The Structure of Argumentation

download The Structure of Argumentation

of 27

Transcript of The Structure of Argumentation

  • 8/12/2019 The Structure of Argumentation

    1/27

    State-of-the-Art: The Structure of Argumentation

    A. FRANCISCA SNOECK HENKEMANS

    Department of Speech Communication, Argumentation Theory and Rhetoric

    Spuistraat 134

    1012 VB Amsterdam

    The Netherlands

    E-mail: [email protected]

    ABSTRACT. In this paper, a survey is presented of the main approaches to the structure of

    argumentation. The paper starts with a historical overview of the distinctions between varioustypes of argument structure. Next, the main definitions given in the various approaches are

    discussed as well as the methods that are proposed to deal with doubtful cases.

    KEY WORDS: argument structure, complex argumentation, complementary argumentation,

    convergent reasoning, coordinatively compound argumentation, cumulative argumentation,

    linked reasoning, multiple argumentation, premise structure, serial reasoning

    1. INTRODUCTION

    The structure of argumentation is an important issue for current approaches

    to the theory and practice of argument analysis. Recurrent questions are:

    How can one lay bare the structure of a complex argument? Which types

    of structures should be distinguished and on what grounds?Laying out the structure of an argument is not only necessary to under-

    standing how arguers defend their positions, but is also indispensable for

    evaluating their argumentation. An overall judgment of the quality of a

    complex argument requires not just a clear picture of individual arguments,

    but also insight into the relations among these arguments.1 If it is not clear

    whether or not an individual argument (or reason) is independent of other

    arguments that make up the complex argument, it cannot be clear just how

    damaging are the consequences of a negative evaluation of that argument.

    In most approaches, at least three types of argument structure are dis-

    tinguished: (1) serial reasoning (or subordinate argumentation), (2) linked

    reasoning (or coordinate argumentation), and (3) convergent reasoning

    (or multiple argumentation). Reasoning is serial if one of the reasonssupports the other. If reasoning is linked, each of the reasons given are

    directly related to the standpoint, and the reasons work together as a unit.

    When each reason separately supports the standpoint (to some degree),

    the reasoning is convergent. A complex argument can combine all of these

    types of argument structure.

    Although at first sight there seems to be agreement on the relevant

    types of distinctions between argument structures, on closer inspection,

    Argumentation 14: 447473, 2000.

    2000 Kluwer Academic Publishers. Printed in the Netherlands.

  • 8/12/2019 The Structure of Argumentation

    2/27

    authors disagree with each other on a number of points. Often the sameterminology is used for distinctions that differ in crucial respects. Moreover,

    the various characterizations of argument structures reflect fundamental

    differences of approach to argumentation. The most important difference

    is that between structural (or logical) approaches, in which attention is

    paid only to the structural aspects of argument structure as they manifest

    themselves in the product of the reasoning process, and functional (or dialec-

    tical) approaches, in which the emphasis is on the process within which

    these structures arise, and the functions the various argument structures fulfil

    in this process.

    Apart from differences in definitions of argumentation structures and

    differences of approach, there is much discussion about the best method

    of analysis, especially in doubtful cases where there is room for more thanone interpretation. For some authors, the ambiguities in the definitions

    and the problems of analysis have been a reason to question the impor-

    tance of the distinctions. They argue that distinguishing between the various

    argument structures is often so difficult in practice that it is better not to

    make these distinctions at all. This is especially true for the notoriously

    problematic distinction between linked and convergent arguments.

    Finally, there is also disagreement about the history of the distinctions,

    in particular about the origins of the linked-convergent distinction. Since

    more insight into the historical backgrounds of the distinctions between

    various types of argument structure might enhance our understanding of

    the current controversies, this survey begins with a historical overview.

    Section 3 discusses the main definitions of linked and convergent argument

    structures (or their equivalents) that are given in current approaches. Section

    4 is devoted to the various methods proposed in the literature to deal with

    doubtful cases.

    2. HISTORICAL OVERVIEW

    Attention to the analysis of argument structure first appeared in American

    debate and logic textbooks in the second half of the previous century. Two

    major influences are identifiable in these textbooks: classical rhetoric and

    Enlightenment rhetorical theorists such as Campbell and Whately.

    Although the concepts are not fully developed, and the contemporary

    terminology is absent, argument structures similar to our present dayconcepts were present in nascent form in both the classical rhetorical

    tradition and the 18th century Enlightenment rhetorics. The functional

    approach, in which the independence or interdependence of arguments is

    determined by the requirements for the burden of proof to be met by the

    arguer, can be found both in the classical and the Enlightenment rhetoric.

    The logical approach, in which the term argument structure refers to

    relationships among premises within different inference types, is promi-

    448 A. FRANCISCA SNOECK HENKEMANS

  • 8/12/2019 The Structure of Argumentation

    3/27

    nent in the work of Enlightenment rhetoricians. They make a distinctionbetween the type of linkage between premises in a deductive argument or

    proof and the way in which premises are combined in an inductive (or

    moral) argument: in the latter argument the premises each separately

    lend some degree of support to the conclusion (and are in that sense

    independent), but they need to be combined to make the conclusion (more)

    probable.

    Classical rhetoric

    In early Roman rhetoric, both subordinate arguments and arguments con-

    sisting of a number of reasons in direct support of a (sub)standpoint were

    distinguished.2 In Ciceros De Inventione, a specific type of complexargument is discussed, the epicheirema, also called thefive-part argument.

    Kennedy (1994) gives the following description of the five-part argument:

    it is in fact a kind of amplification of the Aristotelian syllogism and enthymeme in

    which a proposition (part 1) is supported by a variety of reasons (part 2), then a second

    proposition (what would be the minor premise in a syllogism) is stated (part 3), and

    that is followed by a variety of reasons for believing it (part 4). The fifth part then states

    the conclusion. Such an argument in Greek is sometimes called an epicheirema, liter-

    ally a handful. (120)

    Since part 2 and part 4 consist of reasons in support of another reason,

    they can be seen as serial reasoning or subordinate argumentation. In both

    cases, the support provided by the subordinate argumentation should consist

    of a variety of reasons. Cicero gives the following example of support forthe minor premise Of all things nothing is better governed than the

    universe:

    And then [. . .] they introduce another proof, that is of the minor premise, in this way:

    For the risings and the settings of the constellations keep a fixed order, and the changes

    of the seasons not only proceed in the same way by a fixed law but are also adapted to

    the advantage of all nature, and the alternation of night and day has never through any

    variations done any harm. All these points are proof that the nature of the world is

    governed by no ordinary intelligence. (De Inventione , 1. 59)

    It is not clear whether these reasons are supposed to form one combined

    defence (and are thus a case of linked reasoning) or whether they should

    lend independent support (and are a case of convergent reasoning).

    Also in the classical theory of stasis, the concepts of dependent and inde-pendent reasons seem to have been recognized, albeit it not explicitly. The

    issues that had to be proven in a criminal court case differed depending

    on whether the arguers position was that of the defendant or that of the

    prosecutor.3 In choosing the status for his defence, according to Quintilian,

    the defendant has four options:

    There are four different methods which may be employed in every case, and he who is

    going to plead should study them as first essentials. For, to begin with the defendant,

    THE STRUCTURE OF ARGUMENTATION 449

  • 8/12/2019 The Structure of Argumentation

    4/27

    far the strongest method of self-defence is, if possible, to deny the charge. The secondbest is when it is possible to reply that the particular act with which you are charged

    was never committed. The third and most honourable is to maintain that the act was

    justifiable. If none of these lines of defence are feasible, there remains the last and only

    hope of safety: [. . .] we must evade the charge with the aid of some point of law,

    making it appear that the action has been brought against us illegally. (Quintilian 3, 6,

    8384)

    Each of these ways to defend the accused is, in principle, a sufficient

    defense. It is, however, also possible to combine different types of defenses.

    Quintilian gives the following explanation of the reasons for using such a

    multiple defense:

    There are also a number who are in doubt as to a form of defence which I may

    exemplify as follows: If I murdered him, I did right; but I did not murder him. What,

    they ask, is the value of the first part, if the second can be proved, since they are mutuallyinconsistent, and if anyone employs both arguments, we should believe neither? This

    contention is partly justified; we should employ the second alone if the fact can be

    proved without a doubt. But if we have any doubts as to being able to prove the stronger

    argument, we shall do well to rely on both. Different arguments move different people.

    [. . .] one who is confident of his powers as a marksman may be content with one shaft,

    whereas he who has no such confidence will do well to launch several and give fortune

    a chance to come to his assistence. (Quintilian 4, 5, 1314)

    The prosecutor makes use of the same issues in choosing his main argu-

    ments, but he has to prove all four points in order to make his case:

    The accuser [. . .] must prove that something was done, that a particular act was done,

    that it was wrongly done, and that he brings his charge according to law. (Quintilian 3,

    6, 85)

    In view of the requirements of the burden of proof, the prosecutors main

    reasons for finding the accused guilty must be regarded as interdependent,

    since the success of the argument as a whole depends on the success of each

    of its component parts.4

    A different type of interdependency is at stake in the following fragment

    from the Rhetorica ad Herennium, where the ideal order of the arguments

    is discussed:

    In the Proof and Refutation of arguments it is appropriate to adopt an Arrangement of

    the following sort: (1) the strongest arguments should be placed at the beginning and at

    the end of the pleading; (2) those of medium force, and also those that are neither useless

    to the discourse nor essential to the proof, which are weak if presented separately and

    individually, but become strong and plausible when conjoined with the others, should

    be placed in the middle. (Rhetorica ad Herennium 3, 10, 18)

    Unlike the prosecutors main arguments, some of the arguments that are

    supposed to be placed in the middle are not essential to the proof according

    to the author of the Rhetorica ad Herennium. The reason for their depen-

    dency on other arguments is that they are too weak to stand on their own,

    and they become stronger when combined with the other arguments.

    450 A. FRANCISCA SNOECK HENKEMANS

  • 8/12/2019 The Structure of Argumentation

    5/27

    Enlightenment Rhetoric

    In The Philosophy of Rhetoric (1991/1776), Campbell makes a distinction

    between scientific evidence and moral evidence. One of the main dif-

    ferences between these two types of evidence is a difference in structure:

    whereas scientific evidence consists of a set of interdependent premises

    which together form one argument, moral evidence consists of a number

    of independent arguments:

    scientific evidence is simple, consisting of only one coherent series, every part of which

    depends on the preceding, and, as it were, suspends the following: moral evidence is

    generally complicated, being in reality a bundle of independent proofs. The longest

    demonstration is but one uniform chain, the links whereof, taken severally, are not to

    be regarded as so many arguments, and consequently when thus taken, they conclude

    nothing; but taken together, and in their proper order, they form one argument, which isperfectly conclusive. (45)

    Although each of the reasons given in moral reasoning is independent of

    the others, with each new reason the conclusion may become more likely:

    In moral reasoning, [. . .] there is often a combination of many distinct topics of argument,

    no way dependent on one another. Each hath a certain portion of evidence belonging to

    itself, each bestows on the conclusion a particular degree of likelihood, of all which

    accumulated the credibility of the fact is compounded. (1991/1776: 45)

    Campbell acknowledges that it is possible to give more than one indepen-

    dent proof of the same proposition, but he believes that there is not much

    point in doing this:

    It is true, the same theorem may be demonstrable in different ways, and by differentmediums; but as a single demonstration, clearly understood, commands the fullest con-

    viction, every other is superfluous. (1991/1776: 45)

    Campbell also gives a description of the consequences that will result

    from taking away one or more parts of a demonstration or of a moral

    argument:

    [A demonstration] may be compared to an arch, no part of which can subsist indepen-

    dently of the rest. If you make any breach in it, you destroy the whole. [. . .] [Moral

    reasoning] may be compared to a tower, the height whereof is but the aggregate of the

    heights of the several parts reared above one another, and so may be gradually diminished,

    as it was gradually raised. (1991/1776: 4546)

    Whately discusses two different ways in which reasons may be combined

    in order to produce a stronger argument (and thus may be considered aslinked). A first type of case consists of reasons which separately have

    little or no weight, but which can lend sufficient support to a conclusion

    when taken together:

    The remark above made, as to the force of concurrent testimonies, even though each,

    separately, might have little or none, but whose accidental agreement in a falsehood would

    be extremely improbable, is not solely applicable to the Argument from Testimony, but

    may be extended to many arguments of other kinds also; in which a similar calculation

    THE STRUCTURE OF ARGUMENTATION 451

  • 8/12/2019 The Structure of Argumentation

    6/27

    of chances will enable us to draw a conclusion, sometimes even amounting to moralcertainty, from a combination of data which singly would have had little or no weight.

    (1846: 74)

    In inductions, according to Whately, mentioning just one instance can in

    some cases suffice to justify a general conclusion. But if not, then a number

    of instances are needed to support the conclusion:

    In human affairs [. . .] our uncertainty respecting many of the circumstances that may

    affect the result, obliges us to collect many coinciding instances to warrant even a probable

    conclusion. (1846: 88)

    Instead of combining a number of reasons of the same type (i.e., a number

    of testimonies, signs or instances), it is also possible to produce a stronger

    argument by combining different types of reasons. A second type of linkage

    (implicitly) distinguished by Whately consists of a combination of dif-ferent types of reason, where the one reason is a means of showing that

    specific objections against the other reason (that there are certain

    exception-making circumstances to the general rule or warrant under-

    lying that reason) are not valid:

    the a priori Argument and Example support each other, when used in conjunction

    [. . .]. A sufficient Cause being established, leaves us still at liberty to suppose that

    there may have been circumstances which will prevent the effect from taking place; but

    Examples subjoined show that these circumstances do not, at least always, prevent that

    effect. (1846: 138)

    From his discussion of issues related to the burden of proof (112), it becomes

    clear that Whately also allows for the possibility of putting forward more

    reasons than are required, so as a result the defense as a whole consistsof a number of independent reasons. If the arguer has the presumption,

    then the burden of proof belongs with the other side and putting forward

    further evidence in support of the presumption amounts to giving an

    argument ex abundanti (a superfluous argument).

    It is a point of great importance to decide in each case [. . .] on which side the Presumption

    lies, and to which belongs the [onus probandi] Burden of Proof. For though it may often

    be expedient to bring forward more proofs than can be fairly demandedof you, it is always

    desirable, when this is the case, that it should be known, and that the strength of the

    cause should be estimated accordingly. (1846: 112) 5

    So, there is nothing against doing this, provided it is made clear to the

    opponent that one is doing more than required.

    The early textbooks

    American textbooks on argumentation and debate that started appearing

    in the 50s to a large extent reflected the traditional views on argument

    structure. But there were also a number of new developments. Modern

    tradition has it that Monroe Beardsley (1950) was the first to represent

    the structure of argumentation by using diagrams (consisting of numbered

    452 A. FRANCISCA SNOECK HENKEMANS

  • 8/12/2019 The Structure of Argumentation

    7/27

    statements and arrows indicating support relationships). Beardsley also intro-duced part of the terminology that is still used by informal logicians to refer

    to the different argument structures. He gives the following definitions:

    In a convergentargument, several independent reasons support the same conclusion

    [. . .] In a divergent argument, the same reason supports several conclusions [. . .] A

    serial argument contains a statement that is both a conclusion anda reason for a further

    conclusion (1950: 19).

    Thomas claims credit as the first to distinguish between linked and con-

    vergent arguments rather than as others have thought, Beardsley:

    The distinction between linked and convergent inferences was not drawn by Beardsley,

    not even in his fourth edition of Thinking Straight. It was introduced in the 1973 edition

    of PRNL [Practical Reasoning in Natural Language]. Beardsley represented linked and

    convergent inferential relationships alike, using multiple arrows for (what I call) linkedrelationships, as if all reasoning with multiple reasons were convergent (except that this

    distinction did not exist in his system). The concept of linked reasoning, and the distinction

    between it and convergent reasoning, needed to be added to Beardsleys system of analysis

    before it could be generally applied at all (1986: 457)

    As far as the terminology is concerned, this seems to be correct, but similar

    distinctions between interdependent and independent reasons were already

    made before 1973. One example is Mills (1968), who makes a distinction

    between cases where there is one conclusion with several coordinate

    reasons and cases where one has one conclusion supported by two or more

    pairs of coordinate reasons. The latter type of structure is more likely to

    be a case than an argument. An example of the first type of structure is:

    Conclusion: Our top debaters are top scholarsReasons:

    1. They are John, Mary, and Jim.

    2. John has a 3.9

    3. Mary has a 3.8

    4. Jim has a 3.7

    Mills (1968: 182183)

    An example of the second type of structure, according to Mills, would be

    a case on a value proposition where each pair of reasons consists of a

    criterion and its application.6 Mills also introduces a diagram convention

    for making the distinction between these two types of structures (1968: 183).

    From his examples, it becomes clear that Mills does make a distinction

    between arguments with a structure comparable to a linked argument (the

    first type) and a structure which is comparable to a convergent argument(the second type, where each pair of reasons taken together constitutes an

    independent reason).

    Just as in the classical and enlightenment tradition, for most of the earlier

    textbook authors, the distinction between interdependent and independent

    reasons coincides with the distinction between the structure of premises

    in a deductive argument and the structure in an inductive (or non-deduc-

    tive) argument.

    THE STRUCTURE OF ARGUMENTATION 453

  • 8/12/2019 The Structure of Argumentation

    8/27

    The distinction then is that in a deductive argument the premises needto be taken together to constitute a reason, whereas in an inductive argument

    a combination of reasons is needed to make the conclusion (more or less)

    probable. McBurney, ONeill and Mills (1951) explain that a combination

    of (inductive) analogies may strengthen the support for a conclusion:

    From a series of analogies showing the success of a certain project, we argue that it

    will be successful in a particular case. [. . .] Reasoning from analogy becomes increas-

    ingly strong as the number of comparisons is increased, or in other words, as the analogies

    are cumulated (1951: 108)

    And Ray and Zavos (1966) explicitly compare deductive arguments to a

    chain of links, and contrast this with the more gradual type of support

    supplied by induction and arguments from circumstance:7

    Deductive reasoning is often compared to a chain. The last link in the chain is depen-

    dent not to some degree but totally on the connection of every other link. [. . .] Induction

    and argument from circumstance can be better compared to the relation of the legs of a

    table to the tabletop. Each leg, though cooperating with the others, supports the top

    independent of the other legs (at least to a point). (1966: 101)

    In the early textbooks, the tradition of the classical theory of stasis, in which

    it depends on the requirements of proof that the arguer must meet what

    the resulting structure of his argument is, is also preserved. A good example

    of a textbook in which the latter approach is taken is Windes and Hastings

    Argumentation and Advocacy (1965).

    According to Windes and Hastings, the issues an advocate needs to prove

    in order to establish the guilt of X in an embezzlement case are:

    1. Certain property was misappropriated.

    2. X misappropriated the property.

    3. The misappropriation was fraudulent, i.e., without the consent of the owner.

    4. The property was for the personal use of X (1965: 75).

    In such cases, the issues taken together form one linked argument and the

    advocate has to provide sufficient evidence for each of the issues:

    Usually there is one over-all argument leading directly to the proposition, and the elements

    of this argument are further supported by sub-arguments. The over-all argument states

    the issues and the sub-arguments are contentions in support of the issues [. . .] Each

    one of the issues must be established for the proposition [X is guilty of embezzlement]

    to be established. If any issue is not proved, then the proposition is not proved (1965:

    215216).

    The task of the opponent (the negative advocate) is easier: The other party

    can logically defeat the proposition by defeating one of these issues (1965:

    232).

    Apart from cases where the advocate needs to put forward a number

    of interdependent reasons (corresponding to the issues) in order to defend

    the proposition, Windes and Hastings also discuss the possibility of con-

    structing an argument with independent lines of reasoning:

    454 A. FRANCISCA SNOECK HENKEMANS

  • 8/12/2019 The Structure of Argumentation

    9/27

    Another aspect of case building to note is that independent lines of reasoning may leadto the same conclusion. We may, for example, use three reasoning processes to support

    the conclusion the corn crop of Dullnia is failing.

    1. Dulnia is buying corn on the world market. (Reasoning from effect to cause.)

    2. The testimony of an agricultural expert who visited Dullnia. (Testimonial evidence.)

    3. The presence of drought and poor growing conditions this year. (Cause to effect).

    In this series of arguments, rather than having a clear set of issues to be established, the

    proof of the proposition depends on the number and plausibility of the component argu-

    ments. In such cases, one strong one may be sufficient, but the more independent argu-

    ments which lead to the same conclusion, the more probable is that conclusion. (1965:

    216218)

    For Windes and Hastings, two criteria seem to be relevant in determining

    whether arguments are interdependent or not: (1) whether the arguments

    given by the arguer form part of a clear set of issues that should be estab-lished (the burden of proof-criterion), and (2) whether the argumentation

    as a whole is based on one reasoning process or whether the arguer uses

    different reasoning processes (the different types of argument-criterion).

    3. CURRENT APPROACHES

    Since the 70s, a large number of textbooks have appeared in which it has

    become customary to devote attention to analyzing the structure of argu-

    ments and to ways of portraying this structure in a diagram. With the

    exception of van Eemeren and Grootendorsts (1992) pragma-dialectical

    textbook, all these books are written by informal logicians. It is not until

    the 90s that the subject of argument structure starts receiving a more

    theoretical treatment. In the journalsInformal Logic andArgumentation, the

    distinctions made by informal logicians between different types of struc-

    ture are discussed. Also, three monographs on argument structure appear,

    each of which proposes a dialectical approach to argument structure.

    Since most of the discussion has centered on problems involved in the

    distinction between linked and convergent arguments, I shall restrict my

    further discussion of the literature to these two types of structure.

    The textbook distinctions

    Thomas (1986, 1st edition 1973) was the first to use the terms linked

    and convergent in describing interdependent and independent premise

    support, and his way of portraying these two structures has become standard.

    Although Thomas has had a large influence on all later informal logic

    accounts of argument structure, his own definitions of these two types of

    structures are different from those of many of the other authors. Whereas

    most informal logicians continue the tradition of associating linked argu-

    ments with deductive reasoning and convergent arguments with inductive

    THE STRUCTURE OF ARGUMENTATION 455

  • 8/12/2019 The Structure of Argumentation

    10/27

    reasoning (or other types of non-deductive reasoning), Thomas deviates fromthis tradition:

    The natural-logic concept of linked inference [. . .] bridges the supposed sharp line of

    the traditional induction-deduction dichotomy, severely straining [. . .] the traditional theory

    that induction and deduction comprise two completely different kinds of inference

    [. . .] Both of these supposedly different types of reasoning receive identical diagram-

    matic representations. (1986: 461)

    Thomas gives the following definitions of linked and convergent rea-

    soning:

    When a step of reasoning involves the logical combination of two or more reasons, they

    are diagrammed as linked. (1986: 58)

    When two or more reasons do not support a conclusion in a united or combined way,

    but rather each reason supports the conclusion completely separately and independently

    of the other, the reasoning is convergent. (1986: 60) 8

    Both deductively valid reasoning and inductive reasoning from several

    similar items of evidence are analysed as linked by Thomas. In the case

    of inductive reasoning, the reason for linking the premises is not that each

    reason separately does not give any support to the conclusion, but that

    the support is stronger when the reasons are combined (1986: 59).

    Linked arguments form one single argument, whereas convergent argu-

    ments consist of separate arguments:9

    A convergent argument is equivalent to separate arguments (or evidence coming from

    separate areas) for the same conclusion. (1986: 61)

    The closest formal approximation to the natural-logic portrayal of convergent inferencewould perhaps be two separate deductions leading to two different tokens of the same con-

    clusion. (1986: 459)

    For this reason, convergent arguments are less vulnerable than linked

    arguments to criticism of the acceptability of their premises:

    An important feature of convergent reasoning is that the support given to the conclu-

    sion by each separate reason, or line of reasoning, would remain unaffected even if the

    other [. . .] reason(s) were false. (1986: 61)10

    Pinto and Blairs (1993) definitions closely resemble Thomas distinction

    between linked and convergent arguments. They make a distinction between

    a group of premises that together form one inference and independent

    groups of premises which can be seen as parallel inferences to arrive at

    the same conclusion:

    What makes for groups of premisses which are independentof each other? The fact that

    the premisses work in combination to support the conclusion constitutes them as a set

    or group, and the fact that the premisses of each group are able to provide their support

    without any help from premisses in any other group make them independent of each

    other (1993: 77).

    From their examples it becomes clear that just like in Thomas approach,

    456 A. FRANCISCA SNOECK HENKEMANS

  • 8/12/2019 The Structure of Argumentation

    11/27

    a premise group may both consist of premises that deductively entail theconclusion, and of premises that taken together make a conclusion likely

    or plausible (1993: 111).

    Independent groups of premises should be evaluated separately for

    inference strength:

    you should consider each independent group separately and try to judge how strongly

    its premisses taken together are linked to the conclusion (109).

    Even though independent groups of inferences (or parallel inferences)

    are to be evaluated separately, they are linked up by Pinto and Blair when

    diagramming their structure. This becomes clear from step three in their

    diagramming method:

    Step 3: Make a full diagram showing all the inferences in the reasoning. In doing so,link up any [. . .] parallel inferences (1993: 260)

    Pinto and Blair do not specify what the consequences are for the overall

    evaluation if one of the groups of premises provides strong support, but

    the other weak. The fact that they link parallel inferences in the diagram-

    matic representation, however, suggests that they think that the groups of

    premises should be assessed in combination in evaluating the overall

    strength of the argument.

    Whereas Thomas and Pinto and Blair analyze both deductive and non-

    deductive arguments as linked, Goviers (1992) distinction between linked

    and convergent support more or less coincides with the distinction between

    deductive and non-deductive arguments, albeit that she also considersanalogies as linked:

    In most deductively valid arguments, and in analogies, the support provided by the

    premises is linked, not convergent. There are exceptions, as when a person offers two

    separate premises, both of which deductively entail the conclusion, but this is quite rare.

    There are arguments in which support is convergent and in which the premises do not

    entail the conclusion nor support it by analogy [. . .] These arguments we call conduc-

    tive arguments. [. . .] Some philosophers have referred to these arguments based on

    separately relevant factors as good reasons arguments. The relevant factors provide reasons

    for the conclusion, though they do not deductively entail it. (1992: 308309)

    Linked arguments are thus deductions or analogies, and convergent argu-

    ments are non-deductive (i.e. conductive) unless they consist of two

    premises that each deductively entail the conclusion, but, according to

    Govier, One might also insist that such a case represents several argu-

    ments all with the same conclusion (1992: 356).11

    As soon as a premise can be seen as separately relevant, the argument

    is analyzed as convergent by Govier, even if the premises need to be taken

    together to provide sufficient support to the conclusion. But analyzing an

    argument as convergent has no further consequences for the evaluation: It

    is not the case that each of the premises should then be separately assessed.

    THE STRUCTURE OF ARGUMENTATION 457

  • 8/12/2019 The Structure of Argumentation

    12/27

    Govier claims that the separate premises in a convergent argument shouldalways be treated as a unit:12

    There are [. . .] two basic facts to understand right now about the way in which premises

    support conclusions. First, premises should be considered together in their support. Second,

    they may be linked or convergent. In the latter case, it would be possible to see each

    premise as providing a separate reason or separate bit of evidence, in support of the

    conclusion, but the whole argument requires us to think of these separate strands together,

    as the weight of support accumulates. (1992: 48)

    Groarke, Tindale and Fisher (1997) and Copi and Cohen (1990) give similar

    definitions as Govier of linked and convergent premises.13 Groarke, Tindale

    and Fisher make the distinction in the following way:

    Linked premises work together. Taken independently, they do not support the arguments

    conclusion. Convergent premises do not require each other, for they support the conclu-sion independently of the arguments other premises. (1997: 35)

    If all of the premises but one of a convergent argument were to be taken

    away, we would have a weaker argument, but the premise would still

    provide some evidence for the conclusion. None of the premises requires

    one of the other premises for this to be the case (1997: 36). In other

    words, the premises are independent in the sense that they are separately

    relevant, that is, lend some support to a conclusion by themselves.

    According to Copi and Cohen, premises depend on each other if each

    premise supports the conclusion through the mediation of the other premiss

    (1990: 20). If one of two interdependent premises were true, but the other

    not, the conclusion would have been given no support at all (20). Two

    premises support a conclusion independently if each supplies some warrantfor accepting the conclusion and would do so even in the absence of the

    other premiss. (1990: 19). How independent are the arguments then in a

    convergent argument, according to Copi and Cohen? Do they consider them

    as separate arguments for the same conclusion, as Thomas does? This is

    not the case:

    Emerging practice is to say that it is one argument with two independent premisses. The

    principle seems to be that the number of conclusions determines the number of arguments.

    So by a single argument is meant an argument to a single conclusion, regardless of

    how many premisses are adduced in its support. (1990: 1920)

    However, this remark does not really clarify what exactly Copi and Cohen

    mean by independent premises, since they now introduce a new crite-

    rion for the singleness of an argument: the number of conclusions insteadof the probative force of the individual reasons.

    Johnson and Blair (1994) use a different definition of single argument

    than Copi and Cohen: they speak of a single argument if it contains one

    premise set and one conclusion (1994: 37). Premises belong to one set,

    if they are relevant in combination (p. 36). Convergent arguments consist

    of a combination of two or more premise sets, and therefore of a combi-

    nation of arguments:

    458 A. FRANCISCA SNOECK HENKEMANS

  • 8/12/2019 The Structure of Argumentation

    13/27

    When there are, in effect, two or more distinct independent grounds for a conclusion, thinkof each ground as a separate argument. (1994: 38)

    Convergent arguments can typically be found in a case, which consists

    (minimally) of arguments for the claim and arguments against the claim

    (1994: 247). Each distinct argument for a claim is called a line of

    argument. In cases where the premises constitute several independent

    lines of arguments, the relevance is checked for each premise individu-

    ally, but the sufficiency of these premises should be established by

    considering them as a whole:

    Check the premises individually for relevance. [. . .] Finally, check for sufficiency: Do

    the premises, taken together, satisfy the sufficiency requirement? Do they provide enough

    support for the conclusion? (1994: 269)

    From this comparison of informal logical approaches to argument struc-

    ture, a number of differences emerge. For Thomas, premises are

    interdependent both when no premise separately lends any support to the

    conclusion, whereas the combination of premises does and when each

    premise separately provides some support to the conclusion, but the com-

    bination of premises forms a stronger argument than each premise by

    itself. Both deductive and non-deductive arguments consisting of premises

    whose separate support is too weak for them to stand on their own are

    analyzed as linked by Thomas. Only if each reason alone would be enough,

    if true, to support the conclusion, and if the falseness of one reason would

    not weaken a step of reasoning from the other to the conclusion, the rea-

    soning may be regarded as convergent. Linked arguments are seen as

    equivalent to one single argument, whereas convergent arguments areregarded as a combination of single arguments, which should be sepa-

    rately evaluated for their strength of support. Pinto and Blair seem to employ

    similar definitions of linked and convergent arguments as Thomas. They

    also think independent groups of premises should be assessed separately for

    their strength of support, but they are less clear as to how these separate

    assessments should be used in the overall evaluation of an argument.

    Other authors (such as Govier, Copi and Cohen, and Groarke, Tindale

    and Fisher) employ a more restricted definition of interdependency:

    Arguments are interdependent only when they are separately irrelevant to

    the conclusion and relevant in combination. As soon as premises can be

    seen as separately relevant (as providing some support to the conclusion),

    they are considered to be independent, and thus convergent. This is alsothe case if they can only in combination lend sufficient support to the

    conclusion. To the evaluation of whether the amount of support the premises

    lend to the conclusion is sufficient, it makes no difference whether the

    premises are independent or interdependent in these approaches: they are

    always considered together in their support during the evaluation.

    Johnson and Blairs position seems to be intermediate: on the one hand,

    just like Thomas and Pinto and Blair, they consider convergent arguments

    THE STRUCTURE OF ARGUMENTATION 459

  • 8/12/2019 The Structure of Argumentation

    14/27

    as separate arguments. On the other hand, they claim that the sufficiencyof support should always be assessed by checking whether the different

    premises (or reasons) taken together provide sufficient support for the

    conclusion. It is unclear in their approach what the consequences would

    be for the evaluation of the argument as a whole if one of the lines of

    argument would be flawed but one or more of the other arguments would

    prove to be acceptable.15 In Johnson and Blairs view, the question of

    whether an argument is single or convergent is only important for deter-

    mining whether one should take the premises together in determining their

    relevance, or whether each premise should in principle be relevant by

    itself. The singleness or complexity of an argument does not make any

    difference to the evaluation of the strength of the argument: in either case,

    the premises are considered as a whole. Although Johnson and Blair observethat convergent arguments are typical of a case, in which one should

    respond adequately to relevant objections, they do not devote any atten-

    tion to the connection between the burden of proof and the argument

    structures that result from arguers attempts to fulfil their obligations.

    Although at first sight the definitions given by van Eemeren and

    Grootendorst (1992) seem to resemble the definitions provided by Thomas,

    Johnson and Blair and Pinto and Blair, there is an important difference: both

    the equivalent of a linked argument (coordinatively compoundargumenta-

    tion) and the equivalent of a convergent argument (multiple argumentation)

    are seen as forms of complex argumentation consisting of a number of single

    argumentations:

    Analytically, complex argumentation can always be broken down into single argumen-

    tations. (1992: 73)

    The distinction between coordinative and multiple argumentation is there-

    fore not that coordinative argumentation describes the relations between

    premises within one single argument and that multiple argumentation

    consists of a combination of single arguments, but that the relations between

    the single arguments that constitute these two types of complex argument

    are different. In multiple argumentation, the constituent single argumen-

    tations are, in principle, alternative defenses of the same standpoint (1992:

    73). In coordinatively compound argumentation, unlike in multiple argu-

    mentation, all the component single argumentations are, in principle,

    necessary for a conclusive defense of the standpoint (1992: 77).In pragma-dialectics, unlike in most informal logical approaches, coor-

    dinatively compound argumentation always consists of a combination of

    (independently relevant) arguments, that nonetheless need to be taken

    together for reasons of sufficiency. Coordinatively compound argumenta-

    tion is therefore comparable to the concept of a convergent argument in

    the definitions given by Govier, Groarke, Tindale and Fisher and Copi

    and Cohen. The premise structure of a single argument is not dealt with

    460 A. FRANCISCA SNOECK HENKEMANS

  • 8/12/2019 The Structure of Argumentation

    15/27

    by van Eemeren and Grootendorst in the context of argument structure.Premise structure is seen as an issue of a different hierarchical nature than

    argument structure. Each separate argument in a coordinatively compound

    argumentation can be analyzed as a single argument, consisting of two

    premises, one of which is usually left unexpressed.16

    Van Eemeren and Grootendorst give the following two examples to

    explain the difference between the two structures. Example (1) is an example

    of a multiple argument; example (2) of a coordinative argument:

    (1) Postal deliveries in Holland are not perfect. You cannot be sure that a letter will be

    delivered the next day, that it will be delivered to the right address and that it will

    be delivered early in the morning. (1992: 73)

    (2) Postal deliveries in Holland are perfect. You can be sure that a letter will be deliv-

    ered the next day, that it will be delivered to the right address, and that it will be

    delivered early in the morning. (1992: 77)

    In example (1), each of the single argumentations is in principle sufficient

    to defend the standpoint. What would be the reason for an arguer to produce

    multiple argumentation if one single argument might have sufficed? Van

    Eemeren and Grootendorst mention the following possible reasons:

    It is possible that the speaker tries to cater for various kinds of doubt about his stand-

    point, pertaining to different aspects. [. . .] Multiple argumentation can also be used for

    rhetorical reasons: the profusion of arguments makes the defense appear stronger. The

    speaker gives the impression of having taken account of every possible objection to his

    standpoint. [. . .] Because acceptance is liable to gradations, in producing a multiplicity

    of arguments the speaker may gradually overcome the last remaining morsel of doubt

    in the listener. It may also simply be that after one argument the speaker advances

    another quite different sort of argument in the hope that it will be more successful.(1992: 74)

    In example (2), the three argumentations that support the standpoint that

    postal deliveries in Holland are perfect are interdependent: Only if they are

    taken together, they are deemed to provide sufficient grounds for accepting

    the standpoint. If [any] of the single argumentations proves to be unac-

    ceptable, the entire coordinatively compound argumentation falls apart

    (1992: 77).

    The main reason given by van Eemeren and Grootendorst for putting

    forward a coordinative argument is :

    in many cases, it is not possible to remove all the antagonists doubts by means of one

    argument. When defending his standpoint, the protagonist then has to advance two or more

    connected arguments that, only when taken together, are supposed to constitute a well-rounded and conclusive defence. (1992: 77)

    Each individual argument (or reason) in a coordinative argumentation gives

    some partial support to the standpoint, but the degree of support per

    argument may vary:

    Although the term coordinative may suggest that the argumentation consists of a series

    of arguments that are equally important, this need not be so. It is, for instance quite possible

    THE STRUCTURE OF ARGUMENTATION 461

  • 8/12/2019 The Structure of Argumentation

    16/27

    for one argument to account for, say, 60% of the cogency of the argumentation, the nextfor 10%, and the third for 20%.

    Apart from the fact that the concept of argument structure applies to

    combinations of single arguments in van Eemeren and Grootendorsts

    approach, a second difference with the informal logicians approach is

    that they explicitly connect argument structures with the dialectical situa-

    tion: it depends on the antagonists doubts and the way the arguer attempts

    to deal with these doubts what the resulting structure of his argument will

    be. In this respect, the pragma-dialectical approach can be seen as a con-

    tinuation of the functional aproach of the theory of stasis and debate.

    Theoretical approaches

    In the 1990s, three monographs on argument structure appeared: Freemans

    (1991) Dialectics and the Macrostructure of Arguments, Snoeck

    Henkemans (1992)Analysing Complex Argumentation and Waltons (1996)

    Argument Structure; A Pragmatic Theory. What these three approaches have

    in common are, first, dissatisfaction with the unclear and ambiguous way

    in which the concepts of independent and interdependent arguments have

    been defined in the literature, second, the aims of arriving at more precise

    definitions and giving a functional justification of structural distinctions

    by means of a dialectical approach. Whereas Freeman and Snoeck

    Henkemans each propose a dialectical model of argument structure, Waltons

    contribution mainly consists in a critical discussion of the existing defini-

    tions and tests for deciding whether an argument is linked or convergent.He supports the dialectical approaches to argument structure, and introduces

    a new system of diagramming argument structure.

    Freemans (1991) thesis is that argument structures should be analyzed

    as the results of argumentative processes, i.e. of challenger-respondent

    dialogues.17 His approach to argument structure is to a large extent based

    on the procedural model introduced in Toulmins The Uses of Argument

    (1958). In Freemans view, a theory of argument structure should provide

    a clear-cut demarcation of argument structures. Such a theory should provide

    a rationale for distinguishing different types of argumentative elements and

    structural configurations (1991: 37).

    In his endeavor to provide such a rationale, Freeman takes the basic

    dialectical situation as a starting-point: a respondent develops an argumentin response to the questions of a challenger. The situation is dialectical,

    and not merely dialogical, if the challengers questions are aimed at criti-

    cally testing the claim and if the whole interactive process is clearly

    regimented.

    In the basic dialectical situation, the challenger can ask three types of

    basic dialectical or argument generating questions: acceptability ques-

    tions, relevance questions and ground adequacy questions. Each of these

    462 A. FRANCISCA SNOECK HENKEMANS

  • 8/12/2019 The Structure of Argumentation

    17/27

    questions calls for a specific elaboration of the argument by the respon-dent and results in a different type of argument structure.18

    Linked and convergent argument structures (as well as other types of

    argument structure) are analyzed as answers to different types of dialec-

    tical questions: a linked structure is the result of the respondents answer

    by means of a relevance explaining premiss (9394) to the question

    of relevance (Why is that reason relevant to the claim?). A convergent

    structure results from the arguer giving more evidence for the conclusion

    in response to a ground adequacy question (Can you give me another

    reason?) (95). Then, the result of the exchange is an argument with premises

    that are each independently relevant to the conclusion.

    For Freeman, linked structure is exclusively connected with the question

    of relevance: premisses are linked when we need to take them togetheror they are intended to be taken together to see why we have a relevant

    reason for the conclusion (1991: 94) According to Freeman, the source

    of confusion in the definitions of linked structure in most textbooks is the

    lack of clarity as to whether the arguments are linked because each of the

    premisses by itself is irrelevant to the conclusion, or because each of the

    premisses by itself provides insufficientsupport for it. In addition to the

    normal concept of a linked argument, Freeman introduces a second type

    of linkage, modal linkage. Freeman introduces a special notation for this

    type of linkage. In this way, he thinks, we have a perspicuous way of

    accomodating some intuitions of those who want to link, while still keeping

    modality and relevance issues distinct (1991: 104). In modal linkage,

    each premise constitutes a separate reason for the conclusion, but the pre-

    misses only in combination lend sufficient support to the conclusion. Snoeck

    Henkemans (1994) points out that it is questionable whether Freeman has

    indeed succeeded in achieving his aim of clarifying the distinction between

    linked and convergent arguments:

    This solution merely adds to the confusion concerning the concepts of linked and con-

    vergent structure: an argument can now be (modally) linked and convergent at the same

    time. Another problem is that Freeman ignores the possibility that a convergent argument

    consists of premisses that are not only separately relevant but also have sufficient weight

    by themselves: in his approach, all convergent arguments are modally linked. Instead

    of clarifying the distinction between linked and convergent arguments, his solution amounts

    in fact to drawing a distinction between two types of linked argument. (1994: 320321)

    Snoeck Henkemans (1992) gives a pragma-dialectical analysis of multiple

    and coordinative argumentation as resulting from different types of defen-sive moves aimed at removing different forms of criticism. This analysis

    can be seen as an elaboration of van Eemeren and Grootendorsts (1984)

    pragma-dialectical discussion procedure, in which different types of argu-

    mentation are regarded as functional means to further the resolution of a

    difference of opinion. Snoeck Henkemans (9293) develops a model in

    which rules are given for responding to criticism. The type of criticism

    and the way an arguer chooses to deal with this criticism determines the

    THE STRUCTURE OF ARGUMENTATION 463

  • 8/12/2019 The Structure of Argumentation

    18/27

    form of complex argumentation that will arise in a discussion. When anarguers attempt to get his standpoint accepted on the basis of an argument

    meets with criticism, the arguer can use various types of argumentative

    strategies to make his standpoint acceptable as yet. He can advance more

    arguments to anwer the criticism, he can attempt to counter the criticism

    with new arguments, or he can withdraw the original argument and under-

    take a new attempt at defending the standpoint.

    If the antagonist has criticized the argument for reasons of sufficiency,

    the protagonist can attempt to remove the criticism by supplementing his

    argument with another argument. This can be done in two ways. The pro-

    tagonist can add one or more arguments that in combination with the

    argument given earlier should suffice to convince the antagonist of the

    acceptability of the standpoint (direct defense). The protagonist can alsorefute the counterargument that has been used by the antagonist against

    the original argument (indirect defense). Both types of defense result in

    coordinative argumentation, but in the direct defense case the argumenta-

    tion is called cumulative and in the indirect defense case it is called

    complementary.19 In both cases, the arguments that are put forward must

    be combined, because the arguer can only convince the opponent of the

    acceptability of the standpoint if he succeeds in removing the opponents

    doubt, or criticism regarding the sufficiency of the argumentation.

    In multiple argumentation, the only connection between the arguments

    is that they are all advanced as a defense of the same standpoint. In a

    completely externalized discussion, a protagonist who finds that he is not

    capable of dealing with the antagonists criticism, may withdraw his

    argument and undertake a new attempt to defend the standpoint. The

    argumentation is then multiple in the sense that more than one attempt to

    defend the standpoint has been undertaken, but the final result is in fact a

    single argument. In an implicit discussion this is different. In anticipation

    of a possible non-acceptance of his argument, the protagonist may advance

    a new argument. Each of these arguments is a separate attempt to defend

    the standpoint, and is motivated by the (potential) failure of a previous

    attempt.

    In his approach to argument structure, Walton (1996: xiv) proclaims to

    use methods similar to Freeman (1991) and Snoeck Henkemans (1992), and

    to take a pragma-dialectical approach, although some of his solutions will

    be different. His main goals are to develop more refined guidelines foridentifying linked and convergent arguments and to rescue and refine the

    technique of argument diagramming. Walton discusses numerous examples

    of linked and convergent arguments and develops a classification and ter-

    minology of the different tests used in the literature to determine whether

    an argument is linked or convergent.

    Walton discusses the following five tests (for simplicity, he applies

    them only to two premise arguments):

    464 A. FRANCISCA SNOECK HENKEMANS

  • 8/12/2019 The Structure of Argumentation

    19/27

    1. Falsity/No Support Test: If one premise is false, the conclusion is not given any support.2. Suspension/Insufficient Proof Test: If one premise is suspended (not proved, not known

    to be true) the conclusion is not given enough support to prove it.

    3. Falsity/Insufficient Proof Test: If one premise is false, the conclusion is not given

    enough support to prove it.

    4. Suspension/No Support Test: If one premise is suspended (not proved, not known to

    be true), the conclusion is not given any support.

    5. Degree of Support Test: reasons are dependent when together they make the overall

    strength of the argument much greater than they would considered separately (1996:

    119120, 127).

    Whereas the first four tests are based on the idea of whether the conclu-

    sion is given enough or any support to prove it and are thus absolutistic,

    all-or-nothing kinds of tests (121), the Degree of Support Test is a matter

    of degrees.20

    In order to determine which of these tests should best be used, Walton

    thinks one should consider the purpose of distinguishing between these

    structures, the evaluation of argumentation:

    The critic needs to know whether it is necessary to refute both the [. . .] premises, or if

    it is enough to find fault with just the one, in order for the whole argument to fall down.

    [. . .] From this point of view, the pragma-dialectical viewpoint represented by the van

    Eemeren and Grootendorst conception of the linked-convergent distinction is basically the

    right one, because it is centrally concerned with the question of whether a set of premises

    is sufficient for a conclusion, where sufficient means adequate to resolve the other partys

    doubts in a context of a critical discussion. (1996: 175)

    Walton adopts afunctional interpretation of the linked-convergent distinc-

    tion, meaning that it relates to how the premises of an argument function

    together in supporting the conclusion in a context of dialogue (1996:

    177). Seen from this perspective, the fact that there are many cases where

    it is hard to determine whether an argument is linked or convergent are

    not an indication that the linked-convergent distinction is vague or inco-

    herent: it is possible to see why, in many cases, it is difficult or even

    impossible to determine categorically whether the argument is linked or

    convergent. The reason: there just isnt enough evidence given to enable

    us to determine how the argument is being used in the given context

    (178).

    Waltons conclusion is that the Suspension/Insufficient Proof Test is

    the best test in the sense that it is congenial to our pragmatic theory of

    argument because of its frankly dialectical conception of an argument as

    a dialectical exchange. According to Walton, this test is used by pragma-dialecticians such as van Eemeren and Grootendorst and Snoeck

    Henkemans. Applying this test is problematic, however, since it is extremely

    difficult in practice to determine whether or not the support for a conclu-

    sion is sufficient (180181). Moreover, the fact that this text contextualizes

    the distinction between coordinative and multiple argumentation may be

    positive in itself, but it also makes the test more difficult to apply in cases

    where this contextual background is not available. In such cases, according

    THE STRUCTURE OF ARGUMENTATION 465

  • 8/12/2019 The Structure of Argumentation

    20/27

    to Walton, the Degree of Support Test is most useful to determine whetherthe argument can best be analyzed as linked or as convergent (1996:

    179181).

    Finally, Walton also proposes a new method of diagramming linked and

    convergent arguments, the graph method:

    The most important thing is to see that, by changing from the old notation to the graph

    method, we are freed from always being forced to model an argument as either linked

    or convergent. It is for this reason that the recommendation is made here that we should

    move from the existing method of argument diagramming to the new method of

    reasoning graphs (1996: 187).21

    4. METHODS IN THE ANALYSIS OF DOUBTFUL CASES

    According to Snoeck Henkemans, (1992: 43), the problem of making a

    choice between a linked or a convergent analysis of an argument, is not

    simply caused by a lack of clarity in the definitions, but also by the lack

    of information concerning clues in the verbal presentation. Van Eemeren

    and Grootendorst (1992: 7585) are an exception: they mention a number

    of words and expressions that can be indicative of coordinatively compound

    and multiple argumentation. Examples of indicators of multiple argumen-

    tation are: quite apart from, and then I havent even mentioned the fact

    that, and needless to add that. Indicators of coordinatively compound

    argumentation are: in addition to the fact, when it is also remembered

    that, as well as the fact that.

    Snoeck Henkemans (1992) adds three types of clues to van Eemerenand Grootendorsts indicators of argumentative structure: pragmatic clues

    in the way the arguer has presented the standpoint, dialogical clues in his

    references to criticism, and dialectical clues following from the assump-

    tion that he observes the procedural norms for a critical discussion,

    specifically those norms that are related to the burden of proof. 22 Among

    the pragmatic clues are expressions by means of which the arguer indi-

    cates the force of his standpoint, and the quantifying elements in the

    propositional content of the standpoint: they influence the burden of proof.

    Among the dialogical clues are the arguers acknowledgements and refu-

    tations of counterarguments. Among the dialectical clues are the procedural

    norms pertaining to a critical discussion. One of them concerns the require-

    ments an arguer should meet if his argumentation is to defend theacceptability, or unacceptability, of a speech act; the other concerns the

    requirements he should meet to justify the use of a particular argumenta-

    tion scheme. These two types of requirement enable the analyst to determine

    whether, in order to lend sufficient support to the standpoint, the argu-

    ments that are advanced should be taken together, or whether they should

    be taken to stand by themselves.

    In cases where there is doubt as to whether an argument should be

    466 A. FRANCISCA SNOECK HENKEMANS

  • 8/12/2019 The Structure of Argumentation

    21/27

    analyzed as linked or convergent, informal logicians generally instruct theanalyst to apply the types of tests that have been classified by Walton (1996).

    Groarke, Tindale and Fisher (1997) are an example:

    If you have difficulty deciding whether you should join one premise P to another, ask

    whether the support it provides for the conclusion depends on some other premise(s) (1997:

    36)

    Problems in deciding on the right structure of an argument are generally

    attributed to a lack of clarity in the definitions, and the absence of clear

    and decisive tests. According to Walton, when diagramming arguments,

    there are many cases where there is room for more than one interpretation.23

    This leads to a pedagogical problem in argumentation courses, which is

    compounded by the use of short contextual examples as arguments, con-joined with the use of a test that appears to be precise, decisive, and

    non-contextual (1996: 108). In a functional approach in which it depends

    crucially on the context of dialogue how the structure of a given argument

    should be analyzed, one should be prepared to accept that there will be cases

    where there is not enough contextual evidence to enable a well-founded

    decision (1996: 178). Apart from contextual evidence (evidence concerning

    the type of dialogue, the stage of the dialogue and the burden of proof),

    the analyst can make use of structural evidence (evidence of the type of

    reasoning (deductive, practical) being used) and textual evidence (indi-

    cator words, such as My one reason for believing is this, and my other

    reason is that and This reason, taken along with my other reason [. . .]

    shows that my conclusion is true), which give clues as to how the argument

    is structured. Once all these types of evidence have been collected, therelevant test can be applied as a fourth resource that may help in the iden-

    tification of the argument.

    Some authors propose a special policy for problematic cases. The policy

    which both Thomas (1986) and Nolt (1984: 32) propose for doubtful cases

    is to start with a linked interpretation. To Nolt, the main reason for

    advocating this policy is that it is the most charitable: he thinks that in

    borderline cases an argument is usually strongest if it is regarded as linked.

    Thomas realizes that there may be situations where the policy might not

    be to the advantage of the arguer:

    Probably, you will want to separate reasons, and diagram reasoning as convergent, anytime

    you suspect that one of the lines of reasoning is bad, while the other line of reasoning

    is good (1986: 65)

    Thomas justifies his approach by saying that it is in accordance with his

    Principle of Charity: When unclarity exists, analyze reasoning in whatever

    way it gives the greatest strength (1986: 89). Snoeck Henkemans (1992)

    objects to this solution because she thinks that the analysis and the evalu-

    ation of an argument are mixed up in this policy.24 The question of whether

    the arguers reasons are good or bad, only becomes relevant when the

    THE STRUCTURE OF ARGUMENTATION 467

  • 8/12/2019 The Structure of Argumentation

    22/27

    argumentation is evaluated. In deciding on whether the argumentation isbest analyzed as convergent or as linked, a different issue is at stake:

    The decision on whether to analyze the argumentation as linked or as convergent only

    depends on the degree of support which the premisses, either separately or jointly, lend

    to the standpoint, and should not be confused with an assessment of the acceptability of

    the premisses (1992: 42)

    Van Eemeren and Grootendorst (1992: 8182) propose a policy, called the

    strategy of maximally argumentative analysis, which is the converse of

    Nolts and Thomas policy. In cases that cannot be decided on pragmatic

    grounds, they advise to start by analysing the argumentation as multiple (i.e.

    convergent). They motivate this strategy as follows:

    By way of this strategy, all single argumentations are being given the greatest possiblecredit [. . .] analyzing the structure as multiple can, at the same time, be called more

    and less charitable than analyzing it as coordinative. It is more charitable, because in

    multiple argumentation each individual argument is supposed to have its own, indepen-

    dent argumentative force and, in addition, dropping one unacceptable argument does

    not automatically undermine the whole argumentation. It is less charitable, because in

    multiple argumentation, in principle, all the individual arguments must be separately

    conclusive. More important than charity, however, is from a dialectical perspective that

    the quality of each and every individual argument shall be examined critically (1992:

    8181)

    5. CONCLUSION

    Two approaches to argument structure, the functional approach of theclassical theory of stasis (and of the debate tradition) and the logical

    approach of identifying argument structure with different inference types,

    have continued to exist until present. Dialectical approaches to complex

    argumentation can be seen as a continuation of the tradition of viewing

    argument structures as the result of an arguers attempts to deal adequately

    with an opponents objections in a context of dialogue. Just as in clas-

    sical status theory, it depends on the requirements the arguer should meet

    to provide sufficient support for his standpoint (and thus on the dialogical

    context) whether or not these arguments should be taken together, or seen

    as separate, alternative defenses. In contrast with all informal logical

    approaches (and with Freemans dialectical approach), argument structures

    in pragma-dialectics are always analyzed as combinations of single argu-

    ments. As a result, a large category of linked arguments dealt with by

    informal logicians, i.e. those linked arguments that are equivalent to one

    single argument, falls outside the scope of the pragma-dialectical treat-

    ment of argument structures.

    In informal logic, the logical tradition is still predominant. Informal

    logicians are either concerned with describing different types of premise

    structure within one (deductive or non-deductive) inference, or with dis-

    468 A. FRANCISCA SNOECK HENKEMANS

  • 8/12/2019 The Structure of Argumentation

    23/27

    tinguishing between arguments with one inference and arguments consistingof a number of inferences, that is, with the distinction between single and

    complex argumentation.

    The definitions given in informal logic of the key concepts, linked and

    convergent arguments, differ considerably. The guidelines that are given for

    evaluating linked and convergent arguments also vary from author to author.

    Authors who give a narrow definition of linked arguments, regarding only

    premises as interdependent that need to be taken together in order to be

    relevant, generally advise to always consider all the premises together when

    evaluating the argument, regardless of the structure of the argument. In

    this interpretation of argument structure, part of the rationale for making the

    distinction between interdependent and independent arguments is lost. There

    also seems to be a general (though unsupported) belief among informallogicians that structures consisting of combinations of separate arguments

    are extremely rare, and therefore not worth any serious attention. This belief

    in itself may be traced back to the logical tradition where presenting more

    than one independent proof is, generally speaking, superfluous. In the

    context of our everyday discussions, however, one can rarely be sure that

    a defense that should be conclusive will really convince our opponent and

    one is often at the same time addressing several opponents with different

    views. Presenting a number of arguments which would, if acceptable, each

    provide sufficient support to the conclusion, may then often be the only

    realistic option.

    NOTES

    1 A complex argument is an argument that consists of a number of single arguments for

    a conclusion. A single or individual argument is the equivalent of a reason.2 The Aristotelian sorites can also be analysed as a serial argument in support of the

    major premise.3 In his study of classical status theory, Braet (1984: 55) points out that classical rhetori-

    cians seem to have had some idea of the difference in burden of proof for the defendant

    and the prosecutor, but that they did not devote any systematic attention to this difference.

    The emphasis was always on the defendant. Quintilian is an exception.4 The same difference in obligations for the defendant and the prosecutor is found at the

    sublevels of the argumentation. Cicero (De Inventione 2, 3233) points out that the prose-

    cutor cannot just prove that the accused had a motive for his crime unless he also proves

    that the character of the defendant is consistent with his having such a motive (and vice versa).

    The defendant should show that the accused has never committed any offence, and thisargument will be strengthened if it can be shown that when he had an opportunity of doing

    a dishonest deed with impunity he had no desire to do so (Cicero, De Inventione 2, 35).5 In refuting an opponents argument though, one should be careful not to adduce much more

    than is sufficient to prove ones conlusion, according to Whately, since otherwise ones

    opponents may become distrustful (1846: 165).6 Mills gives the following example: Installment buying is harmful if it encourages people

    to live beyond their means, and it does so; it is harmful if it causes inflation, and it does

    so, etc. (1968: 183)

    THE STRUCTURE OF ARGUMENTATION 469

  • 8/12/2019 The Structure of Argumentation

    24/27

    7 Ray and Zavos (1966: 99) give the following definition of an argument from circumstance:[It] is an argument in which a number of particulars are brought to bear on some point. It

    is usually used to prove the occurrence of some past event.8 According to Thomas, a more important criterion for the independency of premises is

    that the falseness of one reason would not weaken the step from the other to the conclu-

    sion than that each of them individually supports the conclusion: It is possible to have a

    correct convergent diagram in which the result of combining the separated reasons would

    (if this were done) be a stronger argument than either reason provides alone, as long as the

    negation or falsity of the various separated reasons would not decrease the support given

    by the other(s) to the conclusion. (1986: 62)9 Yanal (1991: 139) gives a similar analysis of the distinction between dependent and

    interdependent reasons: What [. . .] is the difference between dependent and independent

    reasons? [. . .] Dependent reasons form one argument; independent reasons form multiple

    arguments.10 The idea that convergent arguments are less vulnerable than linked arguments is all myth

    according to Conway (1991: 155): Even when premises clearly work together, we knowperfectly well that the falsity of one need not necessitate rejecting the entire argument.

    With any sort of argument, when we find that one of the premises are false, we dismiss

    the unacceptable premises and evaluate the support on the basis of those that remain. We

    do not need a notion of convergent support to enable us to do this. Conway concludes that

    we can very well do without the notion of convergence (156).11 It is not clear to which category inductive generalizations belong in Goviers classifica-

    tion.12 According to Conway (1991: 148149), if individual relevance is the criterion for

    convergence, the distinction does not correspond to our intuitions about the separateness

    of premises. Moreover, an argument being convergent on this criterion cannot serve as an

    instruction to evaluate each line of support separately or to evaluate the degree of support

    of the premises for the conclusion in any other particular way. Conway concludes: if there

    is any evaluative importance to the linked/convergent distinction, understood in this way, it

    is not in the area of the degree of support premises give to a conclusion (149).13 Similar definitions of linked and convergent arguments as Goviers are also given by

    Kelley (1988: 687), albeit that he uses the terms additive and nonadditive premises.

    Fishers (1988: 19) definition of independent arguments is somewhat different, since he

    analyzes independent reasons as reasons that are intended by the author to lend conclusive

    support to the conclusion by themselves: reasons may be presented as independently justi-

    fying a conclusion so that if you accept one of the reasons the author expects you to

    accept the conclusion. Nolts (1984) criterion for independence is ambiguous: on the one

    hand he seems to consider arguments as interdependent if they form one inference, and

    independent if they each constitute a separate inference; on the other hand he recommends

    combining inferences to produce a single inference in cases where such a combined infer-

    ence would make a stronger case (1984: 32). In this respect, his approach is more like Thomas

    approach.14 Vorobej (1995b) gives similar defintions as Govier (1992) and Copi and Cohen (1990),

    but also introduces a third category of premise structure: the hybridargument. In conver-

    gent arguments, according to Vorobej, each premise, in isolation, is relevant to or provides

    a reason in support of the conclusion (289). If a set of premises is relevant to the conclu-

    sion, whereas no subset of the premise set is relevant to the conclusion, an argument is

    linked (290). In hybrid arguments, the relation between the premises is asymmetric. One

    premise (or premise set) is relevant on its own and another premise (or premise set) is not

    relevant to the conclusion on its own. The latter premise (or premise set) supplements the

    other premise (set): taken together, the premises provide a better argument for the conclu-

    sion than the one relevant premise (or premise set) alone (292).15 According to Conway (1991: 149), there are many authors who are unclear as to how

    the evaluation of convergent arguments should take place: Remarkably, most who claim

    470 A. FRANCISCA SNOECK HENKEMANS

  • 8/12/2019 The Structure of Argumentation

    25/27

    that the linked/convergent distinction is important for evaluation say nothing about how toevaluate convergent arguments.16 Even though each coordinative argument is analysed as composed of a number of single

    argumentations that each consist of two premises, one of which has been left unexpressed,

    in reconstructing a coordinative argument, one combinedunexpressed premise is made explicit.

    This is done to make it clear that it is only in combination that the arguments are supposed

    to lend sufficient support to the standpoint. For instance, in an example of a coordinative

    argument that is provided by van Eemeren and Grootendorst (1992: 89), two single arguments

    are given for the standpoint that The Telegraph is a good newspaper: 1. The Telegraph

    contains facts, news and background information; 2. The Telegraph is a common-sense

    newspaper. In the diagram of this argument the following unexpressed premise is made

    explicit: A newspaper that combines quality with wisdom is good.17 The basic ideas of Freemans approach were already introduced earlier, in an article in

    Informal Logic (1985) and in his textbook Thinking Logically (1988).18 Freeman thinks that there is independent confirmation for the fact that these questions

    are central to any argumentative exchange, since they pertain to the three criteria for argumentappraisal distinguished by most informal logicians, they parallel Grices (1975) conversational

    maxims and are also represented in Reschers (1977) challenging moves of formal disputa-

    tion.19 The terms cumulative and complementary stem from an unpublished manuscript of

    Pinto and Blair. In this manuscript, they distinguish between three types of interdepen-

    dency: linked premiss sets, cumulative premiss sets and complementary premiss sets.

    Snoeck Henkemans (1992: 94) analyzes linked premiss sets (which consist of a minor and

    a major premise) as single arguments. Explicitizing an unexpressed premise in response to

    an arguers reaction is, according to her, only adequate in cases where there is a problem

    concerning the comprehensibility, not the acceptability of the argument. Since all forms of

    complex argumentation are seen by her as attempts to remove an opponents doubt or

    criticism, and thus to resolve a problem concerning the acceptability of a given argument,

    making explicit the unexpressed premise cannot result in complex argumentation. Cumulative

    and complementary arguments are regarded by Snoeck Henkemans as two different ways

    answering criticism of the sufficiency of one of the other arguments: if the attempt to

    answer criticism consists of adding another argument that supports the standpoint directly,

    the coordinative argumentation is cumulative. If it consists of a refutation of a counter-

    argument, the coordinative argumentation is complementary (1992: 174). These two types

    of support are reminiscent of Whatelys (1846: 88, 138) ways of combining arguments.20 In the discussion of the informal logical distinction between linked and convergent

    arguments, the tests that are used to identify linked and convergent reasoning are often subject

    of discussion. According to Vorobej (1994), most authors propose isolation tests for linked

    arguments (tests that require the analyst to find out whether the premises are relevant in

    isolation, or whether they provide less support in isolation than taken together). The problem

    with those tests is that (in a weak form) they yield the result that any argument with a

    single irrelevant, superfluous premise is linked, and they do not classify all deductively

    valid arguments as linked. The latter is the case because many deductively valid arguments

    contain no premises which are independently irrelevant to the conclusion (149). Vorobejs

    own alternative is to use an elimination test, which says that an argument is linked if the type

    of (positive) support which its premises offer its conclusion would be weakened upon

    elimination of at least one of its premises (1994: 151). On this test, every deductively valid

    argument (with two exceptions) is linked, all analogical arguments are linked, and all argu-

    ments positively corroborating hypotheses. The test classifies as convergent all conductive

    arguments and inductive generalizations. Isolation-relevance tests are comparable to Waltons

    Suspension/No Support Test and the Elimination test is similar to Waltons Degree of Support

    Test. Vorobej acknowledges that his Elimination test cannot do justice to the idea that in a

    linked argument all the premises must be considered together if we are to recognize a per-

    suasive case for accepting the conclusion and that convergent arguments are equivalent

    THE STRUCTURE OF ARGUMENTATION 471

  • 8/12/2019 The Structure of Argumentation

    26/27

    to, and ought to be treated as more than one single argument (156). However, he believessuch ideas are of questionable value insofar as they tend to generate accounts of the

    linked/convergent distinction which seriously run the risk of undermining the significance

    of that very distinction (157). This is because, according to Vorobej, convergent arguments

    consisting of multiple arguments are extremely rare. (1994: 157).21 Apart from the fact that I fail to see the advantage of not having to make a decision

    between a linked or a convergent diagram, I do not think this is correct. Walton explains

    that with the graph method, linked arguments all get the same number, while convergent

    diagrams receive different numbers (1996: 187).22 A difference between the indicators of multiple and coordinative argumentation mentioned

    by van Eemeren and Grootendorst (1992) and the clues offered by Snoeck Henkemans, is

    that the first type of clues indicate how the author intends the text to be understood, whereas

    the second type of clues are clues about how reasons offered in support of a conclusion support

    the argument according to the most charitable analysis.23 Both Walton (1996) and Snoeck Henkemans (1992) mention a similar type of argument

    of which it may be difficult to establish in practice whether it is linked or convergent.Walton calls this type of argument evidence-accumulating(132-133), Snoeck Henkemans

    cumulative coordinative argumentation. According to Snoeck Henkemans (1992: 175), it

    is predictable that multiple argumentation will be particularly difficult to distinguish from

    cumulative argumentation, because it is not always clear whether the arguer may be assumed

    to anticipate that he will have to withdraw some of his arguments, or whether he may be

    assumed to believe that the arguments he has advanced strengthen each other.24 According to Vorobej (1994: 153), classifying arguments as either linked or convergent

    inevitably involves adopting an evaluative point of view during the analysis. The objection

    might therefore be raised that there is little point invoking a distinction between linked

    and convergent arguments as an aid to argument evaluation if the distinction can be drawn

    only afterengaging in that very process of evaluation. But Vorobej thinks this objection is

    not sound. Appraising an argument requires several stages. Therefore, the fact that some

    distinction is drawn by employing evaluative concepts and adopting an evaluative point of

    view does not preclude that distinction from serving a useful purpose in the evaluative

    enterprise (153).25 Vorobej makes a similar observation. He claims that determining whether an argument

    is linked or convergent requires appraising the actual degree of logical support provided by

    (various combinations of) the premises. Although this means that the linked/convergent

    distinction does not operate purely at the descriptive level, the process of evaluation is not

    over once the structure of the argument has been decided on, and the way one sho