How Important is it to be Conservative?
A Critical Tour through Robert Brandom's Making it Explicit1
Abstract. Robert Brandom proposes to equate meaning with inferential role. I argue that on one way of understanding this the claim can be made good. But it is doomed to a conservatism in that it can only allow those inferences to confer contents on expressions whose conclusions merely explicate what is implicitly given in the premises (deductive ones). On another way of understanding it, though, content is to be read off from our ways of altering our attitudes, and thus largely a concept of pragmatics. It is also argued that Brandom's pragmatics, which is based on a duality of commitments and entitlements, should be one of commitments only.
On a common view, semantics is taken to be the study of the content of linguistic expressions, performances, and intentional states. Pragmatics, by contrast, has the job of studying the force or significance of all the things that can have those contents. To account for the 'practical significance' of intentionally contentful states, attitudes, and performances is of particular importance to Robert Brandom's project of 'Making it explicit'. His program is to detail a story of this practical significance of an intentional state or performance in terms of implicit norms that govern the use of linguistic expressions.
The connection of the pragmatic project with the semantic one is that it is held to require an inquiry into the semantic content of those states, attitudes, and performances. Attitudes and performances can only have a practical significance if they have a semantic and indeed propositional content that can determine this significance in context. What we are told about this connection, for example, is this: "[T]he theoretical job of the contents (...) is precisely to determine, in context, the particular significance of being in or attributing the states those contents are associated with."(68, my emphasis; see also 359) I think we can conclude from this that if we look at particular language uses, performances or attitudinal states, there is a sense in which their semantic contents are meant to be fixed (in whatever way they are fixed) already. Also we find that pragmatics "seeks to determine in a systematic way the pragmatic significance of [a] contentful performance"(133), which would seem to assume that the contents are already there for pragmatics to spell out what they consist in.
Brandom's answer to the question what having a content consists in appeals to its inferential role. By looking at the role an expression plays in inferences, we see what its content is. Brandom's project of 'making it explicit' is one of rendering in an evaluable and propositional form the 'inferential commitments' which are implicit in language use. The goal is to put content as conferred by a net of inferential relations open to critical view. Concepts are to be notated in a way that they "wear their contents on their sleeves"(109).
Suppose for example we wish to know what the meaning of the logical connective "&² is. Then one way of answering this question is to tell what rules of inference govern its use. According to Gentzen, these are rules of two kinds, one rule which tells when a sentence containing this connective can be asserted, and another rule telling when the connective can be eliminated from a sentence containing it. Take this as an example that the meaning of "&², called "conjunction², consists in its inferential role: conjunction is whatever plays the inferential role specified by this pair of rules.2
The program of making explicit what is implicit as such is a conservative one. This is the same sense of the word in which Hegel's philosophy might be called conservative (on which more below). While there is nothing thereby wrong with either Brandom's or Hegel's approach, we shall now see this conservatism has a dimension that is seriously problematic.
For once we stipulate in the above way that meaning is determined relative to a specific set of rules (not by virtue of properties of reference, for example), the question of the stability of meaning arises. The same expressions may belong to different fragments of a language when does it preserve the meaning it has in one of them when we go over to the other? Take two systems of expressions S and S' such that SÃS' and such that both contain the logical sign "&². Is an expression containing "&² that plays the inferential role of conjunction in S necessarily also something that plays the same role in S'? It is, if there are no expressions containing "&² which stand in inferential relations in one of the systems in which they do not stand in the other. If this is the case for all expressions in S, we say S' is a conservative extension of S. This means that when going from S to S' inferences that you used to draw in S remain valid, and no inferences involving only vocabulary of S which were not valid in S become valid in S'.
Now if you want to provide a meaning theory for the items in S by depicting their inferential roles in some such way, you will make explicit the inferential commitments implicit in uses of the expressions of S. Whatever is not implicit in the contents of the items in S cannot be of your concern. A necessary condition for this enterprise to succeed is thus that you consider the system S either only statically (you ignore extensions), or, if you consider it dynamically, you will only consider conservative extensions of it. To be conservative is in this sense all important for you if you want to be sure that in going to the extension S' of S you do not when using vocabulary from S incur any commitments different than those you have already implicitly undertaken in S.
But of this you can only be sure if the inferences you draw in S are of a deductive kind. A recurring theme in this paper is that this consequence is disastrous for Brandom's purposes. For the inferential roles which his theory of content will come up with will be restricted to those inferential properties of expressions which are deductive. Deductive inferences are explicative in that they merely explicate what is given in the premises of an inference already. Inductive inferences, by contrast, are defined by the fact that they do not merely explicate a content already given, by go beyond it. By adopting the conclusion of an inductive inference you incur a commitment which is not a commitment implicitly given in the premises. Precisely for this reason different agents may come to very different conclusions as to which inductive inferences should be drawn. I might want to infer from the fact that Gustav is a dog that he is four-legged. You might be more cautious. (We shall later see that conservatism in the case of inductive inferences is in fact irrelevant as a requirement on extensions.)
Thus you cannot make explicit the inferential roles of the contents of premises of inductive inferences. The contents of the conclusions are not implicit in those of the premises. Hence there is no way to make good the claim that contents are inferential roles, or that contents consist in inferences we may draw from them. This would be true if the thesis were restricted to the contents of expressions figuring only in deductive inferences. But since in natural languages expressions figuring in the premises of deductive inferences figure in the premises of inductive inferences as well, the thesis that content is inferential role cannot be upheld for natural languages.
I shall now focus on the nature of the task Brandom sets himself: to understand the notion of the semantic content of an expression. In the start I shall follow Brandom in using the distinction between force/significance and content generally. That is, I will apply it to public performances like promises and queries no less than to attitudes like beliefs and intentions. As I hope will become clear soon, though, in my view the distinction itself is artificial and arises through practical needs, not conceptual ones. Indeed, what I wish to ask is how anything, given that it has the content it has, could ever come to determine a force in the way the above initial quotations suggest.
I want to claim that in the order of explanation, the force, or the pragmatic significance, of a performance or an attitude comes first. 'Contents' are a pale and remote abstraction from the practical field of play where agents are cognitively engaged and linguistically purposeful in their own various ways. Contents arise as theoretical constructs, cut to their appropriate size with respect to defined (e.g., computational) purposes. At times we find it useful to abstract from the vagaries (emotional, valuational, accidental etc.) which the practical dimension of language use involves. Thus in the first place there are believing, desiring, and intentful agents. In the second, there are 'contents' and 'propositions' because we may want to strip their beliefs, desires, and intentions bare of their force.
The structure of 'content' will then by consequence always be relative to a chosen level of abstraction, dictated by practical or specific computational concerns. But if this is so, contents cannot, as our initial quote suggests, have the theoretical job to determine the forces which are enacted in the practical field of play, or the pragmatic significance of the attitudes which players have in it. And they do presuppose an account of it rather than not. Not attitudes presuppose an idea of content; it is contents which owe their existence to the attitudes.
This would mean that we are lacking a motivation for embarking on Brandom's project of 'grounding' a semantic theory, consisting in ascriptions of contents, in a normative pragmatics. Brandom's theoretical ambition is to understand and explain what it is for an expression and attitude to have a content, and what it is for an expression or state to count as meaningful. A theory of necessary and sufficient conditions for the existence of pragmatic significances as arising in communication and social interaction is developed to explain this. But, if what I said about the origin of content as a scientifically useful and well-defined theoretical abstraction is correct, what exactly is it we don't understand about the notion of semantic content?
There are general features to it, which are invoked by large words like 'intentionality', 'objectivity' and 'representation'. But all we can do by way of clarifying what we mean by them is to speak about what the agents or devices to whose states we grant contentfulness do and why, that is, by appealing to their attitudes. Looking at what they do and asking why, no doubt we ascribe contents to their states, at a chosen level of abstraction, and indeed we presuppose that there are those contents which can be ascribed to them, as well as that we know what it is to have a content. But what could result if one day we decided we do not understand this idea of contentfulness any more? What could result if not this that we look again at the social practices from where we abstracted, as I claim, our notion of semantic content?
Is there something theoretically unsatisfactory about assuming or presupposing the idea of content or meaning when doing one's inquiry? Not for the pragmatist, with whom Brandom associates. As I understand the former, he would grant the idea of content and might ask: Given that this communication or experimental test has this content, how does it or ought it to change the current conversational situation or doctrine? Surely not for the linguist. He equally presupposes the idea of content and asks: Given that this particular string of expressions has the content we (or the native speaker) ascribe to it, what features of this particular string of expressions make it have that content? Finally, surely not for the formal semanticist either, whose formal apparatus allows to talk about potential contents a string of signs has with respect to defined set-theoretical structures, but cannot account for the question what it is for an expression in a natural language to have the content it has.
It is a foundationalist's interest that is not satisfied in assuming contentfulness, as all inquirers and sciences do, and requires to make this a topic for inquiry itself. On the theoretical agenda then are the conditions for the very possibility of content, and it is this foundationalist and transcendental (indeed Kantian) question which gives us a good feeling for what Brandom is concerned with.3 It is the transcendental philosopher, not the pragmatist, who asks where the contents come from (as did Heidegger and Wittgenstein who were convinced that not how the world is is the relevant issue but that there is anything rather than not).
Against Brandom I shall argue that to understand basic laws and principles of human interaction doesn't require making a theory of content or meaning a task prior to it. This is to deny that an epistemological account of the dynamics of belief and other attitudes how we change them in the light of rational principles of doing so presupposes that we explain what it is for all of these to have a content. In other words, that semantics is something prior to the task of studying the attitudes. The order of priority is the reverse. We know what we mean by content because we know what we mean by attitudes, and what can give substance to the notion of content is a principled analysis of how we alter our attitudes in the light of experience and evidence. It is here as well, as we shall later see, where substance can be given to the ideas of 'intentionality' or 'objectivity'. A theory of content or meaning is no theoretical goal beyond the theory of attitudes. A notion of content can be a by-product of it if one or other notion of content or meaning is wanted.
I should mention that there is an important pragmatic notion of content, as when we say about a talk, which we heard on a conference, "it had absolutely no content." Clearly the notion of content here is that of relevance, and we know perfectly well what we mean by using this notion. What we mean is that a particular performance makes no difference to our inquiry; we know as much before as we know after. I have no intuitions as clear as those as regards the semantic notion of content.
In a sense on Brandom's own account there is no theoretical use for abstractions like 'content' or 'propositions' to play the role of things which 'determine' the "behavioral significance" (133) of the states and performances which have those contents or express those propositions. For we read that this behavioral significance is nothing but "the difference those states make to what it is appropriate for the one to whom they are attributed to do" (133). According to Brandom, the theory which accounts for this 'difference' is a pragmatic theory, and it is meant to give the contents and propositions which a semantic theory specifies a practical dimension and relevance. Now why separate the semantic study of 'contents' and 'propositions' from the study of those practical differences? The difference an attitude or performance makes to inquiry and reasoning can by itself serve as a notion, not just of practical significance (166), but of semantic content if one is wanted, rather than grounding an independently given one.
If this were what Brandom made us do, what we did was essentially dynamic semantics as conceived in the early eighties.4 The idea there is that rather than defining meaning in terms of 'static' semantic values such as sets of truth conditions, meaning now is defined in terms of 'dynamic' semantic values. These are functions which determine, given a context and a sentence or discourse, what the new context is which arises from enriching the given context with the information carried by the sentence or discourse. Contexts are usually thought of as sets of worlds or possibilities which are, as far as present information goes, compatible with what the agent or agents in those contexts believe.
Meaning then, understood as a function from contexts to contexts, is a matter of belief modification and the alteration of our attitudes. Now, it is true that meaning is a theoretical term, which one may define as one wishes. All the same I see no way to make good the claim that belief change or the alteration of our attitudes is a matter of meaning (alone). Evidently, how you change your mind is, in the general case, a cognitive decision problem constrained by principles of rationality. What you take the meaning of an input sentence to be might be said to be part of the evidence, which goes into your decision as to how to change your mind. But it doesn't determine it. In the general case, where your mental life consists in more than just the running of certain cognitive routines, you make a decision, given your exposure to some input sentence, whether you want to adopt a new belief, give up an old, do both or neither. The result is a cognitive change from which we can read off in some sense what the 'meaning' (better: 'significance') of the input sentence in your situation was. But then it is the context change which explains the meaning and not the meaning which explains the context change. It is on these lines that one might come to the conclusion that both meaning and content can be theoretically understood as by-products of an account of the dynamics of epistemic states and the rational modification of attitudes in the course of communication and inquiry. There are no 'genuinely semantic' notions of 'content' and 'proposition' besides genuinely pragmatic ones.
Now Brandom in effect gives us a version of the dynamic semantics picture of meaning, accompanied however by a richer idea of context. Following David Lewis5, his idea is that at any point in discourse a 'deontic score' is kept for each conversational participant, specifying information as to what his current commitments and entitlements are, information which is kept track of as discourse proceeds. The meaning of an utterance consists in modifying this score, although Brandom says this without specifying, in a systematic way, general principles and rules for how we score in conversational games (neither does dynamic semantics).
Brandom explains what he calls his pragmatic theory of meaning as follows: "Specifying the pragmatic significance of a speech act kind such as assertion requires showing how the transformation of the [deontic] score from one conversational stage to the next effected by such a speech act systematically depends on the semantic content of the commitment undertaken thereby." (186) Now that idea would make the dynamics dependent on a static theory of semantic content.6 This I have essentially claimed can only be reasonable if 'dependent' is read in the weak sense I mentioned. That is, a deontic state change may in part be determined by the semantic content we may wish to ascribe to a speech act, just as how you change your mind in the light of some communication depends on what you take the sentence used in it to mean. But it is surely not determined by it, and thus it would seem that all the hard work lies in specifying systematically rules and principles for changing your deontic score, but all of this would not be part of a theory of meaning but something independent and prior to it. But then, indeed, just a sentence later Brandom affirms that "these scorekeeping attitudes and shifts of attitude [can] be used to define both contents and interlocutors"(187, my emphasis). This now might be read to make the pragmatic theory of significances itself a dynamic semantic theory in the sense that you read off, from the alterations of an agent's attitudes, what the contents (for him) were which a speech act carried.
When Brandom finally writes that "the significance of an assertion of p can be thought of as a mapping that associates with one social deontic score (...) the set of scores for the conversational stage that results from the assertion"(190), he is essentially defining functional semantic values in the sense of dynamic semantics. Content or meaning becomes a function which, given a speech act, determines for each score of a conversational participant the new deontic score which arises by virtue of the speech act's taking place. But note that the problem I raised for dynamic semantics becomes more troublesome here. Of course nothing forbids defining meanings mathematically in the way the last quote suggests. But the result is an incredibly abstract notion of content, perhaps so abstract to be devoid of philosophical interest. What is the theoretical gain to look at an deontic-epistemic state change, subtract the posterior from the prior state, and define the difference to be the 'meaning' of the input motivating the change? The more contentful question seems rather to be what makes an agent choose this particular contextual change in the light of this input rather than that?
To my mind there is nothing wrong with a dynamic account of meaning. But it should be understood in the way that the dynamics are a topic for the theory of rationality, quite independently of semantic concerns. This undertaking may give rise to some notion of content or other. But this will be a theoretical abstraction which to explain we need not despair. And surely, its status will not be that of determining, for a speaker who grasps such a content, what pragmatic or context-changing significance a sentence expressing it has. Sentences are not the right kinds of things to change deontic and epistemic contexts. Nor are linguistic meanings.
I hasten to emphasise that the idea I have started with, that for Brandom semantic content determines pragmatic significance, indicates just one order of explanation for him. The other order is pursued when it is not that contents are presupposed in an account of what people do, but when instead it is what people do that gives the words they are using the contents they have (cf. 133, 134). "[A]ttitudes have the contents they do in virtue of the role they play in the behavioral economy of those to whom they are attributed"(134).
To illustrate what this means consider for example how Brandom's theory of the "inferential commitments² which "concepts² involve applies to pejorative terms like the French word 'Boche'. Brandom says that to use the word 'Boche' involves to commit oneself to the consequences of applying it, which are according to Brandom that one thinks of the person one is applying it to as liable to things like cruelty and barbarism. He clearly assumes in developing this example that it is this inferential commitment which "gives [concepts like 'Boche'] their content" (127), rather than it being the content which determines the permissible or obligatory inferences, or what the commitments are. Commitments are also said to be "content-conferring" (116, 76), and thus I take it are not determined by contents. And the strategy is said to be that of pursuing an inferential account of semantic content "not in abstraction from pragmatism about the norms implicit in the practical application of concepts"(132). Thus the road seems here to be clearly a road from practice to content and meaning, not vice versa. This second viewpoint of course seems considerably closer to the truth in my eyes, but Brandom clearly pursues both directions of explanation, and thinks there is a road back from meaning to use, besides the road from use to meaning.
In what way this is coherent escapes me, for it would seem that either content determines significance or significance determines content, but that you cannot have usefully both. Even if one's study of the pragmatics of communication provides a notion of content, I am not sure one can go on saying that once such a notion is in place it can determine, for a speaker who grasps it, how he should alter his or other's deontic score. This notion of content is a hypostasis, based on the idea that the sentences of a language come with or are such public and shared instructions to change score.
There is a similar problem in Brandom's work with respect to the explanatory relation of the notions of semantic content and inference. Brandom explicitly denies that contents are merely manifested in reasoning, and affirms that they consist in it.7 As I argued in the beginning, here one must distinguish two cases, the deductive and the inductive one. The argument was that in the first case we may agree with Brandom, for then the content of the premises merely unfolds in the inference. So in this case we can say that the content consists in the conclusions we may draw from it. But an inductive inference involves a cognitive leap: in this case we have to justify our going beyond the commitments we have incurred in endorsing the premises. We do this by assuming the premises, and otherwise by appealing to certain contextual factors by means of which the induction can be justified (such as our assessment of the probability of the conclusion given the premises, and its informational value for us at a given stage of inquiry). In this case we cannot say the contents of the premises consist in the inferences drawn. If this were so we could never justify the inference in a way such as the one I have sketched. Any justification of this kind requires the contents of the premises to be there to start with. On this basis we proceed to contemplate and evaluate various hypotheses, one of which may then be our inductively inferred conclusion. It is thus absolutely crucial to separate the deductive and the inductive case in a way Brandom doesn't, and restrict the thesis that content is inferential role to the first. Of course this makes in the general case the notion of content prior to that of inference. This I think is just as it should be.
Like Frege, furthermore, I take it for granted that from a sentence (of a natural or formal language) nothing is ever inferable. For Frege from premises which are not known to be true (i.e., which do not correspond, in my jargon, to beliefs one holds) nothing ever follows.8 It is agents who either draw conclusions on the basis of settled assumptions or otherwise make judgements as to what follows from assumptions adopted for the sake of the argument. The first kind of behaviour we call inference, and we will then understand it in the way that an agent who draws the inference thereby explicitly incurs, if the inference is deductive, a commitment implicit in those already undertaken.
That a commitment to the truth of A&B induces a rational commitment to endorse the truth of A is nothing that seems to have anything to do with language. Language as such never entitles to anything. We may represent an inference in terms of a sequence of sentences, each of which we will take to express a belief or rational commitment of thought. But both that a language as such 'allows inferences', or that a sentence 'follows from another', under one interpretation or under any, are formulations which seem true at best in a derivative sense.
This view, for all its obviousness to me, is an important disagreement with Brandom who sees languages as systems which as such incorporate norms for altering one's attitudes and states of doxastic commitments. It is true that there are certain routines for changing one's mind in terms of new linguistic inputs as when I start, without anything like a genuine decision, believing it is 5PM when somebody tells me "It's 5PM". But there is no question that such a routine entitles me to such an inference. Routines if they exist are useful, since otherwise we would give them up, but they are quite different from justifications based on non-linguistic deliberation.
To sum up what I said so far about the semantics-pragmatics interface, just as there are problems with the first above view on the nature of this interface, there are problems with the second. The first, I have already argued, demands the impossible. Abstractions like content cannot determine what they are abstracted from. But the second misses a direct path which there is from what people do to what they (are committed to) believe. Brandom on his map draws this path as leading from what people do to what their words mean (134), and from there to what they believe. This is a detour in my eyes, since an account of what they believe is needed in any case, and on the direct path the account of meaning can be hoped to be a by-product of the account we give of belief.
What gives us the distinction between the force and the content of an attitude in the first place? Perhaps grammar seduces us to take too seriously the fact that attitude verbs like believe and desire are transitive. If we believe something there has got to be something that is believed, the 'object', as one says, of the belief, which then happily transforms into the 'content' towards which the belief is 'directed'. Be that as it may, let us now consider more concretely the agent who wants to know an answer to an open theoretical question. He faces a cognitive decision problem in that he has a series of options (potential answers) in front of his mind. Among these he wants to determine the optimal one, optimal in that it satisfies his desire to acquire knowledge but respects his concern to avoid carrying something worthless since false into his current doctrine. Why not say, firstly (if a notion of content is wanted), that the content of what he believes is given by the set of epistemic commitments which to endorse, given his doctrine, he rationally has to be disposed to when prompted appropriately?
Secondly, why not say the 'content' of what he wonders about are the potential doctrines he would adopt were he to add one of the potential answers to his open question as a further thesis to his doctrine? We might say, with Brandom, that the content of a hypothesis is the difference it would make to the current doctrine were the agent to add it to it. The question what the contents of the hypotheses or potential answers are is thus answered by reducing their content to that previously defined for his beliefs. These are what the difference is a difference of. The content finally of what the agent values might be given by the potential beliefs he may have concerning how good something would be.
At no point then would we have to speak of propositions or semantic contents. All the contents there are reduce to the contents of beliefs. And these are not propositions or semantic contents but are further understood in terms of rational commitments of thought, which are always relative to a given set of such commitments. But have we cheated? Isn't a rational commitment of thought something that itself has a content, and thus requires us to inquire what those contents, given independently by a genuinely semantic theory, consist in?
We haven't, because indeed I recommend that, at this point, we break up our chain of reductions. If I talk about commitments and I am asked, which commitments, with what contents, I answer by offering these commitments as primitives. So they are the commitments they are. But wait, if you wish to know more about what these contents qua commitments consist in, I'll say that they consist in the set of whatever else is a commitment if they are. Beliefs, qua commitments, will thus be represented as their own consequence sets. Now these consequence sets we can again collect into a set. This completely abstract structure we will call the algebra of potential states of full belief. These states are partially ordered according to whether they are stronger or weaker, an ordering that we may call an implication relation. This relation can be characterised in a purely structural way, that is, without assuming the meanings of any logical operators as given. In fact the latter can be completely characterised in terms of the former. For example, conjunctions are values of a function C applied to elements A and B of the algebra which have certain properties relative to the implication relations: C(A,B) implies both A and B and is the weakest element of the algebra with that property.9
What we get then is a way of spelling out that 'content consists in inferential role' without having to appeal to anything like the internal structure of sentences representing potential states of full belief. Nor do we need introduction and elimination rules to determine the 'meaning' of logical signs occurring in such representations. It is perfectly clear on this picture, though not on Brandom's, that each and every move from one consequence set to another is not a matter of content or 'meaning'. Rather, one is in a state of belief, whose content is given by its consequence set, and tries to justify a move to another. There is no way that the contents of those prior states consist in such cognitive moves. The only inferential moves in which a content (belief) may be said to 'consist' are those in which an agent stays within one consequence set.
Opinions differ as to precisely how much structure our algebra of states of full belief should have. Some have taken it to be a Heyting algebra, some require it to be Boolean, still others wish to remain uncommitted and simply assume a complete lattice.10 In each case, the structure derives from requirements on the rationality and coherence of potential states of belief. Logic, when read off from the algebra, will thus derive from the theory of rationality and not be prior to it.
My primitives, I claim, are qua primitives at most as reprehensible as 'semantic contents', 'propositions', or truth conditions, for all of the latter are much more remote from practice and thus more difficult to give a clear content by linking them to it. And my primitives are at least as good as they since they do the job these things do, and more. For rather than having to link an attitude-neutral and force-neutral notion of proposition or 'content' to its occurrences in action and deliberation where they always already have a force attached to it, we start where agents incur commitments of thought and action right away. Elsewhere I argue the primitive of commitment should also be preferred to Donald Davidson's favourite primitive, the notion of manifest preference.11
Finally, there is a sense in which they are just the primitives which Brandom employs, since although he claims that his "semantic primitives" are "proprieties of material inference"(133), not commitments, they "can be explained in the pragmatic theory as implicit in discursive practice"(133). But what is implicit here boils down to nothing but the commitments implicitly undertaken in reasoning and discourse. So inference is itself further explained in terms of commitment. Moreover Brandom allows meaning determining but non-inferential language entry and exit moves, as deriving from direct perception and action.
Despite this seeming agreement, the disagreement I insist on is that for him commitment is a pragmatic primitive, and that there is a semantic primitive in addition to it which I claim is not needed. (Another point I will insist on is that the study of commitment is not the study of some public conduct or the practice to "treat inferences as correct"(134).)
Still you may ask, what are the identity conditions of these commitments, and of the kinds of commitments? As regards the first question, the question must be made relative to the fineness of the grain with which you want to consider an attitudinal commitment. Given that you have one such commitment, you can always differentiate it into infinitely more. For example, the belief of Inge that she visits her friend soon in Toulouse may split or be further determined as the belief that she visits her boyfriend in a village close to Toulouse on a putatively beautiful Monday towards the end of the month, and so on. Once you have chosen a certain fineness of grain, with due respect to your present interest and concerns regarding Inge, you can go on to make an inquiry whether some commitment or belief with that particular fineness of grain is indeed the one Inge really has. This question concerning the identity of a commitment can be settled as good or as bad as any other question we ask about the identity of something (for example about whether the house we saw yesterday is the house we see today, or whether your taste of this Minestrone is similar to mine).12
As regards the second question, concerning various kinds of commitments, there are, for example, commitments to be disposed to make probability judgements given that you make others. If you think rain is probable to a degree of at least 60%, you are committed to judge lack of rain to be probable to a degree of at most 40%. If you find chocolate is disgraceful you are committed to be disposed to stay away from it as long as you want to be coherently describable as having this view of chocolate. And if you rule out that Betty is the murderess as a serious possibility for your inquiry (say, because she was dead at the time), you are committed to accept it as a necessary truth that Betty is not the murderess.
I answer that indeed we have to take one such kind of commitment as central, and from my exposition it is clear that this will be the commitment of the last specified kind. These are commitments that something is certainly and absolutely true, in other words, that its falsehood is not a serious possibility. So it is such commitments which will be the elements of our algebra. The other commitments do not span separate algebras consisting of other kinds of creatures, but are modelled by distributing values (values of probability, values of utility, various other values) over the one and only algebra of potential states of belief or epistemic commitment.13
This now has led me in the middle of what is in effect one of Brandom's own central concerns: "that belief can be modeled on the kind of inferentially articulated commitment that is undertaken or acknowledged by making an assertion"(157). The first half of this claim I have already endorsed. But I object to the second half, since it is not public assertions that induce commitments. This is immediately clear in the deductive case. And in the inductive case I shall now point out that justifications must be assumed to be intrapersonal ones.
Our abolishment of Brandom's duality of commitments and entitlements means to rule out inferences which Brandom claims are 'primitively good in virtue of the contents of the concepts involved in them' (Brandom's "material inference"). Examples are "There is lightning, so thunder will be heard", or "Gustav is a dog, so Gustav has four legs". As I pointed out before, in such inferences the truth of the premises here doesn't guarantee that of the conclusions. These are consistent with the premises, but there is no question that the premises entitle to them: to infer them requires some courage and boldness on the side of the agent (however little). The inferences are not safe, and this brings about a need to tell when one can or should and when one cannot or should not draw such inductive inferences.14
That inductive inferences involve a risk of error, as seen in the fact that in the above inferences for some reason thunder may not be heard, and Gustav may turn out to have three legs, does not mean one should not draw these inferences. But one should know what one is doing: one should weigh the risk. Consequently the agents for which I make a case here are audacious agents who put the boldness of their quest to rid themselves of their ignorance against their calculated risks of coming to believe the false. They go beyond the commitments they have incurred to incur new ones. These then form an addition to the stock of things one takes for granted while addressing further open questions.
Striving for being faithful to the harmonious interplay of the 'conditions of application' and 'permissible consequences' of the concepts and words which already have acquired the authority of public usage is not central for those agents. Brandom's agents, by contrast, explore the 'real' contents of concepts by making explicit their putative grounds for applying a concept or word and for inferring its material content. To see whether they fulfil the public standards for applying these concepts they put them "into the open as liable to challenges and demands for justification"(127).
But there is something rationally incoherent about an agent who, if he rules out the possibility of Betty's being the murderess, sees Betty's not having murdered the victim as still being open to doubt or liable to challenge. It is simply not the case that he is under some kind of obligation to get involved in the "game of giving and asking for reasons", as Brandom calls it, since he can only rationally be expected to doubt things whose falsehood is a serious possibility for him.
Also disagreement by itself would be no good reason to change the commitments I have incurred, and doesn't factually seem to be taken to be one either. It is true that disagreement may eventually of course belong to the evidence which may lead me to make a suppositional belief revision of the following kind: "I believe A. But what if I were wrong in the end? Then perhaps I could explain something I couldn't explain before", and so on. Note that at this point, where I make a belief revision for the sake of the argument to explore possible consequences, none of my convictions is yet given up (so this is not what the disagreement has led to). And indeed I shouldn't do this without good grounds. It is irrational for me to first express a conviction, A, of mine and then to backtrack upon mere disagreement. But then, if my supposition does eventually lead me to change my mind (I give up my belief and/or adopt a new one), this is not something the disagreement will have caused. For it will be a decision I have to justify not in front of the linguistic community I happen to be involved with, but to myself. The decision to take this step is a decision in need of justification, namely to count something as dubious or false which was reckoned to be absolutely and certainly true.
Note that this change of judgement would have its status and could be theoretically studied as such in complete independence of public or 'external' acts of assertion.15 There is good reason not to focus on public acts when studying rational commitments of thought. I happen to fully believe that Schröder has been elected as the last German Chancellor. I wouldn't think of providing a justification for this, unless someone gave me reason to doubt it, which nobody has. If I went on the street, though, today on 31 January in Berlin, and declared that Schröder had been elected as chancellor, people might actually ask me to justify my claim, given that it presumably would not address any real issue either for them or for me. All the same, what people would not ask me to do in this situation is to justify my belief that Schröder has been elected chancellor. Or if I believed and claimed that Rudy is a crook, and did so publicly and loudly at a ceremony celebrating Mayor Rudolph Guiliani's 60th birthday, people may approach me with looks clearly indicating that I was not supposed to or had no right to make this claim. But even then I do not have to and would not be asked to justify my belief that Mayor Rudolph Guiliani is a crook.
Brandom's claim that "[T]he use of any concept or expression involves commitment to an inference from its grounds to its consequences of application"(126) thus seems wrong. Let us take it that most of us are agents who have, in the course of their inquiries, ruled out certain logical possibilities as serious ones. These are the things we assume for sure and which form the unquestioned basis for all further question asking. For this to be a permissible practice the grounds on the basis of which one of these certainties was or is assertible are irrelevant. Whether I recall them or not does not matter. I am neither committed to know them nor to infer consequences of application from them. That is so precisely as long as I am not conducting an inquiry into their grounds or justifications (which, by our assumptions, I am not).
In sum, I do not have to justify what I assume in my efforts to acquire knowledge my beliefs -, but the decisions I make about which one of the hypotheses I have selected as potential answers to my problem is the correct one my changes of belief. The inference I draw to a conclusion, which I then adopt as a true belief, moreover is something I have to justify only in case this new belief was not entailed by the evidence available to me. And again it is decisive that if I want to count as rational, I am under an obligation to justify this inference to myself, it being irrelevant to this task how others than myself assess my inferences as correct or incorrect.16
The above viewpoint requires me to assert that relative to my present full beliefs the truth of another belief which were to contradict one of them is not a serious possibility. But it is perfectly consistent with it that I can judge, relative to a potential future set of full beliefs, that the falsehood of one of my present beliefs is absolutely certain. For example, relative to the potential future set of beliefs according to which Schroeder was an actor set on the scene by the Americans, I do believe that my present belief in Schroeder's election has turned out false.17
These claims have the important consequence that there are no publicly given 'concepts' or 'inferential norms' which, as far as some public or official standard goes, "ought to be endorsed" (127), or "govern the use of expressions"(134). It is an agent who makes, on the basis of what he has found is true, a decision as to whether he ought to endorse or can justify to himself - an inference, considered as a non-linguistic thing. You may use the expression 'dinner table' where I would use the word 'tiger', not to your advantage presumably, but there would be nothing thereby rationally incoherent about you. You may or may not infer danger or food where I do, but there is no point requiring that you infer food where I infer danger.
Brandom himself takes up at one point (127) (albeit with a marginal interest) what I have been calling the inductive case. The issue Brandom raises is conceptual progress in science and the introduction of new concepts into a language. Both Dummett's and Brandom's question at this point is this: the introduction of the concept, for example the concept "Boche", will come with an inference license which will make permissible inferences which have not been permissible in the language before (such as that a German is cruel and barbarous). The result of this addition is in this sense a nonconservative extension of the language.
This account of what the problem here is presupposes that language is the sort of thing which makes certain things permissible or not, or entitles one to an inference or not. It seems, though, that agents have such powers. And it is not language one worries about when one reflects as to whether one makes a certain inference since what is worrisome is whether one particular belief should be held given that others are. One can wonder: should I believe, given what else I believe that Germans are barbarous? But this is not a question as to whether to use a word or not, although it is a practical consequence of its answer that I may not find it useful to publicly use the word Boche to express the beliefs I have about Germans. But note that even if I used it, there would be no sense in claiming that I am thereby committed to believe that Germans are cruel and barbarous. True, my language use is one of the guides for my listeners for ascribing beliefs to me. Using the word Boche may be found misleading, reproachable in some other way, or indeed very wise depending to whom I am addressing my words. But there is no way that it commits me to believe, or take as evidence, that Germans are cruel and barbarous.18
The central question according to Brandom with respect to a new concept added to a language is whether the new inferences should be endorsed or not, given a present set of inferences 'licensed by the language'. What is surprising is that Brandom doesn't investigate general principles and parameters that could help an agent in making such a cognitive decision. Suppose I don't believe Germans are cruel and barbarous, and I hear an assertion of "Freddy est Boche". If I know Freddy is German and know the word Boche suggests certain inferences about properties that Germans have, what should I do? Change my mind about Germans? Try to change speaker's mind about Germans? Something else? Brandom doesn't tell. Indeed, no matter how close one looks at language, no answer will fall out of this. If content is to derive from change of deontic score, and deontic score is changed by performances, Brandom should have an answer to that question.
Brandom writes "nonconservativeness just shows that it [i.e., the new concept] has a substantive content, in that it implicitly involves a material inference that is not already implicit in the contents of other concepts being employed."(127) That seems to me exactly right, and in fact it points precisely to the unacknowledged need, within Brandom's enterprise, to integrate a systematic theory of the justification of inductive inference in the context of a theory of linguistic meaning. One might even perceive in Brandom's sentence the interesting idea that 'substantive' content derives from, and can be measured in terms of, the degree of surprise, or the degree of novelty which an agent attaches to a belief not entailed by what he believed before.
Inductive inferences, although nobody is ever strictly speaking entitled to them, may still be legitimate relative to a context in which decision-relevant parameters are given and supplied with values. All of these serve to evaluate potential changes in the current doctrine or view, and thus evaluate aspects of the content of an input (such as its probability, relevance, or feasibility). We should note, though, that we cannot expect that in a context with a group of agents the evaluations they attach to a hypothesis conform to one objective standard for evaluating hypotheses. There is nothing that compels the view that agents who share some piece of evidence, or witness the same performance, reach the same conclusions concerning the relevance that this piece of evidence has relative to their common or individual inquiry and/or deontic status. But then, since the 'difference' the evidence makes with respect to the inquiry or deontic score is what Brandom takes to be the measure of its content, it follows that the 'content' of the same piece of evidence can be very different for each of the group members. But this gives us the consequence that there is no theoretical point in the assumption that a 'language' is a treasure house of public and essentially shared meanings and norms that are what induces the changes in deontic statuses. It is the deontic and epistemic dynamics we must start with, and from where we may extract if we find it useful a notion of 'proposition' or 'semantic content'.
Again Brandom might evoke at this point his functional construal of meaning, according to which what a speaker grasps when he is exposed to a grammatical sentence is a rule which, when applied to an epistemic-deontic state, returns a new epistemic-deontic state. So a language would in the end be a treasure house of shared meanings in the sense of such functions. Again I point out that there is nothing wrong with this meaning-theoretic idea mathematically, but I doubt what its philosophical interest can be. The output of the function for a given argument is not intuitively determined by this argument (alone). It is thus not a good notion of sentence meaning. What an agent who grasps such a function grasps is the result of a cognitive decision which, when exposed to the input, he would yet, as it would seem, have to make. All the meaning-theoretic interest would lie in how an agent's given state interacts with a new input to produce such a function.
A consequence of adopting a principled approach to the alteration of our attitudes is that it makes language idiolectical. But there is an advantage to this, namely that it allows Brandom's interesting account of coreference as commitment-preserving substitutability to work a lot better than I think it can otherwise. What is to be a commitment given that something else is one is a question to be decided with reference to the commitments already incurred by the agent to whom the question is addressed. The question whether some substitution of a subsentential constituent for another is commitment preserving or involves an unwarranted commitment-change has then a precise sense. If language, by contrast, is by its essence social, the question when substitution preserves commitment and entitled inference doesn't seem to have a clear sense -, as I shall now explain.
According to Brandom, the content of an expression derives from its behaviour in inferences. But then it would seem that the only notion of content we can make sense of is the notion of the content of a sentence. To extend this notion of content to subsentential constituents, Brandom assumes that the contents of singular terms and predicates can be characterised by indirectly inferential roles. For example, two singular terms have the same content if substitution of one for the other doesn't alter any feature of the inferential role of the sentences they are substituted in. This gives rise to the notion of a substitution inference, which is one in which the conclusion is a substitutional variant of one of its premises (282). Finally, just as there are inferential commitments associated with whole sentences, there are now also substitution-inferential commitments associated with subsentential expressions: they are commitments to the propriety of substitution-inferences and may be expressed, for example, in terms of an identity-sentence. A good substitution inference preserves substitution-inferential commitment. Anaphora, finally, is what preserves substitution-inferential commitment as associated with a singular term. If we add the notion of anaphora to that of inference and substitution, we get Brandom's account of reference, which is interestingly explained in terms of the former.
The point I want to make is not directed against this path to the notion of reference. By contrast, it subserves it by making the whole account applicable to idiolects rather than 'languages'. The point is that it is just not clear what it can mean to think of substitution-inferential commitments as eo ipso social substitution-inferential commitments, as substitutions an individual in a community 'ought' to endorse as commitment preserving. Suppose the most advanced standard of astronomical education in a linguistic community allows for commitment-preserving substitution of 'Hesperus' for 'Phosphorus'. Does the same substitution have to be commitment-preserving for Inge? If so, how many followers would Inge have to gather to impose her diverging standards of substitution on the community? The questions seem senseless (or rather, they seem irrelevant to the theory of meaning), but Brandom provokes them by making substitutional commitments relative to a given community's social linguistic practice, and the notion of a community is vague.
As we see, Brandom is quite concerned to reconstruct a reference relation between words and objects in the world, in a semantics which doesn't take this notion as a starting-point.19 This in fact is to subserve his other major aim, in chapter 8, of reconstructing what he calls the 'representational' dimension of propositional content. One of his reasons for the need to reconstruct the notion of reference is the thought that objectivity is not 'merely a matter of our attitudes', of 'what we take to be true', but of 'how the world is'. Certain proprieties of judgment and inference, Brandom writes, "outrun actual attitudes of taking or treating judgments and inferences as correct.(...). Our cognitive attitudes must ultimately answer to these attitude-transcendent facts"(137).
In my view, this anxiousness about objectivity misses much of the point of Brandom's own attempt to understand attitudes in terms of commitments. Having an attitude, on the present construal, is something quite different from 'taking or treating something to be so-and-so', or 'having a certain picture of things'. We do not act as if our beliefs were true but because we assume they are. That I have full beliefs means that I am committed to the fact that things are in the way I believe they are. This objectivity is a commitment: we strive for living up to it. It is crucial to us to distinguish between what we come to believe true and what we come to believe true erroneously, which is why we try to minimize the risk of doing the latter. The distinction between our take on things and the things themselves in this sense plays a crucial role in our deliberational dynamics by being its driving force. But it is nonetheless a distinction that falls within our attitudes. To make sense of this practical concern to distinguish truth from falsity, though, the notion of an attitude-transcendent facts doesn't seem needed. It is true that, as Professor Brandom pointed out in discussion, he doesn't wish to start with or presuppose a notion of objectivity either. The goal after all is to reconstruct these notions on the basis of an account of attitudes and communication. But even if there is less disagreement here than it seems, it remains that the distinction between true accepted beliefs and false accepted beliefs which each and every individual can give a clear practical content in his deliberations is an intra-personal one. Ultimately I cannot let a community make my cognitive decisions.
So I don't see what more can there be to the idea of objectivity than a set of propositions whose truths I am fully committed to, together with the admission of corrigibility. This is the admission that further information may always force revision, even if I exclude falsehood from my full beliefs now. It doesn't give these beliefs the status of a vision, and doesn't require a project like establishing that the world really is as we hold it is. The only way this project could proceed is by looking whether given evidence which gives reasons to doubt a claim one fully believed should lead to revision or not. But the only outcome this project can have is that we either change or don't change our state of epistemic commitment. So we either end up with a new set of full beliefs or stay with the old one. Whatever there is to the idea attitude-transcendence can thus be cashed in in terms of our epistemic dynamics.
Coming now back to the conservatism issue, Brandom allows and discusses, as mentioned, nonconservative extensions of a language in virtue of scientific progress or conceptual dynamics. In my terminology, these extensions are nothing but potential inductive expansions we may find useful to perform on our current doctrine. In the context of an open question, there will always be many of them, and thus an evaluation of them is needed. I have mentioned contextual parameters is conservatism one of them? Also, is Dummett's and Brandom's concept of harmony relevant, by which they mean that the introduction rules of a concept (the rules telling when to apply it) are not in conflict with their elimination rules (the rules telling which consequences to draw from an application of a concept)?
Harmony, in the present context, translates into the requirement that a state of full belief should be in a reflective equilibrium under rational criteria of self-criticism. For example, I am not in such an equilibrium if I think that not all Germans are barbarous, but are Boche, and think that people who are Boche are barbarous. This inconsistency in which I may find myself would rationally require me to move into another state, again in equilibrium, in which the inconsistency is removed.
It seems to me that the answer to both questions, whether conservativity is relevant and whether harmony is, is clearly no. Neither to be conservative nor to be in harmony with oneself is a requirement of rationality. Of course, if one wants to justify adopting a new belief, or to use an old word in a new way, one needs to take something for granted in providing this justification. According to Brandom, these are the norms of language as existing at present, according to me the settled questions or full beliefs at this time. But suppose the new belief is inconsistent or incoherent with the current norms or present beliefs. Then the agent has to make a decision as to whether to give up his commitments to one of these norms or beliefs. Is conservatism a constraint for the outcome of this decision?
No, for even if his decision is to reject the information that would have led to the new belief and to stay with the old norms or doctrine, the reason that these are the old norms or doctrine is not a good reason for not changing them. Rather, we need a reason of some such kind as that the risk of error with which the new belief comes along is too high for adopting it, no matter how high its informational value may be. Conservatism is no similar such criterion. Equally, if our agent finds himself in a community having a seemingly inconsistent or non-harmonious set of norms, or finds himself having an inconsistent doctrine, he will typically have many options for removing the inconsistency and to adopt a new doctrine. All of these will, if they constitute any improvement, be coherent or in equilibrium. But of course, that they are in equilibrium is not a good reason for endorsing them, for otherwise the agent would have a good reason for adopting just about any of them.
Commitments, I have argued, don't have to be carried into the open to play an accountable theoretical role. So why make it explicit? Let us finally ask how large the scope of the enterprise of making implicit commitments explicit can be. One general problem that limits this scope is that so many connections between concepts and beliefs are probabilistic in nature, and thus not inferential in a strict sense. Probability assignments for example are arguably not even propositional in nature, that is, carry truth-values.20 But then they cannot figure in inferences, in fact on Brandom's own view of inference where inferences involve only propositions as premises and conclusions (cf. 83). Moreover, many performances, such as imperatives, are meaningful although they do not figure in inferences at all.21
But also, so much that is decisive for decision-making and understanding cannot be explicitated at all. The project presupposes that everything which can interest us can be carried into the open by being made explicitly the part of a propositional (assertible, knowable) content.22 But if, for example, probability assignments are not propositional, being implicitly committed to a probability assignment cannot be explicitated in a full sense by describing this fact through a sentence like "His probability for ... is...² since this is a proposition, and the assignment is not. This point extends far beyond probability assignments. A command I make is expressed in an imperatival utterance, but is not 'made explicit' by or can be understood in terms of the meaning of the sentence "I command you to...². The study of this meaning doesn't illuminate the nature of imperatival performances because it presupposes an account of them. Also, the sentence "I believe that p" is regarded by Brandom as an explicitation of my belief that p. But if explicitation is to serve the goal of examining the "consequences and possible justifications"(128) of given implicit commitments, explicitations such as the one Brandom offers here are not the right kind of things to study. To study the practical significance of an attitude is not to study the meaning of a sentence which expresses (belief in) the fact that the attitude is held. This is simply a different thing.
The study of an explicit attitude attribution is thus not the right path to the attitude that is attributed, which has to be treated directly for what it is. In an attitude report, Brandom writes, the "force becomes part of the content"(121, 116). I wonder whether this is not to propose precisely what Brandom says we should not do, namely to assimilate judgement to predication.23 For having a belief means to make a judgement that something is true. But this is not the same thing as to classify something in the sense of predicating, of it, that it is true, or that I believe it, for example in the form of a sentence like "I believe that A" or "A is true". For these sentences express propositions, and thus are true or false, whereas my judgement that A is true is a judgement that A is a consequence or an element of my doctrine or view. This judgement is not again something whose truth or falsity I can wonder about: if I commit myself to the truth of A, there is no question for me whether my judgement that A is true is really true. If it were, I would not have judged it true but still be in suspense. Frege conceived the judgement stroke in the way he did because he insisted on the fact that the force cannot be part of the content. Frege's force indicator is not a meaningful part of an interpreted language, which is why the complex of the proposition and the judgement stroke attached to it does not again determine a proposition.
To generalise this point, the man who thinks it more likely that it will rain than not makes a probabilistic judgement, or has an attitude, both of which we may describe by saying "His probability for rain is greater than .5." But this sentence can perfectly well be understood as a statement of fact, and so doesn't necessarily capture or make explicit what is most crucial to the study of such attitudes: that agents who have them or make judgements corresponding to them decide on the probability of something. This is a cognitive engagement by which they are then committed to act in accordance with the commitment they have undertaken. This active and performative aspect of making a judgement is not captured by making a third-person biographical remark about a person to the effect that it is true of her that she makes a certain probability judgement. Also my first person remark "I believe it more likely to rain than not" does not by itself 'make explicit' this particular belief of mine: it does not do this because it can be understood as a sentence which expresses a proposition describing an autobiographical fact. But it is not an autobiographical fact that an agent who makes such a judgement commits himself to.24
The general point is that a judgement is an act and as such has a performative aspect which, as an object of study for decision theory, can be represented but not studied in terms of a sentence with a propositional content that describes it. (Of course we will have to formulate a sentence about what we want to talk about, but I take it this trivial fact is not the one Brandom has in mind.) A sentence can be used to represent the fact that an act or decision has been made. One can say, of a performance, that it is one, but if we do so, it is out of a practical, not a theoretical need.
Brandom's project seems inspired by Hegel's in many regards.25 Moreover, the notion of 'making it explicit' has its origins as a neo-Platonic metaphor. As we start reading the Phenomenology of Spirit, not other than start listening to the solemn beginnings of the Rheingold, we are meant to think or hear the spirit as it is entirely within itself. It is implicit, but this just means it is on the verge of evolving, of becoming explicit since it cannot remain within itself due to internal contradictions. But this process is not one of course which runs endlessly and blindly into the open; indeed, genuinely there is no future at all, for whatever may become explicit is implicitly all there right from the start. At the end of the Phenomenology what we have come to know is what that is that we knew all along. Beyond this, there is nothing to be known. Subtract from this that in Hegel pride of place is presumably not given to language in the way Brandom's linguistic philosophy suggests, and one sees an early attempt of 'making it explicit'.
Another of Brandom's historical allegiances, Frege, resisted explicitation in a crucial point, though. Frege thought that truth cannot be elucidated, and the notion of the act of judging a proposition true is a primitive that cannot be made the object of theoretical study. It is precisely here where Dummett took objection with Frege, and Brandom follows suit. The objection was that if we take the notion of truth for granted, we cannot hope to do any explanatory work in associating contents with sentences and explain these in terms of conditions for truth. The task we settle on then is "to win through" to a notion of truth26, to gain the right to use it within semantic theory. As Brandom might read this, semantics then becomes a preface to normative pragmatics (in Dummett's case in the shape of proof theory, in Brandom's in the shape of a vast generalisation of this, the theory of commitments and entitlements).
It is of some importance to note, though, as I pointed out before, that a structural account of the attitudes in terms of commitments does not in any way depend on the idea of a 'truth condition' or that of a 'propositional content'. Truth enters in quite another way, namely as a parameter that is to be taken into account in justifying changes of commitments. In the context of an agent's efforts to justify such changes, truth is both a value and at risk. No matter how valuable an answer to one such open question may be, it may, for all we know, always be a false answer. Truth plays an equally important role in assessing the communicative moves of one's fellow speakers. But it should be noted that to tell such a story about the role that truth plays in communication and inquiry is not to elucidate in any way 'what truth is'. It is if we become too attached to a theoretical construct, 'propositional content' or 'meaning', that we think there is such a theoretical need.27
1 Robert B. Brandom, Making It Explicit Reasoning, Representing, and Discursive Commitment, Harvard University Press 1994. (Cited henceforth with page references only.)
2 As is well known, Michael Dummett has considered the idea of trying to extend this simple meaning-theoretic account of logical expressions in the Gentzen-style to any kind of expressions of natural language. This aspect of Dummett's work is the origin of Brandom's enterprise. It should be noted though that Dummett manifests a clear scepticism about the extent to which this alternative meaning-theoretic project can succeed. This scepticism I don't find in Brandom's work. We should note that in Dummett's case the scepticism doesn't merely extend to the idea that any kind of expression comes with meaning-determining introduction and elimination rules. It extends also to the idea to think that the meaning explanations for conjunction and disjunction work properly even as meaning explanations for the inferential use of expressions like "and² and "or² in natural languages. See e.g. the discussion of the "Fundamental Assumption² in Ch. 12 of M. Dummett, The Logical Basis of Metaphysics, HUP 1991.
3 I do not want to fault Professor Brandom for his historical allegiances, but, for the record, I do not believe that there are such necessary and sufficient conditions for calling something an 'agent', or for calling something's states 'contentful'. These questions are, in my view, empirical.
4 It finds no mention in Brandom's book. See Groenendijk, J. and M. Stokhof, "Dynamic Predicate Logic", Ling.& Phil. 14, 1991, or H. Kamp and U. Reyle, From Discourse to Logic, Kluwer 1994.
5 "Scorekeeping in a Language Game², reprinted as Ch. 13 of D. Lewis, Philosophical Papers, 233-249.
6 That the dynamics is ultimately derivative from a static account of meaning is also true of many versions of dynamic semantics such as Hans Kamp's DRT, where for every dynamic semantic value there is a static semantic value, too, depending on the perspective you take. But other versions have sought to show that meaning is essentially dynamic, that is, irreducible to a static account (in my view without success).
7 E.g., "Semantic Inferentialism and Logical Expressivism", Lecture read in Genoa, 26 January 1999, 3.
8 Letter to Jourdain, Wiss.Briefwechsel (eds. G. Gabriel et al), Hamburg: Meiner 1976, 118. I owe my scholarly insights into Frege to Göran Sundholm.
9 See e.g. A. Koslow, A Structuralist Theory of Logic, CUP 1992, Chs. 2-3.
10 See Peter Gärdenfors, Knowledge in Flux, MIT Press 1988, Ch.6, Isaac Levi, The Fixation of Belief and its Undoing, CUP 1990, Ch.1-2, and Bas van Fraassen, "Identity in Intensional Logic: Subjective Semantics", in Eco, Santambrogio (eds.), Meaning and Mental Representation, Indiana University Press, 201-219.
11 See Wolfram Hinzen, "Belief and Meaning", Ms.
12 This also (my understanding of) Isaac Levi's view on the individuation of attitudinal commitments.
13 For one possible development of this view see Isaac Levi, The Fixation of Belief, Ch.2.
14 I take inductive inferences in the way they are understood here as what provides the intended applications for the more recent study of 'non-monotonic reasoning'.
15 In other words, I see no reason to believe that 'judgements are the interiorization of the external act of assertion', as Brandom approvingly quotes Michael Dummett (153).
16 Just imagine the case where you draw an inductive inference, and do so in a linguistic community in which one half approves of your inference, and another half does not.
17 This way of making absolute certainty compatible with a refusal of dogmatism is familiar from the writings of Isaac Levi, see his The Enterprise of Knowledge, MIT Press 1980.
18 Cf. Brandom's claim of the opposite: "Employing an expression with [a content] involves endorsing the inferential commitment from [its associated] circumstances of entitlement to [its] consequences of commitment."(335)
19 In my view intuitions like that beliefs are 'about' something, or that there are singular terms which 'refer' to objects in the world can be accounted for without breaking unanalysed propositions into constituents. See Bas van Fraassen, "Quantification as an Act of Mind", J. Phil. Logic, 11, 1982, 343-369, or A. Koslow, fn. 10, Ch. III,20-21.
20 This view is held by writers in statistics (J. Savage, B. de Finetti), as well as philosophy (see Isaac Levi, For the Sake of the Argument, CUP 1996, Ch. 3, and Bas van Fraassen, "Belief and the Will", J. Phil. 81, 1984, 235-256 for different defences of this view).
21 This point confirms my claim that Brandom puts too much emphasis on his understanding of inferential propriety as a semantic primitive. If we take inference as a semantic primitive seriously, too much becomes meaningless.
22 Explicit sayings, judgings etc., Brandom assumes, are propositionally explicit. Cf. e.g. 135, 283.
23 Judgements are generally assimilated to propositions within model-theoretical semantics, which doesn't recognise judgements as a category distinct from propositions. One of the major motivations for adopting the one and only existing kind of 'inferential semantics', Per Martin-Löf's Intuitionistic Theory of Types (Bibliopolis, Naples 1984), may be that it keeps judgements and propositions apart. See also Wolfram Hinzen, The Semantic Foundations of Anti-Realism, Berlin 1998.
24 This point has been forcefully argued in Bas van Fraassen 1984, see fn.22, and in the same author's "Belief and the Problem of Ulysses and the Sirens", in Phil. Stud. 77, 7-37, 1995.
25 Even stylistic ones, see e.g. 153, top. See also the last bit of p. 198, or the quote on p. 199.
26 See "Realism and Anti-Realism", in M. Dummett, The Seas of Language, Oxford 1994, 474. See also "Truth and Meaning" (1985), ibid., 157, 161.
27 Research leading to the completion of this paper was made possible by the Swiss National Foundation of Science.
back to the page of events