. Whereas, to emphasise, the question relative account only appears to be supported by the examples Schaffer uses once we consider that one's epistemic justification is better in one case than in the other.

The same move can be analogously made to explain why it is that the Millard Filmore case would appear to support a question-relative account of assertion--even though it ultimately does nothing more than to support the idea that assertions with epistemic justification are warranted whereas ones which lack epistemic justification are not.