CRfTIOA, Ru"'. H.'p •••o.m., •••••• d. Filo,oj(.Vol. XX, No. 58 (abril 1988): 43-65

PROBLEMS OF DISCOURSE THEORY

ROBERT ALEXY

Christian-Albrechts- Universitiit, Kiel

There are three categories of problems in discourse theo-ry. The first group of problems arises if discourse theoryis regarded as a truth theory. The second group inves-tigates its practical applicability, and the third groupconcerns the justification of discourse theory. We talkabout the status of discourse theory as a truth theory,when we discuss the relationship between the conceptsof truth and correctness and the concepts of consensus,unlimited discussion and rationality. 1 The problem ofapplicability is at stake, if the discourse theory is ac-cused of being without any content and merely formal,which becomes apparent by the fact that it does notlead to any definitive result.f The problem of iustifi-cation is the problem of establishing conclusive reasonsfor discourse rules and principles.P

I want to discuss just the first two problems, namelythe status and the applicability problems. There will

1 See e. g. H. Scheit, Wahrheit - Di,leur, - Demokratie, Freiburg/Munich,1987, pp. 123 ff.; A. Wellmer, Ethi/c und Dialog, Frankfurt/M., 1986,pp. 70 ff.:l A. Wellmer [n. 1), p. 72.a Compare J. Habermas, "Diskursethik - Notizen zu einem Begriin-dungsprogramm", in: J. Habermas, Moralbewuptrein und /communi/cativuHandeln, Frankfurt/M., 1983, pp, 67 ff.; W. Kuhlmann, Reflexive Letst-begriindung, Freiburg/Munich, 1985, pp. 181 ff.

43

be another limitation to my subject. Both problems areproblems pertaining to all forms of discourse, meaningthat they appear, for instance, in theoretical, practical,and esthetical discourse. I will only consider practicaldiscourse.

I. Discourse Theory as a Procedural TheoryDiscourse theory belongs to the class of procedural the-ories.t According to all procedural theories the correct-ness of a norm or the truth of a proposition dependsupon whether the norm or proposition is or can be theresult of a certain procedure -or not. What is actu-ally the result of a procedure, can be the result of thatprocedure, while this does not apply in reverse. Thecan-version does, therefore, embrace more. It is goingto be the starting point of my considerations. If a isthe exponent of a procedural theory of the can-version,according to which the procedure P is to be applied,then a will answer the question, whether or not a normN is correct, with:

D: A norm N is correct, if and only if it can be theresult of the procedure P.

There are various ways of constructing the procedureP. The differences may relate first to those concern-ing the individuals, and second to those concerning therequirements of the procedure. Third, its character de-pends upon how the procedure is constructed with re-gard to the individuals and the requirements.

As to the individuals, one has to differentiate throughnumber and characteristics. P can be carried out by oneindividual, but in P several or all individuals of a more

• See R. Alexy, "Die Idee einer prozeduralen Theorie der juristischenArgumentation", in: Reckt,theorie, Beiheft 2 (1981), pp. 178 ff.

44

or less wide class may also take part. Concerning thecharacteristics of the individuals, one can work withactually existing or with constructed or ideal individ-uals. The ideal observer theory proposed by Firth5 isan example of a procedure that firstly uses only oneindividual, and secondly an ideal one. Discourse the-ory again is characterized by the fact that an unlimitednumber of individuals, in the state they actually exist,can participate in P.

The requirements can be of various kinds. So it ispossible to establish certain cognitive and motivationalcharacteristics of the individuals, or to fix conditions orcircumstances under which argumentation and decisionhave to take place, or to formulate rules according towhich the procedure has to be performed. The most im-portant differences arise from variations in the strengthof the requirements.

The procedure's character depends on the numberof individuals and the kinds of requirements. For de-termining the character of the procedure it is decisive,whether the possibility of a change in empirical andnormative convictions of the participants is provided ornot. If this is not the case, one can decide on the basisof the empirical and normative material at one point intime. Such a model of decision theory is suggested byRawls in order to choose his principles of justice, whichhe calls "the only choice consistent with the full descrip-tion of the original position't" that can be made fromthe "standpoint of one person selected at random"," Dis-course theory as a model of argumentation theory againis characterized by the possibility that the empirical and

II R. Firth, "Ethical Absolutism and the Ideal Observer", in: Philo,-ophy and Phenomenological Reseerch, 12 (1952), pp. 320 ff.

6 J. Rawls, A Theory 01Judice, Cambridge, Mass., 1971, p. 121.7 J. Rawls (n. 6), p. 139.

45

the normative convictions, as well as the interests of theindividuals, can change because of arguments presentedin the course of procedure. In the following, I will onlydiscuss this version of a procedural theory.

II. Rules of DiscourseThe requirements of discourse theory can, as there isno prescription concerning the individuals, be entirelyformulated by rules. I have tried elsewhere to expressthe system of discourse rules as completely as possible."The system consists of rules which demand, for exam-ple, non-contradiction, clearness of language, empiricaltruth and sincerity, and of rules which express the ideaof universalisability, e. g. by giving everybody the right

. to participate in the discourse and by granting every-body equal consideration in the discourse, and of ruleswhich concern the consideration of consequences, theweighing of reasons and the analysis of the genesis ofnormative convictions.

Against this variety the objection has been raised,that it expressed a "mixed concept of rationality't.?which did not actually serve the clarity of analysis.To this one can answer that practical rationality is acomplex matter. The result of an analysis of a com-plex matter can only be a complex model. The deci-sive question iljJwhether the model can cope with thecrucial problems arising in the relevant realm. Further-more, it has been critically remarked, that some ruleshad a moral content.P This could only be a valid ob-

8 R. Alexy, Theori« der juri,ti,chen Argumentation, Frankfurt/M., 1978,pp. 234 ff.

II O. Weinberger, "Logische Analyse als Basis der juristischen Argu-mentation", in: W. Krawietz/R. Alexy (eds.), Metatheorie juri,ti,cherArgumentation, Berlin, 1983, p. 200.10 See O. Weinberger (n. 9), p. 195.

46

jection, however, if at least one of three preconditionswere true. The first would be that the concept of prac-tical rationality explicated by the rules must not haveany moral content. In my opinion it is at least allowed,and probably even necessary.ito incorporate moral con-tents in a fully-fledged concept of practical rationality.The second precondition would be fulfiled, if those ruleswhich have a moral content, such as those which expressnon-trivial demands for universalisability, were not jus-tifiable. Here I cannot discuss the thesis that they arecapable of a transcendental-pragmatic justification byway of a presupposition-analysis.U At least it can beclaimed that it is uncertain whether they cannot be jus-tified. A third precondition for the success of the objec-tion mentioned above would be that the moral contentof some of the discourse rules would deprive the wholesystem of its practical applicability. This leads one backto the general question of practical applicability.

From the point of view of applicability, the main flawof discourse theory consists in the fact that its rule sys-tem provides no procedure which allows one to cometo exactly one result in a finite number of operations.There are three reasons for this. First, the rules of dis-course contain no prescriptions concerning the startingpoints of the procedure. Starting points are the norma-tive convictions and interpretations of interest of theparticipants, just as they appear. Second, the rules ofdiscourse do not prescribe all steps or argumentation.Third, a number of the rules are only approximatively

11 J. Habermas (n. 3), pp. 93 ff.; R. Alexy (n. 8), pp. 230 ff.

47

fulfilable.P Thus discourse theory does not guarantee adefinite decision in each case.

One might think that this already reveals the unappli-cability of discourse theory. To get rid of this objectionone has to distinguish between real and ideal discourses.Let us have a look at the latter first.

III. The Ideal Discourse

Discourse can be ideal in either some or all respects.Only the discourse which is ideal in all aspects shallbe of interest here. It is defined by searching for ananswer to a practical question under the conditions ofunlimited time, unlimited participation, and completefreedom of constraints by way of achieving completelinguistic-conceptual clearness, complete empirical in-formation, complete ability and willingness to changeroles and complete freedom from prejudice. The con-cept of a discourse ideal in all respects provides manyproblems. The four most important are: the problemsof construction, of consensus, of criterion, and of cor-rectness.

1. The Problem of Construction

The problem of construction results from the fact thatdiscourse theory works with real, actually existing per-sons as participants. A construed prolonging of par-ticipation into unlimited time causes the real, actuallyexisting participants to become partially ideal or con-structed, that is immortal, participants. This is not theonly idealization. One could imagine that a person, who

12 The concept of rule is used in a wide sense here, embracing bothdefinite obligations and obligations concerning optlmalisatlon; on thisdistinction compare R. Alexy, "Rechtsregeln und Rechtsprinzipien",in: Arckill/iir Rechtl- vnd Sozio.lpkilolopkie, Beiheft 25 (1985), pp 13 ff.

48

is immortal, will take part in discourse eternally, andstill learn little or nothing. That is why the conceptof a discourse ideal in all respects implies that its par-ticipants fulfil all rules of discourse completely. Thatmeans that they achieve, regardless of how much time isconcerned, complete clearness, information, ability andwillingness to change role and freedom from prejudices.That this is not actually possible in fact, does not raiseany problems as long as only the concept of ideal dis-course is being discussed. To be taken seriously in thiscontext is the question, whether or not the describedstate is conceptually possible at all. So the questionarises, whether linguistic-conceptual clearness is achiev-able in a discourse with participants from very differ-ent cultures with very different languages. Questionsof this kind cannot be discussed here. At least it isclear that participants in the ideal discourse undergoa nearly total change from real and actually existinginto ideal and constructed participants. This seems tocontradict the basic idea of discourse theory mentionedabove, namely that discourse is a procedure performedby non-fictitious, that is, real individuals. However, con-sidering. that the ideal discourse is not perfect from thebeginning, but rather becomes perfect by a construedpotentially eternal continuation and because of learningprocesses taking place with -at the beginning- realindividuals, this problem can be solved.

2. The Problem of Consensus

With the problem of consensus, the question arises,whether an ideal discourse leads to a consensus aboutevery practical question. Such a consensus does not fol-low logically from the statement, that the conditions ofideal discourse are fulfiled. Consensus about a certain

49

normative question is a substantial matter. The condi-tions listed above have a formal character in relation tothis. Therefore, consensus about every question couldbe guaranteed only if it could be assumed that the em-pirical premise, that the fufilment of the conditions ofideal discourse would make all differences of opinion onpractical issues vanish, holds true. This premise pre-supposes that there are no discourse-resistent anthro-pological differences, which can exclude a consensus inpractical, that is, value-questions, even under the idealconditions listed above. I think this question cannot bedecided. This is so, because there exists no method toforetell the behaviour of real persons under the men-tioned non-real conditions. This means that a guaran-tee of consensus can neither be excluded nor assumed.This has the consequence, that it has to be consideredas possible, that even after a discourse lasting eternally,participants will adhere to norms which are incompati-ble with each other. The result of the procedure wouldthen be both N and oN. This means that, accordingto the definition D given at the beginning, both NandoN would have to be labeled as "correct". The ques-tion resulting from this will have to be dealt with in thediscussion of the problem of correctness.

The question whether a consensus reached after po-tentially infinite duration and under ideal conditionswould be necessarily a final or definite consensus, leadsto a second aspect of the problem of consensus. A stableconsensus would be reached if no additional argumentwhich could lead to a change of the normative convic-tion in question in at least one participant were possible.I think that in this context it does not matter whetherthe class of normatively relevant arguments is finite orinfinite. As long as the participants are not assumedto have been equipped with the characteristic not to

50

overlook any possible normatively relevant argument,even a potentially infinite discourse, in which the par-ticipants achieve the ideal conditions mentioned, doesnot guarantee that a new argument could not destroythe consensus reached for once. Things would be differ-ent only if the infiniteness of discourse were interpretedas actual13 infiniteness in such a way that all possiblearguments were brought up and considered, which shallbe not done here. A consensus could then by definitionno longer be destroyed by a new argument.

Summing up what has been said about the problem ofconsensus, two things should be pointed out: (1) Evenin the case of a potentially infinite ideal discourse itcannot be excluded that there will be no consensus atall. (2) Even in the case of a potentially infinite idealdiscourse it is never certain whether a consensus reachedfor once is final or definitive.

3. The Problem of CriterionThe things discussed above have serious consequencesin regard to the third problem, the problem of criterion.This problem concerns the question to what extent theideal discourse will work as a criterion of correctnesswhen substituted for P in the definition D given above.As the ideal discourse cannot be carried out by def-inition, it can only be used as a criterion, by askingwhether a norm N could be the result of an ideal dis-course. Three problems are of special importance here.

The first problem is caused by the fact that in orderto use the ideal discourse as a criterion of correctness,a procedure, which is essentially a collective endeavourinvolving several persons, must be carried out in the

IS On the concepts of potential and actual infiniteness compare P.Lorenzen, "Das Aktual-Unendliche in der Mathematik", in: P. Loren-zen, Methodi.eher Denken, Frankfurt/M., 1974, pp. 94. ff.

51

mind of one person, and in this sense, monologically.It is of the essence of the practical discourse that sev-eral persons are involved, for the following reasons. Ina practical discourse correct answers to practical ques-tions are being searched for, which concern the inter-ests of several persons. Thus, the correct solution ofa conflict of interests is at stake. For this purpose,the respective actually existing normative convictionsof the participants concerning the right solution are ex-posed to rational scrutiny. In this process, the respec-tive interpretations of the participants' interests and thechanges thereto, caused by arguments, playa decisivepart. How an interpretation of interests eventually hasto be changed because of arguments, is finally every-body's own concern. From this it follows that then, ifthe correctness of the results of discourse depends on thecorrectness of the interpretation of interests, and if thecorrectness of interpretation of interests is a matter ofargumentative investigation, the discourse is essentiallynon-monological. The consequences resulting from thisfor someone who asks himself, whether something canbe the outcome of a discourse, are less fatal than theyappear to be at first glance. Though discourses are es-sentially non-monological, a discourse carried out in themind of one person can still approximate a discourseconducted by several persons. One can never be sureof the arguments, the interpretation of interests, andthe changes in the interpretations of other persons, butit is possible, to a considerable degree, to make wellgrounded conjectures about them. Manifold argumentshave been uttered, by various persons, about almost anypractical question. Everyday life, literature and scienceprovide numerous informations about possible ways ofthe interpretation and changes of interests. From themonological conduct of discourse arises a considerable

52

amount of uncertainty. Nevertheless, for this reason, thevirtual discourse carried out in the mind of one persononly would be an inappropriate criterion only if fromthe uncertainty of a criterion would follow its inappro-priateness.

The second sub-problem of the criterion-problem re-sults from the ideal character of the ideal discourse'srequirements. Neither a real discourse nor a virtual dis-course, carried out in the mind of one person, can everfulfil these demands completely. However, it is possibleto fulfil the ideal requirements approximately. A result,which only approximately meet the requirements of theprocedure, is necessarily an uncertain criterion. But asmentioned above, uncertainty does not imply useless-ness.

The third sub-problem of the criterion-problem iscaused by the internal structure of the ideal discourse.When I discussed the problem of consensus, I statedthat first of all, even in a potentially infinite ideal dis-course, nobody can be certain whether a once-reachedconsensus is finite or definite, and second, it cannotbe excluded that even in such a discourse there is noconsensus at all, meaning that two contradicting normscan be the result of the ideal procedure. The problemof definiteness causes lesser difficulties. A consensuswhich is the result of a potentially infinite ideal dis-course does not grant final certainty but still such ahigh rate of security that it can be accepted as a crite-rion. The problem of contradiction is more sincere. A ,criterion that does not exclude the possibility that twocontradicting norms are correct seems to fail because ofthe most basic semantic demands of the term "correct" .This problem shall be pursued further in the followingdiscussion of the problem of correctness.

53

4. The Problem of CorrectnessConsidering the problem of correctness three questionshave to be discussed: (a) the problem of the concept ofcorrectness, (b) the problem of objectivity and (c) theaforementioned problem of contradiction. I want to em-phasize that my considerations only apply to practicaldiscourses and therefore only to practical correctnessor truth. How far what is said here is transferable tothe theoretical discourse and the problems of theoreti-cal truth must remain open.

a) Concept and Criterion of CorrectnessThe standard objection against the discourse theory,concerning the concept of correctness, is that it con-fuses the difference between concept and criterion. Tocounter this objection one has to distinguish between acriteria-free and a criteria-laden definition of the con-cept of practical correctness. A criteria-free definitionis achieved if one develops, using ideas of Tarski,14 a se-mantic conception of practical correctness which is ori-ented towards the following equivalence: The sentence"X is obligatory" is correct, if and only if X is obligatory.Such a conception of practical correctness enlightens animportant aspect of the concept of practical correctness.However, there are further aspects of this concept whichcan only be grasped by a criteria-laden procedural defi-nition, as suggested here. Both conceptions are compat-ible. They are not in a relationship of competition butin a relationship of complementarity.P For each thereare purposes which justify them.

l' See A. Tarski, "The Semantic Conception of Truth and the Foun-dation of Semantics", in: Philo,ophy and Phenomenological Rerearch, 4(1943/44),p. 343.115 On the thesis that various truth and correctness theories do notnecessarily stand in a competitive relation, but can also stand in a

54

b) The Problem of Objectivity

More important is the second sub-problem of the cor-rectness-problem, the problem of objectivity. It con-cerns the objection that the links between the conceptsof correctness and of truth with those of discourse andof consensus suggested by discourse theory are not ap-propriate. Correctness and truth are something which isobjective. Discourse and consensus again belong to therealm of subjective conviction and mere believing andaccepting.P The discourse theory does not distinguishsufficiently between regarding as true and being true.!?The fact that all agree to a sentence does not mean thatit is correct or true, because all can be mistaken. Thisalso applies to the result of ideal discourses. IS

This objection partly expresses misunderstandings,partly leads to very basic questions. It would be a mis-understanding to think, that according to discourse the-ory something is true already when all think it is true.It is not the consensus that is decisive. But rather theperformance of the discourse procedure. This goes so faras to acknowledge that even in a dissent the contradict-ing opinions can be labeled as "correct" in a sense tobe specified later, provided that they have survived thediscourse procedure. It is therefore incorrect to blamediscourse theory for taking the consensus as a reasonfor correctness or truth.l?

Not the consensus but the perfomance of the proce-

complementary relation, compare O. Hoffe, "Kritische Uberlegungenzur Konsensustheorie der Wahrheit (Habermas)", in: Philo.ophi.cherJahrbuch, 83 (1976), pp. 315 H.HI O. Weinberger (n. 9), pp. 188 H.17 K.-H. Ilting, "Geltung als Konsens", in: 10 neue helle lur plt.ilo.opllie,(1976), p. 36.III O. Weinberger (n. 9), p, 192.III So A. Wellmer (n. 1), p. 72.

55

dure, according to the discourse rules, is the real cri-terion of correctness in discourse theory. The decisivequestion is, whether performing a procedure accordingto the basically formal discourse rules is related to thesubstantial correctness of normative propositions at all.This question formulates the main problem concerningthe relation between procedure and correctness. An an-swer can be successful only if a premise essencial fordiscourse theory is revealed. Discourse theory presup-poses that the participants of discourse, that is, humanbeings, such as they actually exist, are generally ableto distinguish between good and bad reasons for sub-stantial propositions. It assumes that there is, in gen-eral, a sufficient potential or power to form consideredjudgements on the part of the partlcipants.P? This doesnot mean that a sufficient power of judgement is a re-quirement of the procedure.P The relation between thediscourse procedure and the sufficient power of judge-ment on the part of the participants is analogous ratherto that between the constitution of a democratic stateand the ability of its citizens to take part in political,economical and social activities. The latter is not de-manded by norms of the constitution, but presupposedby the constitution. Further, it should be added herethat sufficient power of judgement is presupposed onlyin general. It is one of the purposes of the procedure ofthe discourse to develop this faculty.

If the presupposition of it generally existent sufficientpower of judgement, that is, an ability to distinguishbetween good and bad reasons for substantial proposi-tions, is introduced in this way as a link between pro-

20 The absence of such a "missing link" between the discourse rulesand the correctness of results has often been criticized. Compare e. g.o. Haffe (n. 15), p, 330; K.-H. Hting (n. 17), p, 34.21 On this version see A. Wellmer (n. 1), p. 72.

56

cedure and correctness, then one comes to ask why itis still the procedure that is decisive and not simplygood reasons or sufficient justifications.P The reasonfor this is that there are, at least in practical questionswhich are essentially concerned with the interpretationand reconciliation of interests, no good reasons as such.What actually is a good reason can only be revealed inthe process of discoursive investigation. Applying theconcepts of subjectivity and objectivity, one might saythat the result of the discourse procedure is neither justsubjective nor just objective. It is subjective insofar asit is based on characteristics of the participants. It isobjective insofar as it has proved itself resistant againsta discoursive investigation, resting on the general powerof judgement of the participants. In this way, discoursetheory avoids both the flaws of subjectivistic or rela-tivistic and objectivistic moral theories.

c) The Problem of ContradictionThere remains the third sub-problem of the correctnessproblem, the problem of contradiction, which resultsfrom the fact that it cannot be excluded, that evenan ideal practical discourse can have two contradictorynorms as its results. According to the definition of "cor-rect" given here, this means that the two contradictorynorms can be correct simultaneously. One has to remarkthat this does not imply that one and the same personcan maintain contradictory norms. For the system ofnorms of every single person the postulate of noncontra-diction remains valid. Only incompatible norm systemsof different persons are admitted. The question remainswhether incompatible parts of the norm systems of dif-

22 Compare A. Wellmer [n. 1), p, 70, 72; O. Weinberger [n. 9),pp. 190 ff.

57

ferent persons, if and because they have survived theprocedure, can be equally labeled as "correct".

This would not be admissable, if there were just oneright answer to each practical question,23 independentlyof whether there is a procedure to find or to prove it.Those who adhere to this thesis separate the conceptof correctness from the concepts of justifiability andprobability. From this an absolute concept of correct-ness arises, which has a non-procedural character. Itwould indeed exclude the possibility to label both Nand oN as "correct". The problem is, that the assump-tion that there exists one right answer to each practicalquestion independently of a procedure is an ontologicalthesis, which is not only difficult to justify, but also notvery plausible. Answers to practical questions rest notonly, but essentialy on the interpretation and weighingof interests. It cannot be assumed on this basis thatthere is just one right answer to each practical ques-tion. The thesis of the existence of one right answer toeach question is, at least as far as practical questions areconcerned, a non-justifiable ontological fiction. Such afiction is not sufficient to determine how the term "cor-rect" is to be used.

The fact that the thesis of the existence of one rightor correct answer to each practical question must begiven up, does not entail that the concept of correct-ness has no absolute character at all. It has an absolutecharacter as a regulative idea. As a regulative idea, theconcept of correctness does not presuppose that therealways already exists a right or correct answer to eachpractical question, which only has to be found.24 The

23 See R. Dworkin, A Matter of Principle, Cambridge, Mass./London,1985, pp. 119 ff.

2<1 See I. Kant, Kritik der reinen Vernunft, A 509, B 537: "ein Prinzipiumder Vernunft, welches, als Regel, postuliert, was von uns im Regressus

58

only correct answer rather acquires the character of agoal to be strived for.25 The participants in a practicaldiscourse have to claim that their answers are the onlycorrect ones independently of whether there is one singlecorrect answer, if their statements and justifications aresupposed to make sense. This only presupposes that itis possible that there are practical questions, for whichone single correct answer can be found in a discourse,and that it is not certain what questions these are, sothat it is worthwhile to try to find the one single cor-rect answer to each question. This expresses an absoluteprocedural conception of correctness. It meets up withthe meaning of the term "correct" in ordinary languagewithout any problems.

The problem of contradiction thereby leads to a split-ting of the concept of correctness, into an absolute and arelative one. If both N--,N are results of the procedure,then they are both correct relative to this procedure.i"The absolute procedural concept of correctness againdemands that the search for only one answer is con-tinued. The concept of relative procedural correctnessplays its most important role as far as real discoursesare concerned. It will be made clear within the courseof their discussion.

geschehen soll, und nicht antizipiert, was im Objekte vor all em Regres-sus an sich gegeben ist , Daher nenne ich es ein regulatives Prinzip derVernunft" 0

25 See I. Kant (n. 24), A 644, B 672: "Dagegen aber haben sie einenvortrefflichen und unentbehrlich notwendigen regulativen Gebrauch,nimlich den Verstand zu einem gewissen Ziele zu richten, in Aussichtauf welches die Richtungslinien aller seiner Regeln in einen Punktzusammenlaufen" .28 The conception of relative correctness proposed here is not theonly one possible. A conception oriented towards the idea of differentcognitive processing mechanism has recently been proposed by NoUn-win, "Beyond Truth: Towards a New Conception of Knowledge andCommunication", in" Mind, 96, (1987), pp, 299 ff.

59

IV. The Real Discourse

1. The Discoursive Modalities

To discuss the problem of real discourse it is useful touse a simple model. Assume two persons, at and a2, tryto answer a practical question by means of the proce-dure defined by the discourse rules. At the time tt, atmaintain Nt, and a2 maintains N2, N, and N2 beingincompatible. At the time t2 which marks the end ofthe procedure, for example the following assignmentsof solutions to the participants are possible: (1) Bothagree on Ni, which can be identical with Nt or N2, butneed not be; (2) both reject Ni; (3) at supports N, anda2 supports Nj, i f:. i. It seems to make sense to markthese three cases terminologically. In the first case N,is, relative to the discourse rules, the degree of their ful-filment, the participants, and the timepoint t2, discour-sively necessary. In the second case N, is discoursivelyimpossible respectively. In the third case N, and Nj arerelative to the discourse rules, the degree of their ful-filment, the participants and the timepoint t2, neitherdiscoursively impossible nor discoursively necessary, butonly discoursively possible. It is important to note here,that the picture in time t3 and with participation ofother individuals can look different.

According to the definition D of correctness givenhere, both what is discoursively necessary at the end ofthe procedure as well as what is at this time only dis-coursively possible must be labeled as "correct" . The in-troduction of the concept of discoursive possibility how-ever makes the problem of contradiction less severe. Thefact that both N and .., N can be correct does merelymean that both Nand .., N are possible in a specificway, namely discoursively possible. That both parts ofa contradictory conjunction are possible is logically un-

60

objectionable. What at first glance seems to be a verycounterintuitive result, namely that two contradictorynorms can be equally correct, thereby looses some of itsproblematic character.

2. The Relative Concept of Correctness

The relativity of the concept of correctness raises themost severe of the problems presently discussed. Therelativity of correctness has four aspects: (1) the dis-course rules, (2) the degree of their fulfilment, (3) theparticipants and (4) the points of time.

I shall not discuss the first problem here, the relativ-ity in respect of the discourse rules. This problem doesnot concern the applicability and status problems, butrather the justification problem, not to be investigatedhere. After all, it should be apparent that, and how, thisproblem as well, relates to the problem of correctness.

Second, the fulfilment of some rules is an all-or-noth-ing affair, while others can be fulfiled only approxi-mately because of their ideal character. In the lattercase fulfilment is a matter of degree. This causes thesecond problem of relativity, the problem of relativ-ity in respect to the degree of fulfilment. What is es-sential here, has been said already when dealing withthe criterion problem in discussing the ideal discourse.Merely approximate satisfiability leads necessarily touncertainty of the discourse as a criterion, though, ashas been stated, uncertainty does not imply uselessness.

Third, the relativity in respect to the participantsdoes cause uncertainty as well. Moreover, it leads, atleast as far as real discourses limited in time are con-cerned, to a broadening of the scope of what is onlydiscoursively possible. On the other hand, because ofthe structure of practical problems, uncertainty can-

61

not be avoided. It demands, as shown above, a non-monological procedure. Further, every discourse musthave a starting point. It cannot start from nothing.Its starting points are the factual existing normativeconvictions of the participants, whatever they are. Thediscourse is nothing but a procedure of rational investi-gation of these convictions. In this process any norma-tively relevant conviction is a candidate for a changecaused by rational argument. This restriction to therational structuring of the process of argumentationamounts to an important advantage of discourse theory.A more demanding theoretical alternative might try todetermine the outcome of the procedure. To this end itwould have to establish certain substantial normativetenets as starting premises, instead of merely structur-ing the process of argumentation or decision. This al-ternative has to overcome at least two severe obstacles.First, objections to the substantial starting premises,chosen by the theorist, are usually harder to answerthan objections against the basically formal discourserules. Second, it has to come to grips with the gen-eral objection that the theorist treads ground he shouldrather leave to the participants of the discourse, be-cause his substantial normative convictions cannot beassumed to be generally more correct than those of theparticipants themselves. It should be further noticedthat the role of the participant is open to the theoriststoo at any time.

Fourth, relativity in regard to points of time is in-escapable because of the limits to every real discourse.As far as merely discoursively possible results, that is,dissent, is concerned, the regulative idea of correctnessdemands the non-final character of the results. Withdiscoursively necessary results, that is, consensus, the

62

necessarily non-final character is caused by the neces-sary imperfection of any real discourse.

One may be tempted to think, that considering thefulfilment, participant and point of time relativity oneshould give up the concept of correctness. Instead, onecould suggest that concepts like that of warranted as-sertability, plausibility, justifiability and reasonablenessshould be used. This indeed would have the advantage,that one would not have to go on with a bifurcatedconcept of correctness: that of absolute procedural cor-rectness, which has the character of a merely regula-tive idea, and that of relative procedural correctness,which includes a high degree of uncertainty. As often,the terminological problem reflects serious substantialquestions. At least it can be said, in favour of the termi-nology preferred here, that the colloquial way of speechadmits saying that the adherents of two wellgroundedbut incompatible normative opinions are both correctin their way (relative correctness), in order to continueasking who really is right (absolute correctness), at thesame time admitting that probably nobody is going tofind out. The substantial reason in favour of the termi-nology chosen here is that the discourse leads, if not tocertainty, at least out of the realm of mere opinion andsubjective belief. Taking into account that more cannotbe achieved in practical questions, the use of the conceptof relative correctness seems to be appropriate.

Although many objections against discourse theorycould be answered, the result remains strangely unsat-isfying. On the one hand, the concept of correctnessweakens to a regulative idea, on the other hand it ishighly relativised and overburdened with uncertainties.This must not be the last word, however. It may be true,that the practical value of discourse theory will showits full extent only if it is taken as the basic theory

63

of the institutions of a democratic constitution whichincorporate the idea of discussion.f" But to tackle thesequestions'" would require a new paper.

n Compare the German Federal Constitutional Court, in: BVerfGE5, 85 (197 ff).:18 See M. Kriele, Recht and prakti,che Vernunft, Gottingen, 1979, pp. 30ff.; R. Alexy (n. 4), pp, 185 ff.

64

RESUMEN

En la teorfa del discurso hay tres categories de problemas. Elprimer grupo de problemas se presenta cuando Ia teorfa del dis-curso se considera como una teorfa de la verdad. El segundo grupoinvestiga su aplicabilidad practice, y el tercero se ocupa de la jus-tificaci6n de la teorfa del discurso, Nos referimos al estatuto deIa teorfa del discurso como teorfa de la verdad euando discutimoslas relaciones entre los conceptos de verdad y de correcci6n y losconceptos de consenso, de discusi6n ilimitada y de racionalidad.Nos enfrentamos al problema de la aplicabilidad si a la teorfa deldiscurso se le acusa de carecer de contenido y de ser meramenteformal, 10 cual se manifiesta en el hecho de que no conduce a ningtinresult ado definitivo. El problema de la iu.tificacion es el problemade establecer razones concluyentes para las reglas y los principiosdel discurso.

Aquf se discuten solamente los dos primeros problemas: el pro-blema del estatuto y el problema de la aplicabilidad. Ambos pro-blemas pertenecen a todas las formas de discurso. Se presentan,por ejemplo, en el discurso te6rico, en el discurso practice y enel estetico, etc. En el articulo se considera iinicamente el discursopractice,

65