Transmission of warrant-failure and the notion of epistemic analyticity

In this paper I will argue that Boghossian's explanation of how we can acquire a priori knowledge of logical principles through implicit definitions commits a transmission of warrant-failure. To this end, I will briefly outline Boghossian's account, followed by an explanation of what a transmission of warrant-failure consists in. I will also show that this charge is independent of the worry of rule-circularity which has been raised concerning the justification of logical principles and of which Boghossian is fully aware. My argument comes in two steps: firstly, I will argue for the insufficiency of Boghossian's template which is meant to explain how a subject can acquire a warrant for logical principles. I will show however that this insufficiency of his template can be remedied by adopting what I call the Disquotational Step. Secondly, I will argue that incorporating this further step makes his template subject to a transmission of warrant-failure, assuming that certain rather basic and individually motivated principles hold. Thus, Boghossian's account faces a dilemma: either he adopts the Disquotational Step and subjects his account to the charge of a transmission of warrant-failure, or he drops this additional step leaving the account confronted with explaining the gap that has previously been highlighted. I will then suggest various rejoinders that Boghossian might adopt but none of which—I will argue—can resolve the dilemma. Lastly, I will raise and briefly discuss the question whether this worry generalizes to other accounts, such as Hale and Wright's that aim to explain our knowledge of logic and/or mathematics in virtue of implicit definitions.


Introduction
In this paper I will argue that Boghossian's explanation of how we can acquire a priori knowledge of logical principles through implicit definitions commits a transmission of warrant-failure. In the first section I will briefly outline Boghossian's account 1 followed by an explanation of what a transmission of warrant-failure consists in. I will also show that this charge is independent of the worry of rule-circularity which has been raised concerning the justification of logical principles and of which Boghossian is fully aware. 2 Thus, it constitutes a distinct worry for Boghossian's account. 1 In the third section I will argue for the insufficiency of Boghossian's template which is meant to explain how we acquire a warrant for logical principles. I will show however that the insufficiency of his template can be remedied by adopting what I call the Disquotational Step. In the fourth section I will then argue that incorporating this necessary further step makes his template subject to a transmission of warrant-failure, assuming that certain rather basic and individually motivated principles hold. Thus, Boghossian's account faces a dilemma: either he adopts the Disquotational Step and subjects his account to the charge of a transmission of warrantfailure, or he drops this additional step leaving the account confronted with explaining the gap that has previously been highlighted. The fifth section discusses various rejoinders that Boghossian might adopt, but I will argue that none of these can resolve the dilemma. Lastly, I will raise and briefly discuss the question whether this worry generalizes to other accounts, such as Hale and Wright's [2000: 286 -319], that aim to explain our knowledge of logic and/or mathematics in virtue of implicit definitions. 3

I. Boghossian's Template for A Priori Knowledge of Logical Principles
Boghossian's explanation of our a priori knowledge of logical principles has two steps. The first is based on assuming what might be called inferentialism, the thesis that genuine meaning-constituting rules provide an entitlement to reason according to such rules. Although inferentialism might give us an entitlement to use a certain rule, Boghossian notes that this by itself does not yet explain our belief that certain rules are valid. On his view the subject does not need to know that a certain rule is meaning-constituting in order to be entitled to infer according to it. Therefore a second step is required in order to explain our knowledge of certain presumably a priori principles. Namely, we need to explain how we can be warranted to believe that those principles are valid or true, which is explained in virtue of Boghossian's template. In the following, I will leave aside the issue of whether inferentialism is acceptable and sufficient to provide entitlements for suitably restricted rules and focus on the second step. 4 In his earlier paper 'Analyticity' he suggests that we can explain how to gain a 'warrant for our belief in the elementary truths of logic ' [1997: 358]' by the following (slightly adjusted) template: This template is a meta-language construction. The first premise above is what Boghossian considers to be an implicit definition of C, while C should be regarded as a name for a term, such as 'and' and A as a name for the meaning-constituting inferences for 'and', such as its Introduction and Elimination rules (I-E rules). The second premise is the antecedent of the conditional of the first premise. It is phrased slightly confusingly and I suggest that it be understood such that 'C means what it does' is roughly equivalent to 'C refers as intended by the implicit definition' or 'C refers as intended' for short. 5 I will discuss this ambiguity in due course.
Various problems have been highlighted concerning Boghossian's account. 6 The following criticism is independent of points previously raised in the literature. Before outlining my criticism let me quickly explain the basic notions I will use in my criticism such as 'warrants', 'transmission of warrants', and 'transmission of warrant-failure' and the contrasting idea of a 'rule-circular justification'.

II. Warrants, Transmission of Warrant(-failure),
and Rule-circular Justification I will be rather brief in my explication of the notion of a 'warrant' and only later introduce further complication in the context of possible rejoinders for Boghossian. For the moment, it will be enough to regard 'warrant'-a notion hardly explained in Boghossian's article-as a placeholder for that which turns a true belief into knowledge. The notion of transmission of warrant also needs clarification and can be characterized by contrasting it to the notion of closure of warrant. 7 The latter notion merely claims that if there are warrants for the premises of a valid argument then there is a warrant for the conclusion too. The notion of a transmission of warrant is stronger in that it provides a stricter connection between the warrants for the premises and the resulting warrant for the conclusion. It imposes requirements on the thinker acquiring a warrant for the conclusion in virtue of the premises being warranted. Hence the idea is that a warrant is transmitted if a thinker acquires a warrant for the premises of a valid argument, knows that it is valid and thereby-i.e., in virtue of both components-acquires a warrant for the truth of the conclusion. We also need to clarify the notion of a transmission of warrant-failure. There are different ways in which such failure might occur. One type, which will be important to my argument, can be characterized as a form of begging the question. 8 The basic idea is that in order to hold true one of the premises 5 Also note that I will neglect the minor difference between premise (1) and (2) in that (1) says 'is to mean what it does' and (2) reads 'means what it does'. I will assume that both occurrences are intended to be used interchangeably to warrant the modus ponens inference. 6 Criticism concerning implicit definitions and I-E rules can be found in Prior [1960: 38 -9]. Other problems for Boghossian's account can be found in Enoch and Schechter [forthcoming]. Horwich [1997: 423 -40], Margolis and Laurence [2001: 293 -302], Wright [2001;41 -85], and Williamson [2003: 249 -93]. 7 This distinction was previously drawn in Wright [2002: 330 -48]. I follow closely this discussion. The notion of closure of warrant goes back to Dretske [1970Dretske [ : 1007]. of the argument, i.e., to have a warrant for the truth of the premises, already presupposes having acquired a warrant for the truth of the conclusion. This results in a transmission of warrant-failure, since we can't transmit a warrant for the truth of the conclusion from the premises, because such warrant already has to be acquired in order to ascertain a warrant for the premises.
It is important to note that this charge is independent of the problem already highlighted by Quine [1976a: 77 -106;1976b: 107 -32] and others that to justify logical inference results in a justificatory circle-another form of begging the question-as we already need to presuppose the logical inference in question in order to appreciate the validity of the argument which aims to provide a justification for its validity. Boghossian is aware of the issue of rule-circular justification (RCJ), but he aims to provide an account in which we can regard this rule-circularity as providing a genuine justification. Let me briefly clarify the notion of a rule-circular justification, highlight some differences and then move on to my argument against Boghossian's account.
A canonical statement of the rule-circularity worry is that an argument is unpersuasive if the conclusion of an argument establishes the validity of an inference rule, whose validity is already presupposed in reasoning from the premises to this very conclusion. To wit, both worries (TOWF and RCJ) have to do with the persuasiveness or conclusiveness of an argument, but I think there are various considerations that show that the worry of RCJ is distinct from the charge of a transmission of warrant-failure.
RCJ applies to the justification of inferences, where an inference is used in the argument and occurs in the conclusion. Reconsider the above example with the connective 'and' and its meaning-constituting rules in the template. The conclusion is about the validity of the 'and' rules and the reasoning merely uses modus ponens, so there is no charge of a rule-circular justification possible. The TOWF-worry is in this respect more general as, if it is correct, it will show that the template is wrong for any introduced connective. Furthermore, the charge also applies to statements that might be justified in virtue of the above reasoning which are not inferences. Consider, e.g., an extension of Boghossian's account to mathematical principles such as Hume's Principle (HP), which, if I am right, fails due to a transmission of warrant-failure. 9 There are further differences, but I won't dwell on them here and instead present the main thrust of this paper. In the next section I will further investigate Boghossian's template, as I think it suffers from serious insufficiencies. Important amendments have to be made in order to regard it as being able to at least prima facie provide a 'warrant for our belief in the elementary truths of logic' [Boghossian 1997: 358]. But having these amendments in place will be to commit a transmission failure, as I will show in Section IV.

III.
Step 1: The Insufficiency of Boghossian's Template The issue here surrounds the elusive and inherently ambiguous notion of 'is to mean what it does' of the first premise or 'means what it does' of the second premise, which presumably are synonymous. Still, there are various ways to interpret the phrase. One option is to deflate this notion and accept that even an empty term 'means what it does', namely we can give some descriptive content to it. So take for example the term 'extension' as introduced by the inconsistent axiom Basic Law V. 10 Despite the axiom's inconsistency one can contend that on the deflated version of the second premise the term 'extension' means what it does, in that we can explain the notion with appeal to co-extensiveness. Thus, on this reading, the second premise of Boghossian's template is close to trivial, while putting most of the weight on the first premise, which becomes rather substantial: Why should the mere having of descriptive content be sufficient to figure in valid patterns of inference? Just consider Basic Law V again.
Alternatively, we can-as I will do for the rest of this paper-interpret the phrase as equivalent to 'refers as intended'. This renders the first premise close to trivially true, as what it claims is that, if a term has the intended reference in order to make a pattern of meaning-constituting rules valid, then this pattern will be valid. On a material conditional interpretation this above conditional is a necessary truth. But, the second premise becomes rather substantial as it claims that a term refers as intended. However, on what grounds can we rule out that it is not an empty term? It is this issue that will become prominent in my later discussion, and the main point will be that in order to appreciate or to have a warrant that the term involved refers as intended, we already have to appreciate or have to have a warrant for believing that the conclusioni.e., the meaning-constituting patterns-are valid. The core of my charge consists of two key thoughts that highlight the shortcomings of Boghossian's account. Firstly, the template as outlined by Boghossian has to guarantee knowledge of the validity of a pattern of inference as an object-language statement and not merely as a metatheoretical statementa claim that I will substantiate in this section. Secondly, to grasp the (object-language) content of the second premise-that is the intended reference of the term-an understanding of the term's role in whole sentence such as the conclusion is necessary. Thus, in order to be in a position to know that the term has its intended reference, we need to know that the conclusion has the meaning to make the pattern of meaningconstituting inferences valid. This will eventually lead to a transmission of warrant-failure of Boghossian's template-the details of this latter claim will be expounded in the section labelled 'step 2'. 10 Basic Law V is Frege's famous principle which introduces the term 'extension' (or rather the identityconditions of this term) in virtue of co-extensiveness. Its fame is mainly due to its inconsistency which was first shown by Russell (and independently by Zermelo).
To develop my worries culminating in the claim that Boghossian's account needs further amendments let us look closer at the conclusion of the template, as cited it says: Boghossian's template 1. If C is to mean what it does, then A has to be valid.

A is valid.
But, as already mentioned C and A are names for connectives and meaningconstituting rules, and thus the conclusion is merely a meta-linguistic statement. This however raises an immediate worry, namely how does the meta-linguistic statement concerning the validity of the rule guarantee that the subject has grasped the content of the rule in question? That is, granting that the template provides a warrant for the validity of A for the subject, it is not secured that the subject has also understood the rule he has a warrant for. To wit, the meaning-constituting rules for 'and', or other more complex connectives, might be written in Chinese and, as it happens, my Chinese friend tells me that premise one and two are true without telling me which connective he is talking about. Surely, I would be able to assent to the conclusion and agree such and such signs represent valid rules of inference, without grasping the content of the rule in question. That is, it seems questionable-to say the least-that the template as stated is sufficient to provide the right type of warranted belief in the rules/principle. Therefore, I believe Boghossian's template needs to be adjusted in order to account for such a case (the Chinese Friend example). This is to add a further step which I will label the Disquotational Step (DS). 11 The idea, roughly, is that we can account for the kind of identifying knowledge that is needed in explaining our knowledge of logic or the rules in question if we are warranted to disquote on what Boghossian took to be the conclusion that is (3) above. Thus, I will adjust the above template by adding a further step from the initial conclusion (3) to the new conclusion (4)-i.e. the object-language statement of (3). This further step is what I will refer to as the Disquotational Step. For purposes of presentation and clarity I will use 'and' for A and 'P and Q ) P' for C, while also omitting the second elimination rule for simplicity.
Boghossian's new template 1. If 'and' is to mean what it does, then 'P and Q ) P' has to be valid. 11 This step is not strictly speaking an instance of disquotation. Disquotation normally concerns the truth of a statement and not the validity of an inference rule as in the above template. One option to bring my use of (DS) closer to the standard usage is to reinterpret Boghossian's template in terms of statements of a rule and consequently to use truth instead of validity and so make the additional step a genuine instance of disquotation. However, I chose to remain as close as possible to Boghossian's initial characterization of the template and merely labelled the additional step, which I argued is necessary, a 'Disquotation Step' for lack of a better name. See also footnote 12 for further discussion.
2. 'and' means what it does.
3. 'P and Q ) P' is valid.

P and Q ) P. 12
Adding the last conclusion forecloses the above worry of knowledge by testimony, but it raises the question how we can justify the move from (3) to (4). As we have seen above, the premises as so far understood, are insufficient to provide a guarantee for the Disquotational Step (remember the Chinese Friend counterexample). Thus we have to examine how we can warrant this move and I will argue that, based on other assumptions which I will outline below, this will lead to a transmission of warrant-failure for the new and adjusted template.

IV.
Step 2: Transmission of Warrant-failure of the New Template So, if the stronger conclusion (4) is needed to establish a warranted belief in the truth of the principle in question, how can we justifiably move to this conclusion? That is, what is it that warrants the application of the Disquotation Step-moving from premise (3) to the conclusion-and thus disquote? The worry I will raise in the following is an instance of the transmission of warrant-failure charge. Namely, that under a suitable reading of (2) and (3) we need, in order to justify the Disquotational Step, a prerequisite warrant for (4). This claim is based on four assumptions, which individually are either implicitly held by Boghossian or can be considered as rather minimal claims. In the following I will outline those assumptions and then clarify my worry.

A1: Being warranted to disquote a sentence requires an understanding of that sentence
The idea is that in order to be warranted in making the Disquotational Step it is necessary to understand the sentence being disquoted-this much seems to me a platitude. And in order to have such understanding one has to possess, or rather grasp, the concept that the principle is about, i.e., in our case 'and', hence one has to understand or grasp the content of the second premise. This assumption is introduced so to avoid knowledge by testimony and thereby to 12 As noted above, in order to turn the step from (3) to (4) into a genuine disquotation move we could further adjust the template such that we are not talking about the validity of I-E rules (')' is meant to represent an inference), rather the truth of the statement of the rule. Accordingly (3) would then be represented as 'P and Q j ¼ P' is true. I contend however that nothing hinges on this variation since my argument concerns a structural difficulty with Boghossian's template: what is merely needed is a way to represent the step from being warranted in believing that conjunction elimination is valid (or that the statement of conjunction elimination is true) to being warranted in believing conjunction elimination and (DS) is meant to represent just this move. If, as I argued, this move is necessary for a subject being warranted believing (and grasping the content of) such inference, the following section will outline a transmission of warrant-failure for Boghossian's position. If the step is rejected then I find it hard to see how the template could suffice to explain a subject's warranted belief and thereby grasp of the truths of logic. account for the justification to use the Disquotation Step. The following assumption is rather distinct and more general.

A2: The Context Principle
Basically, the thought encapsulated in the Context Principle (at least in the following rather deflated version) is that the basic semantic unit for our understanding are sentences and, hence, in order to appreciate what reference a term has, one has to understand the role this term plays in various sentences. This version is close to Evan's Generality constraint and we can assume that prima facie it is accepted by Boghossian.

A3: Understanding involves knowledge of meaning
Just as (A1) and (A2), I also consider this claim to be rather benign and generally acceptable. Thus, in the following I won't assume a specific model of what understanding involves (knowledge of truth-conditions perhaps), rather the minimal conditional that-in general-to understand a sentence is to have knowledge of the meaning of the sentence.

A4: Epistemic notion of analyticity
Lastly, (A4) is Boghossian's very own conception of epistemic analyticity. More explicitly he claims that a statement S is epistemically analytic for a person T, if 'the mere grasp of S's meaning by T sufficed for T's being justified in holding S true' [Boghossian 1997: 334]. Thus Boghossian's conception of analyticity, slightly rephrased, involves that knowing the meaning of a statement S, implies a warrant for the truth of S. Although, the onus is not on me to motivate such a conception, there surely are rather intuitive examples that give rise to such a view. Take for example the sentence 'Everything red is coloured'. Here it is prima facie appealing to think that a grasp of the truth conditions of this statement by a person T suffices for T being justified in holding the sentence to be true. 13 With these assumptions in place let me outline my charge of a transmission of warrant-failure in the new template above concerning the added fourth premise. Hence, let us ask what it is to have a warrant for the Disquotational Step. The following argument will show that a necessary condition for such a warrant is that one has to have a prerequisite warrant for the conclusion.
The argument goes as follows: Based on (A1), to have a warrant for making the Disquotational Step one has to have an understanding of the second premise. The idea here is that in order to be justified to disquote one has to be in possession of the concept involved-in our case 'and'-which is just the content of the second premise. However, in order to understand the second premise, that is, to understand the intended referent of A (which is the content of the second premise) we have to understand the role this term plays in appropriate sentences, due to an application of the Context Principle (A2), i.e., we have to have a previous grasp of statements like the conclusion in Boghossian's argument. But, in order to have an understanding of the conclusion it involves, by (A3), knowledge of the meaning of this statement. However, and this is the crux of the argument, the logical principles in question are, for Boghossian, to be regarded as epistemically analytic. Hence, we can assume (A4)-Boghossian's conception of epistemic analyticity-according to which, one can have knowledge of the meaning of a sentence only if one has a warrant for the truth of such sentences. Thus, assuming the transitivity of the conditional, one can only have a warrant for the Disquotational Step in Boghossian's new template if one has a prior warrant for the truth of the conclusion. 14 Therefore, we can confront a proponent of Boghossian's account with the following dilemma. Either his template is inherently insufficient in accounting for our knowledge of certain principles on the basis of being unable to guarantee disquoting the initial conclusion (3). Or he facespresupposing the tenability of the above assumption-a transmission of warrant-failure. This charge is based on his template and his additional thesis of epistemic analyticity, which (so I argued) are incompatible. To put the basic charge rather crudely, we can say that the incompatibility arises because his template requires that understanding a sentence/rule is distinct from knowing, or being warranted in believing, that the sentence/rule is true/valid. Otherwise the template can't transmit a warrant for the conclusion. However, his own conception of epistemic analyticity is such that a mere understanding of the sentence/rule suffices to know, or be warranted to regard, that the sentence/rule to be true/valid. This is the basic tension in Boghossian's account that this argument is aimed to expound.

V. Various Rejoinders to My Criticism
In this section I will discuss three replies that could be made on behalf of Boghossian to my criticism. The first discusses the application of the Context Principle. The second takes up a distinction between internal and external warrants and develops a reply along externalist lines that Boghossian seems likely to adopt. Lastly, I will draw a distinction within the notion of acquiring a warrant on the basis of which a further response can be developed. I will argue that none of these lines of response undermines my argument.
14 For the more formally minded, the argument can be represented using the following abbreviations: DS stands for Disquotational Step; K(a) for knowledge of a; U(a) for understanding a; M(a) for meaning of a; W(a) for warrant/being justified for a and T(a) for truth of a.

A. The Context Principle Revisited
A possible reply against my argument is that the application of the Context Principle does not guarantee that in order to understand the second premise one has to have an understanding of the conclusion. Rather, granting that the Context Principle correctly applies here, it only demands that one has to understand some sentences in which the newly introduced term occurs. This does not necessarily imply that it has to be the conclusion of the template.
Although I think this is a correct observation I don't believe it has much force. The conclusion of the argument is a disquoted version of what is supposed to be the meaning-constituting rules for the new term in question. Hence, to grasp the reference of the term in question, which implies (at least) knowing what role the term plays in whole sentences (Context Principle), one must grasp what role the term plays in its own meaning-constituting rules that are meant to fix the reference of the term. Hence, although I can discern a weakness in the general structure of the argument, the fact that the conclusion is concerned with the meaning-constituting inferences of the term in question should imply that grasping the reference of the term as such, already involves a grasp of the conclusion.
But to reinforce my argument note the following two points: Firstly, if the role of the Context Principle is accepted and one has to appeal to some sentences to grasp the reference of the new term, while excluding the conclusion to avoid my argument, then this implies the following insufficiency: the template-by itself-is not sufficient to account for a subject's knowledge of the logical principle in question, rather the subject has to appeal to other sentences not mentioned in the template in order to understand and thus acquire knowledge of that principle.
Secondly, to focus the discussion we might imagine a scenario in which a character, who barely has a grasp of language and who has no grasp of any logical connectives, is meant to be introduced to these new connectives in virtue of the template. This is a case in which Boghossian's template should be able to generate a warrant for the logical principle for the character in question. The crucial point in this scenario however, is that the character only knows of the meaning-constituting rules for the connectives (on the basis of premise (1)) and thus, if the Context Principle is in place, has to appeal to the conclusion of the argument as, per impossibile, there are no other occurrences of the term he can refer to. Hence, I think that exploiting the structural weakness of my argument won't establish a serious line of reply to my charge of a transmission of warrant-failure.

B. Internal vs. External Warrants
In his various papers on rule-circular justification of logical inference, Boghossian can be interpreted as adopting an externalist position in order to evade the charge of rule-circularity and even to show that such circularity can be used as a kind of justification for our knowledge of logical principles. Boghossian's point-in a nutshell-is that we can still acquire a warrant for an inference despite its circularity because it is not required that we have to appreciate an antecedent warrant for the rule in question. That is, in an externalist fashion, we don't need to first reflectively acknowledge the truthpreserving nature of the inference to be warranted-on the basis of this inference-to be justified in accepting this very same inference.
A specific notion of warrant is assumed for such externalism. Let us therefore distinguish two different statuses a warrant might have. A warrant is internal if the warrant that underlies a certain belief (that a certain inference is valid for example) can be claimed to be had by the subject in virtue of a priori reasoning, self-knowledge, or reflection. A warrant is external if the warrant that underlies a certain belief is not so accessible and thus cannot be claimed through a priori reasoning, self-knowledge, or reflection. Having this distinction in place, maybe one line of response for Boghossian is to argue that, as in the case of rule-circular justification, merely an external warrant is required for the conclusion and not an internal one.
There are large issues lurking in the background of which three questions can be separated for the discussion at hand. Firstly, the very general issue, which I won't discuss here, as to whether externalism in the context of logic and mathematics can be regarded as providing the appropriate type of warrant. Secondly, and more importantly for the issue at hand, is whether an appeal to externalism is suitable to evade my criticism. Is Boghossian in a position to make use of external warrants? That is, is his position that aims to explain the warrant for our beliefs in logic in virtue of the template compatible with merely providing external warrants? And lastly, if it is compatible, how exactly would an appeal to external warrants circumvent my charge of a transmission of warrant-failure, considering that my argument did not appeal to a specific type of warrant.
Let us first consider whether Boghossian can appeal to external warrants. I think there is an awkwardness in interpreting him as invoking only these kinds of warrants in the context of logic. This is because he provides a template to explain how a subject could, on the basis of reasoning through the template, acquire a warrant for our logical principles. Surely, a warrant acquired in this way would be internal in the sense that it can be claimed by a priori reasoning or reflection itself, otherwise it is unclear what the point of his template is. Furthermore, consider that Boghossian appeals to a notion of epistemic analyticity where grasping the meaning of terms suffices to be warranted in holding true certain statements. It seems rather intuitive, also in this case, that the type of warrant can't just be purely external-that is not claimable by reflection or self-knowledge. To use external warrants here would seem to imply that grasping the meaning of the term or understanding a certain expression is spelled out externally, which (at least intuitively) goes against the initial motivation for the notion of epistemic analyticity. Hence, I have strong reservations in interpreting Boghossian's position in a strict externalist fashion.
Nevertheless, even assuming he is merely in the business of providing external warrants there seems no reasonable way out of the transmission of warrant-failure charge unless Boghossian gives up on his template. To wit, a strict externalist should not be concerned with providing a template in the first place, and hence would not accept that there is a transmission of warrant failure-just because the actual source of the warrant for the conclusion is some kind of reliable process and not reasoning based on the template. Hence, for a strict externalist there is no transmission of warrantfailure, just because there is no warrant to be transmitted from the premises to the conclusion. And unless Boghossian wants to give up on the template transmitting a warrant, appeal to external warrants seems no way out. This point has a further interesting consequence for Boghossian's position. Assuming that the template does some work in explaining our knowledge of logic, it seems as if the template is aimed to provide warrants which are internal and thus might be in conflict with his more recent thoughts on the matter.

C. Two Notions of 'Acquiring a Warrant'
Here I want to suggest a last rejoinder by drawing a further distinction within our notion of warrant. This distinction is not concerned with internal and external warrants, rather it has to do with the relation a warrant (internal or external) can have to a subject. Thus, we might distinguish two different statuses for a warrant to count as 'being acquired'.
On the one hand a thinker might be in possession of a specific warrant for a belief. That is-metaphorically speaking-the subject has the warrant within his epistemic field. In contrast a thinker might have a specific warrant merely available without actually possessing the warrant. That is, to use the above metaphor, the subject has the warrant close to his epistemic field ready to be possessed without actually possessing it. Although this distinction might not be as clear-cut as one would like, the intuitive idea is clear and will prove to be useful for this discussion. 15 There might be various ways in which Boghossian could apply this distinction. Here, I will discuss the most obvious option. He might argue that the notion of epistemic analyticity should be revised such that instead of maintaining that a statement S is epistemically analytic for a person T, if 'the mere grasp of S's meaning by T sufficed for T's being justified in holding S true', we should say that 'the mere grasp of S's meaning by T sufficed for T's having available a warrant in holding S true' without necessarily possessing it. This alteration might be used to stop the charge of a transmission of warrant-failure in the following way: In order to make the last step in the above argument I need to assume that what the notion of epistemic analyticity provides is the possession of a warrant, which would then lead to a transmission of (possessed) warrant-failure, as to possess a warrant for the second premise involves already possessing a warrant for the conclusion. But, crucially-so the line of response might go-having acknowledged the above distinction, the notion of epistemic analyticity involved is weaker in that it only makes available a warrant for the conclusion. However, this does not suffice for a genuine transmission of warrant-failure per se, as what we end up with is that the possession of a warrant for the premises will involve the availability of a warrant for the conclusion. But this result-by itselfcan't be sufficient to reject the template, because for every valid argument it will be the case that somehow the possession of a warrant for the premises will involve the availability of the conclusion-otherwise how could there be any transmission of warrant in the first place? I want to consider two lines of response to this thought. First, I will examine whether weakening the notion of epistemic analyticity is intelligible. Second, I will argue that even if it is the case that the possession of a warrant for the premises merely involves the availability of a warrant for the conclusion, there still remains a problem for Boghossian in the special case of his template.
I think that the suggested revision of the notion of epistemic analyticity will weaken this notion to such an extent that the conception of epistemic analyticity hardly does any work in an epistemology of the a priori. To wit, the main motivation for the notion of epistemic analyticity is that a grasp of meaning plays a substantial role in being justified in holding certain sentences to be true. Now, on the weakened version of epistemic analyticity, understanding can't be appealed to in claiming to be in possession of a warrant for holding a statement true. Understanding merely makes a warrant available to a subject, which however, is not sufficient as a genuine justification for holding true certain statements. Thus, on this weakened version of epistemic analyticity a further story, over and above the template and the idea of epistemic analyticity, has to be told in order to explain why a subject (seems to) possess a warrant for a logical principle. Hence, making this revision within the notion of epistemic analyticity will lead to an insufficiency in the account, as put forward by Boghossian, which has to be remedied by appealing to other theses that in turn might threaten to undermine the initial position.
Concerning the second issue, even if we grant that the distinction between possession and availability is intelligible and a revised version of epistemic analyticity is compatible with Boghossian's position, there still remain problems that are reminiscent of a transmission of warrant-failure. My reply will partly rely on appealing to the intuition that in order to have adequately accounted for a subject's knowledge of logical principle in virtue of this template, the subject should be able to reflect on its warrants for the premises and its conclusion and thus be in a position to claim a warrant for them. According to the above rejoinder on behalf of Boghossian, to possess the warrant for making the Disquotational Step involves the mere availability of a warrant-due to the weakened version of epistemic analyticity-for the conclusion. However, from a reflective stance to claim to possess a warrant for this premise also involves a claim to have available a warrant for the conclusion. Thus, we need to ask on what grounds can I make the claim to have available a warrant for the conclusion?
And here is the impasse-in the form of a dilemma-that Boghossian faces. He might show how we can claim to have available such warrant by How, just by stipulating that a certain sentence '#f ' is true where '#' is already understood, and 'f ' is a hitherto contentless expression . . . is it supposed to be possible to arrive at an a priori justified belief that #f ? . . .Well, the route seems relatively clear provided that two points are granted: first that a stipulation of the truth of the particular '#f ' is so much as properly possible [1.] . . . and second that the stipulation somehow determines a meaning for 'f '.[2.] . . . Moreover if the stipulation has the effect that 'f ' and hence '#f ' are fully understood . . . then nothing will stand in the way of an intelligent disquotation [3.]: the knowledge that '#f ' is true will extend to knowledge that #f. In other words: to know both that a meaning is indeed determined by an implicit definition and what meaning it is, ought to suffice for a priori knowledge of the proposition thereby expressed. [Hale and Wright 2000: 295 -6] premise, as it is this claim which justifies disquotation. Thus, importantly holding this premise to be true rules out the kind of Chinese Friend counterexample, where I have knowledge from testimony that the stipulation is true and fixes a meaning for 'f ', without grasping the content of the stipulation. Still, we should raise the more general worry, of how, in detail, the application of the Disquotational Step is justified in virtue of the antecedent of the above conditional. Let us grant Hale and Wright that the stipulation is successful and so true, and assume that it is 'f ''s determined meaning that provides the link to understanding, 16 thus we might focus on the following simplified conditional: 3**. If 'f ' has a determined meaning [refers as intended], then 'f ' and '#f ' are fully understood.
However, (3**) is reminiscent of the problem raised concerning Boghossian's template. Namely, how can a subject be warranted to know what 'f ' refers to without grasping its role in the relevant sentences in the object-language? Hence, the charge against Hale and Wright would be that in order to be warranted to grasp the intended meaning of 'f ' and thereby come to understand 'f ' and '#f ' we already have to have a warrant for #f, assuming the same notion of epistemic analyticity is adopted by them. Hence, although this argument will need further detailed discussion (which has to be postponed to another occasion), as it leaves Hale and Wright much space to manoeuvre, the general idea behind my argument seems prima facie applicable to the approach of Hale and Wright. Thus, to (cautiously) generalize, the charge of a transmission of warrant-failure for templates invoked to found our knowledge of mathematics and logic on implicit definitions by appealing to a notion of epistemic analyticity seems to pose a general threat to the viability of these conceptions. 17

University of St Andrews
Received: June 2004