Is there a burden of questioning?

DOUGLASWALTON

Department of Philosophy, University of Winnipeg, Winnipeg, Manitoba, R3B 2E9, Canada

E-mail: d.walton@uwinnipeg.ca

Abstract. In some recent cases in Anglo-American law juries ruled contrary to an expert’s testimony

even though that testimony was never challenged, contradicted or questioned in the trial. These

cases are shown to raise some theoretical questions about formal dialogue systems in computational

dialectical systems for legal argumentation of the kind recently surveyed by Bench-Capon (1997)

and Hage (2000) in this journal. In such systems, there is a burden of proof, meaning that if the

respondent questions an argument, the proponent is obliged to offer some support for it give it up.

But what should happen in a formal system of dialogue if the proponent puts forward an argument and

the respondent fails to critically question it, and simply moves on to another issue? Is this some kind

of fault that should have implications? Should it be taken to imply that, by default, the respondent has

conceded the argument? What, if anything, should be the outcome of such a failure to question in a

formal dialogue system of argumentation? These questions are considered by examining some legal

cases of expert opinion testimony in relation to rules for formal dialectical argumentation systems.

Defeasible arguments are inconclusive. They hold only tentatively in an investigation,

and are subject to defeat if new evidence should come in that rebuts

the argument or undercuts it by posing criticisms of it (Pollock 1995). A central

problem for argumentation studies is how such a defeasible argument, when it

is good one, should be binding on a respondent.1 Should the respondent be free

to ignore it altogether, with no penalty or loss of probative weight for his side?

There is a burden of proof on the proponent’s side. Should there also be a burden

of questioning (or challenging) on the respondent’s side? If not, it would seem

that defeasible arguments need not be rationally binding on a respondent and thus

may have no real force, weight or impact as reasons in the argumentation in a

dialogue. The problem is how a rational argument put forward by one party in a

dialogue should be binding on the other party. This problem arises for deductively

valid arguments. It is a general one for formal or abstract models of argumentation

of the kind proposed by Hamblin (1970, 1971), van Eemeren and Grootendorst

(1984, 1987, 1992) and Walton and Krabbe (1995). But it arises in a form that is

even more difficult to solve when defeasible arguments are considered.

Legal evidence is largely made up of defeasible arguments (Verheij 1996;

Bench-Capon 1997; Prakken 1997). One common defeasible argument of this sort

is the appeal to expert opinion of the kind often used as testimony in trials (Walton

2 DOUGLAS WALTON

2002). The problem of the burden of questioning can be posed in a pointed way

using illustrations of arguments based on expert opinion testimony. Especially interesting

is the kind of case where strong expert opinion testimony is presented by

one side in court and the other side fails to challenge, to rebut, or even to crossexamine

this hostile testimony.2 What should be said about the kind of case in

which a jury ruled in a manner indicating their rejection of strong and convincing

expert testimony by a credible expert whose testimony was not contradicted and

whose credibility was not questioned? This type of failure to question or challenge

a strong defeasible argument could be seen as simply poor advocacy or poor trial

strategy on the part of an attorney. But could such a failure to respond appropriately

also be modeled as a fault of rational argumentation? The jury’s ruling indicated

their rejection of the appeal to expert opinion, but the failure to question it seems

to imply that the argument was taken to be acceptable. Is this a contradiction, or

logical lapse of a kind that violates a standard of rational argumentation? Should

there be a burden of questioning? Should a respondent’s failure to discharge this

burden, when presented with a strong defeasible argument, be taken to imply his

or her acceptance of the argument?

1. Applying formal models of dialogue to legal argumentation

Evaluating argumentation in a dialogue model in which two parties question each

other and advance arguments directed to those of the other, is an old idea that goes

back Plato and Aristotle, and even before them to the Sophists. But it wasn’t until

the research program of the Erlangen School in Germany that efforts were made

to carry out a systematic program for constructing formal systems based on the

dialogue model.3 This research never caught on, however, and was not carried forward.

But independently, Charles Hamblin (1970, 1971) constructed mathematical

models of dialogue for the practical purpose of providing methods for evaluating

fallacies of a kind that had been for a long time of practical interest to logicians. On

his model there are three fundamental formal components of a system of dialogue

for representing rational argumentation. First, there are two participants (parties),

usually called the proponent and the respondent, or White and Black respectively.

Second, there is a set of moves made by each party in which the party puts forward

a so-called locution, like making an assertion or asking a question. Nowadays these

locutions are called speech acts. By convention, White moves first, and then the two

take turns. Hamblin (1971, pp. 131–132) showed how are these three components

are combined to define the concept of a dialogue as an orderly sequence of moves.

Such a sequence (Hamblin 1971, p. 130) is a triple, _n, p, l_, where n is a number

representing the length of the dialogue (the number of moves so far), p is a participant,

and l is a locution. Accordingly, each move in a dialogue is a triple _n, p, l_.

Hamblin (1971, p. 131) gave the following example to illustrate a dialogue.

DIALOGUE 1. _0, P0,L4_, _1, P1,L3_, _2, P0,L2_

IS THERE A BURDEN OF QUESTIONING? 3

In dialogue 1, at the first move (move zero), the first participant P0put forward

a locution, L4. At the second move, the other participant P1 put forward another

locution, L3. And then at the third move, the first participant P0 replied by putting

forward another locution L2. Thus according to Hamblin’s definition, a dialogue

can be modeled in this way, as a numbered sequence of moves. One type of move

is the putting forward of an argument. Thus the argument in any given case, as

found in a text of discourse, can be modeled not only as an inference from a

set of premises to a conclusion that has properties like deductive validity and so

forth. It can also be viewed as a move that is part of a longer sequence containing

prior moves and subsequent moves. Another notion that Hamblin introduced is

highly characteristic of the types of dialogues he constructed. This is the notion of

a commitment set, or so-called commitment store. As each party makes a move, a

statement (proposition) is inserted into or deleted from a set of statements attributed

to her and kept track of as the dialogue proceeds.

A natural hypothesis is that such a dialogue model could usefully be applied

to legal argumentation. Alexy (1989, orig. German version, 1978) showed how

such a dialogue model can be applied to legal argumentation in an influential book.

He based his theory on rules practical discourse governing argument moves made

by a proponent and a respondent in legal dialogue. For example, one rule is that

a speaker may not contradict himself, while another rule states that whoever has

put forward an argument is obliged to defend it. These rules show how Alexy

had moved to a dialogue model of legal argumentation, founding a program now

being carried forward by a group of researchers in AI and law (Loui 1998). Recent

surveys of this research have shown how important argumentation has become in

AI and Law (Bench-Capon 1997). It has also been shown how dialogue based, or

so-called dialectical models, fit well with new developments in AI, especially with

new systems used to investigate defeasible argumentation in AI (Hage 2000).

Another development was the growing interest in legal argumentation on the

part of those working in AI. Bench-Capon (1995) recognized the crucial role that

argumentation plays in legal justification, and showed how the dialogue format

of argumentation is highly compatible with computer formalization. Hage et al.

(1994) analyzed procedural reasoning of the kind used in argumentation in socalled

hard cases in law. They developed what they call a “dialogical reason based

logic” through the analysis of the reasoning used to justify conclusions in these

hard cases. According to their analysis, legal reasoning needs to be seen in a dialogue

framework that can model an adversarial setting in which there are arguments

on both sides of a case. But in order to accommodate their analysis, they showed

that just thinking of logical reasoning in law as a chain of inferences is not good

enough, and that rules governing the moves of dialogue between the two sides also

need to be taken into account. They conclude (p. 113) however that there is no

one set of dialogue rules governing the argumentation in a case: “there are many

concurring sets of rules that govern particular types of dialogue”. This analysis,

4 DOUGLAS WALTON

Table I. Types of dialogue

Type of dialogue Initial situation Participant’s goal Goal of dialogue

Persuasion Conflict of opinions Prove your thesis is true Resolve or clarify issue

Inquiry Need to have proof Find and verify evidence Prove (disprove) hypothesis

Negotiation Conflict of interests Get what you most want Reasonable settlement that both can live with

Information-seeking Need information Acquire or give information Exchange information

Deliberation Dilemma or practical choice Co-ordinate goals and actions Decide best available course of action

Eristic Personal conflict Verbally hit out at opponent Reveal deeper basis of conflict

IS THERE A BURDEN OF QUESTIONING? 5

along with converging developments in AI and law, pointed the way towards a

dialectical treatment of argumentation.

Gordon (1995) developed a dialogue model of legal pleading, the pleadings

game, used to identify the legal and factual issues of a case (Gordon 1995, p. 109).

The pleading stage is the first in a four-stage series of civil proceedings also having

a discovery stage, a trial and an appeal stage (p. 110). The plaintiff begins by filing

a complaint, and then the defendant may file an answer (p. 111). In the answer,

each of the assertions in the complaint can be admitted or denied, or a motion to

dismiss can be made (p. 111). The pleadings game analyzes and evaluates legal

argumentation within a dialogue format. Lodder (1999) also presented a dialogue

model of legal justification that incorporates features of previous dialogue systems,

including those of Lorenzen, Barth and Krabbe, Hamblin, and Perelman. Lodder’s

book summarizes many of these earlier systems and comments on how various

features of them can be adapted to the study of legal argumentation. It is clear from

Lodder’s work that the dialogue approach to the analysis of legal argumentation

fits the use of AI into modeling legal argumentation.

Legal argumentation can have many different dimensions, and there is no reason

to think any one formal structure of dialogue will fits all contexts. A new characteristic

of formal dialogues introduced after Hamblin’s time is the idea that the

sequence of moves in a dialogue moves towards some goal. The idea is that there

can be different types of dialogues and each type has a different goal. Hamblin only

made a few remarks on such matters. He wrote (1971, p. 137), that formal systems

of dialogue are “information-oriented”, meaning that it is assumed that the purpose

of the dialogue is the exchange of information among the participants. However,

it looks more likely that the formal systems of dialogue constructed in Hamblin

(1970) could better be classified as having a goal of rational persuasion of one party

by the other through the use of a series of steps of argumentation. In these formal

dialogues, the respondent starts out by being uncommitted to some statement, and

the efforts of the proponent are directed towards securing the respondent’s commitment

to this statement through the use of a connected sequence of arguments,

one step at a time. It would appear that Hamblin had identified two different types

of dialogue, one which seems to have a goal of rational persuasion of one party by

the other, while the other type of dialogue is information-oriented. Thus we are led

to the question of whether different types of dialogue can be indentified.

As an entry point into classifying the main types of dialogue, the classification

of the six basic types of dialogue analyzed in Walton and Krabbe (1995) and

Walton (1998) is presented in Table I below. These six types are not the only types

of dialogue that might be encountered in legal argumentation. However, they are

six basic types that have emerged in argumentation theory as the most basic types

for studying informal fallacies, and for analyzing and evaluating argumentation

generally.

As shown in Table I, each type of dialogue has what might be called a communal

goal, and each party in the dialogue has his or her personal goal as a participant.

6 DOUGLAS WALTON

This typology can be applied to any given case in which an argument was used in

a text of discourse, the assumption being that the evaluation of how the argument

was used (well or badly) depends on the type of dialogue in which it was used.

An argument that may be an appropriate means to fulfill one goal may not be

useful to fulfill a different goal that is appropriate for a different type of dialogue.

Indeed, in some instances an argument can be used as a deceptive tactic, or fallacy,

to block or interfere with a goal of dialogue, even if it appears to be appropriate

to the respondent against whom it was used. Fallacies are tricky tactics used to

try unfairly to get the best of a speech partner in a dialogue (Walton 1995). Thus

formal dialogue systems can be used to study fallacies.

The best way to get an intuitive idea of what persuasion dialogue is supposed to

be like is to consider the type of dialogue known in the argumentation literature as

the critical discussion. Recently the view has been out forward that the argumentation

in a trial can be modeled as a critical discussion (Feteris 1999).4 In the model of

van Eemeren and Grootendorst (1984, p. 34), the purpose of a critical discussion

is to resolve a conflict of opinions by means of rational argumentation. On their

account (1984, pp. 85–86), there are two parties (participants) and four stages

in a critical discussion. At the confrontation stage (p. 85), the one participant, a

proponent, advances a so-called “point of view”, and the other participant, a respondent,

casts doubt on that point of view, or in some cases advances an opposed

point of view. A point of view, also called a standpoint, is defined as a proposition

(statement) and an attitude (pro or contra) with respect to that proposition (van

Eemeren and Grootendorst 1992, p. 15). At the opening stage, the two parties

agree to attempt to resolve the dispute by expressing their points of view, and

undertaking to resolve the conflict between them by advancing opposed rational

arguments. During the argumentation stage, each side brings forward arguments

to support his or her own point of view, and each takes turns questioning and

criticizing the arguments put forward by the other side. At the closing stage, the

conflict of opinions is resolved.

The following ten dialogue rules govern all moves made by both parties

during the argumentation stage. This set of rules was stated by van Eemeren

and Grootendorst in their book (1992, pp. 208–209), but can also be found in their

article (van Eemeren and Grootendorst 1987, pp. 284–291), as quoted below.

Rules for a critical discussion

Rule 1. Parties must not prevent each other from advancing or casting doubt

on standpoints (p. 284).

Rule 2. Whoever advances a standpoint is obliged to defend it if asked to do

so (p. 285).

Rule 3. An attack on a standpoint must relate to the standpoint that has really

IS THERE A BURDEN OF QUESTIONING? 7

been advanced by the protagonist (p. 286).

Rule 4. A standpoint may be defended only by advancing argumentation relating

to that standpoint (p. 286).

Rule 5. A person can be held to the premises he leaves implicit (p. 287).

Rule 6. A standpoint must be regarded as conclusively defended if the defense

takes place by means of the common starting point (p. 288).

Rule 7. A standpoint must be regarded as conclusively defended if the defense

takes place by means of arguments in which a commonly accepted scheme of

argumentation is correctly applied (p. 289).

Rule 8. The arguments used in a discursive text must be valid or capable of being

validated by the explicitization of one or more unexpressed premises (p. 290).

Rule 9. A failed defense must result in the protagonist withdrawing his standpoint

and a successful defense must result in the antagonist withdrawing his doubt about

the standpoint (p. 291).

Rule 10. Formulations must be neither puzzlingly vague nor confusingly ambiguous

and must be interpreted as accurately as possible.

In the model of van Eemeren and Grootendorst (1984, p. 86), a successful critical

discussion ends with the resolution of the initial conflict by showing that one

party’s argumentation was successful while that of the other was not. If the conflict

was not decisively resolved (1984, p. 86), “it is unclear whether the discussion

has had any point”. This condition is reminiscent of a trial in law. The purpose

of the trial is to resolve the conflict of opinions and, unless it realizes this goal,

it is not really successful. Trials have all kind of procedural rules that vary from

one jurisdiction to another. But still, the trial itself does have a core dialectical

structure representing rational argumentation of a certain sort. According to Feteris

(1999) the critical discussion model represents a normative standard or framework

in which instances of legal argumentation can be judged as either meeting the

standards of the model or failing to meet them.

A central assumption on which the analysis of this paper is based is that the

persuasion type of dialogue is at the heart of the argumentation in a fair trial. On

this model, the fair trial is supposedly based on a conflict of opinions, and each

side has a central thesis or claim to be proved or cast into doubt. The two opposed

sides are supposed to present the strongest possible argumentation for each side

(Feteris 1999). The model of the persuasion dialogue fits what is often referred to

as the advocacy system, held to be the system of dispute resolution in the Anglo-

8 DOUGLAS WALTON

American trial (Frank 1963). The purpose of the attorney on each side is to win

by presenting a more persuasive argument. But the goal of the trial cannot be seen

as purely adversarial. It is supposed to provide due process, provided by having

a trier (a judge or jury) who listens to the arguments put forward by both sides,

and arrive at a decision by ruling on who presented the more persuasive argument.

The trier resolves the conflict of opinions, not the participants themselves, and the

introduction of this third party makes the trial more than just a persuasion dialogue.

It is a complex dialogue, with many participants.

The most obvious way a common law trial differs from the model of the persuasion

dialogue is that besides the proponent and the respondent, the trial has

additional participants such a judge or a jury, who has powers that influence the

outcome. These participants can allocate a burden of proof, and can assess the

relevance, probative weight and relevance of arguments, based on rules of evidence

and other procedural rules that apply to the argumentation in a trial. For example,

even if one side in a civil case disputes everything the other side says throughout the

case, she still loses if the jury or judge finds for the other side. Hence it is clear that

the common law trial is not simply a persuasion dialogue. The relationship between

the two can be clarified by recalling Wigmore’s distinction (1931) between what

he called the science of proof, or principles of logical argumentation generally, and

the trial rules used to judge argumentation in a judicial tribunal. Wigmore held that

there should be a relationship between these two aspects if the argumentation used

in a given trial is to be evaluated as a rational process of drawing a conclusion

meant to seek the truth about an issue. Some idea of the complex nature of this

relationship is shown in a passage from Wigmore’s Principles quoted by Twining

(1985, p. 156).

1. That there is a close relation between the Science and the Trial Rules

analogous to the relation between the scientific principles of nutrition and

digestion and the rules of diet as empirically discovered and practiced by

intelligent families.

2. That the Trial Rules are, in a broad sense, founded upon the Science; but that

the practical conditions of trials bring into play certain limiting considerations

not found in the laboratory pursuit of the Science, and therefore the Rules do

not and cannot always coincide with the principles of the Science.

3. That for this reason the principles of the Science as a whole, cannot be

expected to replace the Trial Rules; the Rules having their own right to exist

independently.

4. But that, for the same reason, the principles of the Science may at certain

points confirm the wisdom of the Trial Rules, and may at other points

demonstrate the unwisdom of the Rules.

These remarks reveal clearly how the relationship between the general notion of rational

argumentation in a persuasion dialogue is abstract and general, representing

normative rules of a that determine the kinds of moves that can be made in rational

argumentation. The problem is how such abstract rules apply to individual cases

IS THERE A BURDEN OF QUESTIONING? 9

of real argumentation that might occur in an actual trial governed by procedural

rules that apply in a given jurisdiction as interpreted by a judge. Such trial rules,

as Wigmore observed, have their own right to exist independently as part of an

institution. The persuasion dialogue is only an abstract normative (logical) model,

whereas an actual trial governed by trial rules is a particular speech event, a case of

argumentation used in a social or institutional setting with rules that apply within

that jurisdiction.

Another complication is that the persuasion dialogue, at best, only models the

argumentation during the argumentation stage. The way evidence enters into a

common law trial is a multi-staged process. Gordon (1995) has studied the argumentation

mainly in the pleadings stage, where the issue is defined. This stage

poses the conflict of opinions that is supposed to be revolved by the argumentation

that follows. But surrounding this central persuasion dialogue is an elaborate process

of dispute resolution and evidence collection in the common law trial that has

nine stages (Park et al. 1998, pp. 4–8). First is the pre-trial litigation stage, including

discovery, motions, and hearings. The second stage is that of jury selection. The

third stage is the presentation of opening statements to the assembled court by the

attorneys for both sides. At the fourth stage, witnesses are called by the plaintiff,

and then examined by both plaintiff and defendant. At fifth stage, each side has

an opportunity for rebuttal. At the sixth stage, either side can make a motion for

judgment. The seventh stage is the putting forward of closing arguments by each

side that sums up its case. In the eighth stage, the judge instructs the jury on the

law that is the basis for deciding the case. In the ninth stage the jury makes its

deliberations and reaches a verdict. The common law trial is a complex nesting of

dialogues within dialogues, and no formal model of dialogue can encompass all

aspects of it. Still, as exponents of the advocacy system have so often maintained,

persuasion dialogue has a central place.

Woven around the central persuasion dialogue, however, are not only other

embedded types of dialogue, but also procedural rules of various kinds that determine

what is allowed into the central persuasion dialogue as evidence. An

important factor is that not all arguments are admissible in a common law trial,

because rules of evidence lay down requirements on what sorts of arguments can

or cannot be presented. These rules of evidence determine what sorts of arguments

are admitted and whether they are held to be rationally persuasive. There is also

a second type of dialogue embedded in the persuasion dialogue in a trial. It is

the information-seeking dialogue, which enables the collection of facts in a case,

enabling the argumentation in the persuasion dialogue central to the case to be

based on premises that include the relevant information. The persuasion dialogue

is most emphasized as a model of argumentation by lawyers who talk or write about

trials in the Anglo-American system, because the goal of the advocate is to win a

case. What may be ignored is that a persuasion dialogue may go off the track and

come to a wrong conclusion if the argumentation is in it is not based on accurate

information that really represents the facts of a case.

10 DOUGLAS WALTON

There is not enough space here to properly support the thesis that the trial

in the common law system can be seen as a persuasion dialogue, or to fill out

all aspects of how the various types of dialogues in the trial fit together. Nor is

there space to comment on the dialectical framework of the trial in continental

law, in the so-called inquisitorial system. Anyhow, let’s move ahead on the not

implausible hypothesis that persuasion dialogue of some sort is centrally involved

in the common law trial, as appears to be suggested by the remarks of those, like

Judge Frank (1963), who have defended the common law trial as a method for

providing due process on the philosophical basis that it represents an adversarial

system in which each side brings out its strongest and most persuasive arguments

to clash with those of the other side. Even if this hypothesis is granted, there remain

questions on precisely what type of persuasion dialogue it is. There are technical

problems in building formal models of persuasion dialogue, and the study of these

problems have indicated that more than one type of model needs to be considered.

The key technical problem in modeling persuasion dialogue is the management

of commitment retraction (Walton and Krabbe 1995). Commitments need to be

binding, because the whole purpose of the dialogue is to fix commitment. Thus a

participant cannot be allowed to simply retract a commitment at any point, once she

begins to run into some sort of trouble attempting to maintain it. On the other hand,

commitments cannot be fixed permanently, because tolerance for a certain amount

of retraction is necessary. For example, suppose a participant is maintaining a

commitment, but it is then shown by the other participant that this commitment

is logically inconsistent with some central commitment in the first party’s position

she had earlier maintained vigorously. Here, the reasonable thing for the participant

who has been shown to have inconsistent commitments is to retract one or the

other of the commitments that have been cited. The technical problem, as shown

by Walton and Krabbe, is to find a system that is loose or flexible enough to allow

for such retractions but is also tight enough so that a participant can retract any

commitment at any time, even if such a retraction would clearly be at odds with

what she is maintaining, or needs to maintain, as her viewpoint in the dialogue.

The solution to this problem worked out by Walton and Krabbe (1995, p. 126)

is to introduce two distinctive types of persuasion dialogue, PPD, or permissive

persuasion dialogue, and RPD or rigorous persuasion dialogue. PPD allows for

a reasonable freedom of retraction, and also for making of moves, while RPD is

much more restrictive in these matters. In a PPD there are two parties, called White

and Black, who make moves of various kinds (p. 133). There is an initial conflict

description, stating the initial assertions and commitments of both parties (p. 133).

Each party has a commitment store (commitment set), a set of statements (p. 134).

Each move of each party can contain various components: retractions, concessions,

requests for retractions or concessions, arguments and challenges (p. 135). In a

PPD, a move is a six-tuple, and a party can choose to put forward any or all of

these six components at each move. Thus there is a good deal of flexibility in

how a party make a move in PPD. In contrast, in an RPD, a party can only put

IS THERE A BURDEN OF QUESTIONING? 11

forward one of these components at each move. In both PPD and RPD, there are

rules governing how the second party must respond to each type of move made by

the first party. These rules are rigid in RPD, meaning that only a small number of

precisely determine responses are allowed, while in PPD the rules are permissive.

2. A problem in modeling legal argumentation

A general problem of legal argumentation can be posed as follows. What happens

if a strong argument supporting one side in a trial is not rebutted or even critically

questioned by the other side? Does it mean that, inevitably, the argument will be

convincing to the trier, and thus undercut the argumentation of the side who failed

to cross-examine? Or does it just leave us wondering whether the trier even took

the argument into account? To be more specific, consider a case in which a witness

presented testimony that would be very convincing evidence to the jury, but the

other side failed to cross-examine on this testimony. Would this always be a serious

lapse? Should it mean that the side that failed to cross-examine had committed a

fault that was so serious that it ought to be seen as a failure of argumentation of

a kind that always means that one side has not been adequately represented? Or

by default, should it be assumed that the trier accepted the argument, since he

(she/they) failed to critically question it? These questions need not be merely psychological

ones of how a judge or jury might react to such a failure to question an

argument. They can be questions about rational argumentation and how to evaluate

it as successful or not according to standards in a normative model of dialogue.

The modeling of legal argumentation proposed in Walton (2002) accepts

Feteris’ general view that legal argumentation in a trial has the normative structure

of a critical discussion, but takes the analysis even further along. In the Walton

model, the critical discussion is taken to be subspecies of persuasion dialogue,

referring not just to psychological persuasion but to rational persuasion using

reasoned argumentation (Hamblin 1970, p. 264). Rational persuasion refers to the

goals of the two parties, the kinds of moves they make in arguing, questioning

and replying, the rules governing these moves, and the commitments of the two

parties. Basic to the persuasion dialogue, and indeed any model of dialogue, is the

notion of a commitment set, following the analysis of it attributed to Hamblin in

the preceding section. Each party begins with a set of commitments, and these sets

are added to, or reduced, as the dialogue proceeds. The goal of the proponent of an

argument in a persuasion dialogue is to get the respondent to become committed

to a proposition that he was not committed to before, based on premises he is

already committed to. The concept of rational persuasion is essentially defined

by the transition by means of argument from one point to another The one point

is the participant’s initial lack of commitment to a proposition. The other is his

subsequent commitment to it.

Of course, the critical discussion is not a formal model of dialogue. It is a set of

rules defining a type of dialogue. But formal models of persuasion dialogue have

12 DOUGLAS WALTON

been presented inWalton and Krabbe (1995). As Rod Girle once observed, a formal

dialogue can be viewed as a structure typical of a search procedure of the kind

familiar in AI, having three basic components. The first is a start point. The second

is a set of rules that can be applied recursively to the start point, transforming it into

a next point, and that point into a next point, and so forth, producing a sequence.

The third is an end point, or last point in the sequence. In the account of Walton

and Krabbe (1995, p. 149), there are four such rules. Locution rules indicate the

types of permissible moves. Commitment rules govern which statements go into

or out of commitment stores at each type of move. Structural rules determine turntaking

and what kind of reply is required after each move. Win and loss rules

define the participants’ aims in the dialogue.5 There can be all kinds of formal

models of dialogue, some that are strict and highly regulated, and others that are

more permissive, and thus more reminiscent of natural language argumentation.

Note that nowhere in the ten rules of the critical discussion does it say that one

party has to address an argument just put forward by the other, say, by critically

examining it or arguing against it. Is this an oversight, or does it simply reflect the

assumption that each party should be free to adopt any strategy that, in his or her

opinion, might be most successful as a means of advocating his or her viewpoint?

So far, this is an open question.

The analysis of legal argumentation of the kind typical of a trial is based not

just on persuasion dialogue, but on other types of dialogue as well.When a witness

presents testimony in the trial, the question-reply sequence is seen as informationseeking

dialogue (Walton 2002). When the witness is questioned, this is seen

as what is called examination dialogue. ‘Examination’ means not only looking

at something carefully, but also testing it by probing into it by asking critical

questions. Thus on this account, one dialogue can be embedded in another. For

example, a persuasion dialogue can often be much more successful if based on a

prior information-seeking dialogue in which the facts are collected and examined.

Thus, for example, a critical discussion on a topic like euthanasia or abortion could

be more successful if both parties are well informed about the medical facts relating

to the subject. Such a case indicates how a kind of information-seeking dialogue

called expert opinion dialogue is often embedded in a persuasion dialogue. Appeal

to expert opinion is a very common form of argumentation, both in law and in

everyday speech. It has a special form, or so-called argumentation scheme, as

explained further below. For the moment, let us see that an appeal to expert opinion

can often be a very strong form of argument in law.

Expert testimony has become such a dominant form of evidence in Anglo-

American law that many worry that it is overwhelming other kinds of evidence

and that this is a problem. For example, a Canadian judge recently spoke out about

“hired gun” experts who are biased, because they are paid to testify by the defense

or prosecution side in a trial (Dube 2002, p. A1). Standards for expert testimony

in law are still in a process of evolution having been modified through decisions

in Frye, Daubert and Kumho tire (Safer 2002). Frye v. United States (293 F. 1013

IS THERE A BURDEN OF QUESTIONING? 13

(D.C. Cir. 1923) laid down the criterion that scientific expert evidence had to be

based on general acceptance in a scientific field. Daubert v. Merrell Dow Pharmaceuticals,

Inc., 509 U.S. 579 (1993) held that expert testimony must be both

relevant and reliable, including such factors as whether the scientific theory has

been tested, subjected to peer review, whether its rate of error is known, in addition

to general acceptance. Thus appeal to expert opinion is a form of argument that

is an excellent test bed for examining criteria of how argumentation should be

analyzed an evaluated, both in everyday contexts and in legal argumentation. One

special legal problem falling under this heading, however, is especially interesting.

In Anglo-American law, the jury’s role as fact finder means that they are the

judges of whether a witness is to be believed or not. But does this power extend

so far that a jury can reject the testimony of a superbly qualified expert, even if

that testimony has not been contradicted or impeached? It seems that it might. In

a case involving a collision between a motorcycle and a pickup truck (Mechanik

v. Conradi 139 AD2d 857, 527 NYS 2d 586), it was argued that the jury should

be required to accept the plaintiff motorcyclist’s expert testimony concerning the

speed of the motorcycle, on the ground that the defendant did not present any expert

testimony on this issue. The court disagreed, claiming that a jury is free to reject

an expert’s testimony even if it uncontradicted.

There is another issue to be considered, however. What if a motion for a new

trial is made on the ground that, because of the unrebutted expert testimony, the

verdict was contrary to the weight of evidence? In a malpractice case (Calderon

v. Irani 745 NYS 2d 610 3d Dept. 2002), the patient sued the doctor for causing

the spread of her cervical cancer by a misdiagnosis. The following summary of the

case, where P is the patient and D the doctor, is quoted from the account of Hoenig

(2002, p. 4).

P sued D, a gynecologist, for malpractice in causing the spread of cervical

cancer by misdiagnosis. P’s expert, a gynecologic oncologist, testified that P

had an 80 percent chance of long-term survival had the cancer been diagnosed

earlier. Instead, the ultimate spread of the cancer was a “death warrant”. This

causation testimony was undisputed. Nevertheless, the jury returned a defense

verdict finding that there was malpractice but that the deviation was not a

substantial factor in causing the injuries. The trial judge granted P’s motion

for a new trial on the ground the verdict was against the weight of the evidence.

Since the expert testimony attributing P’s decreased survival rate to D’s

negligence was uncontroverted, “the jury had no basis to conclude as it did on

the issue of proximate cause”.

This case shows that if a jury entirely disregards an expert’s testimony, there can be

repercussions in the form of later appeals for a new trial because of this evidence

having been ignored. Indeed, the court in the Calderon case found that, in the

absence of “reasonable justification of the jury’s rejection of the expert’s causation

testimony, the trial court was correct in setting aside the verdict” (Calderon 745

NYS 2d at 612). This case shows that the thesis that jury should have complete

14 DOUGLAS WALTON

freedom to disregard expert testimony is problematic, and that there are reasons

for putting boundaries on this freedom.

The problem posed by Calderon is a specific one for law, but is at the same time

a general one of standards binding rational argumentation. The general problem can

expressed simply through the following scenario. Suppose one party in a dialogue

puts forward an argument, like an appeal to expert opinion for example, and the

other party simply ignores it. He does not critically question it, or try to rebut it,

or respond to it at all, even though it is a relevant argument in the dialogue, and

simply passes on to some other unrelated argumentation. How should we evaluate

the argumentation such a case? Should the respondent’s failure to address the

argument be taken to imply, by default, that he accepts it? Or should it simply

be ignored? Or should there be some repercussions for the respondent? Should

there be some penalty laid on him, or should his side of the argument be judged

to be vulnerable, or open to criticism for this lapse? This problem is a general

one because there are many argumentation schemes, or forms of argument that,

like appeal to expert opinion, represent defeasible types of argument, as opposed

to deductive and inductive forms of argument. Such arguments are best evaluated

on a balance of considerations in a dialogue. Thus the problem is posed of what

rules of dialogue should be used to evaluate them. An argument like appeal to

expert opinion, if it is a rational argument of some sort, should be binding on a

respondent when put forward by a proponent in a dialogue. But how should it be

binding since, if it is not deductively valid, the respondent does not have to accept

the conclusion even if he accepts the premises?

To those of us studying formal models of dialogue in logic and computing, this

unsolved problem will be recognized right away as a central and serious one, but

for those less occupied by problems at this level of abstraction, the problem can be

posed in a more concrete way in legal argumentation. Suppose one side in a trial

brings forward an expert, like a ballistics expert or a DNA expert for example, to

testify in support of her side of the case. Suppose the witness is a genuine expert, is

highly credible, and his testimony is very convincing, but the other side fails to try

impeach him by attacking his credibility, to bring in an opposing expert witness, or

even to try to attack or question his testimony in any way. In Anglo-American law,

the jury is held to have the power to decide whether a witness is credible, or how

much of his testimony to believe. Thus theoretically, there should be no penalty

or curtailment of any sort, if a jury simply ignores an appeal to expert opinion,

even a very strong one. But is this way of ruling very satisfactory in all cases?

Perhaps not, for the other side could possibly appeal, asking for a new trial on the

ground that not all the relevant evidence had been taken into account. Thus this

very general theoretical problem about rational argumentation is also a concrete

and very real problem of legal argumentation. Before returning to appeal to expert

opinion as a special form of argument, it is necessary to situate the problem that

has now been posed in a broader context of formal dialogue systems used to model

rational argumentation. Most vital are the rules governing commitment.

IS THERE A BURDEN OF QUESTIONING? 15

3. The no commitment problem and stability adjustments

Since Hamblin (1970, 1971) first proposed formal dialogue systems as the basis

for a method of analyzing and evaluating forms of argument related to fallacies and

other logical problems, the problem of retraction of commitment has always been

central. Formal dialogue systems, Hamblin showed, are based on what he called

commitment sets, as shown in section one above. These are sets of statements

representing not only the thesis an arguer is trying to prove, but also commitments

she takes on as she makes various moves in a dialogue. Rules for incurring commitments

are fairly clear in many instances. For example, if a participant asserts

a statement, then she is committed to it, and it goes into her commitment set.

Rules for retraction are more problematic however, as indicated briefly already in

section one. In a normal persuasion type of dialogue, a participant needs to retract

if persuaded by a reasonable argument by the other party. But if he can retract

commitments whenever he wants, the other party may have a hard time persuading

her to accept anything, once he sees it can be used against him. For these reasons

the problem of retraction in persuasion dialogue is not easy to solve.

The general problem is to design commitment rules for various types of dialogue

that can regulate retraction in way that is appropriate and productive for that

type of dialogue. Of course, there can be many types of dialogue, and the rules

for any type need to be suited to the particular requirements for that type. But to

at least begin to investigate retraction, persuasion dialogue is the logical place to

start. In a persuasion dialogue, there are two participants, called the proponent and

the respondent. The proponent has a thesis, a statement he is supposed to prove

to the respondent. The respondent’s job is to raise doubts about the proponent’s

attempts to prove her thesis. Or in some instances, the respondent’s job is to prove

a thesis to the proponent, where that thesis is the opposite (negation) of the proponent’s

thesis.6 In such a case, the goal of each party is one of rational persuasion.

A participant carries out rational persuasion by presenting a structurally correct

argument7 to the other participant, an argument that also has the property that the

other party happens to be committed to all the premises. If the proponent presents

such an argument to the respondent, and the conclusion is the proponent’s thesis to

be proved in the dialogue, then the proponent has succeeded. She wins the game,

so to speak, and the original conflict of opinions has been resolved in favor of

the argumentation of one side. At any rate, these are the defining characteristics

of persuasion dialogue set out in the account of it given in Walton and Krabbe

(1995), where the steps taken to deal with problems of retraction of commitments

have already been outlined. Now the problem is to extend these steps to deal with

dialogue systems containing presumptive argumentation schemes.8

A central problem with using formal dialogue systems to model persuasion dialogue

was stated as early asWalton (1984). In a persuasion dialogue, the proponent

has a thesis, a particular statement that it is her goal to prove by means of a valid

argument. The respondent has the role of questioning this thesis. In many cases, the

16 DOUGLAS WALTON

respondent will himself have a thesis to be proved that is the opposite (negation)

of the proponent’s thesis. In a formal system there will be rules that define what a

valid argument is. To win, the proponent needs to present a valid argument to the

respondent with premises that are all commitments of the respondent. This seems

simple enough. But there is a fly in the ointment (Walton 1984, p. 135). If the

proponent simply asks the respondent to accept such an argument, where the conclusion

is the proponent’s own thesis, or even looks like it could be used to prove

that thesis, the respondent is likely to balk. Once he sees his own imminent defeat,

he is going to look for a way out. Hence neither party in a persuasion dialogue will

ever be able to achieve the goal set for him or her. Once he or she gets close to

success, the other party will simply start retracting commitments. This difficulty

for formal systems of dialogue could be called the no commitment problem.9 The

early systems of formal persuasion dialogue (Hamblin 1970, 1971) had no way

of dealing with this problem. Somewhat later systems were devised that tried to

deal with it (Mackenzie 1981, 1990; Walton 1984), but the solutions now seem

ad hoc, and do not seem to generalize very well. Certainly problems of retraction

are fundamentally important for formal systems of dialogue, and are among the

hardest problems to solve.

A better solution to the no commitment problem comes through the device

called a stability adjustment in Walton and Krabbe (1995, p. 147). To explain the

motivation of this notion, it is best to explain in a little more detail the kind of situation

that typically gives rise to this problem. Suppose the proponent has presented

the respondent with an argument that is valid, and all the premises of the argument

are statements that the respondent has previously committed to. It might seem that

the respondent is stuck. He must now accept the conclusion of the argument. But

persuasion dialogue requires some freedom to retract one’s commitments. There

are various reasons for the need for such freedom in a persuasion dialogue (Krabbe

2001, p. 143). One is that the critical discussion, a type of persuasion dialogue,

requires retraction as an essential part. A critical discussion can only be successful

in its aim of resolving a conflict of opinions by rational argumentation if one party

is persuaded by the other to give up his or her commitment to his or her thesis

(van Eemeren and Grootendorst 1992, p. 34). Another reason is that formal models

of persuasion dialogue will be closer to empirical reality if the participants are

free to change their minds once in a while (Krabbe 2001, p. 144). But suppose

the respondent who is presented with a persuasive looking valid argument by a

proponent is free at any point to retract any of his commitments. What he would

normally do, as indicated above, is to immediately retract commitment to one of the

premises. That is a natural move, after all. If he is not committed to the conclusion,

or is even opposed to it, and he sees that the premises imply the conclusion by a

valid argument, how will he normally react? Well, if the argument is valid, he can’t

contest that very well. So given that he does not accept the conclusion, he will soon

begin to think there is something pretty suspicious about one of the premises. The

problem posed by cases of this sort is how to fix the commitments of an arguer.

IS THERE A BURDEN OF QUESTIONING? 17

There are two types of stability adjustment described in Walton and Krabbe

(1995, Section 4.3.3), an external and an internal stability adjustment. The internal

stability adjustment refers to an adjustment of commitments in one’s own arguments

(Krabbe 2001, p. 149). If an arguer has retracted commitment to a statement

A, but earlier had offered some premises in an argument supporting A, then that

arguer will also have to retract commitment to at least one of these premises. For

example, consider the following dialogue on euthanasia. Pam is an exponent of

euthanasia of the kind practiced in Holland, but Roger is against euthanasia.

The euthanasia dialogue

Roger: You have maintained that killing a person is always wrong?

Pam: Yes. I do maintain that.

Roger: Well then, isn’t euthanasia killing a person?

Pam: Of course not. The decision is a voluntary one by the patient. Nobody

is killing this patient. The doctor is merely helping him to die a merciful and

peaceful death.

Roger: Yes, that can be true in many cases, where the doctor can give the patient a

drug that he can take himself. But what about a case of euthanasia where the doctor

has to administer the drug herself, because the patient can’t do it? Such cases do

occur in Holland, don’t they?

Pam: Yes, I understand that they do.

Roger: So that’s a case of euthanasia where the doctor killed the patient. Right?

Pam: I suppose it is.

Roger: According to your own admission then, some cases of euthanasia are

wrong, because you agreed that killing is wrong.

Pam: Hold on a minute. I didn’t mean to claim that killing is always wrong.

Killing can be justified in some cases, for example in war or self-defense.

In this dialogue Pam retracts her earlier commitment to the universal statement

that killing is always wrong, once she sees that this statement, along with other

commitments that she defends, leads to a conclusion that contravenes her view.

Roger’s argument can be represented in the argument diagram of figure 1 below.

18 DOUGLAS WALTON

Figure 1.

If Pam commits to the conclusion that in some cases euthanasia is wrong, this

admission would contravene her fundamental thesis in the dialogue that euthanasia

is a good policy. Thus Roger’s argument shows that something has gone wrong in

her set of commitments as a whole. She sees that she needs to make a retraction

in order to make her commitment set consistent. She needs to make an internal

stability adjustment. She needs to retract at least one of the premises in the linked

argument in Figure 1. In the case above, she retracts commitment to the statement

that killing is always wrong.

This procedure of internal stability adjustment can also be recursive, as can be

shown by extending the case a bit. Suppose that Pam had earlier put forth some

argument to prove that killing is always wrong. She would then have to go back

and retract at least one premise of this argument. Or alternatively, she could start

a different line of argument by retracting commitment to the statement in some

cases of euthanasia the doctor kills the patient. The recursive rule required to build

internal stability adjustment into a dialogue is rule 11.1 (Walton and Krabbe 1995,

p. 152).10 In the euthanasia dialogue, one can see that if Pam does not retract one

of the commitments from her position, Roger can argue that her position is inconsistent.

Thus the internal stability adjustment becomes especially important in ad

hominem arguments, or other arguments of the kind where an arguer is confronted

with some sort of real or apparent inconsistency in her commitment set.

The recursive procedure of retraction can be quite complex in some cases.

If there are many prior arguments supporting this premise in a chain of argumentation,

the respondent will have to go through the whole chain and retract

commitment to at least one premise in each of the linked arguments in it. Krabbe

(2001, p. 155) described the external stability adjustment as follows.

The idea is that a discussant who concedes all the premises of an argument . . .

must also concede the conclusion. . . . Consequently, a participant confronted

with an argument who wished to withhold or retract commitment to the conclusion

of the argument must withhold or retract commitment to at least one

IS THERE A BURDEN OF QUESTIONING? 19

premise. Since this premise may again have been supported by argument, and

so on, this will lead to a recursive procedure of retraction.

The effect of putting a rule requiring an external stability adjustment in a formal

system of dialogue is that commitment becomes “sticky”. You can retract a

commitment at any time, but such a retraction is no longer straightforward or

immediate. To effect it, you may have to carry out many other moves in a dialogue

first. This stickiness tilts the balance of power in a dialogue by giving the proponent

of an argument more power to fix the conclusion into place as a statement the

respondent is committed to. It is not so easy for the respondent to retract commitment,

once confronted with a conclusion he doesn’t like or support. He can do it,

but there is a cost to it.

The external stability adjustment is also important in dealing with the no commitment

problem. The term ‘external’ refers to the other party’s argument (Krabbe

2001, p. 155). An external stability adjustment comes into play when a proponent

puts forward a valid argument to a respondent and the respondent is committed to

all of its premises. The question, as posed by the no commitment problem above,

is whether the respondent can retract commitment to the conclusion all by itself,

without retracting commitment to any of the premises. If an external stability adjustment

is required he can retract commitment to the conclusion, but only if he also

retracts commitment to at least one premise of the given argument. An example is

the following dialogue about tipping. The proponent takes the view that tipping is

a good social practice. The respondent is doubtful about this claim, and is inclined

to be against tipping. He has argued that if any practice has negative consequences,

it is a bad practice.11 And he has argued that tipping has negative consequences.

For example, he argued that tipping leads to social discomfort.

The tipping dialogue

Proponent: Could failure to tip have negative consequences?

Respondent: I don’t see how.

Proponent: Well, suppose I fail to tip a taxi driver, and he takes this failure to

imply that he has somehow failed to provide good service, even though he did his

best.

Respondent: So?

Proponent: Wouldn’t this failure to tip have produced negative consequences?

Respondent: How so?

Proponent: Being upset about an alleged failure is troubling. It’s a negative thing.

20 DOUGLAS WALTON

Respondent: Yes, I’ll have to admit that.

Proponent: So it follows that failure to tip is a bad practice.

Respondent: Well, maybe.

Proponent: So it follows that tipping has to be a good practice.

Respondent: I can’t accept that. I definitely reject that view of the matter!

This dialogue has led the respondent into a position where he appears to be trapped

into a kind of inconsistency of commitments. He wants to reject the conclusion

that tipping is a good practice, but he has agreed to all of the proponent’s premises

leading to that conclusion by a deductively valid sequence of argumentation. If he

really wants to reject the conclusion, he has to go back and indicate which premise

he now wants to retract commitment to. The external stability adjustment requires

it.

In this case, however, there is a weak link in the chain of argumentation. Using

the example of the upset taxi driver, the proponent has shown by a chain of

argumentation that failure to tip is bad practice in some cases. The respondent

reluctantly agrees to this conclusion. But then the proponent claims, “So it follows

that tipping has to be good practice”. This statement is ambiguous. It could be a

universal generalization claiming that tipping is always a good practice. Or it could

be a defeasible generalization claiming that tipping is generally a good practice,

subject to exceptions. If the latter is meant, the general statement is consistent with

the statement that failure to tip is a bad practice in some cases. Only if the former

is meant is there a real inconsistency in the respondent’s commitment set. So in

this case there is a kind of problem or fallacy involved.12 It may look like there is a

need for a retraction and an appropriate stability adjustment. But there may not be,

until the ambiguity is resolved.

4. Presumptive argumentation schemes

The devices of internal and external stability adjustments were put forward in

Walton and Krabbe (1995) as possible solutions to the no commitment problem in

relation to the use of a deductively valid argument used by one party is a persuasion

dialogue to try to persuade the other party. But how could one solve the problem in

relation to other kinds of arguments, like inductive arguments or presumptive arguments

that are not deductive in their structure? Deductive logic is based on forms of

argument like modus ponens and hypothetical syllogism that close off commitment

in the following sense. When a proponent presents an argument having one of

these forms, and the respondent is committed to all the premises, then in order

to be rational he must also commit to the conclusion. Otherwise he contradicts

IS THERE A BURDEN OF QUESTIONING? 21

himself. Thus there is a strong need for retraction when dealing with deductive

arguments. However, in addition to deductive forms of argument, many inductive

and presumptive forms of argument have also been recognized. The problem of

retraction affects these kinds of arguments in a different way in dialogues. Our concern

here will be with the presumptive forms of argument, especially those already

recognized in the literature on argumentation schemes. Some of these forms are

identified in the list of presumptive argumentation schemes given inWalton (1996).

Perelman and Olbrechts-Tyteca (1958) identified even more of them in a less

formalistic way. Hastings (1963) built the first systematic taxonomy, aside from Aristotle’s

account of the so-called topics, or argument commonplaces. Kienpointner

(1992) has set out a fairly comprehensive list of argumentation schemes, including

deductive and inductive forms among many presumptive argumentation schemes.

The list of presumptive schemes in Walton (1996) includes forms of argument

like argument from sign, argument from example, argument from commitment,

argument from position to know, argument from expert opinion, argument from

analogy, argument from precedent, argument from gradualism, and the slippery

slope argument.

One of the rules of the critical discussion (van Eemeren and Grootendorst 1987,

p. 289) provides a way of fixing commitment. Their Rule 7 reads: “A standpoint

must be regarded as conclusively defended if the defense takes place by means of

arguments in which a commonly accepted scheme of argumentation is correctly

applied”. This rule seems to suggest that if the proponent of an argument defends it

by correctly applying an argumentation scheme, then the respondent has to accept

the conclusion. But can the respondent later retract commitment to that conclusion?

If the standpoint has been conclusively defended, it would seem not. If not, then

this rule is a way of fixing commitment. But as noted above, it is very important

that a critical discussion allow for retraction of commitment in some instances.

Hence in order to explore how the critical discussion could be modeled in a formal

dialogue system with commitment rules, the problem of how to formulate precise

rules of retraction must be addressed.

The following dialogue presents an example of an argument based on the argumentation

scheme for appeal to expert opinion.

The cholesterol dialogue

Proponent: Eating food containing high levels of fat causes high cholesterol.

Respondent: How can you prove that claim?

Proponent: Dr. Sheila says so, and she is an expert.

Respondent: What field is Dr. Sheila an expert in?

Proponent: Medicine. She is a doctor.

22 DOUGLAS WALTON

The proponent, it seems, has won this round. The respondent should have to

commit to the claim made in the first move by the proponent. If the argumentation

scheme for appeal to expert opinion has been correctly applied in this case, then

according to Rule 7 of the critical discussion, the argument has been conclusively

defended. Hence the respondent must accept it, and presumably he can’t try to

retract commitment to the argument or to one or more of the premises. But of

course the respondent should have some other options before committing to the

proponent’s claim. He should be able to ask some critical questions. An appeal to

expert opinion could be deductively valid if epistemic closure is achieved, meaning

that the expert can be treated as omniscient. But in the typical case, appeal to expert

opinion should be treated as a defeasible form of argument. Experts can be wrong.

If so, it is the presumptive type of argumentation scheme that should be applied to

cases like the one above.

The structure of how arguments like the one in the cholesterol dialogue should

be evaluated can be based on the argumentation scheme for appeal to expert

opinion formulated in Walton (1997, p. 210). A source is taken to represent a

knowledge base that can be questioned so that knowledge, advice or information

can be extracted from it.

Argumentation scheme for appeal to expert opinion

Major Premise: Source E is an expert in subject domain S containing proposition

A.

Minor Premise: E asserts that proposition A (in domain S) is true (false).

Conclusion: A may plausibly be taken to be true (false).

According to the analysis in Walton (1997), appeal to expert opinion is, in typical

cases, a presumptive form of argument. There is quite a natural tendency to respect

experts, but in many cases, it is best to critically question an expert opinion before

accepting it. Thus appeal to expert opinion should be seen as a presumptive form of

argumentation that is open to critical questioning. The six basic critical questions

matching the appeal to expert opinion, as indicated in Walton (1997, p. 223), are

listed below.

1. Expertise Question: How credible is E as an expert source?

2. Field Question: Is E an expert in the field that A is in?

3. Opinion Question: What did E assert that implies A?

4. Trustworthiness Question: Is E personally reliable as a source?

5. Consistency Question: Is A consistent with what other experts assert?

6. Backup Evidence Question: Is E’s assertion based on evidence?

IS THERE A BURDEN OF QUESTIONING? 23

In the cholesterol dialogue, the respondent asked the field question. The proponent

gave an appropriate and plausible answer. So the proponent has the upper hand for

the moment. But perhaps the respondent could go on to ask other critical questions

on the list. The device of critical questions shows how presumptive appeals to

expert opinion need to be evaluated in a dialectical way. If a given argument meets

the requirements of the argumentation scheme, and the respondent accepts the

premises as commitments, then the respondent should commit to the conclusion.

But the respondent should have the right to ask any of the appropriate critical

questions indicated above. When he does that, the weight of presumption shifts

back from the proponent’s side.

There is a problem here, however.13 The shift in the burden of proof, as described

above, may not always happen. In some cases, it may be that a positive

answer to a critical question can be assumed. In such a case, asking a critical

question may not be enough, and the respondent has to provide an argument to

show why the critical question has a negative answer. Thus for example in a case

of an appeal to expert opinion, the expert may be very credible, and both parties to

the dialogue may know and accept that she is highly credible expert. In such a case,

asking the expertise question may simply not be enough to shift the burden of proof

back against the proponent of the appeal. It is not clear yet how this problem should

be dealt with. Generally, in any given case, one critical question may have much

more impact than others. The others may have already been answered or addressed

quite well in the previous dialogue, or by the common knowledge shared by both

parties in the dialogue. Thus it seems that this shifting of the burden of proof tends

to work only with some critical questions and not others.

An important part of the new research on argumentation schemes is to try to

grasp how the schemes fit into formal systems of dialogue (Reed and Walton 2001,

2002). Many of the most common schemes have a question-reply format of a kind

that seems readily adaptable to a dialogue format. And it seems that some understanding

of how argumentation schemes can be used to evaluate argumentation and

fallacies in natural language texts of discourse will only come through fitting the

schemes into formal systems of dialogue. But many of the most basic questions

regarding such fittings have barely even been asked. Below, some of them are

posed in the form of fundamental problems. It seems to be early in the game to

offer definitive solutions to them, but even so, framing the problems in a clear

and orderly way is a worthwhile first step. Even the first step of trying to classify

different types of argument into deductive, inductive, and some third category, is

highly difficult to take without running into unsettled questions and controversies.

In deductive logic, deductively valid forms of argument like modus ponens and

disjunctive syllogism have the property that if the rational respondent commits to

the premise he must also commit to the conclusion. The problem for deductive arguments,

as indicated above, is that the respondent, at least in a normal persuasion

dialogue, should have the right to retract commitment to one of the premises. One

solution is to require a stability adjustment in such a case before the retraction can

24 DOUGLAS WALTON

be made. But suppose the proponent puts forward an argument that has the form

of one of the presumptive argumentation schemes. A probative weight is transferred

from the premises to the conclusion, but probative weight is defeasible.14

The respondent may be committed to the premises, but that does not necessarily

mean he must also commit to the conclusion. He can ask critical questions. And

indeed, he should ask them. As Prakken (2002, p. 5) noted, most AI and law models

of legal procedure have incorporated formal dialogue systems, but have extended

them by making argumentation in them defeasible. And presumptive argumentation

schemes represent defeasible forms of argument. Thus the no commitment

problem for presumptive argumentation schemes arises in a different form in legal

argumentation that doesn’t seem to be amenable to solution by imposing internal

or external stability adjustments on argumentation in a dialogue.

5. The RIB problem

In Walton (1996), twenty-five argumentation schemes for presumptive reasoning

have been identified and analyzed. Matching each scheme is a set of critical

questions. An argument is evaluated by weighing evidence on both sides at the

given point in a dialogue where the argument was used. If the respondent is

committed to all the premises, and the argument put forward by the proponent has

the form of a known argumentation scheme, then that is a reason for the respondent

to commit to the conclusion. But it is not a conclusive reason. He can still ask a

critical question, and indeed, that is what he is supposed to do. But here too, as in

the case of deductive arguments, there seems to be a no commitment problem.Why

should the respondent have to accept the conclusion, or even pay any attention to

the argument at all? A stability adjustment could be the solution to this problem,

but an even more direct solution seems to be available. This solution is to rule

that the respondent should have only the following three options when confronted

by an argument that has the form of one of the presumptive argumentation schemes.

1. Question one of the premises.

2. Ask an appropriate critical question.

3. Accept the argument (at least tentatively).

The rule that the respondent is bound to respond to an argument with one of

these three options could be called the three-option reply rule or TOR rule.15

5.1. THE TOR RULE

If a proponent puts forward an argument that has the form of a presumptive argumentation

scheme, at the next move the respondent must (1) question one of

the premises, (2) ask an appropriate critical question, or (3) accept the argument

(at least tentatively). The respondent is not allowed to make any other kind of

IS THERE A BURDEN OF QUESTIONING? 25

move, like putting forward a counter-argument, until he has made one of these

three moves.16

The TOR rule seems attractive because it looks like a good way of dealing

with the no commitment problem without having to introduce devices like stability

adjustments. But the TOR rule raises another question. What about the option of

bringing forward another argument that defeats the appeal to expert opinion that

was just put forward by the proponent? Should the respondent have that option

right away? Or should he have to wait? This question expresses what could be

called the right to immediately rebut (RIB) problem.

The RIB problem can be illustrated by adding another move to the cholesterol

dialogue above. Suppose the respondent were to make this move.

Respondent: Your claim is just wrong. It’s the saturated fats that cause high

cholesterol, not the unsaturated fats.

In this kind of case, the respondent has done an end run around the proponent’s

appeal to expert opinion argument. Instead of trying to attack this argument, or ask

further critical questions about it, he attacks the proponent’s thesis directly. But

should he be allowed to do this? It seems initially reasonable that he should be.

The rules of PPD17 in Walton and Krabbe (1995, pp. 133–140) allow a respondent

to bring forward a counter-argument at the next move after a proponent’s

argument. But if the TOR rule were imposed, the right to rebut immediately

would disappear. Thus a big question concerning the relationship of argumentation

schemes to formal systems of dialogue remains open. This is the question of how

binding the argumentation scheme should be in a dialogue. Should it force the

respondent to either accept the argument or critically question it?18 Or should it

leave him free to rebut by posing a counter-argument? These options for dialogue

rules could be called the tighter versus the looser approach.19 Most observers would

possibly opt for the looser approach, since it leaves more freedom to express opposition

to a proposed argument. But there is an argument for the tighter approach.

This argument is based on a holistic approach to argumentation evaluation that can

be explained as follows.

In a dialogue like a protracted critical discussion many defeasible arguments

tend to be put forward by both sides. Each, by itself, gives only a small weight of

evidence for or against the theses at issue globally in the dialogue. It’s only when

you put them all together at the end of the dialogue that you get a decisive tilting of

the burden of proof to one side or the other. The resolution of the issue in a dialogue

doesn’t work by pitting one single argument against another.20 Evaluation should

be holistic, and at the end of the dialogue, when all relevant arguments are weighed

together (Fox and Das 2000). If we adopt this holistic point of view, it provides an

argument for supporting the tighter approach. The reason is that when a proponent

puts forward one argument, it is of no immediate significance that the respondent

pits it against an attacking argument that has the opposite conclusion. For example,

26 DOUGLAS WALTON

suppose the proponent puts forward an appeal to expert opinion for conclusion

A. Suppose at the next move he puts forward an argument from analogy with the

conclusion not-A. Is this a useful or appropriate response? On the one hand, it

seems so, because the respondent is indicating his reason for disagreeing with A.

But on the other hand, it may be irrelevant, because the appeal to expert opinion has

been left dangling, and the argument from analogy may take the dialogue away to

a completely different direction. Also, the opposition between the appeal to expert

opinion and the argument from analogy is unimportant in the end, on the holistic

view. For the issue will be decided when all arguments on both sides are weighed

up together in the mass of evidence collected through the dialogue. What is more

important is that the respondent should address the argument from expert opinion,

indicating whether he accepts it, or can find a weakness in it. Thus the argument

from evaluation at the closing stage of a dialogue suggests that the TOR rule is

justifiable.

Still, even those who like the holistic view may have reasons for questioning

whether it supports the tighter approach. Henry Prakken, in comments to the author,

gave four reasons for questioning this support. The first is that even if the holistic

view is adopted, the dialogue rules should allow the stating of counterarguments at

some point, and what is a more natural point than immediately after the attacked

argument? The second is that logics for defeasible argumentation partly capture the

holistic viewpoint anyway, since they evaluate an argument in light of all relevant

arguments and counterarguments. The third is that there is currently an unresolved

problem of evaluating sets of arguments and counterarguments. Pollock (1995)

has argued that arguments do not accrue, and if there is more than one reason

for a conclusion, they should be combined into one argument. On the other hand,

in everyday dialogues, people often speak as if they added up arguments. Thus

evaluation at the closing stage, while it might work well for different kinds of

formal dialogue, is not a very natural approach to empirical argumentation. For in

real cases of conversational argumentation in natural language, putting forward a

defeater in response to another party’s argument is quite a common move. Finally,

the tighter approach won’t seem to work for legal argumentation, where one argument

is often attacked in court by posing a counter-argument. For example, an

argument based on witness testimony may be attacked by an opposed argument

based on the testimony of a different witness.

There is an additional reason for thinking that the TOR rule is too strict. This

reason is that the respondent, when confronted by any argument put forward by a

proponent, should have the right to make other kinds of moves. He might not want

to question the probative strength of the argument, but he might want to question

whether it is relevant as a move in the dialogue at that point. Or he might want

to question one of the terms used in the argument. For example, he might want to

argue that a term is ambiguous, or overly vague. Or he might want to argue that the

term is emotive language of a kind that needs to be questioned or defended. Surely

this kind of move should be allowed. If so, restricting the respondent’s next move

IS THERE A BURDEN OF QUESTIONING? 27

to asking one of the formatted critical questions (ones formulated as appropriate

critical questions matching the scheme) is too restrictive.

Putting all these reasons for doubting the usefulness of the TOR rule into a balance

of considerations, serious doubts are raised about whether the tighter approach

will really work very effectively, especially as applied to cases of argumentation

in everyday discourse and legal argumentation. It seems like there are too many

factors for questioning the applicability of the TOR rule to make it useful as a

general rule for managing argumentation in dialogues. The TOR rule is too strict

if it bars all other kinds of moves, including counterarguments, until all the critical

questions have been dealt with appropriately or until the respondent concedes the

proponent’s argument. On the other hand, there still seems to be a place for some

rule that takes the asking of the critical questions into account in evaluating an

argument. Even if the dialogue rules allow for counterarguments, they should as

Prakken put it, “in some way force attention to the critical questions of the original

argument”.21 The problem is how to formulate a rule of this sort. But even if such

a rule could be formulated, one that would meet all the objections to the TOR

rule, there are reasons for thinking that it would not be universal. It would only be

appropriate in a special type of dialogue.

6. QPD: A new system between RPD and PPD

The contrast between RPD and PPD in Commitment in Dialogue was useful as a

starting point in examining some important properties of formal dialogues. But it

represents two extreme poles, so to speak. RPD is nicely amenable to formalization,

and it certainly allows the proponent of an argument to pin the respondent

down to some fixed options so he can’t waffle around by refusing to answer, go

off on irrelevant tangents, and use other escape routes. But RPD is too rigid to

represent cases of real dialogue. PPD is much more flexible, and hence much more

realistic as a way of representing cases of real dialogue. But this flexibility introduces

a fundamental problem of pinning down commitment. The respondent has

a lot of flexibility. So when the proponent presents a valid (or structurally correct)

argument and the respondent doesn’t want to accept its conclusion, he can just not

accept the premises. Or if he was previously committed to the premises, he can

just retract commitment to one or more of them now. Thus the proponent has a

problem. How can he “pin down” a respondent who always retracts commitments

when it appears that he might be in danger of losing the argument? Walton and

Krabbe (1995) dealt with this problem by the devices of internal and external

stability adjustments. But these stability rules seem to represent a higher level

of rationality than one might always wish to impose on a dialogue. If we could

get by without the stability rules in PPD, it might be possible to have a simpler

system that would show more clearly how argumentation schemes have an effect

on commitment. One such simple rule discussed above is the TOR rule. It says

that if the respondent accepts the premises of the proponent’s argument and the

28 DOUGLAS WALTON

argument is structurally correct (by having the form of a known argumentation

scheme), then the respondent should either have to accept the conclusion or he

should have the burden of asking an appropriate critical question.22 The TOR rule

gently forces the respondent to take on commitments. Or at least it prevents him

from dodging around by refusing to answer, or going off on an irrelevant tangent.

The implementation of the TOR rule as suggested above is too restrictive, as shown

by Prakken’s doubts about it as a general rule. Could there be some other less

strict dialogue rule that allow for counterarguments yet still draws the respondent’s

attention to the critical questions so that he is rationally required to take them into

account before proceeding further in the dialogue?

The problem is to allow a respondent an appropriate degree of freedom and

flexibility in replying to an argument, while finding some kind of rule that requires

him to take the critical questions appropriate for the scheme into account. Any

system of dialogue having such a rule would not be as free as a PPD type of

dialogue. But then it would not be as restrictive as an RPD dialogue either. It would

be somewhere in between. It could be called a QPD type of dialogue. It is in this

special type of dialogue that something like the TOR rule, or amore flexible version

of it, might be useful.

QPD is a simplified version of PPD (permissive persuasion dialogue), sharing

some features of PPD given in Walton and Krabbe (1995, pp. 133–140). In PPD,

are there are three kinds of commitments. The first two kinds are called concessions

and assertions. An assertion is a commitment that a participant is obliged to defend

if challenged (p. 186). In other words, it has a burden of proof attached. In contrast,

a concession is a statement a participant is committed to only in a weak sense,

meaning that she is not obliged to defend it if it is challenged (p. 186). Those of the

third kind are called dark-side commitments. These are “off-record” commitments.

They represent an arguer’s position in a dialogue, but were not explicitly stated

anywhere in the dialogue by the arguer. There are merely implied by what he said

or didn’t say. In the description of QPD below, the term ’commitment’ is used in a

broad sense, referring to any of the above three kinds of commitments.

The main difference between PPD and QPD is that in PPD, a party can make

several moves concomitantly, so to speak. She can ask a question, for example, and

in the same move, make an assertion and put forward an argument. In QPD, each

move must be made up of only a single speech act (locution). Thus if a party asks

a question, then that is the only thing she can say at that move. She has to wait to

see how the other party replies before making another move. In that respect, QPD

is comparable to RPD (rigorous persuasion dialogue), because a participant’s has

less freedom than in PPD. In PPD, a move can contain any or all of the following

components: retractions, concessions, requests for retractions, requests for concessions,

arguments and challenges. In QPD, each move can contain only one of these

locutions. The following general features define QPD as a type of dialogue.

IS THERE A BURDEN OF QUESTIONING? 29

Features of QPD as a type of dialogue

Feature 1

There are two participants (parties), White and Black, who take turns making

moves. White moves first.

Feature 2

Each has a commitments set, including a set of initial commitments that each starts

off with at the opening stage of the dialogue.

Feature 3

Each party has a special commitment, designated at the opening stage as his or her

thesis. Each party has the goal of proving this thesis to the other party, by using

rational arguments based exclusively on premises that are commitments of that

other party.

Feature 4

Each party can only put forward a single locution at each move.

Feature 5

The TOR rule governs a party’s next move whenever the other party puts forward

an argument that has the form of a presumptive argumentation scheme.

Feature 6

The commitment rules determine the commitments of each party at each move,

depending on what type of move it is.

Feature 7

Each party may demand the resolution of an explicit inconsistency found in the

other party’s commitments.

Feature 8

When one party has successfully proved her thesis, the other party must concede

this thesis at his next move, and the dialogue ends in the first party’s favor at that

move.

These features are very general. Various problems discussed below will suggest

how more specific rules might be formulated. QPD dialogue is generally similar

to a PPD in that both represent persuasion dialogue. In both types of dialogue,

the proponent’s aim is to persuade the respondent by rational argumentation to

commit to her central thesis in the dialogue. The respondent’s aim is either to raise

doubts about the proponent’s line of argumentation or to persuade the proponent

to commit to a thesis that is the opposite of her central thesis. The kinds of rules

30 DOUGLAS WALTON

for both types of dialogue are similar, except for features 4 and 5. In QPD each

party can only put forward a single locution as a move, while in PPD, a party

has a choice of putting forward any or all of six locutions at one move. The other

difference is that, in QPD, if that locution is a presumptive type of argument, the

response of the party is bound by the TOR rule.

A question is whether the rules of QPD allow for what is called backtracking

(Prakken 2002, p. 12). Backtracking is making an alternative reply to the same

argument, or other locution put forward by the other party. An interesting question

is thus posed. Suppose a proponent puts forward an argument, the respondent asks

some critical questions about it, and then the dialogue moves on to consideration of

other arguments. Can the respondent, at such a later point, come back to the original

argument and ask some more questions about it, or even make a counter-argument

aimed at rebutting it? The answer is that backtracking is allowed in QPD.23

Henry Prakken posed some problems by questioning how can certain types of

responses be allowed in QPD, given that a move may contain just one speech

act. Suppose a party wants to concede one premise of an argument but challenge

another, for example. Or suppose he wants to concede the argument’s premise but

ask a critical question. In QPD, he can’t make such moves. Do these restrictions

pose problems? Certainly they impose limits, but it seems like they are limits that

a respondent can live with. The respondent must simply choose which move he

wants to make. If you have to make a concession at your next move, according

to the rules, then that will be your move. But generally in persuasion dialogue,

taking on a commitment can be a risky move. So normally, it would be a priority

to ask questions and put forward challenges, rather than taking on commitments.

In QPD, one would normally ask critical questions first, and then later one could

make concessions, if making them helps one’s strategy. Since backtracking is allowed,

a party can concede one premise and then at a later move challenge another

premise. Thus I think these kinds of problems can be dealt with in QPD. But they

do show that there are certain kinds of moves that you can’t make in QPD that

would normally be allowed in everyday conversational argumentation.

QPD recognizes presumptive argumentation schemes, as well as deductively

valid forms of argument, as binding argument structures. It could be a useful feature

of the QPD type of dialogue that some restriction like the TOR rule could be

applied when the proponent puts forward an argument that fits one of the argumentation

schemes. Suppose the proponent puts forward such an argument at a given

move. In QPD, if the respondent is committed to all the premises of the proponent’s

argument at that move, he should have the option of critically examining the argument

by asking one or more of the appropriate critical questions. As indicated

above, he should have other options as well. But if he fails to take advantage of the

option of asking critical questions, that should have some consequences. Maybe all

that it means is that he has failed to challenge the argument because he agrees with

it, or at least does not want to question it using the formatted questions because

he has other objections and wants to proceed directly to a counter-argument. The

IS THERE A BURDEN OF QUESTIONING? 31

problem is to find some rule that would allow for these freedoms while still taking

into account the respondent’s failure to critically examine the examine in the

standard way. The value of the formatted critical questions derives from the fact

that it is sometimes very hard for an arguer to think up strong counter-arguments

or other objections that may require knowledge of the subject-matter of a dispute.

The questions perform an inventive function by offering a standardized list that

may automatically enable a respondent to search for weak points in an argument.

But as noted above, there is no reason to think that such lists of critical questions

are complete, or that they have to preempt other moves. The problem is to find a

QPD rule that achieves the needed balance.

There is another problem to be considered in QPD as well. What happens when

the respondent runs out of critical questions? At that point, does he have to concede

the conclusion of the proponent’s argument? This question formulates what could

be called the completeness problem for QPD. The answer is that critical questions

for presumptive argumentation schemes can contain critical subquestions. Thus

the respondent should have the right to ask an appropriate subquestion, having

previously asked a main critical question. In principle, therefore, the dialogue could

go on and on, as long as subquestions are there to be asked. On the other hand,

there is nothing to prevent some limit being imposed on the length of such critical

questioning intervals. The list of appropriate subquestions is surely finite. And in

many instances, it may not be all that long.

Another problem stems from the defeasibility of argumentation schemes (or

many of them, anyhow). In such cases, an argument should be subject to defeat

if new counter-arguments come in later that defeat (refute) the argument. Even

though the TOR rule above only required that the respondent “tentatively” accept

the conclusion, it seemed too restrictive because it categorically rules out other

options. But it still leaves some options open. It leaves open the possibility that

the respondent could later retract commitment to the conclusion, even after he has

asked all the critical questions and the dialogue has run along further. This kind of

freedom should be open to the respondent in all cases where he is responding to

a defeasible argument, because a defeasible argument is subject to defeat if new

evidence comes in at any point in a dialogue. Thus there is a general dialogue rule

that applies to all instances of defeasible argumentation.

New Evidence Rule. Once the respondent has accepted all the premises of an argument

that is defeasible and asked all the appropriate critical questions, he may

be rationally bound to tentatively accept the conclusion in some way. But such

acceptance should be regarded as provisional, meaning that he can later retract it,

provided new, relevant information that defeats the conclusion has come into the

dialogue.

The New Evidence Rule would enable presumptive argumentation schemes to have

some impact on commitment, but would at the same time be compatible with the

32 DOUGLAS WALTON

defeasibility of the presumptive argumentation schemes. The new evidence rule

works for presumptive argumentation comparably to the way a stability adjustment

works for deductive argumentation in a persuasion dialogue. It may suggest

a way of coping with the no commitment problem when allied with some qualified

version of the TOR rule, but it is not so drastic that it completely forbids retraction.

It may make a respondent temporarily commit to the conclusion of a proponent’s

argument in some way, even though the respondent can retract that commitment

later in the dialogue provided he makes the right moves.

These problems suggest that some rule comparable to the TOR rule could still

be useful in the setting of QPD dialogue. But it needs to be less restrictive. An

argument fitting an argumentation scheme needs to have some effect in a dialogue

in line with its being a rational argument that the respondent should have to respect.

But it can’t be so binding on the respondent that it unfairly restricts his capability

for making retractions of a kind that are reasonable, or for making other kinds of

moves that should be allowed.

7. The TORC rule in formal dialogue systems

There seem to be two ways to go in setting up a framework for analyzing and evaluating

arguments using argumentation schemes. One is to go with a PPD system that

has internal or external stability adjustment rules. The other is to go with a QPD

type of dialogue that is more restrictive but still fairly flexible in how it allows a

respondent to react to an argument. Of course, a QPD dialogue could have internal

or internal stability built into it as well. An asset of such a QPD system is that it

could be used to make a respondent take on some kind of rational commitment,

even of a very provisional sort, in response to an argument fitting a scheme put to

him, assuming he does not dispute the premises. Such a QPD system represents

a higher kind of rationality than one might always find in many instances of real

argumentation, where critical questions are not considered, or an argument that

should require critical examination is even ignored. In real cases, one might expect

to find all kinds of violations of any rule that would require the asking, or the failure

to ask critical questions, into account. But it seems reasonable in some cases that

a failure to take up the option of asking an appropriate critical question, where a

critical examination is called for, should be seen as a default.

In the end, how these problems need to be solved is by having different systems

of dialogue representing different levels or kinds of rationality. Thus we might have

many different systems, and have to make and justify decisions about whether a

given normative system applies to a given empirical case. This multiple systems

approach is by no means incompatible with the approach that has been taken in

formal dialogue systems to this point. Hamblin (1970, 1971) clearly advocated a

plurality of different formal systems of dialogue for different purposes, and leading

works in the field have followed this approach (Mackenzie 1981, 1990; Barth and

Krabbe 1982). Thus to require the asking of the right critical questions, and to

IS THERE A BURDEN OF QUESTIONING? 33

see the failure to ask them as a deficiency in the respondent’s performance in a

dialogue certainly represents some kind of rationality assumption. But what kind

of rationality does it represent, and what sort of rule could be formulated for QPD

that would concede with the default, but at the same time not restrict other options

that the respondent should have?

The way argumentation is guided by procedural rules for handling evidence in

legal contexts offers a clue on how this kind of problem should be dealt with.When

one side puts forward an argument, say an argument based on an appeal to expert

witness testimony, the other side has the right to cross-examine the witness. But

the other side also has the right to waive the exercise of that right. Sometimes, for

example, the side who has the right to cross-examine will simply say, “No further

questions”. This failure to question may be a lapse, if the appeal to expert opinion

was weak in some respect, and should have been critically questioned to bring out

this weakness. But in other cases, the failure to ask critical questions may turn out

to be insignificant. For example, the side who fails the examine may know that he

has such strong evidence that he will bring forward at a future point in the trial

that the other side’s appeal to expert opinion will fade into insignificance. It will

be rebutted, and thus there is no real need to ask critical questions of the appeal to

expert opinion argument.

We have to inquire further into what is happening in such a case, from a dialectical

point of view. The lawyer waived her right to ask critical questions, because

she could have felt, for a variety of reasons, that asking such questions would not

help prove her ultimate claim (fulfill her onus probandi or burden of proof) in the

case. She should have this right. But when she waives it, she is making a kind

of concession. She is, in effect, conceding the argument from expert opinion, by

not challenging it. This move represents the taking on of a kind of commitment.

She is not committed to the argument from expert opinion, or to its conclusion,

as propositions she advocates. She is not taking on a burden of proof. She should

not have to defend this argument or this proposition if her commitment to them

is challenged. Instead, she is conceding them “for the sake of argument”. Here

it is useful to invoke a distinction between two kinds of commitments made in

Walton and Krabbe (1995, p. 186). A commitment of one type is incurred by

making an assertion. This type of commitment is one that a party in a dialogue

is obliged to defend if the other party requests that she justify it. A commitment

of another type is incurred by making a concession. This type of commitment has

no burden of proof attached. It is a weaker type of commitment, representing a

commitment incurred for the sake of moving the sequence of argumentation along

in a dialogue. It can also be described as a negative kind of commitment. It means

that you agree not to dispute something. It means that even though you may not

accept the statement in question as part of your own position (viewpoint) in the

dialogue, you are willing to waive your right to attack it.

The sequence of argumentation in a QPD dialogue can be structured by using

this negative kind of commitment as a kind of default. If a respondent fails to

34 DOUGLAS WALTON

Figure 2.

exercise his right to critically question a proponent’s argument, then the argument

defaults in the proponent’s favor. This means that the respondent has to temporarily

accept the argument, but only in the sense of negative commitment, meaning he

has agreed for the present not to examine it by critical questioning. However, if the

respondent does exercise his right to ask a critical question, either the proponent

has to provide an appropriate answer or she agrees to temporarily give up the

argument. This means that the argument defaults in the respondent’s favor. So such

an argument can default to the one side or the other, depending on the moves made

in the dialogue. This dialogue sequence of default for an argumentation scheme is

represented in Figure 2.

This weaker or negative kind of commitment that functions as a default is a

useful device because it can help an argument move along without delaying to

dwell on critical examination that may not be turn out to be useful or necessary

as the dialogue moves along through its argumentation stage. Could this form

of commitment be useful also for devising a rule that could deal with the kinds

IS THERE A BURDEN OF QUESTIONING? 35

of problems of critical questioning a retraction of commitment discussed above?

Possibly it can, but another problem needs to be dealt with first.

An underlying problem concerns the apparent differences in the different kinds

of arguments. Deductive arguments make the no commitment problem apparently

simpler, because the list of options for the respondent is shorter. Inductive arguments

have not been explored in relation to the problems posed above, but they may

be more difficult to deal with. Presumptive arguments seem to be the most difficult

of all to deal with, because they are more open-ended. They are defeasible, and

have to be open to new evidence. They need to leave the respondent room for asking

critical questions before taking on a commitment. But they also need to leave him

room for other moves that might also be appropriate. It is perhaps for this reason

that QPD seems initially to be a good fit for presumptive argumentation schemes. It

may be too soon to provide definitive formal solutions to these problems, but there

is one possible solution for presumptive argumentation that should be considered.

This is to apply a TOR type of rule in a QPD dialogue, but make the rule less

restrictive in ways to allow for the options discussed above. It is combined with

the stability adjustment mechanism, and in a way that leaves the respondent room

for making other kinds of moves. This revised version to the TOR rule is called the

TORC rule.

TORC Rule. When the proponent puts forward an argument that has the form of

an argumentation scheme accepted as structurally correct, and the respondent is

committed to all the premises, then the respondent must (i) negatively commit to

the conclusion as presumptively plausible, (ii) retract one of the premises, (iii) ask

an appropriate critical question, or (iv) if he chooses (iii), he can go on asking

appropriate critical questions until he runs out of them. If he chooses option (i) or

(ii), the proponent can call for a stability adjustment.

The expression ’negatively commit’ in clause (i) refers to the weaker kind of

commitment described above, referring to the distinction between assertions and

concessions made in Walton and Krabbe (1995, p. 186). According to the TORC

rule, a respondent confronted with a proponent’s argument can waive his right to

ask critical questions. But if he does, or does not directly examine the argument

by one of the other avenues cited in the TORC rule, he concedes the argument.

This means that he has agreed not to dispute it, for the moment anyhow, and therefore

that he incurs a weak commitment to its conclusion as a proposition that is

plausible, based on the argument that was put forward. The rules of QPD allow for

backtracking, thus leaving room for counter-arguments later in the dialogue. Even

at the next move, the respondent can put forward a counter-argument. But if he does

so, then by waiving his right to critically examine the other party’s argument just

put forward, he weakly concedes the argument. Such a commitment is a negative

one, however. It means that he has agreed not to challenge the argument just now,

in order for the dialogue to move on.

36 DOUGLAS WALTON

How does the TORC rule deal with the general problem of retraction? The

TORC rule prevents the respondent from simply retracting the conclusion of the

proponent’s argument, or one of the premises and with it the conclusion, every

time it looks like he has to accept a conclusion that may be inconvenient for him

or that seems to go against his viewpoint. On the other hand, it still allows for

retraction. It’s just that the retraction has a cost, and cannot be carried out in one fell

swoop. Adopting the TORC rule, as stated above, may not be the only way to solve

the problem of retraction for dialogue systems with presumptive argumentation

schemes. It is a first step that may point the way to better solutions as new formal

dialogue systems come to be developed.

The TORC rule could be modified by allowing for other forms of criticism

as well. In addition to asking critical questions matching a scheme, a respondent

might want to ask critical questions, for example, about the meaning of a term

used in the proponent’s argument. What this type of rule suggests is that when a

proponent puts forward an argument in QPD, the respondent has the right to an

examination interval in which he can ask critical questions. During this interval, he

may also have the right to ask for clarifications, or to request that the proponent

define a key term used in the argument. The function of such an examination

interval is to offer the respondent a chance to critically probe into the proponent’s

argument and ask questions about it, before having to decide whether to accept it

or try to argue against it.

8. Should there be a burden of questioning?

In the case of Calderon v. Irani, appeal court ruled that a jury’s rejection of an expert

opinion “cannot be made arbitrarily” but “must be supported by other testimony or

by the cross-examination of the expert”.24 It was ruled that in the absence of any

reasonable justification for the jury’s rejection of the expert testimony, the trial

court was correct in setting aside the verdict (Hoenig 2002, p. 4).25 But what does

this ruling imply? It seems to imply that the court assumed that the jury must

have rejected the appeal to expert opinion, based on two grounds. One is that

they failed to challenge or to critically question it. The other is that they ruled,

in their verdict, in such a way that can be taken to imply that they didn’t accept the

argument. It appears that the appeal court assumed, judging from the outcome of

the trial, and from the jury’s finding that the deviation was not a factor in causing

the injury, that the jury failed to accept the argument from expert opinion. From

this information, however, we can’t really say for sure what the jury really thought,

or whether they accepted or rejected the appeal to expert opinion. The best that

can be said is that they ruled and drew conclusions in a way that is consistent

with their non-acceptance of it. And since the argument was (evidently) strong,

their non-acceptance of it may suggest that they ignored it, not even taking it into

account.

IS THERE A BURDEN OF QUESTIONING? 37

It is evident from the Calderon case, and similar cases cited by Hoening (2002),

that there is no procedural rule in Anglo-American law that requires the asking of

critical questions or the putting forward of counter-arguments in reply to an appeal

to expert opinion. There is no such binding rule, even if the argument appears to

be quite a strong one. The proponent incurs a burden of proof in putting such an

argument forward. But the respondent, whether jury or a cross-examiner, has no

burden of questioning or challenging such an argument. If the respondent fails to

critically examine the argument, or to try to refute it by putting forward a counterargument,

there is no penalty. There is only the potentially bad consequence that

an appeal might be launched for a new trial.

The same is true of the critical discussion as a type of dialogue, and of the formal

models of persuasion dialogue of the QPD family considered above. There is no

burden of questioning, or obligation to critically question, as it might be called,

that applies to a respondent when he replies to an argument. Adding the TORC

rule to a QPD dialogue would remedy this lack. The TORC rule puts pressure on a

respondent to reply to an argument, once it has been put forward by a proponent,

in one of the four ways allowed by the rule. In effect, the TORC rule creates

a burden of questioning. It makes the respondent at least confront the argument

and react to it some way. Of course, the TORC rule, as stated above, may be too

narrow. Maybe the respondent should be allowed to ask for clarifications of various

kinds, or to pose counter-arguments of various kinds, in addition to the four options

expressed in the current rule. What such a rule requires is that the respondent react

appropriately, by telling the proponent whether he at least provisionally accepts it

for the sake of argument, or if not, or by giving his reason for not accepting it.

The intent of the rule is the respondent should have an obligation to express his

reason for doubting an argument, and for not accepting its conclusion, if he has

such reasons. So should the TORC rule, or some variant of it, be applicable to

argumentation?

As far as the theory of formal dialogues goes, there is no harm in constructing

a QPD system with a TORC rule built in. It represents a special type of dialogue.

The normal persuasion dialogues considered so far in the literature have no burden

of questioning requirement of the kind stipulated by the TORC rule. But QPD with

TORC could be an interesting type of dialogue for some purposes. In contexts of

dialogue where critical examination of arguments is very important, this type of

dialogue might be a good model. Would it be applicable to legal argumentation?

So far, the only evidence that there are special contexts of legal argumentation

where QPD with TORC is a good model are the cases cited above. These suggest

there might be some advantages to having a generally applicable evidence rule in

law that would forbid the ignoring of a strong argument like an appeal to expert

testimony. But then again, imposing such a sanction on a jury or a cross-examiner

might be a burden that would be impractical, or not worth the cost and effort that

it would impose on a trial. Still, such a rule, when invoked, would cut down the

likelihood that an appeal might later be launched on the ground that a weight of

38 DOUGLAS WALTON

evidence had been ignored in a trial. The TORC rule is valuable because it makes

for a better assurance that the respondent is actively taking part in a dialogue. The

need for such an assurance especially important when an argument has the form of

one of the presumptive argumentation schemes. For apart from the TORC rule, it

is difficult to understand why the argument is rationally binding on the respondent.

There are some further questions about the TORC rule. One problem is where to

draw the line between critical questions and counter-arguments. Consider appeal

to expert opinion as an example. The consistency critical question matching the

scheme, as indicated above, where A is the statement asserted by the expert, poses

the following query. Is A consistent with what other experts assert? Of course,

this is just a question, but in some cases asking it may be based on evidence that

some other expert has disagreed with the one cited. Thus there will always be

a practical question of judging whether something is merely a critical question

as opposed to a counter-argument. In principle, however, it is both possible and

useful to make a distinction between merely raising doubts about an argument and

putting forward a counter-argument meant to rebut or refute the original argument.

Pollock’s distinction between arguments that defeat another argument, as opposed

to arguments that merely undercut it, can be cited here as relevant, even though

undercutters, in his sense are more than critical questions.26 Another problem is

whether other questions should be allowed under the TORC rule, as well as the critical

questions matching the argumentation scheme. For example, questions asking

for clarification of a term used in the argument would seem to be appropriate.

These problems suggest that what might be useful is some further consideration

of the purpose of the TORC rule, and how it achieves this purpose. Its purpose is to

be sure the respondent has paid some attention to an argument that was put forward,

and has considered it. It does this by making the respondent reply by indicating

he has made a critical examination of the argument just put forward, before the

dialogue moves on to consider other argument or issues. The rule is meant to

insure that the argument has had some impact in the dialogue. The way it does

this is to get some sense of the respondent’s reaction, by getting some assurance

that he has not just ignored it or thoughtlessly dismissed it. Thus the question is

raised: what is a critical examination of an argument? Some of the problems about

what sorts of replies ought to be included by the TORC rule turn on issues relating

to this question. For example, a question asking for clarification of a term used

in an argument should surely count as coming within a critical examination of

that argument. A counter-argument would not, if it is a new argument that is quite

different from the argument it counters. However it would, if it is part of the critical

examination of the original argument. The suggestion is that the way to move

towards a better TORC rule is to reach agreement on what a critical examination of

an argument should consist in. However, this topic is a serious problem in its own

right (Walton 2002, 174–180) and there is not enough space to seriously discuss it

here. A hypothesis is that critical examination can be defined as a type of dialogue

in relation to the kinds of speech acts that fall under the TORC rule. These include

IS THERE A BURDEN OF QUESTIONING? 39

speech acts like asking critical questions matching an argument, posing questions

that seek clarification of terms used in an argument, and posing relevant counterarguments

that probe into weak points in an argument or indicate it is wrong in

some way.

It is very important to think about these kind of problems, because presumptive

argumentation schemes need to be seen as rational arguments of a different sort

from the traditional models of deductive and inductive rational argument. To say

such an argument is a rational argument means that when brought forward by a

proponent, it is binding on the respondent. This means that it somehow affects the

commitments of the respondent by altering or restricting them in some predictable

way. A commitment rule needs to be formulated that specifies precisely how such

a commitment alteration should take place. The TORC rule is just such a rule. It is

a significant problem to try to formulate some rule of this kind, and to see how it

works in a QPD type of dialogue.

Addressing these problems is necessary in order work towards a general method

for the analysis and evaluation of defeasible arguments, an important goal for both

legal argumentation and AI. A defeasible argument is open-ended in the sense that

it is based on a rule (inference warrant) that is subject to default in the future as

new information comes in. But whether it will default in a given case cannot be

known in advance. For an arguer does not know in advance whether an exception

to the rule will arise in a given case. Thus the dialogue model of argumentation

fits defeasible arguments very well, because any argument in a dialogue can be

evaluated in relation to a future sequence of dialogue moves and counter-moves

that has not occurred yet. The dialogue model fits argumentation in cases of lack

of knowledge and uncertainty very well. And of course, one type of uncertainty

is brought in by the use of defeasible arguments, like those represented by presumptive

argumentation schemes. Such schemes are based on generalizations that

are subject to exceptions that have not arisen yet, but may arise in the future.

Whether the argument defaults or not depends on the evidence that will come in

as the dialogue proceeds. To represent rational argumentation in such a setting, the

respondent who replies to the argument must have the freedom to criticize such an

argument or argue against it. And yet it must have some effect on his commitments

so that it is rationally binding on him. The proponent must be able to use it, at the

end of dialogue, if not before, to build up the evidence she can use to realize her

goal.

Notes

1 I would like to thank the Social Sciences and Humanities Research Council of Canada for a

Research Grant that supported this work, and the two anonymous referees who helped to improve it.

2 The cases I have chosen are from Anglo-American law. I haven’t tried to investigate how the

problem they illustrate arises (or fails to arise) in other systems of law.

3 Alexy (1989, pp. 138–154) outlined these historical developments.

40 DOUGLAS WALTON

4 Of course, the trial ismuch more complex than the simplemodel of the critical discussion. The trial

is more like a play with a large cast of characters. There are the two central parties, their advocates,

the judge or other triers, and so forth. The claim is only that the critical discussion models the central

normative structure of the argumentation within a broader institutional setting that also has to be

taken into account.

5 Prakken (2002, p. 7) calls locution rules speech act rules, and he calls structural rules move

admissibility rules.

6 This distinction between the two basic types of persuasion dialogue is a familiar one in the literature,

although the terminology varies. For example, van Eemeren and Grootendorst (1984, p. 80)

distinguish between simple and complex disputes. A simple dispute is one where the respondent only

expresses doubt about the proponent’s thesis, while a complex dispute is one where the respondent’s

is the opposite thesis. For a more complete analysis, see Walton and Krabbe (1995, pp. 68–70).

7 As shown below, deductive validity can be one important standard of structural correctness of an

argument, but other standards can be considered as well. These other standards are weaker, because

they represent forms of argument that are not conclusive.

8 This problem arose from discussions at a workshop ‘Computational Models of Natural Argument’

at the ECAI Conference in Lyon, France, on July 22, 2002. For discussions that helped articulate

some concerns I would like to thank Subrata Das, Fiorella de Rosis, Floriana Grasso, Antonis Kakas,

David Moore, Pavlos Moraitis, Henry Prakken, Chris Reed, Francisca Snoek Henkemans, and Tangming

Yuan. I would especially like to thank Henry Prakken for making numerous critical comments

in November of 2002 on an earlier draft of the paper. The current version took the form of a dialogue

in which I tried to answer Henry’s critical questions by making a series of revisions to my earlier

arguments.

9 A comparable problem of argumentation was recognized by DeMorgan (1926, pp. 296–297) as

a “common occurrence” related to the fallacy of begging the question: “it is the habit of many to

treat an advanced proposition as a begging of the question the moment they see that, if established,

it would establish the question”.

10 Rule 11.1 says that for each elementary argument of the listener, if the premises and the argument

link from the premises to the conclusion are concessions of the speaker, the conclusion is also a

concession of the speaker.

11 There are some technical issues about negation and types of opposition that there is no space

to adequately discuss here. Notice that the respondent’s concession that not tipping is bad practice

does not necessarily imply that tipping is good practice. We simply assume that the two propositions

‘Tipping is a good practice’ and ‘Tipping is a bad practice’ are opposites (negations) of each other.

But whether this is technically correct as an assumption depends on how ‘good’ and ‘bad’ are defined.

12 The fallacy could possibly be diagnosed as one of amphiboly, based on a syntactical ambiguity.

13 This problem was pointed out by Henry Prakken, in some comments he made on this paper.

14 Prakken (2001, p. 191) takes defeasibility as the same property as nonmonotonicity, when applied

to arguments. A monotonic argument is one that stays correct, to the same standard, no matter how

many new premises are added. For example deductive arguments are monotonic. They stay valid if

new premises are added. A defeasible argument, in contrast, can default if a new premise is added.

15 The TOR rule was first proposed (although not by that name) by the author in a talk, ‘Argumentation,

Dialogue Types, Shifts and Embeddings’, at the Conference on Argumentation in Dialogic

Interaction, Lugano, Switzerland, July 3, 2002.

16 A way of expanding the TOR rule would be to allow for certain other kinds of questions as well.

For example, questions that ask for clarification of terms used in the argument could be allowed. The

rationale is to restrict questioning to the critical examination of the argument, to make sure the weak

points in the argument have been examined before turning to a consideration of other arguments.

IS THERE A BURDEN OF QUESTIONING? 41

17 PPD, or permissive persuasion dialogue, is contrasted with RPD, or rigorous persuasion dialogue.

In the latter moves are tightly constrained by strict rules. In the former, there is more freedom on

what kinds of moves can be made, and on what kinds of replies to a move are allowed.

18 The distinction between strong and weak refutation is useful here. According toWoods, Irvine and

Walton (2000, p. 56), a distinction between the two types of refutation in a dialogue can be drawn

as follows. To say that a proposition is refuted in the strong sense means that the respondent has

shown it to be false. To say it is refuted in the weak sense means that the respondent has shown

that the proponent has insufficient grounds for holding it. The issue can perhaps also be posed

using the distinction between two kinds of argument defeaters called undercutters and rebutters by

Pollock (1995, p. 40). An undercutter is an argument that raises doubts about the acceptability of a

previous argument. A rebutter is an argument that proves or asserts that the conclusion of the previous

argument is false. The issue can be posed in Pollock’s terms in the following question. Should the

respondent be restricted to replies that undercut the previous argument by raising doubts about its

acceptability, or should he be allowed to put forward a counter-argument that rebuts the previous

argument? Henry Prakken has raised some doubts about the equivalence of these two distinctions.

On his interpretation of Pollock’s definition, an undercutter is not just a question. It is an argument

of which the conclusion says that another argument is unacceptable. Since an undercutter is an

argument, on this interpretation, it also has premises. So an undercutter can be attacked in the same

way as other arguments. Regarding an undercutter as just a critical question would leave no room for

this kind of response to it.

19 These questions of how loose or tight a dialogue should be are related to the problem of whether

one can ever pin an arguer down to a formal fallacy (Krabbe 1996).

20 Another approach is that of Prakken and Sartor (1997). They modeled legal arguments about the

strength of other arguments by giving the judge the power to decide conflicts between arguments of

the adversaries.

21 Quoted from comments on an earlier draft of this paper made by Henry Prakken in November,

2002.

22 Actually even this formulation is too strict.When a proponent uses a defeasible argument, like one

that is an instance of the appeal to expert opinion, for example, premises do not give a conclusive

reason for accepting the conclusion. If the respondent is committed to the premises, and the argument

meets the structural requirements of the scheme, it does not follow that the respondent has to become

committed to the conclusion. All it means is that if the respondent is committed to the premises,

then he should become committed to the conclusion as a plausible statement. The strength of this

commitment should be determined by the strength of his commitment to the premises and by the

strength (plausibility) of the argument link between the premises and the conclusion. This subtlety

will be discussed and taken into account in section five below.

23 Indeed, as Henry Prakken pointed out, it would not be possible for the respondent to make a

counter-argument to a previous argument unless backtracking were to be allowed.

24 Calderon, 745 NYS 2d at 612.

25 This account is a paraphrase of the court’s words, quoted by Hoening from Calderon, 745 NYS

2d at 612.

26 It appears that Pollock intends undercutting arguments to be defeating counter-arguments in a way

that makes them stronger than critical questions.

References

Barth, Else M. and Krabbe, Erik C. W. (1982). From Axiom to Dialogue: A Philosophical Study of

Logics and Argumentation. Walter de Gruyter: Berlin.

Bench-Capon, T. (1997). Argument in Artificial Intelligence and Law. Artificial Intelligence and Law

5: 249–261.

42 DOUGLAS WALTON

DeMorgan, A. (1926). Formal Logic. The Open Court Company: London. Reprint of the original

Taylor and Walton edition of 1847.

Dube, F. (2002). Judge Criticizes ‘Hired Gun’ Experts. National Post (Canada), November 18, 2002,

A1 and A5.

Fox, J. and Das, S. (2000). Safe and Sound: Artificial Intelligence in Hazardous Applications. MIT

Press: Menlo Park, CA.

Feteris, E. (1999). Fundamentals of Legal Argumentation. Foris: Dordrecht.

Frank, J. (1963). Courts on Trial. Atheneum: New York.

Gordon, Thomas F. (1995). The Pleadings Game: An Artificial Intelligence Model of Procedural

Justice. Kluwer Academic Publishers: Dordrecht.

Hage, J. (2000). Dialectical Models in Artificial Intelligence and Law. Artificial Intelligence and Law

8: 137–172.

Hage, Jaap C., Leenes, Ronald and Lodder, Arno R. (1994). Hard Cases: A Procedural Approach.

Artificial Intelligence and Law 2: 113–167.

Hamblin, Charles L. (1970). Fallacies. Methuen: London.

Hamblin, Charles L. (1971). Mathematical Models of Dialogue. Theoria 37: 130–155.

Hastings, Arthur C. (1963). A Reformulation of the Modes of Reasoning in Argumentation. Ph.D.

diss., Evanston, IL

Hoenig, M. (2002). Products Liability: Jury Rejection of Uncontradicted Expert Testimony. New

York Law Journal 228: 3–6.

Kienpointner, M. (1992). Alltagslogik: Struktur und Funktion von Argumentationsmustern.

Fromman-Holzboog: Stuttgart.

Krabbe, Erik C.W. (1996). CanWe Ever Pin One Down to a Formal Fallacy? In Johan van Benthem,

Frans H. van Eemeren and Frank Veltman (eds.) Logic and Argumentation, 129–141. North-

Holland: Amsterdam.

Krabbe, Erik C. W. (2001). The Problem of Retraction in Critical Discussion. Synthese 127: 141–

159.

Lodder, Arno R. (1999). Dialaw: On Legal Justification and Dialogical Models of Argumentation.

Kluwer Academic Publishers: Dordrecht.

Loui, Ronald P. (1998). Process and Policy: Resource-Bounded Nondemonstrative Reasoning.

Computational Intelligence 14: 1–38.

Mackenzie, J. (1981). The Dialectics of Logic. Logique et Analyse 94: 159–177.

Mackenzie, J. (1990). Four Dialogue Systems. Studia Logica 49: 567–583.

Park, Roger C., Leonard, David P., and Goldberg, Steven H. (1998). Evidence Law. West Group: St.

Paul, MI.

Perelman, C. and Olbrechts-Tyteca, L. (1971). The New Rhetoric: A Treatise on Argumentation. In

J.Wilkinson and P.Weaver (trans.), 2nd edn, University of Notre Dame Press, Notre Dame, 1971

(First published, as La Nouvelle Rhetorique, in 1958).

Pollock, John L. (1995). Cognitive Carpentry. The MIT Press: Cambridge, MA.

Prakken, H. (1997). Logical Tools for Modelling Legal Argument. Kluwer Academic Publishers:

Dordrecht.

Prakken, H. (2001). Relating Protocols for Dynamic Dispute with Logics for Defeasible Argumentation.

Synthese 127: 187–219.

Prakken, H. (2002). Models of Dispute Resolution: A Formal Framework and an Application.

preprint at www.cs.uu.nl/staff/henry.html, 1–69.

Prakken, H. and Sartor, G. (1997). Argument-based Extended Logic Programming with Defeasible

Priorities. Journal of Applied Non-classical Logics 7: 25–75.

Reed, C. and Walton, D. (2001). Applications of Argumentation Schemes. In Proceedings of the

OSSA Meeting.

Reed, C. and Walton, D. (2002). Diagramming, Argumentation Schemes and Critical Questions. In

Proceedings of the ISSA Conference.

IS THERE A BURDEN OF QUESTIONING? 43

Safer, Jay G. (2002). Evolving Standards on Expert Witnesses. New York Law Journal 228: s4–s12.

Twining, W. (1985). Theories of Evidence: Bentham and Wigmore. Weidenfeld and Nicolson:

London.

van Eemeren, Frans H. and Grootendorst, R. (1987). Fallacies in Pragma-Dialectical Perspective.

Argumentation 1: 283–301.

van Eemeren, Frans H. and Grootendorst, R. (1992). Argumentation, Communication and Fallacies.

Erlbaum: Hillsdale, NJ.

Verheij, B. (1996). Rules, Reasons, Arguments: Formal Studies of Argumentation and Defeat.

Doctoral diss., University of Maastricht.

Walton, Douglas N. (1984). Logical Dialogue-Games and Fallacies. University Press of America:

Lanham, MD.

Walton, Douglas N. (1995). A Pragmatic Theory of Fallacy. University of Alabama Press: Tuscaloosa.

Walton, D. (1996). Argumentation Schemes for Presumptive Reasoning. Erlbaum: Mahwah, NJ.

Walton, D. (1998). The New Dialectic: Conversational Contexts of Argument. University of Toronto

Press: Toronto.

Walton, D. (2002). Legal Argumentation and Evidence. The Pennsylvania State University Press,

University Park.

Walton, Douglas N. and Krabbe, Erik C. W. (1995). Commitment in Dialogue: Basic Concepts of

Interpersonal Reasoning. State University of New York Press: Albany.

Wigmore, John H. (1931). The Principles of Judicial Proof. 2nd edn, Little, Brown and Company:

Boston.

Woods, J., Irvine, A., and Walton, D. (2000). Argument: Critical Thinking, Logic and the Fallacies.

Prentice Hall: Toronto.

Retirado de: http://www.kluweronline.com/issn/0924-8463/contents