Dialogical Features of Presumptions: Difficulties for Walton’s New Dialogical Theory

Abstract

According to Douglas Walton, the concept of presumption relates to both logical and dialogical components. Logically, a presumption is the conclusion of a presumptive defeasible inference. Dialogically, the function of a presumptions to shift the burden of proof to the respondent in order to move the dialogue forward when the proponent, due to an objective lack of evidence, cannot present a sufficiently persuasive proposition. Presumptive status, assigned only at the argumentation stage of dialogue, is provisional: a particular presumption stands until further counter-indications are found. This paper points out some possible difficulties pertaining to Walton’s dialogical model of presumption. Firstly, once a presumption's typical practical function is made precise, Walton’s idea of presumption starts to lose its conceptual cohesion, thereby becoming applicable in the opening stage of dialogue, as well. Secondly, cohesion is lost also in practical contexts, since Walton's "seat belts" example might indicate that presumptions are not evidentially defeasible if the practical stakes are raised high enough. Most importantly, contextual evaluation may reveal that presumptions do not shift the burden of proof in any specific or genuine way.

1 Introduction

This paper focuses on the most recent version of Walton’s theory of presumption. In it, presumptions are defined both in logical, as well as dialogical terms. In Sect. 2.1, I briefly sketch Walton’s logical model of presumption and afterwards present his dialogical model. In Sect. 2.2, three crucial dialogical features of presumptions are explained (i.e. the basic dialogical function, the typical practical function, and the typical justification), and the guidelines for the contextual evaluation of presumptions are provided. However, in the critical part of the article I pose three arguments to question Walton’s theory. In Sect. 3.1, I argue that a more precise definition of its typical function makes the concept of presumption unexpectedly broad and heterogeneous, thereby casting doubt on Walton’s idea that presumptions operate only at the argumentation stage of dialogue (and not, for example, at the opening one). In Sect. 3.2, I propose that once the idea of presumption is analysed in practical contexts, it continues to lose its conceptual cohesion. That is, even though they are meant to be conclusions of defeasible inferences, it appears that some of Walton’s practical presumptions need not be evidentially defeasible. The final argument in this paper (Sect. 3.3) seeks to show that an appeal to Walton’s dialogue-types in the evaluation of presumptions may reveal that they fail to shift the burden of proof in any genuine way. In other words, I contend that his theory of argument evaluation fails to preserve the defining feature of Walton's presumptions.

2 Walton’s Theory of Presumption

Douglas Walton’s theory of presumption is undoubtedly one of the most systematic, comprehensive, and ambitious contributions to this domain of research. Originally, presumptions were legal devices, and ever since Roman law they have been used mainly to establish a proper division of labour in the courtroom.¹ Even though Walton’s aim is to construct a model applicable to the context of everyday argumentation (and, within this context, different dialogue types), this model remains legally inspired in at least one respect—the defining feature of presumptions is to allocate or shift a burden of proof in a dialogue. However, this characteristic has been analysed and substantiated in different terms as Walton’s theory developed over time.

His early theory (i.e. his speech-act theory of presumption) treated presumption as a speech act between assertion and assumption (Walton 1993, p. 125). This meant that once a proposition is introduced in a dialogue through a specific speech act of presuming, the proponent is not obliged to provide proof for its acceptance, but the respondent, if in disagreement, has the burden of rebuttal. The theory explained quite precisely what should count as the preconditions for introducing presumptions, as well as what should be the dialogical effects of presuming,² but some scholars (e.g. Freeman 1996) have claimed that its normative component was not constructed solidly enough.

The subsequent stage of Walton's theory (which might be called the schematic theory of presumptive reasoning) is almost fully developed in his book Argumentation Schemes for Presumptive Reasoning (1996). This stage is clearly linked with the previous one—a whole section of the book is devoted to the speech-act conditions that help define presumption (Walton 1996a, pp. 28–30).³ However, presumptions are now also associated with presumptive reasoning, which is analyzed in terms of argumentation schemes. Walton further developed and refined his schematic theory in Argumentation Schemes (2008), co-authored by F. Macagno and C. Reed.

In the most recent stage, Walton (2008a, b, 2014) defines presumptions as conclusions of defeasible modus ponens inferences, consisting of a presumptive rule and presumption-raising fact. According to this new dialogical theory of presumption, presumptions still have the dialogical function of shifting the burden of proof, but this time it is not (only) because dialogical conditions allow them to be introduced via a distinctive kind of a speech act, but (also) because they can be interpreted as conclusions of inferences. In other words, the dialogical theory proposes that presumptions are to be defined at two different levels, dialogical and logical, and that, although the ultimate conditions for introducing and evaluating presumptions might be dialogue-dependent, presumptions may also be justified on the basis of defeasible modus ponens inferences.⁴

In short, we could claim that the speech-act theory was useful primarily in defining presumptions in terms of dialectical obligations (burden of proof), the schematic theory contributed by providing a more developed account of the rebutting conditions for various kinds of presumptive arguments (primarily by investigating critical questions), while the dialogical theory aims to introduce a logically based, and thereby normatively more appropriate, account of presumptions. Of course, previous remarks are only approximations. This presentation of how Walton’s theory developed must, unfortunately, remain the tip of the iceberg; its primary function is to provide a relevant context for understanding the arguments in this paper.

The present paper focuses primarily on the most recent stage of Walton’s theory, namely his dialogical theory of presumption. After a sketch is made of the logical part of this theory, the dialogical part will be briefly presented and three critical remarks will be offered. These remarks should show that Walton’s theory, under scrutiny, remains unclear in dealing with some of the key features of presumptions.⁵

2.1 The Logical Dimension of Presumptions

Walton’s new dialogical theory proposes that presumptions are to be defined at two different but interconnected levels, namely, a logical level and a dialogical level. In principle, a presumption is the “modal status (or property) of a claim (or proposition) indicating that the burden of proof with respect to that claim rests with anyone who would reject it” (Godden and Walton 2007, p. 314). This proposition is the conclusion of a presumptive inference.⁶ Presumptive inferences can be reconstructed as defeasible modus ponens (Walton 2008a, p. 235, 2014, p. 110). Unlike its well-known deductive counterpart, a defeasible modus ponens allows the conclusion to be taken to be false even when the conditional premise is taken to be true, and the antecedent is established (Walton 2008a, p. 212, 2014, p. 87). Accordingly, the crucial features of a presumptive inference are its defeasibility and the provisional character of its conclusion. Thus, a presumptive inference involves a presumptive rule (i.e., the defeasible conditional) and a true antecedent (i.e., a proposition asserting the obtaining of the presumption-raising fact) as premises, and the presumed proposition as the conclusion (Walton 2008a, b, 2014).

According to Walton, presumptive inferences are tightly linked to arguments from ignorance (Walton 1993, 2001, 2008b). This is made evident when the example of the presumption of innocence is considered. Presumption of innocence is a rule of criminal law, and is considered a paradigmatic example of presumption among most argumentation scholars, despite the fact that their approaches are focused mainly on non-legal presumptions (e.g. Godden and Walton 2007; Macagno 2010; Macagno and Walton 2012; Walton 1993, 2001, 2008a; Rescher 2006; Freeman 2005, etc.). According to Walton’s terminology, a presumptive rule (“Anyone charged shall be presumed innocent until proved guilty”), is applied to a presumption-raising fact (“John has been charged”) and gives rise to a presumption (“John shall be presumed innocent until proved guilty”).⁷ Finally, the premises of a presumptive inference can be reconstructed and made explicit, but this neither means that the presumptions necessarily arise inferentially, nor that the reasons for presumptions, in everyday argumentation contexts, need to be explicit from the very beginning (Godden and Walton 2007, pp. 335–336).

2.2 The Dialogical Dimension of Presumptions

Apart from their logical (inferential) structure, presumptions also have distinctive dialogical (dialectical) features. In Walton’s (2014, p. 110) account, these dialogical features “mark off presumption as representing a distinctive kind of reasoning”. Walton discusses presumptions in terms of their basic dialogical function, their typical practical function, their typical justification, and their context dependency.

2.2.1 The Basic Dialogical Function of Presumptions

First of all, when a proponent introduces a presumption in dialogue, the burden of proof rests with anyone who would reject the presumed proposition. The “burden of proof” is, needless to say, primarily a dialogical notion.⁸ Therefore, presumptions are defined in terms of dialogical obligations or, more precisely, in terms of a specific distribution of dialogical obligations. This is a crucial point, because while “there may be other factors or argumentative devices affecting the burden of proof, presumptions are explained in terms of their effect on the burden of proof” (Godden and Walton 2007, p. 334). This dialogical effect provides a functionalistic definition of presumptions. According to Walton, the nature of presumption is, actually, its dialogical function—presumption is defined in terms of shifting or allocating the burden of proof in such manner that the proponent is spared and the respondent is burdened. It is important to note, however, that this does not make “presumption” simply another name for the well known phenomenon, namely burden of proof. “Presumably p” and “burden of proof for ~p” cannot be interchangeably used because “presumption functions only at the argumentation stage, whereas burden of proof is set at the opening stage, has effects at the argumentation stage and is vitally important in determining when the closing stage has been reached in a given case” (Walton 2008a, p. 226, 2014, p. 99, emphasis added). In other words, “burden of proof for ~p” does not entail “presumably p” if, for instance, this burden of proof is set at the opening stage of the dialogue.

Although the question “What is a presumption?” can be answered in terms of basic dialogical function (e.g. “It is a move in the dialogue that allocates/shifts the burden of proof in a certain way”), it still seems reasonable to move the question one step forward and ask “Ok, but what kind of move does that?” In this respect, Walton’s theory has developed over the years. It began with describing a presumption as a “distinctive kind of speech act half way between assertion and (mere) assumption (supposition)” (Walton 1993, p. 125). But more recently, as we have seen, Walton has claimed that presumptions shift the burden of proof because they can be interpreted as conclusions of (unexpressed) defeasible inferences. Having accepted this recent view, there are still two important questions to be discussed: (1) What dialogical problems is this allocating/shifting the burden of proof meant to solve? and (2) What kind of normative support (foundation/justification) does a proposition need in order to allocate/shift the burden of proof in this specific manner (i.e. in order to become a presumption)? The first question will be answered in terms of the typical practical function, and second one in terms of the typical justification of presumptions.

2.2.2 The Typical Practical Function of Presumptions

The basic dialogical function answers the question “What do presumptions do in the dialogue?” Another important question is “What do presumptions do for the dialogue?” What is the final desirable effect of the basic dialogical function? The answer to these questions is given by the idea of a typical practical function of presumptions.⁹

The typical practical function has much to do with the dialogical circumstances in which presumptions are typically applied. While the typical context (dialogue-type) of the use of presumptions is that of a persuasion dialogue (as explicitly stated in Walton 2008a, p. 235, 2014, p. 110), the circumstances of its application actually presuppose a specific state of particular persuasion dialogues. That is the situation (during the argumentation stage of the dialogue) when the crucial issue (regarding p) cannot be resolved on evidential grounds and there is also a practical need for continuing the dialogue. According to Walton, presumptions typically come in these sorts of circumstances—their typical practical function is to enable us to move forward when the persuasion dialogue, from an evidential standpoint, comes to a dead end, and we still need to act. This effect is achieved due to the basic dialogical function which enables us to take p to be true on a provisional basis (i.e. even though the proponent cannot provide enough evidence in its favour), until or unless some satisfactory indication of ~p is found (i.e. unless the opponent satisfies the burden of rebuttal), and ultimately gives us a chance to move on with the dialogue.

Thus, in situations where the state of evidence is not satisfactory, it might be unnecessary, unreasonable, and too costly to require the proponent to back up his proposition evidentially. This demand may unnecessarily block the dialogue. The solution is to accept the proponent’s proposition provisionally, as a presumption able “to remove this potential blockage and enable the dialog to move forward from a given point during the argumentation stage. In practical terms, the function of presumption is to save time and money and effort in communication” (Walton 2008a, p. 211, 2014, p. 115). This view has been the cornerstone of Walton’s theory of presumption for years (see Walton 1993, p. 138, 2001, p. 156; Godden and Walton 2007, p. 326; Macagno and Walton 2012, p. 291 etc.). Nevertheless, Walton does not claim that these (evidential) circumstances are the only ones in which presumptions are to be used, but rather that they are the typical ones.

2.2.3 The Typical Justification of Presumptions

The typical practical function of presumptions is also firmly set on normative foundations. Namely, presumptions are typically justified both on evidential (epistemic) and practical grounds, meaning that they simultaneously have epistemic and practical underpinnings. It is precisely the combination of evidential and practical reasons that justifies the presumptive rule, and then the justification is transmitted to the presumption itself by means of a defeasible modus ponens. Stating that presumptions are argumentative tools primarily used in lack-of-evidence situations does not imply that presumptions lack any base of an epistemic kind. Rather, it means that the evidence (or the set of the proponent’s evidential reasons) is not adequate to meet a certain level of epistemic justification (according to the proponent’s standard of proof) as required for the acceptance of the presumed proposition, at a particular point during the argumentation stage of the dialogue. But the presumptive inference becomes “sufficiently strong, with the practical justification counted in, to shift a burden of production to the respondent’s side” (Walton 2008a, p. 235, 2014, p. 110), therefore enabling the dialogue to move forward.

2.2.4 Contextual Considerations

However, the final evaluation of whether some proposition can be justified in being presumed or not, has to do with the goal of the particular dialogue in which this proposition was introduced. More pointedly, it depends on the collective goal of the type of dialogue that this particular dialogue belongs to. This is the method for evaluating all arguments, as proposed by Walton and Krabbe in Commitment in Dialogue (1995) and Walton himself in The New Dialectic (1998). Walton introduces six different dialogue types (persuasion, deliberation, inquiry, negotiation, information-seeking, and eristic) with six different collective goals, and “each type of dialog is used as a normative model that provides the standards for judging how a given argument should be correctly used in a given case” (Walton 2007, p. 23).¹⁰ This view not only indicates that presumptions can be used in many different contexts of everyday argumentation, but it also shows that they can be evaluated from many different normative perspectives. Different types of dialogues would provide different normative frameworks and, depending on the type in question, the required levels (and kinds) of justification for p, as well as the retraction conditions for p, may vary. A description of how dialogue types affect the evaluation of arguments is given in Walton’s recent book Burden of Proof, Presumption and Argumentation (2014). It may be especially relevant for this topic, since the link between dialogue types and evaluation is described here in terms of burden of proof, which, as we have seen, is closely related to the concept of presumption.

3 Some Difficulties of Walton’s Dialogical Model of Presumption

In the following section I argue that Walton’s theory, in the current state, remains unclear in dealing with some of the key features of presumptions: the dialogue stage at which they operate, defeasibility (as requirement for the usefulness of presumptions) and, finally, their basic dialogical function. For this, three main arguments are presented.

In Sect. 3.1, I argue that two similar, but also relevantly different practical functions of presumptions can be recognized in Walton’s theory. The first consequence of this recognition is that distinctly different sorts of arguments (different both in their dialectical antecedents/levels of justification, as well as their potentials in producing certain dialogical effects) are now placed in the same category. In this way, Walton’s concept of presumption begins to fragment. More importantly, if the aim of presumptions is also to prevent the dialogue from making undesirable regression, then they may also serve as dialogical starting points. However, if this is the case, then presumptions may also operate at the opening stage of the dialogue (not only at the argumentation stage, as Walton insists) which makes them conceptually closer to the burden of proof.

In Sect. 3.2, my aim is to show that once we start to analyse Walton’s idea of presumption in practical contexts, it only continues to lose its conceptual cohesion. There are at least two different sorts of practical presumptions, the first one exemplified by the presumption of innocence, and the second one instantiated by the "seat belts" example. At first glance, they look very similar, but under scrutiny they seem to differ both in their evidential antecedents, as well as evidential defeasibility. That is, even though defeasibility and provisional character are meant to be crucial features of presumptions, the "seat belts" example seems to suggest that some practical presumptions are neither evidentially defeasible nor provisionally introduced.

Finally, in Sect. 3.3, I seek to show that the defining feature of Walton’s concept of presumption, a shift in the burden of proof, fails to be preserved  by his own contextual theory of argument evaluation. If typical presumptions shift the burden of proof only when evidential and practical reasons are conjoined (explained in Sect. 2.2.3), then a typical presumption cannot perform its basic dialogical function in the persuasion dialogue since its practical premises do not directly contribute to achieving the collective goal of this normative dialogue type. On the other hand, a typical presumption may not perform its basic dialogical function from the perspective of action-oriented dialogue, since it can be interpreted as a conclusion of regular practical argument. However, if it is only a conclusion of regular practical argument introduced for practical purposes, there is nothing unusual in the way the burden of proof is allocated or shifted. Therefore, an appeal to Walton’s dialogue-types in the evaluation of presumptions may reveal that presumptions, if they are to be considered “good,” cannot relieve the burden of proof, or at least they cannot do so in any genuine way.

3.1 Two Practical Functions of Presumptions?

When dialogical theory is closely examined, it may be recognized that Walton anticipates at least two substantially different accounts of presumptions, regarding both their dialogical antecedents (and levels of justification), as well as their practical functions. In the typical case, the dialogue has become stuck and it can go forward only if, at that point, interlocutors agree to accept p on a provisional basis. Interlocutors realize that p cannot be supported on evidential grounds that are “good enough;” but, with the helping hand of certain practical reasons, the presumptive status is assigned, p provisionally meets the standard of proof, becomes provisionally accepted, shifts the burden of proof to the respondent, and finally enables the dialogue to progress. Let us call presumptions raised in these circumstances and serving this function “typical presumptions”. However, Walton’s theory seems to tolerate a whole other class of presumptions that are tailored to achieve the similar, but relevantly different, function of preventing the dialogue to regress. I call this class of presumptions “shared (or initial) presumptions.”¹¹ The existence as well as distinguishing features of shared presumptions can be reconstructed from the following passage.

If a claim made always had to be backed up by an argument, there is the danger of wasting time and energy on claims that nobody would dispute, or perhaps that nobody in the dialogue should have any reason to dispute or right to dispute. For example if the claim is already accepted by all parties to the dispute and nobody has any reason not to accept it, there should be no need to put forward an argument to back it up, even though it might be possible to do so. This is where the notion of presumption is useful (Walton 2008a, p. 227, 2014, pp. 100–101, emphasis added).

This quotation indicates that, in Walton’s view, a proposition p can also become a presumption due to the fact that p has already been accepted. In contrast to typical presumptions (where presuming p is, in fact, making p provisionally accepted), here presuming p becomes possible precisely because p has already been (previously, initially) accepted by the interlocutors. Provisional acceptance which seems to be an intended effect of typical presumptions, is now the very precondition of shared presumptions. This indicates that typical and shared presumptions include relevantly different dialogical (dialectical) antecedents. That is, while typical presumptions are introduced in the context of dialectical uncertainty, when there is no agreement regarding the proposition in question, shared presumptions are possible only when such agreement exists, i.e., in the context of dialectical certainty. In the case of typical presumption, interlocutors agree that there are no evidential reasons “good enough” to accept p. In the case of shared presumptions, interlocutors clearly agree that there are “good enough” (evidential) reasons to accept p and, oddly enough, it is precisely this agreement that makes their acceptance of p presumptive.

Besides having different dialogical antecedents (and levels of evidential support), these two types of presumptions also have different potentials to generate dialogical effects. For instance, shared presumptions can hardly enable the dialogue to move forward once it becomes stuck. That is, if the dialogue has reached a dead end despite all available moves (not involving typical presumptions)—and shared presumption p must have been among these moves (being initially shared and thus already accepted by the interlocutors)—then we might ask how bringing p forward can be of any help in saving the discussion. If the dialogue became blocked despite the fact that p was presumed, then introducing p once again can hardly make the dialogue move forward. Therefore, it seems that shared presumptions cannot perform the typical practical function of presumptions (as defined in Sect. 2.2.2). However, shared presumptions can help us to avoid an undesirable regress of dialogue, as explained in the passage quoted above. In other words, at the argumentative stage of a persuasion dialogue (to stick with Walton’s idea that presumptions operate only at that stage), presumptions can perform at least two different practical functions, namely to enable the dialogue to make necessary and desirable progress, and to prevent the dialogue from making unnecessary and undesirable regress. These functions may be similar in many aspects, which is presumably why they are not explicitly distinguished in Walton’s theory. However, once this difference is taken seriously, Walton’s theory of presumption may face certain difficulties.

Firstly, Walton’s concept of presumption becomes unexpectedly broad and, more importantly, starts to lose its cohesion. If “presumption” refers both to arguments whose conclusions are so dialogically uncontroversial (certain) that it would be almost pointless to introduce any reasons for them, and to arguments whose reasons would indeed be appreciated (but are known to be insufficient) and whose conclusions are taken to be uncertain, then the concept of presumption ceases to mark a distinctive class of argumentative manoeuvres. However, it may still be argued that all these manoeuvres, although heterogeneous, share the feature of “being provisional.” But this offers little help, since two classes of presumptions are also provisional in two different senses. The provisional character of typical presumptions p seems to be conjunction of two separate requirements: (1) interlocutors that accept p do not believe p is supported by “good enough” evidential reasons, and (2) they are willing to reject p as soon as sufficient evidence shows the contrary.¹² Shared presumptions cannot be provisional in the same sense, because they satisfy only the second condition of this conjunction. That is, when someone initially presumes “The climate is changing” at a conference about climate change, his presumption is provisional only because he is willing to reject it, if somebody provides good reasons for doing so. However, it cannot possibly satisfy the first condition: it is not the case that the participants of the conference accept “The climate is changing”, while believing that there are no good evidential reasons for that acceptance. The second requirement of “provisional acceptance” alone (shared by both classes of presumptions) is so undemanding that broadness and heterogeneity of the concept of presumption seem to be unavoidable consequences.

Secondly, if two classes are recognized, the concept of presumption not only becomes broader and heterogeneous, but also widely applicable. That is, it is unclear why shared presumptions should operate only at the argumentation stage of dialogue and not, for example, at the opening one. Their function of preventing the dialogue from making undesirable regress seems compatible with the idea that presumptions serve as dialogical starting points for which no reasons are, at least initially, required. If we needed to back up every claim, then dialogues would be “regressing” before they even started, they would get stuck before they even began, and hence they would never reach the argumentation stage. This is the well-known problem of the “infinite regress,” and it is usually a sceptic’s instrument. A common solution to this problem, as proposed by scholars in argumentation theory,¹³ is to suggest that some claims should be accepted, at least presumptively, at the start, in order to get the dialogue going. Taken in this sense, presumptions would not only be something that happens during the argumentation stage of the dialogue, but they would also make provisional grounds that dialogue is built upon. Since Walton recognizes that shared presumptions are means of prevention, and this prevention of the regress can also be needed initially (to set the starting points at the opening stage), it is not really clear what supports the idea that “presumption functions only at the argumentation stage” (Walton 2008a, p. 226, 2014, p. 99, emphasis added).

The first theoretical consequence here is straightforward: if there is a possibility that presumptions operate at the opening stage, then questions regarding consistency of Walton’s theory may be raised and further clarifications will be needed. Another consequence may be that the concept of presumption is now one step closer to merge with the concept of burden of proof. As stated by Walton (2008a, p. 226, 2014, p. 99), the differences between burden of proof and presumption include that burden of proof is also set at the opening stage, and is crucial in determining whether, in a particular dialogue, the closing stage has been reached. Of course, even if some presumptions operate at the opening stage, this still will not make the concepts indistinguishable. However, the tendency of presumption getting conceptually closer to burden of proof, we might interject, could seriously limit its theoretical potential. That is, the closer one concept gets to another well-defined concept, the lesser theoretically useful it becomes.

The possibility that Walton would accept this view, but not be eager about this convergence, may be anticipated from his critique of Cronkhite’s theory: “The problem with Cronkhite’s model is that … it deprives the notion of presumption of any useful job in the theory of argument. … If presumption is only the counterpart to burden of proof, then there is no need for a separate concept, and theoretical questions concerning presumption are answered simply by the theory of burden of proof” (Godden and Walton 2007, p. 324). To be completely clear, we should say that presumption still has not become only the counterpart to burden of proof in Walton’s theory. However, these concepts may be much closer in dialogical theory than Walton himself would be ready to admit.

3.2 Are Practical Presumptions Always Evidentially Defeasible?

As noted in Sect. 2.2.4, presumptions can be used in many different everyday contexts and can be evaluated from many different normative perspectives. Levels of standards of evidence needed to introduce certain proposition as a presumption, as well as to retract (rebut/reject) it may vary from one discourse to another, depending on which kind of goals are promoted in that particular discourse.¹⁴ However, we might ask what happens if the standard of evidence set in the discourse prompts the question: “How could the (presumed) proposition ever be evidentially defeated?” or “How could its introduction ever be described as provisional?”

In the following text I will try to show that when Walton’s presumptions are analysed in practical contexts, a proposition can also be presumed without being evidentially defeasible and/or provisional (neither in the first, nor in the second sense of “provisional” discussed in Sect. 3.1). Since defeasibility and provisionality are meant to be crucial, characteristic features of presumptions, Walton’s concept of presumption continues to fragment. The situation is similar to one found in epistemic context of persuasion dialogue (explained in Sect. 3.1): there are at least two classes of practical presumptions, and they seem to differ both in evidential antecedents, as well as evidential defeasibility and tentativeness. To see how this is possible, let us consider Walton’s “seat belts” example. At one moment during a flight, turbulence started and the pilot required the passengers to fasten their seat belts. When the turbulence passed the pilot announced that the seat belts were no longer necessary. A few minutes later he changed his mind for no discernible reason (except precaution, to “err on the side of safety”), and announced that the passengers should fasten their seat belts again. In Walton’ words,

Although he reported that he had no evidence that there would be further turbulence, he acted on the presumption that there might be, by telling the passengers to fasten their seat belts again. Thus he could be said to have made a presumption, even though there was no evidence, or no objective evidence at any rate, supporting the truth of the proposition that he accepted as a presumption (Walton 2008b, p. 63, emphasis added).

Now, we may all agree that the pilot’s reasoning is well intended and useful. But is it useful in a specific way in which presumptions are meant to be useful? In other words, is this useful move also a useful presumption? When Walton’s own criteria are applied, the answer to the first question is not very clear. That is, there are “three elements that make bringing forward a presumption a useful move. First, there is some evidence to prove the claim; second, there is lack of sufficient evidence to prove the claim; third, there is no evidence to disprove the claim” (Macagno and Walton 2012, pp. 278–279). The seat belts example satisfies only the second condition, but in a trivial way (only because the first one is not satisfied): if there is no evidence to prove the claim, then there is no sufficient evidence to prove the claim.¹⁵ With respect to the third condition, we might contend that there is evidence to disprove the claim—if the pilot has all available evidence to determine whether turbulence will happen, and nothing in the evidence suggests that it might, then, we might say, the evidence strongly suggests that turbulence will not happen. This would make the third condition unsatisfied as well. Therefore, if we accept that the pilot is really presuming p, then p may be a useful presumption even when (1) there is no objective evidence (thus, no sufficient evidence) for p, and (2) there is indeed evidence to prove ~p. Is this enough to say that Walton’s example of a useful presumption cannot satisfy his own conditions for the useful introduction of presumptions? Previous analysis would suggest so, and this is indeed my view.

However, the above analysis is not nearly strong enough to dismiss the “seat belts” example as an instance of useful practical presumption. That is, we could argue that the three evidential elements for the usefulness of presumptions have slightly different meanings when applied in practical contexts, since in these contexts evidence does not play a leading role. This may even result in recognizing the “seat belts” example as useful practical presumption of the most regular kind, tightly linked with the paradigmatic “presumption of innocence.” I will now present this line of thinking, but also argue against it.

Practical (or non-epistemic) contexts behave differently than epistemic ones when it comes to evidential criteria for introducing and rejecting presumptions. We might contend that in some non-epistemic discourses practical goals set the evidential threshold for the acceptance of p so low that p can be initially presumed without any evidence in its favour. At the same time, these practical goals set the evidential threshold for the rejection of p so high that p may be accepted despite any overwhelming evidence to the contrary. The presumption of innocence may appear to be an obvious example here: the accused (let us call him John) will be initially presumed innocent even when (1) there is no evidence in favour of his innocence and, more importantly, although (2) there is an abundance of evidence indicating his guilt.¹⁶ In this respect, the paradigmatic example of legal (and practical) presumption seems to be in the line with the “seat belts” example discussed above. However, this analogy is not as obvious as it appears. That is, even though both presumptions (“Presumably, turbulence will happen” and “Presumably, John is innocent”) are not evidentially motivated, they still have relevantly different evidential antecedents.

Turbulence presumption has a negative evidential antecedent as there was an abundance of relevant evidence proving the contrary. In this sense, it was introduced a posteriori—the pilot is evidentially well informed, but chooses to presume p against all available evidence. On the other hand, the presumption of John’s innocence has no relevant evidential antecedent—there is no relevant (legally recognized) evidence proving innocence/guiltiness at the point when presumption is raised. In this sense, presumption is introduced a priori, on the basis of a legal rule, with no connection to evidence relevantly proving or disproving John’s innocence/guiltiness.¹⁷ Thus, no relevant and strong evidence proving “John is not innocent” was, strictly legally speaking, acknowledged before he was presumed innocent. In the turbulence case, however, the relevant and strong evidence for “Turbulence might not happen” was acknowledged and, nevertheless, the presumption “Turbulence might happen” was made.¹⁸ This gives us reason to believe that the turbulence presumption and the presumption of innocence treat evidence differently from the very beginning, and therefore the latter offers little help in claiming that the former is just a regular case of practical presumption.

More importantly, these presumptions differ not only in their initial treatment of evidence, but also in their interpretation of the final role of evidence. Here, the presumption of innocence is clear, unlike the turbulence presumption. The first one ultimately answers to evidence, and the standard of evidence needed for its rejection is well defined. That is, if the evidence introduced during the trial proves beyond reasonable doubt that John is guilty, then the presumption of innocence is ultimately rejected in John’s case. That is the whole point of the approach to “Consider every defendant innocent, until proved guilty”.¹⁹ In the turbulence case, however, it is not at all clear what standard of evidence is needed in order to reject “Presumably, the turbulence might happen.” The evidence, although overwhelming and known, is either irrelevant here or the evidential standard for rejection is set so high that it is practically unreachable. That is, if nothing in the pilot’s evidence (and he has all the evidence possibly available at this level of scientific and technological progress) indicates that the turbulence might happen (as explicitly said by Walton), then everything in evidence suggests that the turbulence will not happen. In this particular case, the standard of lack of evidence for p and the standard of evidence confirming ~p might be two sides of the same coin.²⁰ If “nothing justifies p” can be translated into “everything justifies ~p” in this particular case, then not only does some evidence to the contrary exist here, but all the evidence possible indicates the contrary. If this is indeed the case—if everything in the evidence indicates that there will be no turbulence, and that is still not enough to reject the presumption “Turbulence might happen”—then we may say that certain practical presumptions are not meant to be evidentially defeasible,²¹ or that the level of evidential justification that the negation must meet is so high that it is practically unreachable.

If Walton’s presumptions are connected with arguments from ignorance (as noted in Sect. 2.1), it is not in fact intuitive to talk about presumptions in this case. However, even if we put this intuition aside we can still ask ourselves: How could we ever describe the introduction of “Turbulence might happen” as something provisional? It makes more sense to claim that p (“Turbulence might happen”) is presumptively accepted because the possibility of its negation ~p (“Turbulence might not happen”) is being actively and categorically disregarded on the grounds of safety—there is nothing provisional either in the pilot’s confidence regarding p, or in the world’s stubbornness in suggesting ~p. At any rate, p cannot be provisional either in the first or in the second sense of “provisional” discussed in Sect. 3.1. As we have seen, both senses of “provisional” share one common (necessary) feature—the presumption should be rejected if sufficient counterevidence is found. Here, pilot has all the counterevidence possibly available, but the presumption still stands.²²

The point is straightforward here. If we take Walton’s three criteria for presumption usefulness seriously, then the “Turbulence might happen” can hardly be a useful presumption. If we, however, still insist on the contrary, then Walton’s set of criteria calls for additional clarification given that some useful presumptions are hardly provisional and/or defeasible. While this would not make Walton’s theory inconsistent,²³ it would prompt at least two interrelated questions: (1) What theoretical benefits can be gained by such a heterogeneous category and (2) what are the grounds on which all these different precepts may be called “presumptions?” Bearing in mind that Walton is trying to provide a theoretical notion of presumption which, in addition, aims to be normative, I fail to see how heterogeneity of this kind can be helpful in providing theoretical clarity, as well as guidance for good (presumptive) reasoning. Walton’s presumptions do not seem to form a single class that can be interpreted, analysed, and evaluated in a single way. Rather, there are many, very different elements that fall under Walton’s category of presumption, and these must be treated differently if they are to be correctly theorized. My argument here is basically in line with Rescher’s, who claims that the legal practice of categorizing irrefutable stipulations as presumptions is “unhelpful because it throws together items whose nature and function are altogether different” (Rescher 2006, p. 5).

With respect to the second question, we may try to explain the heterogeneity of the concept in terms of a “family resemblance.” Some concepts function successfully even when its instances share no particular distinguishing feature. However, it is not clear whether family resemblance presents the best model of how theoretical concepts ought to be constructed. For example, Williamson (1994, p. 87) argues that “[f]amily resemblance concepts appear susceptible to sorites paradoxes” and, using the paradigmatic example, he asks: “What is to stop us from gradually extending the concept of game to any activity we choose, such as nuclear warfare?” Accordingly, even if we assume that Walton’s concept of presumption is meant to function in this way, family resemblance does not necessarily provide grounds to make its heterogeneity legitimate. On the contrary, it can even be used to point more clearly in the direction of problematic (even paradoxical) implications that the concept of presumption may have, and that would need to be resolved, if interpreted in this way.

However, although Walton’s presumption may, in the end, turn out to be family resemblance concept, Walton may not have assumed this from the very beginning. He may still hold that every presumption (whether dialogically certain or uncertain, shared or typical, provisional in the first sense or the second or, maybe, not provisional at all, raised against the evidence or in no relation to evidence, etc.) must shift the burden of proof to the respondent, while sparing the proponent. Hence, the basic dialogical function may be one common feature of all presumptions, causing them to form a single class which is then analysed and evaluated in a single way. In the following section I propose that even this view, which would unite all sorts of presumptions in the “non-family resemblance” way, is not really tenable once Walton’s contextual normativity is taken into account.

3.3 Shifting of the Burden of Proof and Contextual Evaluation

The final argument in this paper seeks to show that the basic dialogical function of presumptions, their one distinguishing and defining feature, remains undetected when contextual considerations are applied in the evaluation of presumptive arguments. Appealing to Walton’s normative types of dialogues may reveal that presumptions do not shift the burden of proof in any genuine, characteristic way. Therefore, if we accept his general method of evaluating arguments, his very definition of presumption (and supposedly the one common feature of all presumptions) may come into question.

First of all, if presumptions are introduced in persuasion dialogues (as Walton claims in 2008a, 2009, 2014, etc.), we should analyse and evaluate them in accordance with the norms of that type of dialogue. In Walton’s (2007, p. 23) account, norms for the evaluation of arguments are derived from the collective goal of the dialogue type. The collective goal of a persuasion dialogue is to resolve the conflict of opinions on whether some final proposition is true (while the individual goal of every participant is to persuade the other that a proposition is or is not true). Hence we should consider as legitimate only those argumentative moves that help to determine whether the final proposition is true or false.²⁴ From this normative perspective, practical reasons for accepting a proposition do not seem to be relevant. They may not necessarily undermine the collective goal of the dialogue, but the fact that the acceptance of some particular proposition is practically justified should not be considered in resolving the main issue, namely whether p is true or not. Even if we are aware of certain practical reasons, they should not be counted in during the final evaluation. The only justification to be considered should be given by evidential reasons. If these reasons cannot be found, we should categorize the move as mere assumption. In that case, nothing happens with the burden of proof given that it falls on neither the proponent nor the opponent. Typically, evidential reasons are indeed found, but they are not strong enough to meet the required standard of proof. In this case, if we are consistent in presupposing the normative framework of a persuasion dialogue in the evaluation, we should just interpret the argument as evidentially weak. There is nothing special or new here, as the burden of proof for the conclusion is not met and therefore cannot and should not be shifted to the opponent’s side.

On the other hand, Walton (2009, p. 2) claims that presumptions can be used in action-oriented (deliberation) dialogues. In this kind of dialogue, interlocutors collaborate to decide the best available course of action (Walton 2007, p. 23). Typically, they put forward proposals and try to defend them as the best means for solving the practical dilemma. This context and presumption get along naturally since presumptions are typically brought forward because of a practical problem (the dialogue has become stuck), and they can be interpreted as proposals to provisionally accept some proposition as true (since that could help the dialogue to move forward). In this kind of normative context, the practical reasons supporting presumptions become relevant in evaluating presumptive arguments. The crucial point is that, just as in the case of persuasion, nothing unusual happens with the burden of proof. We are now dealing with the regular practical argument where presumption is introduced as a proposal, and it is backed up by practical reasons. If the practical reasons are good enough, then the burden of proof shifts in the most regular way. If the reasons are not good enough, the burden of proof will not be satisfied and it will not be shifted to the opponent’s side. We may present the previous dilemma in the standard form.

• P1. Typical presumption p is based upon weak evidential premises;

• P2. If evaluated from epistemic perspective, arguments with weak evidential premises cannot shift the burden of proof to the opponent’s side;

• P3. Typical presumption p is based mostly upon practical premises, and used primarily for practical purposes;

  ∴C1. Presumption p may be treated as a conclusion of regular practical argument introduced in regular practical context [from P3];

• P4. In regular practical context, the burden of proof is initially allocated on the proponent’s side, and it is shifted to the opponent’s side in the most usual way, depending on whether practical premises are “good enough”;

• P5. Practical premises do not directly contribute to the collective goal of epistemic context;

  ∴C2. Practical premises should not be treated as relevant when evaluated from the epistemic perspective [from P5];

• P6. Presumption p is used either in epistemic or in a practical context;

  ∴C3. Presumption p should be evaluated from an epistemic or practical perspective [from P6];

  ∴C4. If evaluated from epistemic perspective, presumption p does not shift the burden of proof to the opponent’s side [from P1, P2, C2 and C3].

  ∴C5. If evaluated from practical perspective, presumption p allocates the burden of proof on the proponent’s side, and the burden of proof is shifted to the opponent’s side in the most usual way [from C1, P4 and C3].

  ∴C6. Therefore, from the perspective of argument evaluation, if presumption p is to be a conclusion of a “good” argument, it either (1) does not shift the burden of proof to the opponent’s side or (2) allocates the burden of proof to the proponent’s side and shifting it in the most usual way [from C3, C4, and C5].

The final conclusion (C6) directly contradicts the basic idea that presumptions do not allocate the burden of proof to the proponent’s side, but shift the burden of proof to the opponent’s side instead. Thus, from a purely evaluative perspective, presumptions do not seem to work as Walton suggests.

However, they might work if evaluated in normative contexts where practical and evidential reasons are, although of different kinds, both treated as relevant. Accepting that kind of mixed discourse as a context of analysis and evaluation might indeed save Walton’s notion of presumption. But this saving might be too costly, since it would require accepting the concept of mixed discourse, which is criticized and held questionable at both the practical and theoretical levels. The role of this concept in Walton’s theory is to provide a certain link between, on the one hand, well-defined and highly regulated normative dialogue types and, on the other, various sorts of dialogues that can be found in ordinary argumentation but cannot be classified as corresponding to any dialogue type in particular. However, as van Eemeren et al. (2010, p. 123) point out, the concept of mixed discourse is not that easily applied, since Walton fails to provide clear criteria of when to interpret a particular dialogue as a composition of some ideal dialogue types (i.e. as an example of mixed discourse) and when not to do so. More importantly, as argued in Goodwin (2007, p. 78), even if it were clear that a particular dialogue was a case of mixed discourse, we do not really know which norms to apply while evaluating the arguments in this discourse. That is, Walton sees norms as a function of the collective goals of the dialogue and, since this dialogue includes different collective goals, it must consequently be normatively ambiguous. Therefore, in order to make the evaluation (and definition) of Walton’s presumptions work, we should adopt a normative perspective that, in Walton’s theory, cannot really work.

However, besides the notion of mixed discourse Walton’s dialogue theory includes the concept of dialectical shift, used for similar purposes. A dialectical shift happens when one type of dialogue gradually or instantly, due to another collective goal that became relevant in the meantime, turns into a different type of dialogue (Walton 1998, p. 198). Thus, we might assume that we are initially engaged in a persuasion dialogue, defined by its collective goal and rule-governed by the corresponding argumentative norms; but suddenly it reaches a dead end. Since it is of great importance to resolve the issue, we are confronted with the practical dilemma of which proposal we should accept in order to continue the original dialogue successfully. Persuasion is temporarily put on hold, and we engage in a deliberative meta-dialogue. Presumption is brought forward as a proposal in that meta-dialogue. Under this interpretation, however, the same problem arises again: the presumptive argument appears to be a regular practical argument evaluated from the perspective of a deliberation dialogue, and there is nothing special or genuine in the way it shifts the burden of proof.

In conclusion, once we accept Walton’s method of argument evaluation (assuming six dialogue types and the concept of dialectical shift), it is not clear why the basic dialogical function of presumption, as a defining and distinguishing feature, should be accepted. From an epistemic perspective, presumptions cannot perform their defining function, and from a practical one, this function is performed in the most undefining way. It also seems that, in order to make this function (definition) work, we should accept the normative perspective (i.e. mixed discourse) that, in Walton’s own theory, does not seem to work very well.

4 Conclusion

Walton’s new dialogical theory of presumption offers more than enough insight to point us in the right direction and motivate investigations of our own. Accordingly, it should be appreciated and studied most carefully. Its basic points can be summed up as follows. At the logical level, presumption is seen as (1) the conclusion of an inference that has (2) the form of a defeasible modus ponens and is linked with (3) the argument from ignorance. On the dialogical side, presumption is a move that (4) happens only at the argumentation stage of dialogue and is (5) based on both evidential and practical grounds. Its basic function is to (6) shift the burden of proof from the proponent to the respondent in order to (7) help the dialogue progress (either by enabling continuation or preventing regression), and for the latter purpose, (8) it is introduced on a provisional basis. In this sense, presumptions should typically be (9) subject to retraction if counter-indications (evidence) arise. Finally, (10) they should be evaluated from the contextual perspective including different normative dialogue types.

The main problem with Walton’s theory is, as I see it, that it leaves many relationships between these features unclear, not always as a result of not dealing with them but also as a consequence of certain oversights and ambiguities. At the dialogical level, accepting (7) makes reasons for claiming (4) unclear, makes (8) ambiguous and, consequently, causes the concept of presumptive argument to refer to a very heterogeneous class of argumentative moves. Furthermore, when practical contexts are taken into account, it seems that some presumptions (e.g. in the “seat belts” example) fail to satisfy (9) and do not fit the ideas behind (8), fragmenting the concept of presumption even further. Finally, Walton’s own method of evaluation (10), linked with (5), poses the question of why we should even accept (6) in the first place, thereby questioning the one defining feature of presumption. The concepts of mixed discourse and dialectical shift offer little help here.

My final and overall point is straightforward and quite modest. Walton’s approach, at this stage, has problems on the dialogical side which, mostly due to the oversights and lack of precision, weaken his theory. We are left with the concept of presumption so heterogeneous and broad that its general theoretical usefulness becomes unclear. My intention here in identifying this weakness is to point in the direction of desirable clarifications. This does not mean, however, that Walton’s intuitions are generally wrong or that his work on presumptions has been heading in the wrong direction. The problem may rather be that his work has been heading in too many directions at once.