1 Introduction

Pragma-dialectics is a predominantly, but not exclusively, dialectical approach to argumentation theory. Although criticisms of the distinction of three basic perspectives in argumentation theory, logic, rhetoric, and dialectics, and their use to classify theoretical proposals in that field have been many and varied, that division remains a useful, if not almost obligatory, starting point. By saying that the predominant orientation of pragma-dialectics is dialectical, I mean that it focuses on argumentative procedures. The introduction in 2002 of the concept of strategic maneuvering represents the passage from standard pragma-dialectics to extended pragma-dialectics, which integrates a rhetorical dimension into a dialectical framework. The rhetorical perspective is characterized by its focus on argumentative processes and the effectiveness of arguments as instruments of persuasion. As van Eemeren and Houtlosser put it, “rhetoric is the theoretical study of the potential effectiveness of argumentative discourse in convincing or persuading an audience in actual argumentative practice” (2006: 383).

What about the logical perspective? Biro and Siegel (1992, 1997, 2006, 2008) criticize pragma-dialectics for lacking normative epistemic criteria for evaluating the logical quality of arguments:

all of these accounts, while they focus on the epistemic status of the conclusion, fail to focus on the epistemic relationship between premises and conclusion. In so doing, they in fact altogether fail to provide a theory of argument, since an argument is essentially a matter of such relationships. (1997: 284)

Indeed, when one surveys the development of pragma-dialectics, one gets the impression that the logical dimension has been blurring, up to van Eemeren and van Haaften (2023), where they state that the two dominant approaches in argumentation theory are the dialectical and the rhetorical, and the central issue in argumentation theory is the relation between dialectical reasonableness and rhetorical effectiveness (Op. cit.: 341, 343). They thus imply that there is no logico-epistemic dimension to the study of argumentation, or that if there is one, it is incidental.

I will try to show that the logical perspective—the study of arguments as products, according to Wenzel’s (2006) well-known and disputed characterization—is not well integrated into pragma-dialectics. Blair has already argued that “the Pragma-Dialectical handling of logic does need some sorting out” (2010: 1), for reasons partly coinciding and partly diverging from mine. I contend that in order to account for the normative dimension of argumentative practices pragma-dialectics must incorporate the logical perspective, agreeing in this with Blair, Biro and Siegel. However, I will contend that the integration of the logical perspective into pragma-dialectics requires the adoption of a conception of logic as a theory of reasons (what I call ‘reasonism’), as opposed to the traditional conception of logic as a theory of inferences (what I call ‘inferentism’), subscribed to by Blair, Biro and Siegel, thus distancing myself from their recommendations.

Here is a brief tour of the paper to follow. In Sect. 2 I will explain the differences between dialectic and rhetorical rules, which are rules of action, and logical rules, which are standards for judgment. In Sect. 3 I will argue that although logical rules have no place in a decalogue of dialectical rules, the rules of validity and argument scheme express an underlying conception of logical validity. Consequently, changes in the formulations of these rules entail changing concepts of validity, which I discuss in Sect. 4, to show that neither is adequate. To propose an alternative, in Sect. 5 I examine the critical discussion model, to show that the integration of the logical dimension requires thinking of logic as a theory of reasons, not as a theory of inferences, and in Sect. 6 I explore some consequences of this shift.

2 Rhetorical, Dialectical and Logical Rules

We can find at least three types of rules in argumentative practices: rhetorical rules, dialectical rules, and logical rules.

Rhetorical rules are instrumental to promote the ends pursued by arguers, that is, those who present something to someone as a reason for something else. Their empirical justification is that following them facilitates the achievement of the ends pursued by the arguer through argumentation. Here is an example of a rhetorical rule:

[RR] It is advisable to include a summary of what is going to be discussed, listing concisely the most important issues, since the audience will be more inclined to listen to the speaker if they know from the beginning what is going to be discussed. (Martín Jiménez 2019: 20; my translation)

Martín Jiménez states the rule and gives a reason for accepting it, preceded by the connector ‘since’: if you include a summary, the audience will be more inclined to pay attention to you, which is a necessary condition of persuasion. We could say that following a rhetorical rule is in the interest of the arguer.

Dialectical rules are also rules of action, which prescribe, prohibit, or permit a participant in a discussion to perform an action in certain circumstances. Unlike rhetorical rules, dialectical rules are rules established by convention to organize argumentative exchanges for critical purposes. The pragma-dialectical Obligation to Defend Rule is an example of a dialectical rule:

[ODR] Discussants who advance a standpoint may not refuse to defend this standpoint when requested to do so.

The validity of such dialectical rules has, according to van Eemeren and Grootendorst (2004: 17), a double justification. First, following them helps to organize the discussion efficiently (problem validity). Second, they have been previously accepted by the participants, either explicitly or by the fact of agreeing to participate in that practice (intersubjective or conventional validity). Actually, that double justification answers two different questions.

  • —Why is it convenient that whoever advances a standpoint in an argumentative exchange should, if asked, defend it with reasons?

  • —Why should I defend my standpoint by giving reasons?

The answer to the first question is that observing that rule facilitates achieving the ends of the discussion, or, as van Eemeren and Grootendorst put it, “The proposed procedural rules are valid as far as they really enable the discussants to resolve their differences of opinion” (van Eemeren & Grootendorst 2004: 16). This first question can be asked by the argumentation theorist or by the participants in a forthcoming debate when they are agreeing on its rules. The second question, however, has to be put in the mouth of a participant, at a given moment of the debate, and in that case, the answer is “because we have agreed to do so”.

Logical rules, on the other hand, are perceived as neither conventional nor instrumental by participants, unlike dialectical and rhetorical rules, and in that sense, it can be said that logic is “an external” constraint that imposes an intrinsic rationality on argumentative exchanges (Blair 2010: 12). Logical rules are standards for judging whether an argument or argumentation is valid, using ‘valid’ as an undefined logical evaluation term. So throughout this article, “valid” does not mean formally valid. Logical rules are therefore not rules of action and are more like clauses of a definition. Consider the archetypal example of the modus ponens rule. There are multiple formulations of that rule, and in some of them the modus ponens rule looks like a rule of action (e.g. “The consequent of a conditional can be asserted if its antecedent is asserted”, Ferrater Mora 1965: 220). Copi, Cohen, and McMahon offer a standard formulation in logic textbooks:

[MPR] A syllogism that has one conditional premise and one categorical premise is called a mixed hypothetical syllogism. […] Any argument of this form [the categorical premise affirms the antecedent of the conditional premise, and the conclusion affirms its consequent] is valid and is said to be in the affirmative mood or modus ponens. (Copi et al. 2014: 284)

Or, more perspicuously:

[MPR] Any argument that has as premises a conditional and its antecedent, and as conclusion the consequent of that conditional, is valid.

MPR does not prescribe, prohibit or permit any action to anyone, but rather seems to be a clause in some definition of ‘valid argument’.

3 The Logical Dimension of Pragma-dialectical Rules

The pragma-dialectical decalogue is a set of dialectical rules, and hence of action rules, in which logical rules, which are not action rules, have no place. Nevertheless, any set of action rules that purports to define some kind of argumentative reasonableness must prescribe participants to propose arguments that they consider valid (in the negative, prohibit them from using arguments that they consider invalid).

The pragma-dialectical decalogue is not an exception. As the formulation of the rules I am interested in has changed over time, I will take as a reference the formulation in the comparatively recent Argumentation Theory: A Pragma-Dialectical Perspective (2018: 59–61). The pragma-dialectical decalogue is, then, composed of the following procedural rules.

R1. Freedom Rule: Discussants may not prevent each other from advancing standpoints or from calling standpoints into question.

R2. Obligation to Defend Rule: Discussants who advance a standpoint may not refuse to defend this standpoint when requested to do so.

R3. Standpoint Rule: Attacks on standpoints may not bear on a standpoint that has not actually been put forward by the other party.

R4. Relevance Rule: Standpoints may not be defended by non-argumentation or argumentation that is not relevant to the standpoint.

R5. Unexpressed Premise Rule: Discussants may not falsely attribute unexpressed premises to the other party, nor disown responsibility for their own unexpressed premises.

R6. Starting Point Rule: Discussants may not falsely present something as an accepted starting point or falsely deny that something is an accepted starting point.

R7. Validity Rule: Reasoning that is in an argumentation explicitly and fully expressed may not be invalid in a logical sense.

R8. Argument Scheme Rule: Standpoints defended by argumentation that is not explicitly and fully expressed may not be regarded as conclusively defended by such argumentation unless the defense takes place by means of appropriate argument schemes that are applied correctly.

R9. Concluding Rule: Inconclusive defences of standpoints may not lead to maintaining these standpoints and conclusive defences of standpoints may not lead to maintaining expressions of doubt concerning these standpoints.

R10. Language Use Rule: Discussants may not use any formulations that are insufficiently clear or confusingly ambiguous, and they may not deliberately misinterpret the other party’s formulations.

‘Sound’ is another term of logical evaluation: an argument is sound iff it is valid and its premises have a certain designated property (acceptability, assertability, truth, or whatnot; see discussion in Freeman 2004). If, as Biro and Siegel claim, logic is fundamentally about the epistemic relations between premises and conclusions, what that property may be is irrelevant to the present discussion. I maintain, with Habermas, that logic also studies the epistemic relations between arguments: logic studies “the structures that determine the construction of individual arguments and their interrelations.” (Habermas 1984: 26, italics added).

Rules 6, 7 and 8 can be called ‘logical rules’ in that they require participants to use arguments that they consider sound—or at least prohibit them from using arguments that they do not consider sound. So, these rules express an underlying concept of soundness that we can formulate as follows:

An argument P1,…,Pn, so C is sound just in case either (1) each Pi, 1 ≤ i ≤ n, is an accepted starting point, and either P1,…,Pn, so C is formally valid as it stands, or it results from the correct application of an appropriate argumentation scheme, or (2) each Pi, 1 ≤ i ≤ n, is either an accepted starting point or the conclusion of a sound argument offered during the discussion, and either P1,…,Pn, so C is formally valid as it stands, or it results from the correct application of an appropriate argumentation scheme.

Note that the second conjunct of (1) and (2), “eitherP1,…,Pn, so C is formally valid as it stands, or it results from the correct application of an appropriate argumentation scheme”, is, in turn, a definition of a valid argument: an argument is valid iff either it is formally valid or it results from the correct application of an appropriate argumentation scheme.

4 The Changing Pragma-dialectical Definition of Validity

As I have anticipated, the Validity Rule and the Argument Scheme Rule have known different formulations (and even numberings!) throughout the history of pragma-dialectics, leading to different notions of validity. So far as the “logical rules” are concerned, I distinguish three periods in the evolution of pragma-dialectics, that, for reasons that will become apparent later, I call ‘conventionalist’, ‘inferentialist’ and ‘dualist’.

4.1 The Conventionalist Period of Pragma-dialectics

The conventionalist period predates the formulation of the Validity and the Argument Scheme rules. In Speech Acts in Argumentative Discussions (1984) the logical rules in force are not prior to the critical discussion, but prior to the argumentation stage, since in a critical discussion the participants must agree at the opening stage which logical standards they accept (van Eemeren & Grootendorst 1984: 167). Although van Eemeren and Grootendorst interpret this agreement as a choice between given logics or logical systems associated with different criteria of validity (Op. cit.: 128), what lends normative force to the logical rules is the agreement of the participants, not the fact that they appear in a pre-established catalog of logics. The rules of discussion are proposals that only come into force when they have been accepted by the parties, and when that happens, the rules acquire the status of conventions to which the parties are committed throughout the discussion (Op. cit.: 163). In later works logical rules are not mentioned among the commitments agreed upon by the parties at the opening stage, which refer only to the procedural rules of discussion and propositions that do not require justification. In the pre-1992 articles collected in van Eemeren 2015 I have not found any mention of logical rules within the opening stage.

My main criticism of this ephemeral conventionalist proposal is that logical systems provide rules of implication (consequence having), not rules of inference or reasoning (consequence drawing), as Harman (2002) and Woods (2016) have convincingly argued.

If there are any principles of inference or reasoning, they are normative principles about when it is rational or reasonable to reach a certain conclusion. Principles of implication are not normative (outside of deontic logic) and do not have a psychological subject matter (outside of the logic of belief). (Harman 2002: 171)


If Harman is right, it raises an important question for argument. We would seem to have it that since conditions on consequence-having don’t direct all the traffic for consequence-drawing, the same might well be said of argument. (Woods 2016: 104)

Thus, the choice of a logical system says nothing about whether a standpoint has been successfully defended or not. Van Eemeren and Grootendorst, on the other hand, admit that there are arguments that seem intuitively valid but whose validity cannot be demonstrated in any of the available logics (Op. cit.: 128).

4.2 The Inferentialist Period of Pragma-dialectics

The inferentialist period begins in 1992, with the publication of Argumentation, Communication and Fallacies, and ends in 2004, the year of publication of A Systematic Theory of Argumentation, that inaugurates the dualist period, which extends to the present day.

During the inferentialist period, Logical rules 7 and 8 were formulated as follows (van Eemeren & Grootendorst 1992: 159, 169):

Rule 7. A party may not regard a standpoint as conclusively defended if the defense does not take place by means of an appropriate argumentation scheme that is correctly applied.

Rule 8 In his argumentation a party may only use arguments that are logically valid or capable of being validated by making explicit one or more implicit premises.

Although Rule 7, as it stands, states a sufficient, but not necessary, condition, van Eemeren and Grootendorst point out that:

In case there are enough mutually acceptable starting points and argumentation schemes and it is perfectly clear what they are, it is, in principle, possible to answer the question whether an argumentation constitutes a conclusive defense for a standpoint. If both the identification procedureFootnote 1 and the testing procedureFootnote 2 produce a positive result, the standpoint has indeed been conclusively defended. (Op. cit.: 159)

Thus, if a party defends a standpoint by correctly applying an appropriate scheme to mutually accepted starting points, it may be regarded as having defended it conclusively. But then, rule 7 provides necessary and sufficient conditions for validity (i.e., all the conditions for a conclusive defense that are not related to the acceptability of the premises), and therefore rule 8 seems dispensable for the purposes of logical evaluation. In fact van Eemeren and Grootendorst argue that every fully explicit argument is logically valid, which deprives rule 8 of any normative value.

The logical validity of the argument concerned is then more or less automatically guaranteed because making unexpressed premises explicit starts with formulating the ‘logical minimum’ that links the explicit premise in a logically valid way with the conclusion. (Op. cit.: 169)

The procedure for identifying the hidden premises of an argument begins by formulating the logical minimum or conditional associated with the argument, if premises then conclusion. In an inferentialist vein à la Brandom,Footnote 3 van Eemeren and Grootendorst warn that the associated conditional only makes explicit a commitment made when proposing the argument, without adding anything new, and therefore is superfluous and cannot be regarded as a premise (Op. cit.. 64). Making the logical minimum explicit serves to identify as a hidden premise the pragmatic optimum, which results from generalizing the logical minimum as much as possible without imputing implausible commitments to the arguer. Their words imply that the logical minimum follows deductively from the pragmatic optimum, and therefore, by adding it as a hidden premise, the argument becomes logically valid. Thus, the Validity Rule ceases to be a logical rule and becomes a guide for argument reconstruction. In fact, one of Blair’s criticisms is that the pragma-dialectical explicitization procedure is limited in scope for being too deductivist (2010: 7).

Thus the Validity Rule is, in this period of pragma-dialectics, a kind of argumentative maxim. Just as participants in a communicative exchange presuppose that their interlocutors observe the conversational maxims of quantity, quality, relation, and manner, and apparent transgressions enable them to grasp what the speaker implies without saying it (Grice 1975), those who participate in a critical discussion assume that everyone proposes logically valid arguments, and apparent transgressions of this assumption enable them to identify the premises they imply. So, the Validity Rule, like the Gricean maxims, is descriptive, not normative.

Another consequence of the above is that van Eemeren and Grootendorst adhere to a generalist conception of arguments,Footnote 4 according to which to argue is to apply a general rule or principle, viz. the pragmatic optimum, which plays a role similar to Toulmin’s warrant. Van Eemeren and Grootendorst also point out that the pragmatic optimum usually turns out to be a useful indicator of the argument scheme being used in the argument (1992: 232, 219).

If I am right, at this period pragma-dialectics adheres to a sort of avant la page inferentialist conception of formal logic (Making it Explicit was published in Brandom 1994), according to which formal logic studies the inferential apparatus of our concepts that serves to make explicit the commitments of acts of asserting. Therefore, formal logic is not properly normative, and does not provide an alternative criterion of validity to that provided by the theory of schemes. Let us use an example from Van Eemeren and Grootendorst (Op. cit.: 62):

  1. 1.

    Bart is Dutch.

  2. 2.

    Therefore, Bart will love cheese.

From an inferentialist approach to logic, whoever asserts ‘Bart is Dutch, therefore Bart will love cheese’ implies (i.e., indirectly expresses) ‘if Bart is Dutch, then Bart will love cheese’. The function of the conditional is to make explicit, in the form of a declarative sentence, the inferential commitment implicit in the argument (Brandom 1994: xix).

The valid argument resulting from this addition [i.e. of the logical minimum] has the form of modus ponens. So the logical minimum amounts to connecting the pieces of information that are already there: All it does is to state explicitly that it is permitted to infer the given conclusion from the given premise.

Pragmatically, this is not enough. From the very fact that he advances this particular argumentation for his standpoint it is already clear that the speaker assumes that this conclusion follows from this premise. The logical minimum contributes nothing new and is, therefore, superfluous. (Van Eemeren and Grootendorst 1992: 64)

Thus, what “follows” from ‘Bart is Dutch’ and ‘if Bart is Dutch, then Bart will love cheese’, is that the speaker asserting those declarative sentences is committed to Bart will love cheese, not that those premises provide sufficient support for the conclusion. Van Eemeren and Grootendorst (Op. cit.: 60) also warn that someone who claims that ‘Dutchmen love cheese’ and ‘Bart is Dutch’, commits herself to Bart will love cheese without the need to explicitly state it. It would therefore be incorrect and misleading to say that the premises ‘Bart is Dutch’ and ‘if Bart is Dutch, then Bart will love cheese’ lend support to the conclusion that Bart will love cheese because the corresponding argument is formally valid.

Blair claims that the logic used in the explicitness procedure makes it incompatible with the use of the theory of argument schemes in the analysis and evaluation of arguments (2010: 1). I believe, however, that the problem does not lie in the use of formal logic in the process of making explicit the hidden premises, but in the assumption of a generalist view of arguments. Van Eemeren and Grootendorst consider that ‘Dutchmen love cheese’, or some other general statement, depending on the context, is indeed a hidden premise of the argument. They justify this because addition as a premise of the pragmatic optimum makes the argument valid and, unlike the addition of the logical minimum, does not transgress the communication rule ‘Be efficient’. But, on the one hand, if the one who argues that Bart will love cheese because Bart is Dutch implies—that is, communicates—if Bart is Dutch, then Bart will love cheese, then his argument is already valid. On the other hand, the fact that the pragmatic optimum does not violate the rule of communicative efficiency because it is not obvious that the speaker is committed to it is a reason not to consider it part of his argument. So the pragmatic optimum cannot in any way be considered a hidden premise.

Be that as it may, in this inferentialist period of pragma-dialectics the only criterion of validity properly speaking is the correct application of an adequate argument scheme.

An argument P1,…,Pn, so C is sound just in case either (1) each Pi, 1 ≤ i ≤ n, is an accepted starting point, and, once added the implicit premises Q1,…, Qm, the argument P1,…,Pn, Q1,…, Qm, so C is an instance of an appropriate argument scheme applicable to the case at hand, or (2) each Pi, 1 ≤ i ≤ n, is either an accepted starting point or the conclusion of a sound argument offered during the discussion, and once added the implicit premises Q1,…, Qm, the argument P1,…,Pn, Q1,…, Qm, so C is an instance of an appropriate argument scheme applicable to the case at hand.

4.3 The Dualist Period of Pragma-dialectics

The rules of Validity and Argument Scheme of Argumentation Theory: A Pragma-Dialectical Perspective (2018) appear already, with minor differences, in A Systematic Theory of Argumentation. The pragma-dialectical approach (2004: 193–194):

R7. Reasoning that in an argumentation is presented as formally conclusive may not be invalid in a logical sense.

R8. Standpoints may not be regarded as conclusively defended by argumentation that is not presented as based on formally conclusive reasoning if the defense does not take place by means of appropriate argument schemes that are applied correctly.

The rules of this period are dualist because they distinguish two types of arguments, with different associated standards of validity. The difference is that in A Systematic Theory an argument belongs to one or the other class depending on whether it is presented as formally conclusive or not, while in Argumentation Theory it depends on whether the argument is explicitly and completely expressed or not.

Other authors have proposed criteria similar to the one proposed in A Systematic Theory to distinguish two types of arguments. For example:

Before going on to see both types of arguments in more detail, it is convenient to insist that their difference lies exclusively in the pretension of the speaker. Deductive arguments are characterized by the fact that they claim that the truth of the premises makes the truth of the conclusion certain, while in inductive arguments it is claimed that the premises support the conclusion only to a certain degree. But in principle, and except for conventions which we can always adopt, nothing formal or structural distinguishes deductive arguments from inductive ones; the difference is intentional, it lies exclusively in the intentions of the speaker with respect to the intended sense in which the conclusion follows from the premises. (Díez & Moulines 1997: 39; my translation)

But as far as I know, there is no precedent to the criterion proposed in Argumentation Theory.

The use of disjunction in the clauses of the following definition, already stated in Sect. 3, evidences its dualistic character:

An argument P1,…,Pn, so C is sound just in case either (1) each Pi, 1 ≤ i ≤ n, is an accepted starting point, and either P1,…,Pn, so C is formally valid as it stands, or it results from the correct application of an appropriate argumentation scheme, or (2) each Pi, 1 ≤ i ≤ n, is either an accepted starting point or the conclusion of a sound argument offered during the discussion, and either P1,…,Pn, so C is formally valid as it stands, or it results from the correct application of an appropriate argumentation scheme.

In a footnote to A Systematic Theory, van Eemeren and Grootendorst attribute the change in the formulation of the Validity Rule and the Argument Scheme Rule, entailing the shift from inferentialism to dualism, to Erik Krabbe, while confirming my interpretation of the explicitization procedure in the previous period:

For the pragma-dialectical analysis of unexpressed premises, see van Eemeren and Grootendorst (1992: 60–72). According to this method, identifying an unexpressed premise involves first validating the reasoning as an intermediary heuristic step in the reconstruction procedure and then determining the “pragmatic optimum” that may in the context concerned be regarded as the unexpressed premise (which can result in an argument that is, strictly speaking, not logically valid). Largely as a result of Erik C.W. Krabbe’s useful comments in describing the reconstruction procedure in this way, and in phrasing Commandment 7 in the way we did, we deviate in some respects from recent descriptions as given in van Eemeren, Grootendorst, and Snoeck Henkemans. (van Eemeren & Grootendorst 2004: 194, fn.20)

The dualist stance is unsatisfactory. When someone asserts ‘Bart is Dutch, therefore Bart will love cheese’, she commits to ‘if Bart is Dutch, then Bart will love cheese’, so that making it explicit adds nothing, and it is difficult to see how this move can affect the logical properties of the argument. But according to the dualist description, if I say ‘Dutchmen love cheese and Bart is Dutch, so Bart will love cheese’, my argument must be evaluated with formal logical criteria, and if on the contrary I say ‘Bart is Dutch, so Bart will love cheese’, the standard of evaluation is conformity to an appropriate argument scheme, although in both cases I would have expressed exactly the same argument. Even worse, in the second case we are entitled, if it is not clear what the instantiated argument scheme is, to add the unexpressed premise ‘Dutchmen love cheese’ (van Eemeren & Grootendorst 2004: 194; 2018: 56). Blair (2010: 14) had rightly argued that if pragma-dialectics wanted to adopt a dualist, or better pluralist, approach to logic, it had to revise its procedure of making explicit unexpressed premises.

5 Beyond the Basic Model of Critical Discusion

The outcome of Sect. 4 is that the inferentialist account is preferable to the conventionalist and dualist accounts, although the unforced assumption of a generalist position is problematic. Diagnosis done, let’s look for the remedy: how can the logical dimension be accommodated within the framework of pragma-dialectics?

The ideal model of critical discussion, which includes the pragma-dialectical decalogue, is designed for comparatively simple varieties of a certain type of argumentative exchange; viz., a persuasion dialogue, in the terminology of Walton & Krabbe (1995). The idea is that, with appropriate adjustments, the critical discussion model can be applied to more complex persuasion dialogues as well as to dialogues of any other type.

Specifically, the ideal model of critical discussion is initially conceived for single and non-mixed discussions. A discussion is single if the standpoint at issue pertains to only one proposition; otherwise it is multiple. A discussion is non-mixed if only one of the parties has a standpoint and acts as the protagonist of that standpoint; otherwise it is mixed (van Eemeren 2018: 42). I will argue that the adaptation of the basic model of critical discussion to mixed discussions requires a deep revision of the “logical” rules in the Decalogue, which entails a shift from an inferentist to a reasonist conception of logic.

Traditionally logic focuses on the relationship that makes considerations support a position on an issue. That relationship can be explained in terms of inferences or in terms of reasons. So, in argumentation theory there are two paradigms of logic: ‘inferentism’ and ‘reasonism’. For inferentism, ‘P supports C’ means C can be inferred from P or C follows from P. Consequently, a good argument is one in which the conclusion follows from the premises, and logic studies the conditions of validity of different kinds of inferences (deductive, inductive, etc.). For reasonism, ‘P supports C’ means P favors the conclusion C. Consequently, a good argument is one that gives a good reason, and logic studies the dialogical construction of reasons. Since there can be both good reasons for and against C, the fact that P favors the conclusion C does not authorize us to infer C from P without further ado. Obviously, inferentism is by far predominant.

Inferentism should not be confused with inferentialism, alluded to by the label ‘inferentialist period’. Inferentism is a philosophical thesis about the nature of arguments, while inferentialism is a thesis about the content of linguistic acts that replaces the relation of representation between language and the world by that of inference to explain how our linguistic acts acquire content (see fn.3).

In a non-mixed discussion there is a clear division of labor: the antagonist questions and the protagonist is obliged to respond with reasons. Almost all descriptions of the argumentation stage in non-mixed discussions make it clear that only the protagonist argues (i.e., presents something as a reason for something else), while the antagonist merely asks questions, formulates doubts, and accepts or rejects the reasons given, reactions that are subsumed under the label ‘critical responses’ (1992: 35; 2004: 61; 2018: 37). The description in Argumentation. Analysis, Evaluation, Presentation differs from the standard description in that ‘critical responses’ is replaced by ‘objections’:

In the argumentation stage the protagonist defends his or her standpoint against the sometimes persistent criticism of the antagonist by putting forward arguments to counter the antagonist’s objections or to remove the antagonist’s doubts. (van Eemeren, Grootendorst & Snoeck Henkemans 2002: 25)

Although van Eemeren, Grootendorst and Snoeck Henkemans do not explicitly define ‘objection’ (which does not even appear in the subject index of the book), their examples suggest that objections are counterarguments. But such examples always occur in mixed discussions. A passage in Argumentation Theory: A Pragma-Dialectical Perspective also mentions ‘counterclaims’: “the protagonist has to advance argumentation that responds methodically to the questions, doubts, objections and counterclaims put forward, or supposed to be entertained, by the other party” (van Eemeren 2018: 23–24). Again, no definition of counterclaim is provided.

I understand by ‘contraargumentation’ an argumentation in opposition to previous argumentation, and by ‘counterargument’ an argument in opposition to previous argument. Van Eemeren and Grootendorst insist that counterargumentation turns a non-mixed discussion into a mixed discussion (1992: 21, 31 fn.; 2004: 165; 2018: 106). I conclude, then, that in a non-mixed discussion only the protagonist argues, and therefore there is no counterargumentation. Counterargumentation appears with mixed discussions, in which all parties have the right to argue.

To further analyze the pragma-dialectical approach to counterargumentation I distinguish terminologically three basic types of counterargumentation (for details, see Marraud 2020a; Leal & Marraud 2022: 311–328).

  • To object to an argument is to claim that one of its premises is not assertable. Therefore, an objection to an argument A is an argument whose conclusion is incompatible with some of the premises of A. E.g.: A. This patient has a streptococcal infection, so presumably this patient needs penicillin treatment. CA. Infection diagnosis is only based on symptoms, no clinical tests have been performed.

  • To rebut an argument is to claim that its premises do not support its conclusion. Therefore, a rebuttal of an argument A is an argument whose conclusion is incompatible with the logical minimum of A. E.g.: A. As a physician and medical ethicist, I am opposed to any form of physician assistance with a patient’s suicide. The Hippocratic Oath clearly states: “I will neither give a deadly drug to anybody if asked for it, nor will I make a suggestion to this effect.” CA. There was a time and place for the Hippocratic Oath to work, it doesn’t mean it’s always appropriate for all situations far into the future.

  • To refute an argument is to claim that a statement contrary or contradictory to its conclusion can be asserted. Thus, a refutation of A is an argument whose premises are consistent with the premises of A, whose conclusion is inconsistent with the conclusion of A, and which is presented as being as strong as, or stronger than, A. E.g.: A. This patient has a streptococcal infection, so presumably this patient needs penicillin treatment. CA. But this patient is allergic to penicillin, and that rules out penicillin.

Pragma-dialectics holds that the basic type of discussion is the non-mixed single discussion, and the other types (non-mixed multiple, mixed single, and mixed multiple) are to be analyzed as combinations of discussions of the basic type (van Eemeren, Grootendorst & Henkemans, 2002: 8; van Eemeren & Grootendorst 2004: 120; van Eemeren 2018: 98). To see how this principle affects the treatment of argumentation, let us consider my previous example of objection.

Dr. Quarry: This patient needs penicillin treatment.

Dr. Home: Oh, really? Why?

Dr. Quarry: This patient has a streptococcal infection.

Dr. Home: Well, the diagnosis of infection is only based on symptoms, no clinical tests have been performed, so you can’t assert it.

The standpoint of the main discussion is whether the patient should be treated with penicillin. Dr. Quarry takes a positive position and Dr. Home takes a neutral position, so the main discussion is non-mixed. When Dr. Quarry gives a reason to support his standpoint and Dr. Home rejects it, a subdiscussion arises as to whether the patient has a streptococcal infection. This subdiscussion is mixed, because Dr. Quarry adopts a positive attitude while Dr. Home adopts a negative attitude.

Recall that the starting hypothesis is that in a non-mixed discussion, the protagonist builds a chain of inferences that ends at the standpoint to answer antagonist’s questions and doubts. This model is reasonably well suited when the antagonist’s critical responses include objections and rebuttals, inserting mixed subdiscussions into the main discussion. The role of the antagonist is then to examine the starting points of the protagonist’s argument and check the validity of the inference links.

Pragma-dialectics analyzes a single mixed discussion as the combination of two parallel elementary discussions, in which each party develops its own chain of inferences ending in the standpoint it defends, which the other party examines raising questions, objections and rebuttals. These two chains of reasoning are not integrated into a single argument, and thus the parties do not interactively construct a single complex argumentation (what I have called in Leal & Marraud 2022: 38 a macro-argument). This description fits what Blair (2012) calls ‘quasi-engaged dialogue’ and illustrates with Harman's and Thomson's Moral Relativism and Moral Objectivity (1996). The book is in four parts: in the first Gilbert Harman defends a version of moral relativism, in the second Judith Jarvis Thomson does the same with her version of moral objectivism, in the third Harman responds critically to Thomson’s argumentation, and finally Thomson responds critically to Harman’s argumentation:

The conclusion that Harman draws from his defense of moral relativism is presumably incompatible with the conclusion Thomson draws from her defense of moral objectivity, yet the two parts of the book in which each author defends his or her conclusions do not engage each other at all. The authors engage in “dialogue” in the respect that they speak to (opposite sides of) the same issue, but it is a “nonengaged” dialogue because, except incidentally, they do not argue for or against, or question, each other’s arguments. Only in the second part of the book does each co-author take up and argue against the case that each had made in the first part. And even there the dialogue is not completely engaged, since there is no communication between the co-authors about their respective refutations of the other’s case. In sum, part of their dialogue is “non-engaged,” and part is only “quasi-engaged.” (Blair 2012: 238)

Refutation changes things radically, because refutation involves weighing the strength of opposing arguments and is the way in which two opposing argumentations are integrated into a single argumentation. Argumentation is no longer a chain of inferences but a network of considerations with an overall argumentative orientation.

The evaluation of a network of considerations is quite different from the evaluation of a chain of inferences and requires taking into account both the “vertical” relations, premises-conclusion, and also the “horizontal” relations, between arguments, expressed by connectors such as ‘but’ or ‘moreover’. The “logical” rules of pragma-dialectics only examine the link between premises and conclusion, and the Decalogue contains no indication of how to weigh the strength of arguments. Weighing shows that the mere fact that A is a reason for B does not authorize inferring B from A, and dissociates conclusion from argument, linking it to argumentation—that is, with the whole constellation of considerations and counter-considerations adduced in the exchange—in an unmistakably holistic manner.

A holistic approach radically changes the problem of identifying unexpressed premises. On an atomistic view, the logical propertiesFootnote 5 of an argument are completely determined by the properties of its parts (premises and conclusion) and the relationships between them, whereas for holism they also depend on contextual elements that are not part of the argument. Thus, it can be said that for atomism logical properties are intrinsic, context-independent, properties of arguments, whereas for holism they are extrinsic, context-dependent properties (Marraud 2020a, 2020b, 2023; Leal & Marraud 2022). The argument: ‘I’d better not stay any longer, for because I’m so tired you must find me boring company’ (Van Eemeren et al. 2002: 74) presupposes that I can leave or that this is not the last chance for a long time to see each other. For an atomist, that is a reason to believe that these are hidden premises, whereas for a holist they are contextual factors that make the argument conclusive on a given occasion and not on another. Thus, formal validity and soundness are intrinsic properties of arguments, while the corresponding defeasible notions are extrinsic properties since its validity in a particular case depends on conditions of exception or rebuttal that are not part of the argument.

6 A Place for Reasons in Pragma-dialectics

Adding weighing rules implies dropping inferentism in favor of reasonism, because the fundamental difference between inferences and reasons is that the latter is a weighed notion, while the former is not (Lord & Maguire 2016: 4). There are better or worse reasons, but a conclusion is not more or less inferable from a set of data.

An argumentation is a network of mutually relevant considerations that, as a whole, has an argumentative orientation on an issue, i.e., it favors a position on the issue under consideration. This means that to evaluate logically an argumentation the mutual relevance of its constituent considerations (expressed by connectors ‘but’, ‘although’, ‘besides’, etc.) must be checked and its overall orientation determined. It is not enough to examine the inferential relationships of one or more of these considerations separately.

Although pragma-dialectics, in its present state, is inferentist, there are indications that a reasonist conception of logic fits better with its general approach. These clues are threefold: the notion of coordinative argumentation, the pragma-dialectical concepts of argument and conclusion, and the separation of the stages of argumentation and conclusion.

The pragma-dialectical typology of argumentative structures (van Eemeren & Grootendorst 1992, Chap.7; 2004: 120–122; van Eemeren et al. 2002: 64–66) is hardly compatible with an inferentist view of logic. Pragma-dialectics distinguishes four basic argumentative structures: single, multiple, coordinative, and subordinative. Although there are different types of coordinative argumentation, the one that interests me now is the one made up of several arguments (reasons), which separately are too weak to conclusively support the standpoint. That each of the coordinated arguments instantiates an appropriate and applicable argument scheme says nothing about the combined strength of those arguments. Although no account of argumentative force is provided, the notion of coordinative argumentation presupposes that it makes sense to say that the coordinative argument ‘A and moreover B’ has more weight than the argument A. Coordination cannot be explained simply by saying that C follows from A and moreover B, but not from A, because ‘following from’ is not a comparative notion.

The traditional definition of argument, which can be found in any logic textbook, formal or informal, states that an argument is a compound of premises and conclusion. Despite their popularity among philosophers, these are not the concepts of argument and conclusion of linguists, nor, as we shall see, of pragma-dialecticians. The following paragraph illustrates the linguists’ use of ‘argument’ and ‘conclusion’:

Argument and conclusion are correlative terms. The “argument—conclusion” relationship is expressed, accurately by expressions such as those listed below. If necessary, “is” may be replaced by “is presented as such by the speaker” (as in line 1, etc.).

The Argument

The Conclusion

— is a consensual statement, or presented as such by the arguer)

— is a dissensual, challenged, disputed

statement

— is more likely than the conclusion

— is less likely than the argument

— is the cognitive starting point in deliberative

argumentation

— is the end point in justificatory argumentation

— is the end point of deliberative argumentation

— is the starting point in justificatory

argumentation

— expresses a reason

— is in search of a reason

— does not carry the burden of proof

— carries the burden of proof

— is oriented towards the conclusion

— is a projection of the argument

— (in a functional perspective) determines legitimizes the conclusion

— (—) determined, legitimized by the argument

— (in a dialogical perspective) accompanies the answer given to the argumentative question

— (—) is the proper answer to the argumentative question

(Plantin 2018: 64)

Pragma-dialectics uses ‘argument’ and ‘conclusion’ in their linguistic sense, not in their logical sense, which generates confusion when handling logical concepts such as ‘valid argument’ or ‘argument scheme’:

The utterances advanced in the argumentation are reasons, or, as we prefer to call them, arguments relating to a standpoint. It is their function that makes arguments and standpoints different from other utterances […]. In the communication between language users, with a standpoint, a point of view is expressed that entails a certain position in a dispute; with an argument, an effort is made to defend that position. (Van Eemeren & Grootendorst 1992: 13-14)

Argumentation theory is the study of the use of arguments (i.e. reasons) to convince others by means of a reasonable discussion of the acceptability of the (evaluative, prescriptive or descriptive) standpoint at issue in a difference. (van Eemeren & van Haaften, 342-343)

In the PD framework […] ‘Argumentation’ refers to the constellation of propositions employed in support of—and not including—a standpoint, whereas ‘argument’ corresponds to a single reason. (Juthe 2019: 476)

Here ‘argument’ is opposed to ‘conclusion’ and assimilated to ‘reason’, and, evidently, a reason is not a compound of premises or data and conclusion. For pragma-dialectics, the conclusion is, first and foremost, the conclusion of an argumentation, not the conclusion of an isolated argument:

the words used for argumentation in most western languages denote a phenomenon that is primarily characterized by being a process (“I am in the middle of my argumentation”) and at the same time a product (“Your argumentation does not look very strong”), by being associated with the defence of a standpoint (which is itself not part of the argumentation)Footnote 6 and by being instrumental in maintaining reasonableness. (van Eemeren 2018: 2)

Since an argumentation is a constellation of statements presented to defend a standpoint, the standpoint is the conclusion of the whole argumentation, not the conclusion that can be inferred or follows from some of the arguments that contained therein. This is the sense that ‘conclusion’ has in sentences such as “the conclusion of the debate was…” or “after discussion, they came to the conclusion that…”.

To differentiate the argumentation stage from the concluding stage confirms the above remarks, because the participants draw a conclusion after arguing, i.e. after proposing and examining the relevant considerations:

The concluding stage of an argumentative exchange corresponds to the stage of a critical discussion in which the parties establish what the result is of an attempt to resolve a difference of opinion. […] In practice, it is usually only one of the parties that puts the conclusion into words, but if the other party does not accept this conclusion, no resolution has been achieved. (van Eemeren & Grootendorst 2004: 61-62)

In the concluding stage of a critical discussion the protagonist and the antagonist determine whether the protagonist’s standpoint has been properly defended against the critical responses of the antagonist. (van Eemeren 2018: 37)

In an inferential approach, the protagonist constructs and proposes chains of inferences ending in her standpoint, which the antagonist must accept or reject step by step, and therefore the conclusion cannot be dissociated from the argumentation itself.

7 Conclusion

I have shown that, as far as its logical dimension is concerned, three successive periods can be distinguished in pragma-dialectics: conventionalist, inferentialist and dualist. Although the dualist interpretation of logic in the last period is incoherent, and thus the prior inferentialist interpretation is preferable, it is not fully satisfactory either, because it does not allow to account for all forms of counterargumentation—in particular for refutation. I have argued that to account for counterargumentation one must go beyond “vertical” premise-conclusion relations, and attend to “horizontal” relations between arguments, expressed by connectors such as “but” or “in addition”. Thus, counterargumentation theory leads from a conception of logic as a theory of inferences to a conception of logic as a theory of the dialogical construction of reasons. Finally, I have advocated that a reasonist conception of logic fits best within the framework of pragma-dialectical theory.