‘Habermas’s Moral Cognitivism and the Frege-Geach Challenge’ 2005, European Journal of Philosophy13: 3, 319-45.
Habermas’s Moral Cognitivism and the Frege-Geach Challenge
James Gordon Finlayson This paper consists in two parts. In the first part I show that the Frege-Geach challenge, an objection to certain non-cognitivist moral theories, also applies to Habermas’s discourse theory of morality, even though Habermas is an avowed moral cognitivist. The problem lies with Habermas’s conception of cognitivism and the thesis there is only an analogy between truth and rightness, on which it rests. In part I this thesis is assessed in the light of Habermas’s original theory that truth can be understood pragmatically as a validity claim, and in the light of his recent revision to his original conception of truth. In the part II I consider four possible responses to the Frege-Geach challenge that Habermas might make, and provide a kind of cost-benefit analysis of each. I assess whether, and if so how, Habermas can meet this challenge to his moral theory without sacrificing any of the central tenets of the programme of discourse ethics.
1. Habermas’s Conception of Moral Cognitivism. For over thirty years Jürgen Habermas has been developing and defending a normative moral theory, the discourse theory of morality. It is the central component of the wider programme of discourse ethics, which in turn is a component of his social theory.1 From before Moral Consciousness and Communicative Action through to his debate with Hilary Putnamand the recent collection of essays entitled Truth and Justification Habermas has not wavered from his claim that the discourse theory of morality is a cognitivist theory.2 The following five theses encapsulate the cognitive assumptions of discourse ethics.
i. Moral statements and judgments (and also moral emotions and feelings) have a rational content which is amenable to reason. Moral reasoning can be correct or incorrect.3
ii. Morality is a form of knowing, in the sense that moral norms and moral statements are amenable to justification, and that moral reasoning can lead to substantive conclusions.4
iii. Moral knowledge (reasoning, discourse) has a social function which is to resolve conflicts of action and to establish the basis of social order.5
iv. Morality is a learning process which can be analysed into different stages according to an invariant and universal logic of development. The process has both an individual and a social dimension.6
v. The programme of discourse ethics aims to provide a justification of the moral standpoint (or the moral principle) which can convince the moral skeptic, provided s/he is rational.7
Habermas does not treat the question of whether cognitivism or non-cognitivism is correct as a metaethical side-issue. On the contrary, he maintains that a cognitivist account of morality is essential to the defence of the ‘significance of moral intuitions’ and a necessary condition of the phenomenological adequacy of moral theory; in other words non-cognitivist accounts of morality fail to do justice to important and characteristic features of morality in modern societies.8 Finally, he maintains that, absent a cognitivist and universalist moral theory there is no adequate answer to the two fundamental questions of social theory: How do social agents in modern societies resolve conflicts of action and how is social order established?
Although it is central, Habermas’s conception of moral cognitivism is nonetheless unusual and easily misunderstood. To avoid confusion, I shall make a terminological distinction between two kinds of cognitivism, a broader conception of moralcognitivism and a narrower conception of metaethicalcognitivism. According to this terminology, moral cognitivism is the position that moral statements (sentences, utterances, judgments, claims, norms) are amenable to justification; metaethical cognitivism, by contrast, is the thesis that moral statements (sentences, utterances, judgments, claims) are truth-apt, and that some at least are literally true.9 Habermas’s position is slippery in that, whilst he makes great play of embracing moral cognitivism (and rejecting moral non-cognitivism), he nonetheless rejects metaethical cognitivism. For example, he often denies that normative sentences can be true (or false) ‘in the same sense in which descriptive sentences can be true or false’.10 Habermas contends that moral utterances make validity claims to rightness and that they do not make validity claims to truth. On my terms Habermas is a metaethical non-cognitivist, because he holds that strictly speaking no moral utterance or statement aspires to truth.
For the moment, I shall postpone discussion of the vexed question of what ‘rightness’ [Richtigkeit] means in this context.11 It suffices to think of rightness as the analogue to truth in domain of moral discourse. Habermas’s conception of moral cognitivism rests on the thesis that there is (only) an analogy between truth and rightness.12 This thesis is the corollary of his denial that moral statements and judgments are truth-apt and that some are literally true, the corollary of his denial of metaethical cognitivism.
Habermas is understandably resistant to being labeled a non-cognitivist, for he does not want discourse ethics to be lumped in with various other metaethical non-cognitivist moral theories; e.g. Ayer’s and Stevenson’s emotivism, Blackburn’s projectivism and Gibbard’s norm-expressivism, with which it otherwise has little in common.13 However, it is evidently only moral non-cognitivism that Habermas rejects. All the above deny, like Habermas, that a moral judgement or statement contains a truth value, that it expresses a belief that aspires to truth. They maintain instead that a moral judgment expresses an attitude or the affective endorsement of a system of norms or something of that ilk.14
Note that the usual motivations for rejecting metaethical cognitivism are abstract philosophical considerations in metaphysics and epistemology: for example that norms and values don’t exist in the sense that they would not appear in an absolute description of the world (or a natural scientific inventory of physical reality), and that we would need a special faculty of moral intuition for detecting them if they did. Habermas is no exception here. His main objection to metaethical cognitivism is that to grant truth to some moral statements is to raise the spectre of moral realism. Moral realism, as Habermas understand it, involves the positing of mind-independent (or language-independent) moral values and properties, and thus threatens to burden moral theory with an outlandish ontology.15 Note also that metaethical non-cognitivists do not dispute that moral discourse has a surface similarity with propositional logic, or that it behaves in syntactically and grammatically similar ways; they only dispute what is going on underneath. Once again Habermas agrees. He claims that he ‘has never doubted that valid moral statements behave logically just like true descriptive statements’.16 Finally, metaethical non-cognitivism is not usually taken to have any more empirical justification or any more phenomenological appeal than the opposed view. Few, if any, metaethical cognitivists defend their position on such grounds.17 On this point, as we have seen, Habermas is a notable exception. He thinks that the phenomena tell in favour of moral cognitivism and against metaethical cognitivism.18 2. Habermas’s Conception of Truth To understand Habermas’s conception of moral cognitivism we have to understand why he denies metaethical cognitivism and why he thinks that there is (only) an analogy between truth and rightness. His reasons for so doing lie in his pragmatic theory of meaning, and it is to this we now turn. In the early seventies Habermas embarked on an ambitious project that takes the pragmatic theory of meaning as a way of solving certain long-standing problems in social theory. The pragmatic theory of meaning takes the question of ‘what it is to understand a speech-act’ as the guide to understanding the meaning of linguistically interpretable actions.19 As a theory of language use, it construes speech-acts (or utterances) as instances of communication. Communication, in turn, is understood not as a game of representation ‘with which the interlocutors inform each other about their beliefs and intentions’, but rather as a response to ‘the need to coordinate the action-plans of independently deciding participants in action,’ a need which structures modern forms of life.20 A central thesis of the pragmatic theory of meaning is that the concept of truth can be understood as a kind of validity claim.21 a.Truth as a Validity Claim
Habermas’s conception of truth as a validity claim is roughly this. Whenever a speaker asserts p, she unavoidably (if only implicitly) claims that p is true. The truth of what is asserted has a necessary pragmatic connection with the reasoned understanding/agreement [Verständigung] of participants in an ideally prosecuted discourse, which he calls ‘rationally motivated consensus’.22 This view can be captured with the following formula:
T-C For any assertion ‘p’: if p is true, then p is amenable to rationally motivated consensus.23
The truth content of the assertion ‘p’ is embedded at either end in the speech situation. On the left, the speaker is pragmatically committed to the truth of what is asserted. She shows this commitment by undertaking to ‘redeem’ or to ‘make good’ her validity claim to truth, i.e., to justify the assertion (what is asserted, namely p), when challenged, questioned or just not understood by the hearer. On the right, by the very act of making an assertion the speaker must suppose that every participant in discourse can assent to p on the basis of the reasons that could be (or have been) adduced in support of it. The modal claim in the consequent is filled out by the idealising presuppositions of argumentation or rules of discourse, which, according to Habermas, a speaker cannot but invoke, even if only implicitly, by the very act of making a sincere speech-act. These rules ensure that the discourse is conducted in logically and semantically respectable manner, that no-one with relevant information is excluded, that everyone has equal access to the discourse, and that no view wins out by dint of deception or force, but only by the rational ‘unforced’ force of better reasons.24 A discourse is ideally prosecuted and the resultant consensus rationally motivated when these rules are adequately followed.
Habermas’s conception of truth as a validity claim is an element of a pragmatic theory of meaning that remained essentially the same from the early 1970s to the mid 1990s. As recently as 1995 Habermas can be found saying:
The pragmatic conception of justification opens the way to an epistemic concept of truth, which can help us out of the well-known difficulties with the correspondence theory. With the truth predicate we refer to the language game of justification that is to the public redemption of validity claims. […] I am interested...in the possibility of conceiving truth, purified of all connotations of correspondence, as a special case of validity, whilst explicating the general concept of validity [Gültigkeit] in terms of the discursive redemption of validity claims.25
There are two separate and important claims here. The first is that the concept of truth is epistemic: truth can be exhaustively analysed into the concept of ideal acceptability in discourse. The second is that truth can and should be understood as a specification of an underlying generic notion of validity [Gültigkeit] i.e., that it is some kind of construct of a single underlying pragmatic norm of correctness. Unfortunately, Habermas does not say nearly enough about what the underlying generic concept of validity is. He assumes, though, that whatever it is, it hooks up with discursive consensus along the lines of the T-C conditional.
V-C For any utterance ‘p’: if p is valid, then p is amenable to rationally motivated consensus.26
It is this basic notion of validity, rather than the concept of truth and of truth-conditions, which is taken as primitive in Habermas’s pragmatic theory of meaning.
Not “truth” but an epistemically understood generalised concept of validity [Geltung] in the sense of “rational acceptability” is the master concept of the pragmatic theory of meaning.27
The validity claim to truth (T-C) is to be understood as specification of the more basic underlying notion of validity (V-C).28
b. Validity Claims to Rightness and the Analogy with Truth
Of course, there is more to the pragmatic theory of meaning than this, but we must return to the analogy between truth and rightness that is the focus of our concern. According to Habermas moral utterances make validity claims to rightness not validity claims to truth. The thought is this: in making a moral utterance, I tacitly endorse the underlying norm of action. Just as I commit myself, in the very act of asserting ‘p’, to the truth of p, so when I utter the sentence, ‘theft is wrong’ I endorse the underlying norm, do not steal. In short, theoretical utterances (such as assertions) and moral utterances have a similar underlying structure, and perform a similar pragmatic function, namely that of eliciting consensus in discourse. Habermas’s distinction between theoretical and moral discourse sits flush with the distinction between the validity-claims to truth and rightness. A theoretical discourse just is a discourse in which validity claim to truth is made good, and similarly for moral discourse.
Initially Habermas argues that construing truth epistemically as a specification of the pragmatic notion of validity brings to light the analogy between truth and rightness. The long passage cited above continues as follows:
Thus we open up a conceptual space in which the concept of normative, here in particular moral validity, can be housed. The rightness of moral norms (and of general normative utterances)…can then be understood in analogy to truth.29
On the view stated here, the fact that truth and rightness can both be understood as epistemic concepts shows that each is species of the genus validity that functions as a general norm of correctness. The concept of rightness Habermas must have in mind can thus be captured with a slightly different formula.
R-C For any norm n: if n is right (valid), n is amenable to rationally motivated consensus.30
The rightness to consensus conditional (R-C) is present in all of Habermas’s various versions of the discourse principle (D) and the moral principle (U).31
At this point an ugly difficulty raises its two heads. What is this thing called rightness that is supposed to be analogous to truth? ‘Rightness’ and ‘right’ [Richtigkeit and richtig] are ambiguous. To begin with there is an ordinary non-moral sense in which the adjective right [richtig] means something like correct or justified. But this cannot be how Habermas is using the term. Rightness must be different from the generic norm of correctness of which it is supposed to be a specification. The most obvious way in which it might differ is in being a special norm of correctness or justification germane to the sphere of moral discourse. However, that difference is difficult to uphold. Suppose we replace the general term ‘right’ by a morally laden term such as ‘morally correct’ or ‘morally justified’. In that case the R-C conditional presupposes the relevant concepts of moral correctness or moral justification, and is no longer doing its job of elucidating the meaning of moral utterances. This second usage points to an entirely different sense of ‘right’, a substantive moral sense as in ‘moral right and wrong’.
It is by no means easy to separate these two senses of rightness. After all, reasoning or arguing in the ‘right’ way (making formally correct deductions and inferring correctly from the evidence) should lead one to discover or establish what is ‘right’ (morally required or permitted) or wrong (morally proscribed). The relevant point here is simply that ‘rightness’ is an ambiguous term, and so long as we don’t know which meaning of rightness is in play, it is very hard to understand and to assess the thesis that rightness is analogous with truth.
c. Habermas’s Revised Theory of Truth
By 1996 Habermas has been persuaded that the concept of truth cannot, after all, be epistemic.32 Two main doubts seem to lie behind this change of view. The first doubt is that, though participants in discourse can (indeed must) assume that the ‘rational acceptability’ of a proposition ‘under the most ideal conditions’ is ‘sufficient evidence’ for its truth, they are never entitled to truth, since they cannot know when all the relevant evidence is in and when all the relevant reasons have been weighed. Actual discourses, Habermas acknowledges, remain epistemically ‘provincial in respect of future knowledge’.33 The second doubt comes from Habermas’s unshakeable conviction that all knowledge is fallible. If all knowledge is fallible, then not only unshakeable convictions, but even discursively justified theories or beliefs (i.e. those that are ideally acceptable in discourse) may turn out to be false.34 But exhypothesi things that are discursively justified are not supposed to turn out false. The idealisations were supposed to strengthen the notion of justifiability sufficiently to make it into a reliable correlate of truth, so that truth could be understood as an epistemic concept. Nevertheless, these two doubts eventually convinced Habermas that truth is non-epistemic or ‘justification-transcendent’ after all, and he makes various alterations to accommodate the revision.35
In my view Habermas concession that truth is non-epistemic is a significant revision, but its ramifications are hard to discern. Habermas initially argues that an epistemic concept of truth makes the analogy between truth and rightness evident. So one might think that the concession that truth is not epistemic would undermine his confidence in the analogy. Not at all. In fact the reverse is the case. Habermas takes the concession that truth is not epistemic after all as more grist to his mill. Recall that Habermas is a staunch anti-realist about morality, and that he assumes that a non-epistemic concept of truth, i.e., one that outstrips even idealised acceptability in discourse, has a close affinity with metaphysical realism.36 Hence, from his perspective, the fact that truth can transcend justification is another good reason to deny that any moral statements can be true, and it is this denial that motivates the claim that there is (only) an analogy between truth and rightness. So in Habermas’s eyes the concession that truth is not epistemic is no reason to abandon the thesis that truth is analogous to rightness, regardless that truth is now ‘justification-transcendent’ and not epistemic, whilst rightness is ‘justification-transcendent’ and epistemic.
The real puzzle is why Habermas ever thought an epistemic concept of truth had any outlandish ‘ontological connotations’ in the first place, and why he originally felt he had to resist metaethical cognitivism in order not to pushed into embracing moral realism. Surely the last thing an epistemic truth theorist should be worried about is the ontological cost (in terms of ‘facts’, ‘truth-makers’ or ‘states of affairs’ of the claim that propositions, even moral ones, can be true.)
The answer is that Habermas always had other grounds for rejecting the package of metaethical cognitivism plus realism, grounds that were quite independent of his conception of truth. Habermas has a deontic conception of morality. On his view morality is not a matter of the perception and description of evaluative features of the social world, but a procedure for justifying norms that is anchored in our everyday practice of communication. Principle (U) is supposed to capture the everyday practice of discourse or argumentation that arises whenever a validity claim to rightness is challenged or loses its presumption of validity.37 According to Habermas, both the nature of moral norms and the grammar of moral language tell in favour of his theory. Norms are not natural facts but social rules that regulate and stabilise the behavioural expectations of agents in the lifeworld. Unlike natural facts these social rules are partly constituted by the discursive practices of the linguistic community, and by the attitudes and beliefs of moral agents towards them, attitudes and beliefs that are themselves formed discursively in the course of socialisation.38 These rules are commonly expressed as imperatives such as ‘do not kill’, ‘do not steal’ etc., of which truth and falsity cannot be predicated. In ordinary language one does not predicate truth of an imperative – the sentence ‘do not kill is true’ makes no sense. These are some of the moral phenomena Habermas thinks tell against metaethical cognitivism.39
So Habermas’s initial denial of the metaethical cognitivist thesis that moral judgements and statements express beliefs that aspire to truth, and that at least some are literally speaking true, makes sense in the framework of the discourse theory of morality, regardless of whether he holds an epistemic or a non-epistemic conception of truth.
3. The Frege-Geach Challenge to Discourse Ethics The minor detour through Habermas’s conception of truth shows the rationale behind the denial of metaethical cognitivism and the thesis that there is an analogy between truth and rightness. Now I want to show how Habermas’s discourse theory of morality is vulnerable to a well known objection to non-cognitivist moral theories. In his Frege-inspired paper on assertion of 1965, Peter Geach identifies a puzzle which has come to be known as the Frege-Geach problem.40 Consider the following arguments:
Argument AArgument B
Premise 1: If stealing is wrong, attempting to steal is wrong. P→Q
Premise 2: Stealing is wrong. P
Conclusion 3: Attempting to steal is wrong.Q
On the face of it, both A and B are valid instances of modus ponens, but in the case of B we can explain why it is a valid piece of reasoning by appealing to the notion of truth, and showing that the inference preserves truth. According to the metaethical non-cognitivist the premises and conclusion of A, unlike B, are not so much as truth-apt. So, in spite of the fact that they have an identical logical structure, the validity of A cannot be explained in the same way as the validity of B, as a truth-preserving inference. The Frege-Geach challenge is to invite metaethical non-cognitivists to explain what makes argument A valid, without presupposing the notion of truth and truth preservation. The assumption is that cognitivists who have recourse to a semantic notion of validity defined on the notion of truth can offer a more convincing explanation. For example, cognitivists can explain that the assertion ‘stealing is wrong’ in premise 2 is the same as the one embedded in the conditional in premise 1. Non-cognitivists cannot say this. Instead they say, roughly, that the phrase ‘stealing is wrong’ expresses an attitude of disapproval towards stealing. The trouble is that the antecedent of the conditional ‘if stealing is wrong…’ certainly does not express such an attitude. So there seems to be an equivocation of meaning between the two occurrences of the linguistically identical phrase, ‘stealing is wrong’ in each premise, in which case the argument is not a valid one.
A variety of responses to the Frege-Geach challenge are open to metaethical non-cognitivists. One response is to offer a logic of attitudes that mirrors propositional logic. Projectivists and expressivists, for example, may answer that the conditional in premise 1 is itself an attitude, namely a second order attitude, an attitude towards the fact that one attitude – disapproval of stealing – is coupled with another – disapproval of attempting to steal.41 In this case, any moral agent who has the complex second-order attitude expressed in the first premise and the simple first order attitude expressed in the second can logically reach the conclusion 3. Of course, whether or not the agent actually has the attitude expressed in 3. will be a matter of socialization and upbringing, not just a matter of logic. This last point raises the question of whether an agent who accepts 1 and 2 (and has the relevant attitudes) but nonetheless rejects 3 is guilty only of a moral failing (i.e. of lacking the attitude of disapproval towards attempted theft) or guilty also of a rational failing.42
Since Habermas accepts that truth-bearing (theoretical) discourse and non-truth-bearing (moral) discourse behave in syntactically similar ways, and since he denies that moral statements admit of truth, he faces the Frege-Geach challenge. However, Habermas must give a different kind of answer [that is different] from that given by the expressivist and the projectivist, for he does not agree that a moral statement is at bottom the expression of an affect or attitude.43 The implication of Habermas’s view that moral utterances make validity claims to rightness is that moral argument aims at something other than truth, namely ‘rightness’. ‘Rightness’, on this view, serves as the norm of correctness proper to moral discourse that can be used to distinguish good moral reasoning from bad in the same way that truth does for assertoric discourses. Habermas can then argue that moral agents who accept premises 1 & 2 but reject 3 are not just in the wrong morally (insofar as they have the wrong set of attitudes), they are guilty of a rational failing because they have violated the norm of correctness proper to moral discourse. This is the kind of answer Habermas must give, if he is to defend his conception of moral cognitivism properly.
Unfortunately this answer does not meet the Frege-Geach challenge squarely. No-one disputes that moral discourse is syntactically disciplined in respect of negation, conditional construction hypothesis, inference etc. The problem is that given the supposed underlying differences, i.e., that the inference in B is truth-preserving while the one in A is not, why is there a perfect homology between the two arguments? Given that one type of discourse aims at truth and the other at rightness, why is there syntactic homology between the two regions of discourse? The onus of explanation here lies squarely on the metaethical non-cognitivist, for the cognitivist has a ready answer.44
The Frege-Geach challenge opens up a new set of problems for discourse ethics. The focus of Habermas’s theory has been on the consensus-bringing function of speech and the action coordinating role of validity claims. Habermas has always considered the ‘logical-semantic’ analysis of language, including the truth conditional theory of meaning to be too narrow and too abstract to explain the social function of speech.45 Whilst that may indeed be the case, it does nothing to stave off the controversy. Regardless of his general arguments for a pragmatist rather than a semanticist account of meaning, Habermas willingly drags discourse ethics into semantic controversy by denying metaethical cognitivism.
The trouble is that his usual tactic of gesturing towards the analogy between validity claims to truth and validity claims to rightness cannot explain the syntactical homology between moral and theoretical argumentation.46 The thesis that there is only an analogy between truth and rightness presupposes the denial of metaethical cognitivism, which gives rise to the Frege-Geach challenge in the first place. So long as Habermas denies metaethical cognitivism, it remains a mystery why there is a syntactic homology between the two regions of discourse.
II I am now going to examine four possible responses to the Frege-Geach challenge. Strictly speaking only the first is offered as a solution, because it accepts the terms of the problem. The next three are best seen as different attempts to avoid the problem by embracing metaethical cognitivism or agnosticism. In order to assess in which direction discourse ethics should be altered, I offer a cost benefit analysis of each response.
1. Deriving Truth and Rightness from the Pragmatic Conception of Validity Habermas might attempt to solve the problem by developing his conjecture that the underlying pragmatic conception of validity is the master concept in the theory of meaning. He could argue that truth-bearing (theoretical) discourse and rightness-bearing (moral) discourse exhibit the similar syntactic features they do in virtue of being derived from a single underlying pragmatic notion of validity.47 Clearly there is an intuitive sense in which participants in moral discourse or argument, indeed anyone who passes reasoned moral judgments aim at being correct, and at finding out through their reasoning what is morally right; but this intuition does not take us far. It does not show that moral discourse is governed by a single aim or norm of appraisal distinct from truth, let alone that this aim is the inherently ambiguous concept of ‘rightness’ rather than, say, coherence or consistency.
To my mind the whole idea that the very fine-grained syntactic features of assertoric discourses can be explained as necessary requirements of reaching intersubjective consensus in discourse is wrong-headed. That said Habermas is by no means alone in the attempt to find a pragmatic and social ground for the syntactic structure of assertoric discourses, to make the concept of truth pay its way in the social world. Robert Brandom’s semantic inferentialism – a combination of pragmatism, rationalism and logical expressivism – in which Habermas has taken a keen interest, sets out to replace the question of what truth ‘is’ by the question of what we do when we treat something as ‘true’.48 However, it is doubtful whether any of Brandom’s arguments can take Habermas further down the path to his own thesis that truth can be understood as a specification of a more primitive concept of validity. For one thing, in spite of all the brilliance and ingenuity, not to mention the daunting length of Making it Explicit, Brandom has not yet provided a convincing, clear and workable pragmatic account of good inference that can replace the semantic notion of validity. If anything, his work only testifies that the best explanation we currently have of what a good inference is, still relies on the concept of truth and truth preservation.49 For another, Brandom disputes Habermas’s key theses that there are two distinct meaning-critical validity claims to truth and rightness, and that moral discourses aim at rightness not truth, and that both truth and rightness are specifications of an underlying concept of validity. He agrees with Habermas only insofar as he holds that material inferences – the normative force of good reasons – are not best explained as disguised formally valid logical inferences.50
So, neither intuition, nor the best available pragmatist account of a good inference, supports Habermas’s conjecture that validity [Gültigkeit] is the primitive concept of the theory of meaning. The first response is thus economical only in the sense that it is backwards compatible with the theory of communicative rationality and the pragmatic theory of meaning. However, it is not economical at all if we factor in what it would take to vindicate those theories as he currently conceives them.
Finally, anyway, there may be a knock-down argument against it. Habermas includes ‘logical-semantic rules’ such as the law of non-contradiction, and the requirement of consistency among the first level of the rules of discourse.51 In helping himself to this law he may simply be helping himself to the semantic notion of validity defined on the notion of truth. If so, the very idea that truth is a specification of validity [Gültigkeit] is a non-starter. If not, he must demonstrate the pragmatic credentials of the law of non-contradiction. This shows that the first response should be discarded, because, if it is not viciously circular, it takes too many hostages to theoretical fortune.
2. Showing that Truth is a Construct of Moral Discourse. The moral of the first solution is that it is probably not feasible for discourse ethics to meet the Frege-Geach challenge and for Habermas to leave discourse ethics, together with the pragmatic theory of meaning and the theory of communication on which it rests, untouched. We ought then, to take a different tack and consider how, and at what cost, discourse ethics can be altered to prevent the Frege-Geach problem from arising. One relatively simple rearrangement would be for Habermas to accept metaethical cognitivism, whilst interpreting the truth predicate as it applies to moral statements as a construct of moral discourse. The task of discourse ethics would be to show that the concept of truth (in the sphere of moral discourse) can be exhaustively understood as the discursive acceptability of a norm in accordance with the moral principle (U). On this view, Habermas could argue that for any moral judgment, such as that killing is wrong, the truth predicate can be correctly applied (e.g. it is true that killing is wrong) when and only when the judgement expresses a norm (such as, do not kill) that all participants could assent to in a discourse in accordance with principle (U). By this means Habermas can circumvent the Frege-Geach problem.
The advantage of this response is that it enables Habermas to enjoy the theoretical benefits of metaethical cognitivism whilst avoiding what he sees as the metaphysical and epistemological expense of moral realism – of having to posit independent moral facts and a special capacity to detect them. By construing the truth of moral judgments as the discursive justifiability of the salient underlying norm Habermas makes moral truth compatible with his constructivist version of moral rightness.
However, one disadvantage of the second response is that it is incompatible with the pragmatic theory of meaning: it complicates the thesis that there are three distinct meaning-critical validity claims to truth, rightness and truthfulness, and messes up the neat triadic architectonic of Habermas’s theory. On this view, the validity claim to rightness becomes a special case of the validity claim to truth. Another disadvantage is that the second response requires that Habermas give one explication of truth in the domain of theoretical discourse and another in the domain of moral discourse. It may grant him the possibility of a unified, truth-conditional semantics, but it at the price of two kinds of truth. A third, even more serious disadvantage is that this response is flatly incompatible with Habermas’s concession that truth is not an epistemic concept and that truth and justification (even ideal acceptability in discourse) can come apart.
Finally, the second response faces a similar problem to the one which beset the first solution. The principles of discourse ethics (D) and (U), and the pragmatic conception of validity, presuppose the rules of discourse, among which Habermas includes the requirement of consistency and the ‘logical-semantic’ law of non-contradiction.52 This poses a dilemma. If the rules of discourse presuppose the concept of validity defined on the notion of truth, the whole idea that the discourse theory of morality can serve as an elucidation of the truth predicate for moral discourse is a non-starter. The thesis that pragmatic validity [Gültigkeit] is the basic concept of which truth is a specification is undermined, along with the thesis that rightness not truth is the relevant norm of appraisal for moral argument. If the rules of discourse do not presuppose a more primitive notion of truth, then we are again left asking after the pragmatic grounds of the law of non-contradiction. It seems then, that the second response fares no better than the first.
3. Metaethical Cognitivism and the Deflationary Theory of Truth. There is a third possible response to the Frege-Geach challenge similar in spirit to the second but theoretically neater. It is that Habermas adopt a deflationary theory of truth instead of the discourse theory of truth.53 That would allow him to embrace metaethical cognitivism and thereby avoid the Frege-Geach problem without [thereby] incurring any realist metaphysical cost. Deflationary theories of truth typically take the equivalence schema
the proposition that p is true if and only if p
to be necessary and sufficient for the function of the truth predicate and to provide an adequate theory of truth in the absence of any further substantive account of what truth is or what truth does.54 Were Habermas to adopt a deflationary conception of truth he would be free to argue that moral judgments and statements are true simply by virtue of satisfying the equivalence schema, e.g.,
the proposition thatmurder iswrong is true if and only if murder is wrong.
No ‘ontological connotations’ and no metaphysical commitments follow.55
The third response, then, like the second, disarms the Frege-Geach challenge by embracing metaethical cognitivism. The benefits of the third response are clear enough. The deflationary theory of truth is compatible with Habermas’s moral constructivism. Habermas can grant truth-aptitude and indeed truth to moral statements, embrace constructivism about rightness, whilst avoiding any realist metaphysical baggage. (Of course he can’t avoid some constructivist baggage, but he has that anyway.)
It has two main advantages over the second response. First, it allows Habermas a single conception of truth across the different spheres of discourse. Second, it assigns the discourse theory of morality an altogether more appropriate task, to interpret the thought on the right hand side of the moral biconditional, namely that murder is wrong. In the light of principle (U) Habermas could explain that ‘murder is wrong’ means that there is a norm (do not murder) to which everyone affected can jointly and freely assent on the basis of their interests in an ideally prosecuted discourse. Discourse ethics, rather than explicating the concept of truth in the moral domain, explicates the meaning of the moral terms (such as ‘wrong) that occur in the sentences that are plugged into the equivalence schema,.
One disadvantage of the third response is that it conflicts with the idea that truth and rightness are distinct meaning-critical validity claims, and would therefore require that revisions be made to the pragmatic theory of meaning. Another is that it requires a stricter division of labour between semantic questions of truth and meaning, and metaphysical questions concerning the nature of physical, social and moral reality. Actually these are only disadvantages, when viewed from Habermas’s perspective. In my view discourse theory only stand to benefit from these revisions. As regards the first, there are anyway compelling arguments against the idea that there is a plurality of distinct meaning-critical validity claims.56 As regards the second, Habermas’s tendency to run the largely metaphysical question of realism together with semantic and epistemological questions only lands him in difficulty.57
There is a less tractable problem with the third response. On Habermas’s deontic conception of morality norms are the nexus of social order, and norms are usually expressed as imperatives such as ‘do not kill’. Habermas thinks that the fact that ordinarily one does not predicate truth of imperatives is evidence against metaethical cognitivism. As a metaethical cognitivist convert Habermas would have to argue that the fact that in ordinary language the truth predicate is not attached to imperatives, and that therefore imperatives cannot be plugged into the equivalence schema, does not disprove metaethical cognitivism. He may even continue to allow that the grammar of imperatives is evidence against cognitivism. However, he would be able point to other evidence that supports the new position; for example, the fact that moral discourse has all the syntactic features of assertion, and that the truth predicate can be attached to moral judgments such as, ‘It is true that stealing is wrong’. At present Habermas holds that such a sentence is just an elliptical way of expressing the presumption that the norm ‘do not steal’ could be accepted by all in a discourse according to principle (U). However as a metaethical cognitivist Habermas would be free to hold that it is a genuine application of the truth predicate. So, although the grammar of imperatives may tell against the metaethical cognitivism, the grammar of moral judgements do not. All in all the phenomena of morality to be taken into account, including ordinary moral language use, are multifarious and indeterminate, and do not point in the direction of any single metaethical theory.
4. The Agnostic Response Whereas the third response gets out of the Frege-Geach challenge by embracing metaethical cognitivism and deflationism about truth, the fourth attempts to hold off the challenge by remaining agnostic about what truth is and whether to grant truth-aptitude and indeed truth to moral statements.
One reason Habermas rejects metaethical cognitivism is that he fears that to allow that moral statements can be true is to invite moral realism. Since he abhors moral realism, he denies metaethical cognitivism. We have already seen several reasons to doubt that there is such a close connection between metaethical cognitivism and the kind of realism to which Habermas is opposed. I have just been arguing that Habermas need have no such qualms about embracing metaethical cognitivism, if only he adopted a deflationary theory of truth in place of the discourse theory. It is worth asking whether Habermas need involve the programme of discourse ethics in any philosophical disputes over the nature and function of truth. John Rawls, who famously recommends that political theory avoid so far as is possible getting involved in philosophical controversy criticizes Habermas for dragging the discourse theory of morality and his conception of political legitimacy into the philosophical mire by advancing ‘a general account of meaning, reference and truth or validity both for theoretical reasons and for several forms of practical reason.’58 Richard Rorty (for rather different reasons) advises philosophers generally not even to attempt to fill in the blanks in statements such as ‘if p is true, then_’, advice that Habermas signally ignores, since his pragmatic theory of meaning does exactly what Rorty warns against.59
Following the albeit rather different suggestions of Rawls and Rorty, the pertinent question is whether any of the specific aims of the programme of discourse ethics, namely to elucidate and justify the moral standpoint and principle (U), and to defend moral cognitivism, require answers to questions about the nature and function of truth? The answer is that they do not. The question of what truth is, is no more germane to the goals of discourse ethics than it is to moral philosophy in general. There is no question but that Habermas’s can maintain all of the five central theses of his conception of moral cognitivism without so much as attempting to answer the question of what truth is. Habermas’s discourse theory of truth is at most part of the background theory which unites discourse ethics with Habermas’s social theory, not part of discourse ethics itself.
Agnosticism about truth (i.e. refraining from answering what truth is or does) might save Habermas some unnecessary and painful intellectual labour, but does not itself constitute a response to the Frege-Geach challenge. Habermas could conceivably combine agnosticism about the nature truth with metaethical cognitivism, i.e., allow that moral statements can be true (or false) on any account of truth. There is no logical inconsistency here. Admittedly though, it would be odd for someone who is so adamant that moral statements can be true, not to hold any views about what it is for a statement to be true. This prompts the question of whether Habermas can afford to go the whole hog and to remain agnostic about metaethical cognitivism too? If he extended his agnosticism to the question of whether metaethical cognitivism or non-cognitivism is correct, Habermas could prevent the Frege-Geach challenge from arising, for, as we have seen, the challenge arises from the denial that moral statements can be true. This extended agnosticism, agnosticism about truth and about metaethical cognitivism, is what I call the agnostic response.
In order to assess the tenability of the agnostic response we must consider whether it would conflict with any of the aims of the discourse theory of morality. There is nothing to prevent Habermas from prescinding entirely from questions concerning metaethical cognitivism whilst arguing that moral discourse aims at establishing universally acceptable norms. Equally there is no problem with his defending his deontological conception of morality, for he need not enter the debate about whether the grammatical form of imperatives in ordinary language has any semantic repercussions. Would agnosticism about metaethical cognitivism pose any threat to his conception of moral cognitivism? Such agnosticism is definitely compatible with theses ii-v. The only one of Habermas’s moral cognitivist assumptions that poses any real problem is the first thesis, that moral reasoning can be correct or incorrect.
Fortunately, the problem is more apparent than real. The agnostic can of course allow that moral agents ordinarily assume that moral reasoning can be correct or incorrect. There is no dispute about the phenomenology nor about the syntactic features of moral discourse. The problem arises for non-cognitivists only when they offer their explanations for these facts. Since the agnostic has not ruled out that the relevant norm of the correctness of moral reasoning is the semantic notion of validity defined on the concept of truth, he is no worse off in this regard than the metaethical non-cognitivist. Of course Habermas the agnostic would have to allow that if metaethical non-cognitivism turns out to be the case, then some other norm of correctness, which perfectly mirrors the semantic notion, must be operative in moral discourse, in which case further explanation is required to account, among other things, for the syntactic homology between theoretical and moral reasoning.
What this shows is that the question of metaethical cognitivism versus non-cognitivism (but not moral cognitivism versus non-cognitivism) is, in spite of Habermas’s assumption to the contrary, a side-issue on which he can afford to remain agnostic without imperilling the programme of discourse ethics in any significant way.
The main benefit of the agnostic response is that it spares Habermas lot of painful intellectual labour. Unlike the first two responses it does not take any needless theoretical hostages to fortune (such as the thesis that truth and rightness are specifications of a single underlying pragmatic notion of validity), and unlike the third it does not presuppose that a controversial philosophical theory about the nature of truth, such as the deflationary theory, is correct. Finally, though it does not exactly solve the Frege-Geach problem, it nonetheless keeps it at arms length, for the problem only arises once Habermas sides with the metaethical non-cognitivists in refusing to grant truth-aptitude and truth to moral statements.
So what would be the costs to discourse ethics of opting for the agnostic response? One cost is that Habermas would have to suspend judgment on his present theory that truth can be understood as a validity claim, and that truth and rightness are specifications of an underlying pragmatic notion of validity. In itself that is no bad thing given the difficulties these theses land him in. The trouble is that they form part of theory of communicative rationality and the pragmatic theory of meaning which are the background to Habermas’s social, moral and political theory. So whilst this fourth option may be beneficial to the discourse theory of morality it will have the effect of fragmenting Habermas’s overall theoretical programme. Habermas may find that disturbing, since he conceives his overall theory as an interdisciplinary research programme whose individual pieces ‘fit together as it were like those of a jigsaw puzzle’ (VE 506). The trouble with that simile is that there is only one way that all the pieces of a jigsaw puzzle can fit together, and the point of a jigsaw puzzle is that it does all hang together. Perhaps it would be more fruitful, as well as more accurate, to think of Habermas’s programme as being composed of theoretical cells. Each cell is more or less self-sufficient. For example, Habermas’s social theory, discourse ethics and democratic theory have their own internal aims, their own distinct guiding questions. However, they can be combined in various ways. For example, the discourse theory of morality, together with Habermas’s theory of law, can help social theory answer the tricky question of how social order is established. On this model discourse ethics would not be so closely tied to, and not so dependent on, the theory of communicative action and the pragmatic theory of meaning.
It might be objected that I have overlooked an unacceptably high cost that the agnostic response exacts of the programme of discourse ethics itself. Surely, if Habermas adopted the kind of agnosticism countenanced here, he could not offer an account of the meaning of moral utterances, since that account would require that he either affirm or deny metaethical cognitivism. And without an account of the meaning of moral utterances Habermas cannot really claim to have elucidated the moral standpoint.
I think this objection is misplaced, although we need to retrace our steps in order to see why. Discourse ethics is concerned with the pragmatic meaning of moral utterances. Habermas’s pragmatic theory of meaning investigates the connection between validity and rationally motivated consensus, and thus the action coordinating-function of speech. Once Habermas gives up the untenable conjecture that truth is a kind of validity we can see that it is in fact entirely possible for him to elucidate the pragmatic function of valid norms without broaching questions about the truth-aptness or truth of moral statements. Moreover this re-jigging is necessary given Habermas’s revised conception of truth.
First recall the formula that captures Habermas’s elucidation of the truth predicate as a specification of the validity-consensus conditional.
T-C For any assertion ‘p’: p is true p is amenable to rationally motivated consensus.
Second, recall Habermas’s admission that truth can outstrip even idealised acceptability in discourse. Earlier we observed that, given the assumption about the propinquity between cognitivism and realism, this admission provides Habermas with an additional reason for denying metaethical cognitivism. Since the thesis that rightness is analogous with truth is a corollary of that denial, Habermas thinks the admission that truth is not epistemic solidifies his position rather than undermines it. But Habermas’s assessment was based on the dubious assumption that there is an internal relation between metaethical cognitivism and moral realism. So we should re-examine the implications of Habermas’s revision to his theory of truth.
Habermas concedes that truth is not epistemic on the grounds that there are true statements which cannot be justified (even in an ideally prosecuted discourse) and there are ideally justified statements which are not true. In that case the T-C formula does not, in fact, hold. There is no ‘necessary’ connection of any kind, (even the formal pragmatic kind) between truth and rationally motivated consensus. Habermas’s concession brings to light the mistaken assumption of the pragmatic theory of meaning, namely that the pragmatic connection with consensus flows from the truth of p, when in fact it flows from the reasons for believing p to be true.
T-C* For any assertion ‘p’: if there is reason to believe that p, then p is amenable to rationally motivated consensus.
It just looks as if the connection with consensus flows from truth, because to believe that p is to believe that p is true. If we make a similar substitution in R-C we can see that the same point holds in respect of the ‘rightness’ of a norm.
R-C* For any norm n: if there is reason to endorse n, then n is amenable to rationally motivated consensus.
The difference is that the rightness of a norm (in the sense of its justifiability) is constituted by the reasons for it.60 So in the case of R-C* we can either say that the pragmatic connection with consensus flows from the norm’s rightness, or that it flows from the reason that justifies the norm; it amounts to the same thing.
The various consequences that unravel from Habermas’s concession that truth is not an epistemic concept have very significant implications for the theory of communicative rationality and for the pragmatic theory of meaning, and not the ones Habermas takes them to have.61 What this analysis of Habermas’s revised position shows is that it was a mistake to have presented rightness as an analogue of truth in the first place. Rightness is an analogue not of the truth of an assertion, but of the reasons for making it. The alleged pragmatic analogy obtains not between truth and ‘rightness but between the reasons that justify assertions and the reasons that justify the norms. Habermas’s pragmatic conception of validity is a relation between reasons and discursive consensus.
Now we can bring this analysis to bear on our objection that the agnostic response would require that Habermas refrain from offering any account of the meaning of moral utterances. That is not the case. Habermas can indeed afford to remain agnostic on the question of metaethical cognitivism and nonetheless offer a account of the meaning of moral utterances that explores the pragmatic connection between validity and consensus, once that relation has been rethought in the light of Habermas’s concession that truth is not epistemic.
5. Conclusion Of the four responses to Frege-Geach problem, we should reject the first two for the reasons already stated, if they are not circular they are unfeasible. The second and third responses remain possible courses of evasive action. The third commits Habermas to a deflationary theory of truth and to metaethical cognitivism, and requires Habermas to make suitable revisions to the pragmatic theory of meaning in order to accommodate these changes. It is less confusing than his current position in that it allows Habermas to be a cognitivist in all senses of the term, to be both what I have here called a metaethical and a moral cognitivist. The fourth response enables Habermas to steer clear of such controversial philosophical commitments, whilst defending his conception of moral cognitivism and moral constructivism.62 Choosing between these options is not easy. As a metaethical cognitivist Habermas is free to acknowledge that the correctness and incorrectness of moral reasoning is no different in kind from the formally valid inferences of logic, but he has to take on the task of defending a controversial deflationary theory of truth. As an agnostic, he has a lighter theoretical load, and can better handle the fact that different moral phenomena, such as the grammar of moral imperatives on the one hand and the syntactic structure of moral reasoning on the other, appear to point in opposite metaethical directions. Either way, both these responses are consistent with the central aims of the programme of discourse ethics, and are preferable to his present position, which ties the discourse theory of morality too closely to the problematic pragmatic theory of meaning.† James Gordon Finlayson
Department of Philosophy
University of Sussex
REFERENCES Ayer, A. J. (1936) Language Truth and Logic,London: Gollancz.
Blackburn, S. (1971) ‘Moral Realism’ in Morality and Moral Reasoning, ed. John Casey,
(1984) Spreading the Word, Oxford: Oxford University Press.
(1988) ‘Attitudes and Contents’ in Ethics 98, pp. 501-517.
Brandom, R. B. (1998) Making it Explicit: Reasoning, Representing and Discursive Commitment, Cambridge Mass.: Harvard University Press.
(1997) ‘A Précis of Making it Explicit’ in Philosophy and Phenomenological
Research,LVII, pp. 153-6.
(2000a) Articulating Reasons, Cambridge Mass.: Harvard University Press.
(2000b) Rorty and his Critics, Oxford, Blackwell.
Davis, F. D. (1994) Davis F. ‘Discourse Ethics and Ethical Realism: A Realist Realignment of Discourse Ethics’, European Journal of Philosophy, 2:2, pp. 125 -43.
Devitt, M. (1984) Realism and Truth, Oxford: Basil Blackwell.
Divers, J. & Miller, A. (1994) ‘Why Expressivists About Value Should Not Love Minimalism About Truth’ Analysis 54.1, January, pp. 12-19.
(1995) ‘Platitudes and Attitudes: a Minimalist Conception of Belief’ Analysis 55.1
January, pp. 37-44
Finlayson, J. G. (2000) ‘Modernity and Morality in Habermas’s Discourse Ethics’ Inquiry 43,
(2002) ‘Making Moral Sense: Beyond Habermas and Gauthier’ Ethics 112: 4, pp. 828-32.
Frege, G. (1970) ‘Negation’ in Translations from the Philosophical Writings of Gottlob
Frege, eds. P. Geach and M. Black, Oxford: Basil Blackwell, pp. 117-137.
Geach, P.T. (1965) ‘Assertion’ Philosophical Review 74, pp. 449-65.
Gibbard, A. (1990) Wise Choices, Apt Feelings Oxford: Oxford University Press.
Mazouz, N. (2002) ‘Diskursethische Varianten’ Deutsche Zeitschrift für Philosophie 50: 1,
Gunnarsson, L. (2000) Making Moral Sense: Beyond Habermas and Gauthier, Cambridge:
Cambridge University Press.
(1979) Communication and the Evolution of Society, London: Heinemann.
(1984a) Vorstudien und Ergänzungen zur Theorie des Kommunikativen Handelns, Frankfurt a/M:
Suhrkamp. (VE )
(1984b) Theory of Communicative Action, vol. 1., Boston: Beacon Press. (TCA1)
(1992a) Autonomy and Solidarity,ed. P. Dews, London: Verso. (AS)