The topic of my paper is an aspect of the relation between Robert Brandom's Between Saying and Doing (henceforth BSD) and his Making It Explicit (henceforth MIE), and the philosophical projects developed in them respectively. One of the many merits of BSD is that it provides a metaperspective on what is going on in MIE. One could almost jokingly say that the theory of BSD can be made to serve a goal we might call Making ‚Making It Explicit' Explicit. My aim is to critically reassess (in the light of BSD) what MIE has to say about the status of a particular kind of pragmatic explicative vocabulary, namely vocabulary for intentional interpretation (propositional attitude ascription), or intentional vocabulary for short. Thus I will offer an exercise in what the dead mighty Germans called thinking critically mit X gegen X, and which they regarded, rightly I think, as a most philosophical form of praise.
I want to begin with a very short recapitulation of what the project of MIE, as I understand it, actually is. Its central goal is to understand genuine intentionality (to borrow John Haugeland’s term), which is the capacity of systems to have non-derived mental states that have content expressible by that-clauses, i.e. propositional attitudes. Another way of putting the point would be to say that having genuine intentionality in this sense just means having thought. This is what we ascribe to systems whose practices we can only make sense of by taking the intentional stance to them. At least that is what Brandom seems to be saying when he writes that a system counts as exhibiting this specific type of intentionality »insofar as the differential responsiveness of the system to the results of its own performances is essentially mediated by states whose functional role in the feedback process can be understood only by taking them to be propositionally contentful, that is, by specifying them in an intentional vocabulary« (BSD, p. 183; italics mine). Now I know that it would not be literally correct to say this, but there seems to be a peculiar kind of VP-necessity claim involved here:1 In 1 In the present context, I assume acquaintance with the vocabulary of meaning-use-analysis that Brandom introduces in BSD. »VP-necessity« is not an official element of this vocabulary. order to count as genuinely intentional, one might be tempted to say, it is not enough that intentional vocabulary be sufficient to specify a system or the practices it is involved in, for intentional vocabulary is sufficient (can be used) to specify almost anything, including the behaviour of thermostats and iron bars. It rather seems to be the case that a system exhibits genuine intentionality if and only if it is necessarily specified by intentional vocabulary.2
What are the essential features of genuine intentionality which a pragmatist account would have to make intelligible? It seems to me to be in line with Brandom’s thinking to say that there are (at least) three such features:
(a) Rationality: Philosophers like Dennett and Davidson have stressed that to ascribe intentional states to a system is to describe it in the light of the constitutive ideal of rationality. To explain and predict a system’s behaviour in the light of propositionally contentful intentional ascriptions is to rationalize its behaviour, to make sense of it by ascribing states to it that would count as reasons for the system to behave in the way it does. Having genuine intentionality is therefore the same as being at home in the space of reasons, as Sellars put it. What constitutes the identity of a contentful state or expression is not its causal position in the natural world, but its inferential position in a rationally connected web ob contentful states.
(b) Objectivity: Propositional content can be characterized in terms of truth conditions, and even if a semantics does not use the notion of truth as its starting point, in the end it has to arrive at the point of being able to explain what it is that makes propositional contents represent states of affairs. This amounts to the same thing as being able to explain the representational character of propositional content. What is propositionally contentful necessarily has a representational aspect, it represents things as being a certain way. Nothing that does not display that aspect would be recognizable as expressing a proposition.
(c) Reflexivity: A system having thought, minimally, would be one whose behaviour can be interpreted as displaying states with content. But that is not enough. Not every understandable behaviour is understanding behaviour. Being minded demands not only displaying, but understanding content (meaning). The rationality and objectivity of the contents that are exhibited in the practices of the system, that is, must not exist only in the eye of a beholder. They must exist, be there exhibited in the practice, for the systems engaging in those practice itself.
Now the most fundamental theoretical commitment of MIE is a commitment to pragmatism. For a pragmatist, the question »What is genuine intentionality?« is transformed into the question »What is it that systems we treat as genuinely intentional are capable of doing that allows and necessitates specifying their practices in intentional terms?« That is, what are the necessary and sufficient conditions for a practice to count as instituting or exhibiting genuine intentionality? At this point, pragmatism aligns itself with the linguistic paradigm in the philosophy of mind, which claims that the practices on the basis of which we call some beings genuinely intentional are discursive practices. Having language, according to this tradition, is constitutive of having thought. Having language itself, in turn, is to be understood in terms of being 2 This would not have to be taken to mean that it would be impossible to specify a genuinely intentional practice in non-intentional ways, only that the non-intentional ways of specifying it would be conceptually dependent on the intentional one. capable of using linguistic symbols. Thus, the project of MIE is to specify what one has to do in order to count as speaking a language. I will call any set of discursive practices which suffices to confer on its practitioners genuine intentionality (including its central features specified above) an Autonomous Discursive Practice (ADP). Thus it is a condition of adequacy on a normative-pragmatist account of intentionality that it be able to say what it means for a practice to institute inferential relations, the objectivity of content, and the reflexivity of its practitioners, and thus makes it an ADP. 2
The aim of MIE, we can now say, is to make explicit what having language is. Now what does making it explicit mean? Brandom’s standard characterization of the relation between the implicit and the explicit is to say that, by making it explicit, we become able to explicitly say what before we could only implicitly do. This corresponds to the idea that making it explicit is codifying a piece of implicit knowing-how in the form of a piece of propositional knowing-that. But certainly this way of putting it gives only a very broad and unspecific idea of this central notion of the theory. E.g., is every act of talking about what we do an act of making it explicit, or is it being able to say what counts as doing what we do, or even becoming able to do whatever it is that we do by saying something? I suggest to understand BSD as offering the conceptual tools make this a little clearer in the following way:
(a) The fundamental idea, as I take it, is to define specification as a relation between a vocabulary V1 and a practice P2. V1 is said to be sufficient to specify P2 (to be VP-sufficient) if and only if it allows one to say what one has to do in order to count as engaging in P2. Thus, the instructions a tennis coach gives to a novice to the game, telling him what counts as a proper service, what counts as winning a set, and so on, might be said to specify the practice of playing tennis.
(b) Building on the notion of specification, we can then introduce a precise notion of making it explicit as follows: Let us introduce the practice P1 that is PV-sufficient to deploy the vocabulary V1 (that is, engaging in P1 counts as using V1). Further, let us assume that P1 can be elaborated (by algorithm or by training) from P2. Thus PPsufficiency holds between P2 and P1: PP-sufficiency holds »when the capacity to engage in one sort of practice or to exercise one sort of ability is in principle sufficient for the capacity to engage in other practices, or to exercise other abilities« (BSD, p. 33). If these conditions obtain, the vocabulary V1 used to specify P2 isn’t just any old vocabulary. It is dignified in a certain way, because everything one needs to do in order to deploy and understand V1 is in principle contained in P2 (in the sense that it can be elaborated from it). Thus I suggest that we should say that V1 makes P2 explicit if V1 is elaborated from and explicative of P2. I will call vocabulary of this sort explicative vocabulary.
(c) We arrive at a special case of the relation of making it explicit if we introduce the further assumption that P2 (the practice being made explicit) is itself a discursive practice, that is a practice sufficient to deploy a vocabulary V2. In that case, the resultant of V1 being VP-sufficient to specify P2, and of P2 being PV-sufficient to deploy V2, is a VV-sufficiency relation between V1 and V2, which means that using V1 allows one to say what one must do in order to count as saying the things expressed by using V2. In other words, V1 is a pragmatic metavocabulary for V2. We might call this metadiscursive specification: saying what counts as saying something else. Furthermore, since P1 is elaborated from P2, we can say that – in a way – we always already understood everything we need to understand V1 if only we understood V2. Thus, V1 is »semantically transparent« towards V2. Therefore it »can legitimately be appealed to as an auxiliary elaborating vocabulary in semantic analysis« (BSD, p. 49) of V2. In this case, V1 is not just a metavocabulary to V2, it is a metavocabulary that is elaborated from and explicative of P2. It stands in a pragmatically mediated VVsufficiency relation to V2 Brandom calls »LX«.3 Thus it is LX explicative vocabulary. If V1 is an LX explicative vocabulary for any discursive practice, it is universally LX explicative vocabulary
Now, let us apply these specifications to the project of MIE. We can then say that its aim is to develop a universal semantic metavocabulary, powerful enough to specify discursive practices as such. It must satisfy the following conditions:
(1) In order to be a universal semantic metavocabulary (one for discursive practice as such), it has to be VP-sufficient to specify any language whatsoever, that is, any Autonomous Discursive Practice (ADP).
(2) In order to be a legitimate semantic metavocabulary, it must be shown to be possible to extend, by elaboration, any ADP into a set of practices that are PVsufficient to count as deploying the semantic metavocabulary.
Conditions (1) and (2) guarantee that the metavocabulary is elaboratedexplicitating for any language, or universally LX explicative vocabulary.
(3) In order to be a philosophically illuminating semantic metavocabulary, it must be conceptually independent of the vocabulary of intentional interpretation that it is supposed to explain. It must not, on pain of circularity, make use of intentional locutions, e.g. specify certain performances within the practices as acts of expressing thoughts or as acts of saying that p, for it is precisely the question what it means for a practitioner to even have a thought, and therefore what it means for any of his performances to be expressive of a thought. What is needed, therefore, is an account that says what it is that makes a certain practice discursive in the first place, without presupposing that the practitioners are genuinely intentional beings. Thus, all in all, the project is, in Brandom’s words, »specifying in a non-intentional, non-semantic vocabulary what it is one must do in order to count as deploying some vocabulary to say something, hence as making intentional and semantic vocabulary applicable to the performances one produces (a kind of pragmatic expressive bootstrapping)« (BSD, p. 78 f).
In MIE, Brandom’s general term for universally LX explicative vocabulary –
vocabulary being explicative of any discursive practice – is logical vocabulary. Logical
vocabulary in this broad sense does not only comprise logical vocabulary proper
(logical operators, quantifiers), but also normative, modal, and intentional vocabulary.
In the terminology of BSD, the term »logical vocabulary« is reserved for logical
vocabulary in the narrow sense, which is just one species of the genus of universally LX
explicative vocabulary. I will stick to the broad use of MIE unless otherwise
indicated. Now, if logical vocabulary is explicative vocabulary for any ADP, it follows
that, for any ADP, the practical capacities sufficient to deploy logical vocabulary are
not PP-necessary for the capacity to engage in the ADP. In other words, an ADP (a
language) does not necessarily include logical vocabulary (although by definition any
ADP includes practices that are PP-sufficient for the potential elaboration of logical
vocabulary). This is what Brandom calls the »two-stage or layer cake picture of the
relation of logical to non-logical vocabulary«
Let’s have a closer look at this: Practitioners in a language game (in the sense of an ADP) have a certain know-how that consists in their being able to calculate the consequences of each move in the game for the normative statuses of every player, licensing or ruling out certain further moves. These calculations are what their understanding of the moves in the practice consists in. Brandom’s by now famous metaphor for this process is to keep the score of the game. Now scorekeeping itself is implicit in the ADP: The way the score is kept by a player just manifests itself in what normative attitudes any player adopts toward her fellow players, what she takes them to be entitled and committed to. No logical vocabulary, no device of explicit scorekeeping is needed for this practice to work. The introduction of logical vocabulary allows practitioners not only to treat others as being committed to certain norms, but to treat them as committed by saying that they are. But this, according to the official picture, only makes explicit the norms – hence, in the appropriate practices, the contents – that were, so to speak, always already there, implicitly conferred on performances by the roles they play in an appropriately structured normative practice. According to the Layer Cake Picture, ascribing intentional states to practitioners and propositional contents to their performances should be understood as making explicit the norms that were already implicitly instituted in a practice of the suitable sort.
Now I want to focus on a special kind of logical, i.e. universally LX explicative vocabulary, namely the locutions we use to engage in explicit intentional interpretation of speakers and thinkers, i.e. propositional attitude ascription vocabulary like »says that«, »believes that« and so on. I will call this type of vocabulary intentional vocabulary for short. Brandom’s claim is that this type of vocabulary is a type of logical, i.e. universally LX explicative vocabulary. More precisely, it is a type of pragmatic (as opposed to semantic) logical vocabulary, i.e. it belongs in a box with other types of normative vocabulary (as opposed to modal vocabulary and logical vocabulary in the narrow sense). It is explicative of the pragmatic interpersonal relations between normative attitudes and normative statuses that constitute any ADP, and »[i]n virtue of this explicitating expressive role, propositional-attitude-ascribing locutions deserve to count as logical vocabulary« (MIE, p. 499). It follows from this that what was said about logical vocabulary in general must also be true of intentional vocabulary. Thus, an ADP must be interpretable as being discursive even before intentional locutions are introduced into (elaborated from) it. The practitioners must be interpretable as saying things even before they can say that they or others say things. Brandom is very clear about that: »[S]corekeeping (and so linguistic practice generally) does not require that one be able explicitly to attribute deontic statuses – to say (assert) that someone is committed to the claim that p. The logical locutions whose expressive role is to make the adoption of such pragmatic attitudes explicit in the form of claimable contents – propositional-attitude-ascribing vocabulary such as the regimented ‚…is committed to the claim that …’ or its vernacular correlate ‚…believes that…’ – form an optional superstratum whose expressive role can be understood in terms of what is implicit in ground-level linguistic practice, but which is not required for, or presupposed by, such practice.« (MIE, p. 629) 4
If what I have said so far is correct, Brandom is committed to the following theses: (1) Something is a piece of logical vocabulary if and only if it is a piece of universally LX explicative vocabulary.
(2) If something is a piece of intentional vocabulary, it is a piece of logical vocabulary.
(3) An ADP does not necessarily include logical vocabulary. (4) Therefore, an ADP does not necessarily include intentional vocabulary.
I will now try to argue that claim (4) is to be rejected. What I want to claim is that the use of intentional vocabulary is a necessary element of any autonomous discursive practice itself. Pace MIE, these locutions do not form an optional superstratum whose expressive role can be understood in terms of making explicit discursive contents that were already implicit in ground-level discursive practice. Rather, these locutions are not expressive, they are co-constitutive of discursive commitments, and hence of any autonomous discursive practice. In other words, practices sufficient to deploy intentional vocabulary are PP-necessary for any ADP, just as, e.g., inferential practices.
The argument for this claim runs as follows: The Layer Cake Picture implies that it must be possible to account for the institution of objective deontic statuses and conceptual contents in terms of practical deontic attitudes of practitioners who do not yet have the capacity for higher-order normative attitudes. But understanding a practice as an autonomous discursive practice – attributing genuine discursive intentionality to the practice and its practitioners – is specifying a set of practices which confer propositional contents on performances for the practitioners. It is attributing a grasp, on part of the scorekeepers, of objective discursive commitments with representational contents. This means attributing to a scorekeeper a grasp of the difference between commitments another scorekeeper or she himself has acknowledged and those she has consequentially but unknowingly undertaken, that is, the capacity to attribute differing attitudes (acknowledging / undertaking) toward statuses (commitments). Without the capacity to attribute normative attitudes toward commitments on part of a scorekeeper, there is no genuine intentionality, because being able to attribute this – to others or to oneself – is what manifesting a grasp of the difference between objective normative statuses – what someone has bound herself by – and normative attitudes – what she merely takes himself to be bound by – consists in. Therefore, without concepts and expressions of these attributions and statuses (that is, without logical concepts in Brandom’s sense of the term) on part of the scorekeeper, genuine intentionality could not be ascribed. But possession of these concepts is manifested in explicit practices of rationally engineering, updating, and rectifying discursive commitments either socially (this is highlighted in MIE, Chapter 8) or self-reflectively (this is highlighted in BSD, Lecture 6). But either of these practices depends on the availability of practical performances that are in themselves acts of self- or other-ascribing discursive commitments. The capacity to explicitly ascribe normative statuses therefore is a necessary condition for there to be such statuses instituted in the practice. Thus, a scorekeeper can only be credited with genuine intentionality if she is engaged in logical practices of explicitly keeping the score, that is using explicit ascriptions of attitudes and commitments. But, since objective contents and propositional attitudes can only be said to be conferred on performances in a practice that institutes conceptual norms, norms that are objective for the practitioners who are bound by them, it follows that we can attribute the conferral of propositionally contentful normative statuses only to practices whose practitioners are full-scale interpreters, that is, fully logical at least in the sense of having the capacity to use intentional vocabulary.
Therefore, there could not be a grounding layer of pre-logical autonomous linguistic practice, instituting objective norms with propositional contents, while the second explicative layer merely made explicit the norms implicitly constituted on ground level. Rather, before the advent of intentional vocabulary, the norms on ground level could not be understood as instituting contents at all. No practice in which the norms are purely implicit could institute objective, i.e. propositionally contentful normative statuses and thereby confer genuine intentionality on its practitioners and propositional contents on their performances. Language in the full sense of the term cannot be understood in layer-cake-style. There may have been norms on ground level, but no conceptual norms (not even implicit). Before we made »it« explicit, »it« wasn’t genuinely intentional, that is propositionally contentful. What makes the norms conceptual, so to speak, is their explicitation. Going intentional, in other words, demands going explicit. This means that the attribution of genuine intentionality is justified if and only if the interpretees are full-scale interpreters too – interpreters who can attribute genuine intentionality themselves, who can, that is, explicitly keep the score. The answer to the question, then, what warrants – indeed, necessitates – taking and interpreting a practice as an autonomous linguistic practice, conferring conceptual states and contents on its practitioners and their performances, is that the practice be PVsufficient to deploy the very vocabulary that is used to specify it – intentional vocabulary.
Another, much more traditional way of making this point would be to say that if you want to account for full-scale objectivity, you need to account for full-scale subjectivity too. Put in Kantian terms, there can be no propositional thought, no thinking of objects, no genuine intentionality, without the conceptual capacities needed to attach the »I think« – and, of course, as Davidsonians would add, the »she thinks« – to one’s thoughts. Brandom, of course, being a Kantian and a Davidsonian, is the last person on Earth who needs to be lectured about this. But I wonder if pointing this out shouldn’t in fact make the conclusion I reached above more than palatable to him. 5
If this kind of reasoning is accepted, it follows that one of the premises of the argument ascribed to Brandom above has to give. I suggest that we should give up (3) and admit that logical vocabulary at least of the pragmatically explicative type is PPnecessary for any ADP.
Now, in reaction to a brilliant paper by Daniel Laurier, who makes more or less the
argument I have rehearsed here, Brandom is close to admitting that (3) should be
given up,4 but adds: »The claim that ascriptional locutions are in-principle
latecoming expressions that make explicit features of practices intelligible as
autonomously discursive in advance of their introduction could be acknowledge to be
mistaken without upsetting the general methodology or architectonic of the project.«
I want to cast a little doubt on the entitlement for this nonchalance. Let me admit right away that certainly, even if what I have said were correct, the walls of inferentialism would not come tumbling down. But I think the view presented here does have significant consequences, first for the general architectonic of MIE as it is presented in that book itself and also in some very recent of Brandom’s articles, which, if I am right, must count as serious self-misrepresentations; secondly for some basic claims of MIE and BSD. For reasons of space limitations, I will only mention the latter ones:
Obviously, the view presented here quite dramatically changes the conception of an Autonomous Discursive Practice. I have argued that any ADP must be PVsufficient to deploy intentional vocabulary, and that it follows from this that it must contain practices PV-sufficient to deploy logical – in the broad sense, i.e. universally LX explicative – vocabulary.
It also requires changing our minds about what explicative vocabularies of the
pragmatic type, i.e. normative and intentional vocabularies, are. Contrary to the
official picture, I would claim, they turn out not to be metavocabularies of any ADP, but
PP-necessary sub-practices of any ADP. Note that they can still be called explicative
vocabularies. For it can still be maintained that there is a set of basic practices they
specify from which the practice of using that type of vocabulary may be elaborated.
But before that, to put it simply, that set of basic practices was not an ADP. Thus
vocabulary of that type makes something explicit. We may even say that it makes
commitments and entitlements explicit. But before the advent of that type of
vocabulary, these commitments did not qualify as discursive commitments (thus, the »it« in
»making it explicit« here does not refer to something conceptual). Elaborating
practices sufficient to deploy intentional vocabulary is not introducing a metavocabulary
for an ADP, it is instituting one. Making it explicit, in this case, is therefore not, as
Brandom used to say, »a process of transformation of what in virtue of its role in that
process becomes visible as a content that appears in two forms, as implicit and then as
explicit«
Finally, let me add a speculation of what all this might mean for our picture of semantic (modal/logical in the narrow sense) explicative vocabulary. The argument I have presented only applies to normative and intentional vocabulary. Thus it would be unwarranted to generalize the claim that normative and intentional vocabularies are necessarily a subset of any ADP to all types of logical vocabulary. However, it is 4 In fact, he seems to be admitting it – as Laurier notes – in MIE, e.g. on p. 640. 5 Maybe this is behind his uncharacteristically defensive (but still, if I am right, unwarranted) claim in BSD that intentional vocabulary »may or may not be a sub-vocabulary of the autonomous vocabulary« (BSD, p. 183). certainly not implausible to assume, as John McDowell does,6 that the ideas assembled here, if correct, cannot leave untouched the account of that other type of universally LX explicative vocabulary, given how narrowly entwined they are even in Brandom’s own account of their relation in Chapter 6 of BSD. However, more arguments would have to be rehearsed in order to say something sensible about this, so it must wait for another occasion.