INTENTIONALITY, NORMATIVITY AND COMMUNITY
Facta Philosophica, 4, 1, 2002
For a number of philosophers today, there are quite close connections between the concepts of intentionality, normativity and community, which can, in broad outline, be formulated in the following way. The property that a creature has of having intentional states – states with an intentional content, such as, paradigmatically, beliefs, desires and other propositional attitudes, is not a purely natural property, a property that a system can have by instantiating properties such as neurophysiological structure, biological features or dispositions. It is also, at least in part, a property which involves normative features. Beliefs, thoughts, and propositional attitude contents are such that they can be evaluated as true, false, justified or unjustified, rational or not, and concepts and inferences are such that they seem to commit us to other propositions or to other concepts. Given that a creature can be credited with certain intentional contents, there are other contents that she ought to have. What is true of intentional mental contents seems also true of the semantic contents of linguistic acts, i.e of linguistic meanings: to mean something by an expression is to be able to appraise correct or incorrect uses of it. All these italicized terms signal the normative features that go with mental or semantic contents, which seem to be, to take up Sellars’ phrase, “fraught with ought”. Now this normative dimension seems to be such that it cannot arise for an individual in isolation. For a norm, as it is ordinarily conceived, or a rule, must be essential sharable: if it can be recognized by an individual, it must be recognizable by others, and liable to sanctions and corrections by others. For if there were not the possibility for the norm to be checked by others, if norm or rule-following where a purely private matter, then the norms or rules would not be objective, as they seem to be, if the normative features or intentional and linguistic contents are supposed to be real. So the norms have to be social, and the conditions of thought and meaning are themselves social. The normative features of thought and meaning, and their social character, seems to constitute the main reason that we have for rejecting naturalism, the view that our normative endowment is either merely an appearance or can be reduced to natural and causal properties.
Or so it seems, according to a line of thought which has been made familiar in contemporary philosophy by writers like Wittgenstein, Kripke, Sellars, Burge, Davidson, Putnam, Mc Dowell and Brandom, to name just a few. This line of thought, as Philip Pettit (1993), Charles Taylor (1985) and Mc Dowell (1994) have reminded us, is not new. It belongs to the lore of nineteenth century romanticism, with writers like Herder, Humbolt, or Hegel and of many philosophers of the hermeneutic tradition, who reacted against the naturalism of the Enlightenment. In many ways, the kind of holism that this tradition exemplifies, according to which having thoughts, having language, being rational, being human and being a member of a community are mutually interdependent properties, stands against what McDowell calls “bald naturalism”, the view that thought, meaning and action are only parts of the natural and causal order.
Now, is there a middle way between the kind of semantic and social holism that sprung from the romantic tradition and which survives today within contemporary philosophy and the kind of semantic and social atomism that is so characteristic of the opposing modern tradition? It is not the place here to describe and evaluate the chances of a sophisticated naturalism in these various fields. But I think that we can have a grip on what is at stake by concentrating upon the kinds of connections that I have just outlined between the notions of intentionality, or normativity and of community. What I want to do here is to try to understand these connections, and to try to show that they are not so easily connected as they seem. In particular, although I shall accept, with some qualifications, the view that there are intrinsic normative features of mental and semantic contents, I shall deny that these features are so obviously social as the preceding characterisation seems to imply. In pointing out that there is layer of normativity in thought which does not to depend upon the social character of the norms involved, I intend to block arguments to the effect that sociality and community are necessary or a priori conditions of thought.
In the first section of this paper, I describe a framework for articulating claims about “the normativity” of contents. In the second section, I examine some of the views which have led philosophers – in particular Brandom - to claim that the norms of thought are essentially social, and point out their difficulties. In the third section I confront these with Pettit’s (1993) views. In the last, concluding section, I sketch a taxinomy of the various sources of normativity in thought.
1.A framework : normative conditionals
As the italicized expressions mentionned above seem to suggest, it is difficult to deny that our talk about mental and semantic properties (content properties, for short) is pervaded with normative notions. Two sorts of questions, however, are raised by such remarks. (1) Granted that our current talk of ideas, thoughts, beliefs, meanings, and other mindful entities is pervaded with normative vocabulary, what is exactly the nature of this vocabulary and of the prima facie “norms” or normative properties which contents are supposed to have? One often talks about such norms as truth and beauty. Presumably, these are epistemic as well as aesthetic norms. But there are, in the usual taxonomies, plenty of other norms: there are moral, legal, social norms as well, but also, maybe, rules of etiquette, professional deontologies, etc.. Do they all enter in our normative talk of ideas or beliefs? Or is there a special subset of norms pertaining only to talk of mental and semantic content? (2) One often says that the prima facie norms of ideas are “constitutive”, and this seems to mean that they are in some sense essential or intrinsic. But is it the case? Given that our normative talk about ideas seems prima facie obvious, does it imply that it registers real and substantial properties of contents, or is a mere prima facie feature of our talk? The question here is similar to the familiar question about moral properties: our ordinary talk about moral properties of agents and acts ascribes, on the face of it, plenty of such properties, such as “good”, “just” or “obligatory”, but do these terms denote genuine properties of the acts or agents? Should we take this discourse at face value, as realists about moral properties suggest, or should we discard it as a mere appearance, as expressivists and other subjectivists about morality want to claim?
In order to get a grip on these questions, let us try to frame our prima facie normative ascriptions under the form of particular imperatives, which I shall call normative conditionals. Again, prima facie, the following conditionals seem to be correct for the meanings of particular words:
(M1) If “dog” means DOG, then one ought to mean DOG by “dog”
(M2) If “dog” applies to dogs, then one ought to apply “dog” to dogs
(M3) If “+” means “PLUS”, then one ought to apply “+” to the triple (68, 57, 125)
We could frame also normative conditionals for thoughts and beliefs, such as:
(C1) If one believes that P, then one ought to believe that P is true
Imperatives such as (M1)-(M3) and (C1) seem to be obvious, and, as Dennett remarks, not even worth spelling out. Indeed they seem to hold a priori, in the sense that an agent would in some sense contradict herself if she failed to accept the consequent when she accepts the antecedent. This is particularly clear for (C1), as the famous Moore’s paradox testifies: P, but I believe that not P. The usual gloss on the oddity of such a statement is that someone who asserts that P implies, or implicitates, that she believes that P, which contradicts her simultaneous saying that she believes that not P, or that P is false. The point is usually put in terms of sayings or assertions, but it can be put in terms of belief or thought: to believe that P is to believe that P is true, and an agent who would recognize that she believes that P could not fail to recognize that she believes that P is true. The presence of the “ought”, or of a similar deontic term, in the consequents of these conditional indicates that there is a normative constraint entailed by the very ascription of a meaning to a word or of a belief to a person. But what is the nature of this constraint? There are, as I remarked, all kinds of oughts: moral, legal, esthetical, etc. Presumably, the ought in question is not moral, nor legal. If “dog” means DOG in English, there is no moral obligation that I incur in using the word with this meaning, and there is no legal rule enforced which would, for instance entail that I would deserve legal sanctions if I do not mean DOG when I use “dog”. And there is no ethical, and not even any aesthetic, obligation for me to believe that P is true, when I believe that P. This is because ethical or moral rules usually pertain to practical judgments, or to acts that an agent is supposed to be free to do, whereas normative constraints such as (M1)-(M3) and (C1) are not such that the agent is free to do something. When someone means that P by “P”, or when someone thinks that P, he seems to have no choice but to conform to the rule of meaning, or to the recognition of truth. In other words, the normative constraints with meaning and belief conditionals are not constraints of the kind that we are free to break, because, unlike actions they are not voluntary or intentional. This feature traces back to the essential unvoluntariness of belief, and its asymmetrical nature with respect to desire: I cannot decide at will to believe that P, at least under normal circumstances (Williams 1971). This why normative conditionals of the kind above are different from normative conditionals of the form: If you promised to return the book , then you ought to return it. This marks off the realm of epistemic or cognitive normative constraints from the realm of practical or ethical normative constraints. This is also why we are tempted to call such normative constraints as those indicated by (C1) as constitutive: a belief would not be a belief if it were not considered by those who have it as a true belief, and a word would not have any meaning if it were not recognized by those who use it as liable to correct applications. So our normative conditionals, with respect to meanings and beliefs are different from ethical conditionals such as P1. This is what underlies our intuition that when I believe that P, there is nothing that I am committed to do. The “ought” in question is not a practical “ought”, but a theoretical one.
Now someone could try to argue that epistemic constraints can, in some cases, be practical ones. For instance one could argue, with some writers, that, contrary to the appearances, some beliefs could be got at will, that there is a special “ethics of belief” which would allow us to comply by specific ethical norms of thinking, or, in a Humpty Dumptian fashion, that all meanings are governed by intentions of the speakers, and hence that meaning that P is an intentional act. There are arguments to this effect, and we cannot exclude from the outset that they might be plausible. If so, theoretical norms would reduce to, or would have some overlap with, practical norms. This is a further question, pertaining to the general architecture of the normative domain. But I shall take here for granted that the constraints expressed by our meaning and content conditionals are specifically epistemic and cognitive.
If we grant this point, we can try to frame other normative conditionals in the epistemic domain. Here are some familiar ones:
(C2) If one believes that P, then one ought to believe that P is justified, or that
there are reasons to believe that P
(C3) If one believes that P and Q, then one ought to believe that Q
(C4) If one believes that Fa, then one ought to believe that there is an a , that
to be capable to believe there are Fs, that Fb, that Ga
(C5) If one believes that this is cat, then one ought to believe that it is a furred animal
Such conditionals are usually held to be obvious in virtue of fundamental properties of thoughts. C2 can hold because, it is said, a thought is a thought, a belief is a belief because it can be justified, or located somewhere in the “space of reasons” (Sellars). C3 can hold because beliefs are governed by an ideal of integration or coherence with other beliefs, which can be inferred from them, and from which it can be inferred (Davidson). C4 can hold because beliefs and thought are structured by the concepts that they involve, and concepts are subject to such constraints as Evans’ (1982) “generality constraint”. C5 can hold because concepts has a holistic structure, such that one cannot have a belief without having other beliefs, inferentially related to the concepts involved (again Davidson). Indeed, these constraints are those which have been held to be essential by many writers who claim that there is a constitutive normative dimension in meanings and mental contents. And it is the apparent obviousness of conditionals such as (C2)(C5) which underlies most of the claims that “content is normative”. These are examples of constitutive epistemic norms.
But it is also at this point that these “constitutive” normative features fail to be obvious. For is it really the case that when I think that P I ought to consider myself as committed to the justifiedness of P ? After all, I might just believe that P because I like it to believe that P. The norm in question is not obvious. Aren’t there whimsical, so to say, beliefs, which I could fail to recognize as justified, even though I have them? Suppose, for instance that I like it to think that my favourite team will win, although I have no good reason to think this. At best, the norm will be aesthetic, or patriotic, and it is not obvious that it has anything to do with such properties as truth and justifiedness. Or take the case of C3: am I obliged to believe that Q when I believe that P and Q? After all, if I have otherwise good reasons to discard the truth of Q, it would make no sense for me to infer this from my previous belief. Or, with respect to C5, am I committed, even in an epistemic sense, to believe that this cat is furred if I believe that this is cat?
The same point can be made from Lewis Caroll’s celebrated paradox of Achilles and the Tortoise (Carroll 1895). Achilles presents to the Tortoise a conditional of the same form as those that we have just considered:
(C6) If you believe that P, and if you believe that if P then Q, then you ought to believe that Q
But the Tortoise refuses to detach and to believe the consequent. And in one respect, she is right: why should she believe that Q, if she has otherwise good reasons not to believe that Q? As Gibbard reminds us (Gibbard 1996), conditionals with an ought (or some other deontic term) in their consequent but with no deontic term or no normative vocabulary in their antecedent are not valid. This is why normativity conditionals of the (C2)-(C6) kind have something fishy in them. An ought seems to be derived from an is, a transition that we have learnt to suspect strongly. Things, however, would be much better off, if the antecedent of the conditional contained a deontic term. And this is just what Achilles, at least on one way of understanding the story, seems to claim with respect to C6:
(C6’) If you believe that P, and that if P then Q, and given that if one believes that P, and that if
P then Q, then one ought to believe that Q, then you ought to believe that Q.
Here, normally, the Tortoise should detach. As well known, she doesn’t. But this is because she fails to grant the normative force of the ought contained in the antecedent. But this is another problem, to which Carroll’s paradox owes its specificity. If we leave this problem aside, and if we suppose that normative terms have normative force, there is nothing wrong in (C6’) as the expression of a normative constraint. There is nothing wrong with conditionals such as (C2)-(C3) if we reframe them in such a way that they contain a normative condition in their antecedents, to the effect that in normal circumstances, or in ideal conditions, if their antecedent holds then their consequent does too. For instance, in normal circumstances, if one believes that P and Q, one should believe that Q, or if one believes that if P then Q, and P, normally, one should believe that Q. Of course, as in the case above, one could have other reasons not to believe that Q, but in the absence of those other reasons, one should believe that Q. Normative conditionals, therefore, are correct –setting aside Tortoise-like reluctance to grant this - only if some mention is made in their antecedents of a constraint which is itself normative.
In other words, there is nothing wrong, and no appearance of an illicit derivation of an ought from a is in claims like the following :
(M1’) If one ought to mean DOG by “dog”, then one ought to mean DOG by “dog”
(C1’) In normal circumstances, if one believes that P, then one ought to believe that P is true
(C2’) If a rational subject believes that P, then he ought to believe that P is justified, or that
there are reasons to believe that P
(C3’) If a rational subject believes that P and Q, then he ought to believe that Q
(C4’) In normal circumstances , if one believes that Fa, then one ought to believe that there is an
a , that there are Fs, that Fb, that Ga
(C5’) In normal circumstances, if one believes that this is cat, then one ought to believe this is
a furred animal
But now the mention of normal or ideal circumstances in the antecedents seems to transform such conditionals into mere trivialities. They seem to be a priori and obvious indeed, but at the price of being platitudes. I believe, however, that it is to something like platitudes of this kind that the philosophers who have claimed that there is a normative dimension in meaning and mental content want to attract our attention to. Their point is that, provided that the circumstances are normal or ideal, provided that the normal conditions of ascriptions of meanings and thoughts are fullfilled, the very fact that a speaker means something by a term, or that a thinker has a thought, entails that he ou she ought to apply the word correctly, or to have other thoughts. Indeed it hardly amounts to the mysterious claim that one could derive a normative constraint from natural facts about speakers, or psychological facts about thinkers, since this normative constraint is already there. The point is that to recognize a thought as a thought is by itself to locate it in the “space of reasons”. Take, for instance, Davidson’s claim that mental concepts have a place within a “constitutive ideal or rationality” which has no “echo” in physical theory. In the terms that I have used, this amounts to saying something like:
(C7) If the rational norms of interpretation are in place, when someone believes that P, then he
ought to believe that P is true, and the propositions which logically follow from P, and have
the concepts which are holistically related to those in P, etc.
(and similarly with other kinds of normative conditionals).
If we reframe in this way the thesis that there are normative properties of meaning and mental contents, this thesis looses its mysteriousness. It amounts to one of the conditions of thought that Pettit points out in his treatment of this issue: that in order for a subject to be thinking subject, to be able to follow rules, there have to be normal or ideal conditions (Pettit 1993: 92; 1999). This gives us a sense in which the constraints indicated by the normative conditionals are “intrinsic” or “constitutive”: the sense in which if the normal or ideal conditions hold, then there are these constraints. But two questions arise, which echo our previous questions (1) and (2) above: what is the nature of the “norms” involved? and how can we know that these conditions arise, i.e that they are really intrinsic, but not contingent features of thought? Many of the writers who defend the view that meanings and mental contents are intrinsically normative claim that these conditions arise only when the thinkers are not isolated individuals, in other words that they arise only within a social setting. Correspondingly, they claim that the intrinsic norms in question are, by nature, social, or that the normal conditions of thought are social. To this thesis I now turn.
2. Brandom on thought and social norms
There are a number of arguments to the effect that there is an essentially social kind of normativity in meaning and mental content. Some are derived from what is often called a form of “social externalism” about thought, of the kind made popular by Burge (1979). Many are derived from Wittgenstein’s considerations about rule following, as interpreted by Kripke (1981), others from Davidson (1991) “triangulation” argument about the intersubjectivity of thought. I shall consider here only those of Brandom.
Brandom (1994) proposes an argument to the effect that meaning and mental contents are normative. His view is cast under two assumptions: a certain conception of meaning holism based upon an inferential role semantics, and an assertion theoretic conception of meaning. Briefly put the first view (which can, in contemporary philosophy, be traced back to Sellars 1963) says that the meaning of an expression in a language and of a concept in the thought of an individual is the “inferential” or “conceptual role” that it plays within a set of inferences involving that expression of concept. The second view, which is closely associated to the first, says that the meaning of a sentence is not a matter of truth conditions, but a matter of the conditions in which it is asserted and justified, in particular within inferences. This amounts to a form of meaning holism, since to understand a word or a concept an individual must be able to master a whole range of other concepts and to be able to recognise a whole range of sentences in which these words and concepts are employed, and of inferences involving these sentences. Now these inferential roles and assertion conditions could be conceived only as internal to an individual’s psychology (for instance as features of his “language of thought”). But this is not the way Brandom conceives of them. According to Brandom, not only are meaning and thoughts individuated by their inferential relations but these relations are themselves determined by social relations within a community. The individual who makes an inference between certain beliefs , or who uses a certain concept has to recognise that these inferences or concepts are correct or incorrect in terms of the proper circumstances of validation of the inferences or the correctness of applications set by the community to which he belongs. In other words, the individual has to take an attitude of commitment , which is a normative attitude, towards the concepts that she uses, the inference that she makes, and she takes these attitudes as those that others would themselves take. So an individual’s commitments are determined by their social interactions with other people, speakers of the same language, and sharers of the same thoughts and attitudes.
The basic idea is here very close to Wittgenstein’s conception of rules: the conceptual and inferential roles are not simply regularities within the psychology of an individual or reliable responses to stimuli. The very uttering of a sentence, and indeed the very having of a thought is the expression of a meaning or of a thought only if it can be evaluated as correct or incorrect in terms of appropriate conditions. Brandom distinguishes several levels within these evaluations (Esfeld 1999, 2: 25):
- commitments : an assertion of kind p commits one to a number of others. For instance asserting that it rains in Paris commits one to the assertion that the streets are becoming wet;
- entitlements : an assertion of kind p entitles you to other assertions. For instance, asserting that it rains in Paris entitles you to the assertion that the temperature will go down.
- precluded entitlements: assertions of kind p precludes an entitlement to other assertions. For instance asserting that it is raining in Paris precludes your from claiming that it is sunny.
Brandom takes our linguistic practice as a form of what he calls deontic scorkeeping: members of a speech community are the scorekeepers of each others’ commitments and entitlements.
This is a plausible description of the “normative pragmatics” within a speech community, and in fact it bears a number of similarities with some other accounts such as Grice’s, or Lewis’(1983) (who uses the very notion of “scorekeeping in a language game”), or of a number of speech act theorists’ views on meaning. The originality of Brandom position comes when he claims that the very individuation of a thought, and in fact the very having of a thought by an individual is dependent upon the kind of “scorekeeping attitudes” that the individual has towards others and that others have towards him or her. Thoughts are thoughts, and meanings are meanings, only because they are ascribed by individual to others, or interpreted. In this sense, Brandom’s position is close to Davidson’s “interpretationist” view of thought and language, or, as he himself calls it, “phenomenalist”: a thought is a thought only if it is can be ascribed by someone. When one ascribes a thought p to a person, one ascribes to her the fact of being committed to q, r , etc. , and one recognises her commitment. In turn, the person herself has the thought only in so far as she recognises this commitment. Again in other terms, the content of a belief is determined by the content of the beliefs that the person should have, where the “should” has to be shared by other interpreters of one’s speech and thought.
This “expressivist” conception of thought implies a form of social holism (Esfeld 1999, ibid.). An individual can only have beliefs within the context of her having other beliefs. The inferential context of a belief is the use of the sentence which expresses it by the whole community. So it is the whole community who, by its shared commitments fixes the content of assertions and beliefs. The order of explanation of mental contents adopted by Brandom is thus the following:
(1) The content of propositional attitudes is determined by the inferential roles
that they have within an individual’s psychology
(2) these inferential roles are normative (they are rules, which involve prescriptions
(3) the normativity of these roles derive from the deontic “scorekeeping”
attitudes of the community; inferential content is social content
(4) hence the contents of propositional attitudes are “normative” because they are social.
The same argument is run for meanings, and indeed, on this view the contents of thoughts derive from the meanings of the sentences they express; normativity of thought and publicity of meaning go hand in hand. Now, let us accept the inferential semantics (a), and let us accept also that these inferential roles are “normative” (for the moment without argument, but I shall spell out below some reasons to grant this). Are we bound to accept (c) and the conclusion (d)? I do not think so. (c) and (d) presuppose that commitments are shared. But what allows us to say that the “deontic statuses” and the commitments of our expressions and beliefs have to be shared by the members of a social community? For these commitments to be shared, the person who ascribes a certain deontic status to a semantic item must herself recognise this status, but, on Brandom’s view, she can recognise this status only by acknowledging herself , and by sharing the commitments that she describes. In this way, the ascriber of a thought makes “explicit” the commitments which are only “implicit” in the practice. As Brandom says : “ Practioners take or treats themselves and others as having various commitments and entitlements. They keep score on deontic status by attributing those statuses to others and undertaking them themselves.[my italics] (1994: 166). Communication implies that one shares these normative commitments, and that one takes responsibility just as others take responsibility for them (ibid.p.175). But is this right?
We can recast Brandom’s conception of the normativity of meaning and mental content by considering normative conditionals of the kind above, for instance:
X says (means, thinks) that p ® X should assert p
Such conditionals express ascriptions of thoughts and meanings by an interpreter. But as we saw, such conditionals would be problematic, for the consequent contains a deontic term (“should ”) whereas the antecedent does not. By what feat can the simple description of a thought entail anything “normative” about it? As we saw nothing would be wrong if the antecedent itself contained already a normative vocabulary, in the manner of:
(1’) X says (thinks) that p and is committed to p ® X should assert p
But if an interpreter, who attributes contents of the form (1’) recognizes commitments on the part of the person interpreted, does it follow that he himself must share the commitment, that is :
(1’’) X says (thinks) that p and is committed to p ( and I, the ascriber, am
committed to p) ® X should assert p ?
The parenthesis, which Brandom’s quotation above seems to imply, is by no means justified. It is one thing to attribute a certain normative attitude, and it is another thing to share it. There is a difference between describing something as a normative practice, as an external observer, and being engaged, oneself, in that practice, that is sharing these commitments (Gibbard 1996). I submit that the reason why Brandom moves to the conclusion that propositional contents are normative, categorically or instrinsically, is that he detaches the consequent in (1) and (1’), which he is not allowed to. Hence the so-called “normativity” of mental content is not intrinsic, or categorical, but only extrinsic, or hypothetical.
One could think that Brandom attempts to reformulate, in his own terms, one of the necessary conditions for the interpretation of thought that Davidson (1984) proposes, namely something like the “principle of charity”: thoughts could not be interpreted, if interpreters did not presuppose that they share a large amount of true and coherent beliefs with the people that they interpret. In this regard, interpreters would have to share what Davidson calls “norms of rationality” with the individuals that they interpret. But Davidson never goes so far as to say that both must share the same norms. His own “norms of rationality” only play a very general role in the practice of interpretation, and they are not intrinsic to the contents attributed: they have a very high profile, which does not imply anything like rules, conventions, or “deontic statuses” shared by interpreters. It is thus is far from clear that it is a necessary or a priori feature of thought that they have to social.
3. Pettit on the conditions of intentionality and thought
On the face of it, Pettit’s approach in The Common Mind bears many similarities with the views that I have examined in the previous section. Like Kripke, he makes of rule following one of the basic conditions of thought. Like Davidson, he insists on the fact that these conditions imply the existence of interaction with others, mostly in contexts of communication. And like Brandom, he emphasises that the normative constraints on thought are such that they can arise only within a public space of reasons (see Esfeld 1999). He takes these points to lead to a form of social holism. But in many respects Pettit’s views depart from some basic assumptions of the upholders of the normativty of meaning and mental content, and a somewhat different picture emerges.
First Pettit distinguishes between intentionality and thought, between being an intentional system and being a thoughtful creature. On the one hand, to have intentional states, and to be an intentional agent “is to interact with a perceptual environment under the control of intentional regularities, at least in favourable circumstances; in particular...it is to interact with the environment in accordance with minimally rational regularities of a theoretical and practical kind.”(1993: 54). Broadly speaking, to have intentional states, in particular beliefs and desires, is to instantiate a functionalistic psychology (ibid, 43). Animals and robots, in this respect, can instantiate intentional regularities. On the other hand, to be a thinking subject is not only to be able to have intentional states, but to be able to have intentional states about the contents of one’s intentional states. Pettit calls this “intentional ascent”(ibid., 60). Other terminologies talk about second-order intentional states or metarepresentations. In particular “an intentional system is a thinking or thoughtful subject just in case it has the capacity to act intentionally with a view to promoting in itself conditions such as those of evaluative and deliberative, inductive and deductive, rationality. Otherwise the system is non-thinking, or thoughtless.” (1993: 57). This distinction is not only important in itself, but also for our purposes here. For many writers who hold the thesis of the normativity of mental content either do not make explicitly the distinction between first-order and second order intentional contents, or presuppose that the notion of mental content which has the alleged normative features that they use is the second-order one. And in fact, up to now, I have myself ignored this distinction. A example of the first kind is Brandom, who simply equates having an intentional state with content and having this state in virtue of normative features. For instance he says: “Intentional states and acts have contents in virtue of which they are essentially liable to evaluations of ‘the force of the better reason’.”(1994: 17). An example of the second kind is Davidson, when he famously argues that in order to have a belief one must have the concept of a belief (1982). Now if normativity and rules come in somewhere, it is presumably not at the level of intentionality per se, but at the level of thought, when a subject becomes conscious of his intentional states and practices “intentional ascent”. For a subject who merely instantiates intentional regularities in his beliefs and desires need not take the regularities that he instantiates as rules or norms for her thinking and acting. The distinction between regularities and rules is quite important here. Regularities can be instantiated quite passively by a creature with intentional states, without this creature getting a perspective upon her intentional states. But thinking, according to Pettit, requires more, it requires something like being able to “shape” one’s own intentional regularities:
“In order to pursue the shaping of our intentional states, we have to be able to identify propositions and propositional elements as norms by which to guide ourselves. These norms give us our bearings as we seek out the rational things to believe, the rational things to desire, and the rational things to do.” (1993: 190)
What are the conditions of this “shaping”? At least it involves intentional “ascent”, and the capacity to understand what it is for the first-order states that one has to be rational (1993: 58). In other words the subject must be able to view himself as instantiating the regularities, not simply following them (remember Quine’s (1972) distinction between a behaviour that fits a rule and a behaviour that is guided by a rule). But sometimes Pettit seems to say that thinking is something like an intentional act: “to think is to act intentionally with a view to having intentional states that satisfy certain conditions: for instance the conditions of deliberative, inductive and deductive rationality, and ultimately, a condition involving truth.”(1993: 67, see also p.68). Do that mean that we could, in some sense, come to believe something intentionnally at will? No, for Pettit distinguishes here between believing that P, which is unintentional and judging that P, which can be. (1993: 59-60). In other words, thinking comes at the level whereby a subject not only entertains a certain intentional content, but also is able to take an attitude towards it. Pettit’s distinction here is in the neighbourhood of the distinction that some writers have made between believing and accepting , but I shall not deal with this here (Cohen 1992).
Where does it leave us with the thesis that there is an essential normative dimension in mental content? This thesis is often found baffling, as we saw above with some formulations of our normative conditionals: why on earth would the mere entertaining by a subject of a certain belief content commit her to anything, and be “normative” in this sense? Why should the mere fact that someone has a thought imply that he or she ought to think or to do something? The answer is that this mere fact implies nothing of the kind, and that there is no derivation of an ought from a is. In this sense, our normative conditionals are just false. But from the moment where the subject is able to take a stance on this fact, and to be able to understand what it requires, i.e when he becomes a thinking subject, he cannot but recognize the conditions under which his first-order intentional states can be optimal (even if they actually fail to be so). He enters the space of reasons. In that respect the normativity thesis is correct, and the normative conditionals formulate a priori requirements on thought.
But the preceding reasoning, would probably be rejected by Brandom. For he says that it can never be a mere fact, natural or causal, that someone has a belief. On Brandom’s view, it is normative, or evaluative, “all the way down” (Brandom 1994: 44). And it may be one of the implications of Kripkenstein’s view: following a rule is never a mere fact. But I think that it is at this point that there is also a difference betwen Pettit’s approach and that of the writers examined in the previous section. For he claims that there can be a definite “fact of the matter” about which rule a subject is following. This involves his own solution to Kripke’s rule following challenge.
Pettit formulates the following genetic account of the formation of rules and thought, leading to his own version of social holism (Pettit 1998: 176-182; see Esfeld, 1999, 3: 10):
1.Being able to think involves the capacity to use voluntary signs in representations of how things, as the subject believes, are.
2. In order to represent a property – or other entity- voluntarily, the thinker must be able to identify that property and must be able to see it as something that they can try, fallibly, to register.
3. Thus the capacity to think requires the thinker to have at best a consciously fallible criterion for determining whether or not the property is present in a given case.
4. How does a human thinker register the presence of a property that in their repertoire is semantically basic, i.e not defined by other properties? They cannot use the fact that they are disposed to apply the predicate in a given case as a criterion for the presence of the property; they could not even think of the property as something that they can try but fail to register.
5. A human thinker might be able to in principle to use an idealised version of this predicative disposition in a criterial role, and so there is no argument in principle against the abstract possibility of the solitary thinker.
6. But in actual fact, the human thinkers are not solitary in that way; they use a socially shared, predicative disposition as an identifying criterion.
7. The social holism thus supported is a deep, not superficial, changeable feature of human thinkers; it explains how human conversationalists can claim to know – to know immediately, not to derive – what they each have in mind with the use of certain words.
Pettit grants that our use of signs for thought is based upon a “capacity” or a disposition, so he rejects the Kripkensteinian bar on such notions. He also supposes that individuals who share the same biological equipment share, roughly, the same dispositions. A lone individual is able to follow rules in idealising his own dispositions or capacities. But individuals also share dispositions to cooperation, which are reinforced by sanctions of the community. When these conditions are in place, rules are followed, and we can know which ones are. Not only does such a view actually use the very notion of disposition in the genealogy of rule following, but it also accepts that solitary individuals can follow rules. Now as a matter of fact we are not solitary: we engage in rule following within communities. So there is no necessity, either causal or a priori, of the existence of the community for the existence of rules (contrast with Esfeld 1999, who holds that it is a matter of necessity). In other terms, Pettit’s “social holistic” view is merely a contingent thesis, which he takes to be compatible with the collectivist version of social holism and with a form of individualism, which limits the attribution of thoughts and meanings to a solitary individual. This is, in my view, a coherent combination, but it limits seriously the argument which underlies the so-called community view attributed to Wittgenstein, for it is not clear that this view is not collectivist. The transcendental argument according to which thought is possible only within a community is blocked, since this condition becomes, in Pettit’s hands, only contingent. So the “transcendental”, a priori, character of this argument is lost. Moreover, the way Pettit understands the emergence of norms form a biological disposition to cooperation does not prevent us from deriving norms in a biological way, in the style of contemporary accounts of the emergence of cooperation (Axelrod 1984). Norms can, in the end, be reduced to idealised dispositions. So it is hard to take such a genealogy as an argument in favour of the view that there is something essential or intrinsic in normativity, which would make it irreducible to natural dispositions.
It seems to me that this analysis gives us a coherent combination of theses, and that it alleviates a number of the difficulties that we have found with the view that there is an essential normativity in meaning and content and that this normativity is essentially social. First, because we have to distinguish intentionality from thought, it does not say that all kinds of contents are “normative” or involve normative properties. Only those which occur at the level of thought involve these normative elements. Second, it does not claim that there is a transcendantal argument to the effect that intentionality per se, nor even thought is essentially social. As Pettit says, his “interactive thesis” for holism (the thesis that a human being can follow a rule on the basis of interpersonal or intrapersonal interaction with others) is weaker than the kind of interactive thesis that the community view, the triangulation argument, or Brandom’s social-inferential view of thought imply, for it is only the disjunctive thesis just mentioned: “There is no formal incoherence in the idea that [the commonability of rules] should be realised in the absence of other people.” (1993:181).
But while these claims seem to mark off Pettit’s view from those examined in section 2, he nevertheless puts forward a modal claim: “The rules followed by a human thinker are not followed on the basis of such intrapersonal interaction alone; they must [my italics] be followed on the basis involving interaction with others” , and he adds that this interpersonal interaction is bound [my italics] to involve social interactions”.(1993: 181) These modal terms seem to reintroduce the necessity thesis, but Pettit also seems to retreat from it at other points by claiming that “all the evidence suggests that as a matter of fact[my italics] the rules that people follow in thinking are commonable” (183) and that “interpersonal interaction facilitates [my italics] commonability”(188). Here “facilitates” could be opposed to “is a necessary condition for”. I prefer to side with the second, rather than the first kind of formulations. Why?
My reason from saying this is that it seems to me that when Pettit comes, in the chapter 4 of The Common Mind, to his argument in defense of social holism and when he formulates his view of the emergence of thought as the “shaping of our intentional states” (1993: 190 as quoted above), Pettit tends to overlook one point which he had raised in his discussion of the requirements for the passage from intentionality to thought (p.54-57). There he tells us that there are certain conditions that it is “in an agent’s interest” that his intentional states should fulfill (54), such as: having true beliefs, having rational beliefs, shaped by inductive evidence and deductive requirements, having desires founded on his true beliefs. Indeed, these “conditions” are just those which derive from the functional architecture of an intentional being, and in his this sense, they are just causal or natural requirements: this is the way we, like other intentional creatures, have been shaped by Mother Nature. But Pettit also tells us that it is possible for an intentional system to act intentionnally with a view of having – or at least increasing the probability it will have – intentional states that do satisfy conditions like those of evaluative and deliberative, inductive and inductive, rationality”(ibid. 57). As we saw, it there that thought, and normativity, comes in. The “conditions” in question derive from our functional psychology: beliefs are such that they “aim at truth” , desires are such that they aim at satisfaction, and have a reverse “direction of fit” from that of desires (see e.g. Smith 1995), and both are such that they combine in appropriate ways to cause actions, in normal circumstances. But this is just the way we have formulated at the outset our normative conditionals (C1 for instance). I do not see any reason why we could not say that a subject who meets these conditions obeys certain basic norms governing his intentional contents. Of course these “norms” are not such that the agent could break them, or choose not to follow them. They are just the conceptual, or epistemic norms that I alluded to in section 1. With Peacocke (1992) we can call them “normative liaisons” which govern the contents that a subject entertains, and we can specify them, in his way, by laying down “possession conditions” for the basic concepts that they exemplify. They may not be conscious, and can be implicit or tacit. An intentional subject will become a thinking subject when he begins to be disposed to develop habits of inference according to those norms, and from these habits he begins to understand them, and act consciously according to them. So I would say that in between the layer of intentional states that a creature can have, and the layer of conscious thought that is characteristic of the full blown thinking subject, there is an intermediary layer of constitutive norms of a normal epistemic agent. The recognition of this intermediate layer is not only important in itself, but also for the argument that I have tried to put forward, that there is a level of thought which is not necessarily social. For I see no reason why an intentional creature, who becomes conscious of the epistemic norms that she tacitly uses, would have to interact with others to develop them. She might as well develop them on a purely intrapersonal basis. They need not even be linguistically shaped. And they need not even imply intentional ascent, or the conscious self-attribution of beliefs. Let us try to illustrate these points.
You come back home in the evening, and look after your keys in your pocket. You don’t find them. You remember suddenly that that you have slammed the door behind you when you left your flat in the morning and you infer that you have left them on your desk before leaving in haste. In this little familiar episode, you have a belief, that you keys are in your pocket, which is disconfirmed by your evidence. You store from memory an information that is relevant to the falsity of the first belief, and you infer another one. Although I agree with Davidson’s description of such episodes as involving beliefs about beliefs (you believe that P, discover that not P, hence are surprised that not P) , and in this sense a surprise (Davidson 1982), but I do not see why the subject, in such a case, need self ascribe to himself a belief in the form: “I believed that P, but now I come to believe that not P, hence I should believe instead that not P”. The processes of inference from one belief to another, with the aid of memory, involve a belief revision, and it involves normative liaisons of support (the absence of the keys in your pocket is evidence that they are not there, the remembered belief that you slammed the door is a basis for your inference that you left your keys on your desk, etc.). But it need not involve “intentional ascent” in the sense in which the subject would portrait himself as believing that P. And the beliefs, although conscious, are quite automatic, and not linguistic. Actually even higher animals can have similar primitive forms of belief revision. So at least in this form, a primitive form of normativity can enter in their beliefs, even if they certainly cannot think in Pettit’s sense. Now suppose that an individual reaches intentional ascent, that he comes to “shape” his own intentional states and the norms of belief confirmation and revisions. He will be able to reflect on his own beliefs, and to evaluate them. But is it necessary for that that he communicates with others? No, he can come to compare his former beliefs with his later ones, and, so to say, assess his former self in the light of his later self. I do not see why such a capacity could not be reached by a lonely individual. 
Actually, Pettit need not disagree with these points. When he describes what he calls the “conversational stance” (Pettit & Smith 1996), he acknowledges that there are, for instance for beliefs, “belief-relevant norms that apply, by everyone’s light to any conversational interlocutor” and which he characterises exactly in the same way as the cognitive norms that I have spelled out here. (1996:433). He adds, in a way which is quite congenial to my point here :
“Such norms are not like the conventional norms of behaviour that might be recognized in a given club or circle: they are not norms, such that subjects can be imagined deciding whether to embrace them or not, depending on their contingent atttitudes or aligments. To be a thinker who believes in certain determinate contents is to be subject to norms like: believe that P if an only if P, believe that P if an only if all the evidence points to P, and believe that P when P is entailed by some of the things you believe, and is not inconsistent with anything else you believe! The relevance of such norms – such evidential norms as we will call them- is inescapable.” (1996: 433-34)
I welcome this, but where I disagree is with the suggestion that such norms are made possible by the activity of conversation, and that they are in place only in thinking. Pettit in fact acknowledges that “conversation”, as he uses the term, need not imply exchange between different people, but that it may involve only one person.( ibid.432). But the word “conversation” suggests something linguistic, or some form of “discursive scorekeeping” in the manner of Brandom. And I have claimed that these norms might well be in place before intentional ascent and rule following. So I suggest that Pettit might have recognized, at an intermediary level between intentionality and thought, the relevance of such norms of belief, and made clearer that these cognitive norms are not intrinsically social. Such norms form the basis of our claims to knowledge, a paradigmatically normative notion. But even if we tend to think of knoweldge as a social, it need not be.
The confusion between these two concepts of norms, the “social” and the “cognitive ” one is responsible for the fact that norms are usually taken only in the first sense of rules intentionally followed, and upon the model of practical norms, which have consequences for action. Indeed a large number of norms are of this type, and language conventions are thought along these lines. But with respect with what I have called conceptual or cognitive norms, these characteristics are not in order. It is not a matter of choice for me, when I have belief, that it aims at truth, when I make an inference, that my beliefs be implied by others. This clarifies, I think, one of the questions that I raised at the beginning if this paper: in what sense does my having of a thought commit me to anything? The answer is that, in so far as my thought is simply a psychological attitude that I have, it is a mere fact that I have it, and it commits me to nothing. But in so far as it is a thought content, it has constitutive conceptual and inferential characteristics, which, when I attend to them, I can’t fail to acknowledge. Such characteristics hold as a matter of conceptual necessity.
If this is correct, there are, as the normativity thesis says, certain norms which are intrinsic to thought. But, contrary to what the full blown holistic versions of Wittgenstein’s arguments imply, these norms are not intrinsically social. Such a conception, however, is not incompatible with Pettit’s, for there is no reason why our constitutive norms for thought and concept would not have their origin in natural dispositions to act (according to his “ethocentric” point of view) which could, under specified conditions become social, and which would, as a matter of contingent fact, depend upon our social interactions.
4. Conclusion : three grades of normative involvement
I have tried to argue here that although we must accept the view that there is a normative dimension in meaning and mental content, we must not accept all the implications which are usually associated with this view, in particular the implication that the normative properties of meaning and thought are necessarily social, in the strong holistic sense (indeed almost collectivist) that some of the versions of Wittgenstein’s rule-following considerations (especially Brandom’s) tend to promote. Contrary to appearances, this seems to me to be a view which is compatible, if not sometimes to the letter, but at least to the spirit of Pettit’s minimal social holism about thought. The point is, as a matter of fact, having a mind is essentially a social matter, but that it need not be.
By way of conclusion, I would like to sketch a quick picture of the sources of normativity, by distinguishing three grades of normative involvement, in the manner of Quine’s (1953) famous distinction of three grades of modal involvment in logic. At the bottom level, there are intentional systems, which belong to nature (or sometimes, as with robots, to artefactual nature). No normativity nor rules arise here, there are just regularities of a causal kind, instantiated in our functionalistic psychology. But the features of this psychology, the causal roles that our mental states occupy, rest on normal conditions which a creature can come to understand and to which it can try to conform. This provides for a first layer, or grade of normativity, which comes with the epistemic liaisons of our concepts and attitudes. At this level, an individual is not necessarily conscious that he follows rules, or conforms to norms. But he can reach “intentional ascent”, and become conscious of these first-order norms, by becoming a thinking subject. There he reaches not only mentality, but also thought. This means that he reaches a second grade of normative involvement. Here full blown normativity enters. But he can also, at a later stage, come to compare the norms and rules that he follows. For instance, he may question the rule that belief must be shaped by evidence, and he may accept that in some cases, evidence is not a necessary condition for belief, by attending to such considerations as those that James puts forward in The Will to Believe, or as those that Van Fraassen puts forward when he claims that scientists should attend to the conditions of acceptance of their theories. In other words, the subject might confront his epistemic principles with others, and deliberate about which are optimal. He will form belief and desire policies. At this third grade of normative involvment, we reach higher-order norms or principles. It is a matter of controversy whether these higher order principles will converge under fundamental norms, such as the search for truth, or not. It seems difficult to envisage that normative discussion, at this third stage, can occur without language and communication, and without a social setting. But if the progress up to the higher levels of normativity is probably a progress in sociality and commonability, it is not clear to me that, at the lower stages, sociality and community are necessary conditions for thought, although they probably, as a matter of fact, are.*
Axelrod, R.1984 The evolution of cooperation, New York : Basic Books
Brandom, R. 1994 Making it Explicit , Harvard : Havard University Press
Burge 1979 “Individualism and the Mental”, Midwest Studies in Philosophy, 73-121
Carroll L ( C.L. Dogson), 1895 “What the Tortoise said to Achilles”, Mind , 104, 416, 695-711
Cohen, J. 1992 An Essay on Belief and Acceptance, Oxford, Oxford University Press
Davidson 1982 “Rational Animals”, in Le Pore, ed. Actions and events, Blackwell, Oxford
1985 “A Nice Derangements of Epitaphs” in Le Pore ed. Meaning and interpretation
Blackwell, Oxford, 433-446
1991 « Three Varieties of knowledge », in A.P.Griffith , ed. A.J. Ayer : Memorial
Essays, Cambridge, Cambridge University Press, 153-166
Dennett, D. 1995 Darwin’s Dangerous idea, New York: Little Brown
Engel, P. 2000 “Wherein lies the Normative Dimension in Mental Content?” Philosophical Studies, 100,
Esfeld, M. 1999 Holism in the philosophy of mind and in the philosophy of physics
Habilitationsschrift, Universität Konstanz
Gibbard, A. 1996 “Thought, norms and discursive practice: a review of Brandom’s Making it
Explicit, Philosophy and Phenomenological Research
Horwich, P. 1998 Meaning, Blackwell, Oxford
Kripke 1981 Wittgenstein on rules and Private Language, Blackwell, Oxford
Lewis, D. 1983 “Scorekeeping in a language game” , in his Philosophical Papers, vol. 1, Oxford,
Oxford University Press
Mc Dowell, J. (1994) Mind and World, Harvard University Press
Peacocke, C. 1992 A Study of concepts, Cambridge, Mass, MIT Press
1996 “Entitlement, self-knowledge and Conceptual redeployment”, Proceeedings of
the Aristotelian Society, XCVI, 117-158, repr. in Peacocke 1999.
1999 Being Known, Oxford : Oxford University Press
Pettit , P 1993 The Common Mind , Oxford : Oxford University Press, second ed 1996 (quoted after
the 1st ed)
& Smith, M. 1996 “Freedom in Belief and Desire”, Journal of philosophy, XCII, 429-449
1998 “ Defining and Defending Social Holism »,Philosophical Explorations, 1, 169-84
1999 “A theory of normal and ideal conditions”, Philosophical Studies, 96, 21-44
Quine, W.V.O. 1953 “Three grades of modal involvement”, in The Ways of Paradox, Havard: Havard
University Press, 1976
1972 “Methodological Reflections on current linguistic theory”, in Davidson,
D. and Harman, G. eds, Semantics for natural languages, Dordrecht: Reidel
Sellars, W.1963 Science, Perception and Reality, London: Routledge
Smith, M. 1994 The Moral Problem, Oxford : Blackwell
Taylor, C. 1985 Philosophical Papers, II, Cambridge : Cambridge University Press
 See for instance, Dennett’s remarks about our ordinary notion of “idea” (Dennett 1995: 363-364)
 The notion of a normativity conditional is elaborated here after the remarks by Gibbard (1996) and Horwich (1998, ch.8), who formulate in a similar fashion the ascriptions of normative properties to meanings and mental contents. See Engel (2000)
 This reasoning is expressed, for instance, by Brandom 1994, p. 624 : “Only we discursive (that is concept mongering) creatures can take ourselves and others to be bound by the norms that are our concepts.
This is the idea that is followed out in the deontic scorekeeping pragmatics presented here. The idiom in which the account of discursive commitment is expressed is normative throughout. Propositional contents are understood in terms of their explanatory role in specifying proprieties of claiming, judging, and inferring – in general in terms of the roles they play in the game of giving and asking for reasons. what it is for something to state of express a fact is explained in normative terms, and what it is for something to be stated or expressed is explained in turn by appeal to that practice.”
 See also Gibbard 1996. Esfeld (199, 2: 279) denies that Brandom has to say that the ascriber must share the commitments. But the quotation above seems to imply it, and it seems necessary for his view that the “normativity” of thought and meaning is “social”.
 Actually Davidson denies that such notions must play a role in understanding a language (Davidson 1984) . The shift comes probably from the fact that Davidson’s theory of interpretation is formulated in terms of a truth conditional semantics, rather than in an assertion-theoretic and inferential role semantics.
 Pettit himself alludes to Peacocke’s possession conditions p.108, 213 as “congenial” for his purposes.
 A similar example is given by Peacocke (1996) p.129.
 Certainly intentional ascent and thought would involve what Peacocke calls “ conceptual redeployment “ (Peacocke 1996),i.e the very same concepts that he deployed with his first-order beliefs. But Peacocke point is also that such redeployment can also occur in a more primitive form of thought, and that iti need not involve language.
 Thus I agree with Peacocke on the following point: “Social elements are not fundamental in an account of the point and importance of knowledge. Knowledge as something which involves a relation of the sort I have outlined to judgments which are both rational and successful is something which is already of value and of importance to us. The value can be elucidated, and indeed exists for the individual thinker considered in isolation, without mentioning the importance for us of saying that someone else knows something.” [Peacocke 1999: 36]
 It seems to me that Horwich’s (1998) discussion of the “norms of language” conflates these two senses of the notion of norm which I intend to distinguish. Horwich wants to show that the so-called norms of meaning and of thought are not intrinsic nor essential, but derivative and contingent from some basic aims of communication and thought, which are factual (for instance it is useful to have true beliefs). But what he says applies to intentional norms that people follow, not to the constitutive epistemic norms that I deal with here.
 At this point, the subject can become responsible for his beliefs. Pettit actually recognizes this third grade of normative involvement in Pettit and Smith 1996.
* For comments on this paper and on previous versions I am grateful to Claude Panaccio and Martin Montminy, with whom I discussed this at the 1999 congress of the Canadian Philosophical Association, to Philip Pettit, Maureen Sie, Marc Slors, Michael Esfeld, Jan Bransen, and other participants at the conference “The Common Mind” in Rotterdam, July 2000, and to the participants at the Summer School on “Normativity” in the University of Parma, July 2000, especially Paul Boghossian, Allan Gibbard and Kevin Mulligan. Many thanks too to Paul Dumouchel,, Uskali Maki, Paolo Leonardi and Marco Santambroggio who gave me the occasions to elaborate these ideas.