In daily life, anything can go wrong in discussions: Some discussions hardly get off the ground, others progress with the utmost difficulty and yet others are derailed even before an argument is put forward. During a now infamous meeting of the European Council of Ministers, the German Social Democrat Member of European Parliament (MEP) Martin Schulz suggested in a response to a speech by the Italian Prime Minister Silvio Berlusconi that the slow progress made by Italy on a number of pan-European judicial measures was indicative of the “virus of conflict of interests” spreading beyond the borders of Italy. Berlusconi replied: “Mr. Schulz, I know a producer in Italy who is making a film about Nazi concentration camps. I will suggest you for the role of a “kapo.” You would be perfect!” (European Parliament, sitting of 2 July 2003).
This is a preview of subscription content, log in via an institution to check access.
Access this chapter
Tax calculation will be finalised at checkout
Purchases are for personal use only
Notes
- 1.
Schulz himself proves to be quite capable of making this kind of unreasonable remarks himself. On May 23rd, during a visit of the Dutch Prime Minister Jan-Peter Balkenende to the European Parliament, the Dutch MEP Erik Meijer defended the Prime Minister by stating that more countries would have voted no to the constitution, had they been allowed to vote. Schulz immediately accused Meijer of practicing National Socialism: “Du verkaufst reintheoretischen Quatsch. Ist das holländischer Nationalsozialismus?” (You are uttering theoretic nonsense. Is that Dutch National Socialism?).
- 2.
Locke says of this argument: “The fourth is the use of evidence that is derived from the fundamentals of knowledge or probability (…). Of the four, (…) only this is a contribution to the increasing of our knowledge” (Essay IV: iii).
- 3.
It should be noted here that Locke uses the term ad hominem in a particular, specific way and that his concept of this type of argumentative move differs greatly from the modern view. Locke calls argumentation ad hominem when it is based on the concessions that are made by the opponent. In this view every argumentation is ad hominem – and it is not in itself unreasonable.
- 4.
When exactly the term gained its negative connotation and from whom is difficult to ascertain (see van Eemeren & Grootendorst, 1993).
- 5.
Aside from this, in the standard approach to fallacies and in informal logic there is not really any general agreement on the use of terminology. “Circumstantial” sometimes stands for accusations of inconsistency in general: “given certain alleged facts about one’s opponent’s background, behavior, prior commitments, or other circumstances, it is inconsistent for that opponent to accept (or reject) a particular point of view” (Brinton, 1995, p. 214). The tu quoque variant seems to be restricted by some to responses to criticism or behavior (“you also do it yourself”) and does not belong to accusations of inconsistency. What is termed “circumstantial” by Copi and in the pragma-dialectical approach is called “poisoning the well” by others. This confusion in terminology is symptomatic of the chaos noted by Hamblin that reigns in the standard approach to the fallacies.
- 6.
There are indeed authors, including Kahane (1971) and Carney and Scheer (1964), who appear to support this idea in practice; these authors treat the tu quoque variant, for example, as a separate fallacy.
- 7.
In this respect, their approach is, at first sight, similar to the pragma-dialectical one in distinguishing between sound and fallacious “strategic maneuvering” depending on the context of the specific “argumentative activity type” in which the argumentative moves concerned are made, but when considered more carefully the pragma-dialectical approach is fundamentally different (see van Eemeren & Houtlosser, 2007).
- 8.
Govier’s approach is virtually in accordance with the tentative analysis of Woods and Walton (1989). They suppose that there is a fixed relationship between authority argumentation that they call ad verecundiam and the ad hominem. As a general rule, they state: “The correct rejection of an argument for its having committed the fallacy of ad verecundiam involves the non-fallacious use of an ad hominem” (1989, p. 71).
- 9.
It is confusing that in most of these approaches the term ad hominem is not only used for the fallacy but also for the reasonable form. It is therefore necessary to continually refer to the “reasonable ad hominem” and the “unreasonable ad hominem.” We prefer to use “personal attack” as a neutral term and “ad hominem” for every fallacious personal attack.
- 10.
Considered from a pragma-dialectical perspective, a speech act that is brought forward in a critical discussion constitutes either a contribution to the resolution of the difference of opinion (as a consequence of which the speech act concerned is deemed by definition reasonable), or no contribution in the sense that it makes the resolving the difference of opinion more difficult or blocks it (as a consequence of which the speech act concerned is deemed by definition unreasonable). From this theoretical perspective, “reasonableness” is therefore an absolute, binary term. From an empirical perspective, however, it is meaningless to present respondents with a binary choice (“Is according to you the discussion contribution reasonable or unreasonable?”) since such a double-option question unavoidably leads to the introduction of a huge amount of “random error,” or noise – assuming that the respondents start out from a gradual, non-binary concept of the term “reasonableness.” And this is certainly substantially the case: Even a five-point Likert-type scale, as is apparent from a pilot, contains in the view of the respondents too few classes to allow the expression of their gradual, differentiated concepts of “reasonableness.” Therefore, in all our research reported in this volume, a gradual scale consisting of 7 classes was chosen. The implication of this method of questioning is that the centrally posed question of our research into the conventional validity of the pragma-dialectical discussion rules can be answered not in binary, absolute terms, but in gradual terms.
- 11.
The post-experimental session included investigation of whether the respondents had worked in a serious and concentrated fashion, whether they had found out about the purpose of the study, what they found wrong about fallacies in general and the specific fallacies under investigation in particular, and so on. The post-experimental session took place standardly in the way reported here; the purpose of this was to examine whether and to what extent the participating respondents might be characterized as naive so that possible alternative explanations of the study results could be excluded. Since 2002, argumentation learning has been a compulsory part of the Dutch language subject at upper secondary level (VWO/HAVO), so research into judgments on fallacies as in the research in question is at the present time no longer possible, at least not with strictly comparable groups of respondents in upper secondary (high) school (4 and 5 VWO/HAVO): Today these students can certainly no longer be regarded as naive.
- 12.
The way in which hypothesis 2 has been formulated strongly invokes the suggestion that it should be possible to model the judgments of the respondents in a uni-causal manner in terms of politeness – but we rightly oppose such a thing. In this chapter, we will in various ways try to demonstrate that the reasonableness judgments are based primarily on rational-critical grounds, and to a much lesser extent on considerations concerning politeness.
- 13.
As we revealed in Chapter 2, in their research Bowker and Trapp present the fragments in isolation, which leads to methodological interpretation problems. In all the research presented in this volume, the discussion fragments are presented in a well-defined context: Sometimes the explanation of the contexts is extremely extensive and the description of this occupies a full page (as is the case in all the research in the context of the freedom rule), sometimes a brief description of the location/situation in which the discussion takes place and of the people that are participating in the discussion suffices (for example: “Man and woman in a vehicle” in the investigation of the argument scheme rule).
- 14.
To obtain more insight into this issue, one would have to calculate the covariance matrix of the scores given by the respondents, starting from the facet design applied (see Table 3.1), and then have to analyze this with hypothesis-testing factor-analytical techniques (via LISREL for example). The central point in this analysis is whether a one-factor model “explains” the matrix better than, for example, a three-factor model (in which each factor is indicative of a type of discussion context). We have indeed tried to perform such an analysis, but attempts at this ran aground, due, amongst other issues, to the non-positive definiteness of the matrices to be analyzed.
- 15.
In all the research to be reported here, based on internal and external validity considerations, a multiple message design was used.Based on practical considerations, the number of messages varied from 2, the absolute minimum, to 6. The use of such designs has far-reaching consequences for the statistical analysis of the data: The customary, standard variance-analytical techniques are in such cases no longer applicable, which in general leads to a more complex form of data processing.
- 16.
The 48 dialog fragments were, at least in this research, constructed by an instructed doctoral student who was specialized in pragma-dialectics, but had no knowledge of the specific research hypotheses. Such a “blind” approach avoids E-bias.
- 17.
The constructed fallacies – and this is standard for all our research – constituted in all cases clear cases of the category concerned. Borderline cases over which differences of interpretation might arise amongst the respondents which would threaten the internal validity were therefore excluded.
- 18.
This element was explicitly added to the instruction since from a pilot it was apparent that respondents found it confusing that they did not know whether the discussion partners were indeed speaking the truth. Aside from this, in all the research reported here a pilot study was performed as standard to investigate whether the instruction was actually clear, whether the fragments presented were in all respects adequate, to determine the time required, and so forth.
- 19.
Note that these non-significant results are also incompatible with an alternative explanation of the research results in terms of an practice or learning effect. In order to further exclude such effects, statistical tests were performed on possible differences between the first 5 discussion fragments and the last 5, insofar as these fragments were at least of the same type. No statistically reliable differences were found, which indicates the absence of learning effects.
- 20.
Under the assumption of a repeated measurement design in which the random Replication factor (with 12 levels) is nested within the fixed factor “Is/Is Not Violation of the Freedom Rule” (with 2 levels), and is crossed with the random Respondent factor, “quasi F ratios” must be calculated (as proposed, for example, by Clark (1973)). All F ratios reported in this volume are of this type. For the sake of convenience, we report these F’s not in the conventional manner (as: F’ – F-prime), but simply as F. The degrees of freedom for such quasi-tests are not precise, but must be approximated (by, for example, the method developed by Satterthwaite (1946)). This approximation method is precisely the reason that in the three reported F ratios, although the number of respondents, number of replications and number of levels of the fixed facts are exactly the same, differences in the numbers of degrees of freedom (for the denominator) are found. Effect sizes are reported in this volume as ES (effect size); these refer to the proportion of variance of the total (explained by all variance sources within the specified statistical model) variance that is associated with a certain effect, under abstraction of the variance source that is associated with the “intervening” variable Respondent (the variance sources for the interactions of respondents with the other effects in the model are certainly involved in the calculations of ES). Since the variance-analytical data technique employed here (applied to a “mixed design”) rests on rather heavy statistical assumptions that are in practice often difficult to check (such as those of sphericity and independence), an alternative data technique such as multilevel analysis (in which as a whole such heavy assumptions are not made) have preference, but due to the relative unfamiliarity of this latter data technique we apply the conventional, more trusted and customary analysis of variance.
- 21.
In this case, just as was the case with hypothesis 1, quasi F’s had to be calculated. A statistical model constituted the starting point for this, in which the random Replication Factor (with 4 levels) was nested within the interaction of the two fixed treatment variables (i.e. the type of ad hominem and the type of discussion context) and was crossed with the random Respondent Factor (this is typically a situation that gives rise to quasi F’s; crossing of two random factors). For this statistical model, the enumerator of the quasi F for (for example) effect A consists a composite of (MS(A) + MS(Rep(AB) × Respondent)), while the denominator consists of (MS(Rep(AB)) + MS(Respondent × A)).
- 22.
If the t tests, in which averages are contrasted (under abstraction of individual variance between the instantiations/representations), are replaced by quasi F’s, the conclusions are identical (for example: both the quasi F for the 36 fallacies (F(2,94) = 0.55) as for the 12 sound fragments (F(2,72) = 0.73) identically indicate no difference between the two sexes). In addition, in virtually all the research reported in this book, the variable “gender” is included, but nowhere is there any significant difference encountered between men and women, either in the judgment of fallacies or in the judgment of sound argumentation. We will make no further comments on this variable “gender.”
- 23.
Also in this research (n = 24; k = 24), fallacies were found to be far less reasonable than non-fallacies (fallacies: 3.43 (0.64), non-fallacies: 5.26 (0.72)) and also in this study the now known order of reasonableness was encountered: The direct attack was judged as the least reasonable (2.99 (0.76)), followed by the indirect personal attack (3.47 (0.94)) and finally the tu quoque variant (3.82(0.88)). Also now the fragments in a scientific discussion context were most strictly judged (average 2.63(0.79)), while no differences in reasonableness were found between the other two discussion contexts (Domestic context = 3.78(0.72) and Political context = 3.88 (0.82)).
- 24.
In all these cases in which the motives of the respondents were also requested, a coding scheme had to be developed. To discourage ad hoc coding, as a rule such a scheme – other than in the research reported here in which verbal interviews were involved – was set up on the basis of an random sample consisting of 10% of the entire material.
- 25.
The starting point in the (fallacious) tu quoque variant is that whoever is not consistent can also not be right. Whoever preaches what he does not himself practice is of course inconsistent, but this does not automatically mean that his standpoint is therefore incorrect. In order to be able to determine the acceptability of this standpoint, the arguments brought forward must be judged. The point that the non-fallacious variant of the tu quoque attack revolves around is that it is absolutely not a fallacy when it concerns a contradiction of a standpoint that someone has adopted within the discussion – on the contrary, looking for inconsistencies is even a necessary part of the judgment.
- 26.
When something exceptionally unpleasant is said about someone, as is the case in this example (“I really find you extremely untrustworthy”), this does not yet mean by definition that there is also an argumentum ad hominem, even if it constitutes an argument that is aimed at discrediting the person concerned. Whoever wants to defend the standpoint that B is untrustworthy can do little other than recount negative actions, characteristics or behaviors of B, but once again this is in itself absolutely not a fallacy.
- 27.
Aside from this, note that the contrast between the fallacious and reasonable variants from a methodological point of view is not as distinct as would be desirable: For the reasonable variants, the impoliteness is located in particular in the standpoint brought forward, in the fallacious variants in the response to this standpoint or in the argumentation that is brought forward on the request of the antagonist. But it is difficult to do otherwise: It would be pointless to have the respondents judge the in one case on the “reasonableness” of the standpoint, and in the other case on the “reasonableness” of the argumentation brought forward.
- 28.
The last two scales therefore functioned as “distracters” in order to conceal the objective of the experiment.
- 29.
An example for clarification. Suppose that a researcher wishes to investigate the effects of an experimental didactic in (initial) reading by young readers, and sets up an experiment for this. Starting readers in one class are assigned to the experimental didactic, those in the other class are subjected to the control didactic. Suppose that after the passage of time it appears that the readers with the experimental didactic achieve higher reading scores than the pupils with the control didactic, then this advance could alternatively be explained by initial differences in intelligence between the pupils in the experimental and the control condition (intelligence of course depends on or co-varies with reading skills). If, however, the experimenter has the IQ’s of all the pupils available, he can keep the effect of these under control by including intelligence as a co-variate in the statistical analysis; co-variant analysis eliminates all initial differences in intelligence between the two groups (in a statistical sense), making use in this of the correlation between the co-variant (in this case intelligence) and the dependent variable (in this case reading skill).
- 30.
A somewhat problematic aspect attached to the method of analysis of covariance applied here may not go unreported. The contrast performed in Table 3.8 (significance testing for the difference with and without removal of the effect the co-variate) is less pronounced than desirable. As a result of computational problems, the analysis of covariance could not be performed on the original 12 fallacies and 12 sound argumentations, but on an random selection of 18 argumentations in total. Therefore results occurring by chance, although highly improbable, cannot be excluded.
- 31.
The domestic and political domains in which there is often “practical argumentation” appear in their nature to lend themselves better to the construction of discussion fragments with evaluative and appealing standpoints, while in a scientific discussion context descriptive standpoints prevail.
- 32.
The 72 text fragments contained a fallacy 54 times (3 types of fallacies × 3 types of discussion fragments × 3 types of standpoints, each represented two times) and 18 times sound argumentation (in which the three types of standpoints were evenly divided). It appeared difficult to construct fragments in which there was a combination of a descriptive standpoint and a tu quoque fallacy, especially within the domestic discussion context. A purely “you jest” for example as a response to a description of an actual state of affairs often seems highly affected; this artificial character disappears, however, in fallacious responses in which the antagonist points to inconsistencies of the protagonist with the aim of excluding him as a serious discussion partner on these grounds.
- 33.
The data within the two groups were analyzed independently of each other – with of course the exception of the statistical test of the difference between both groups. In each of both the groups a statistical model was adopted in which the random factor Replication was nested within the interaction of Type discussion context × Type fallacy × Nature of the standpoint and was crossed with the random factor Respondent. The resulting F’s are of course again quasi-F’s in which both the numerator and the denominator consist of composites of MS terms. Thus the quasi-F with which the influence of the fixed factor “Nature of the standpoint” is tested consists of a numerator with the composite MS (Nature of the standpoint) + MS ((Replication within Type discussion context × Type fallacy × Nature of the standpoint) × Respondent), and of a denominator with the composite MS (Replication within Type discussion context × Type fallacy × Nature of the standpoint) + MS ((Nature of the standpoint) × Respondent).
Author information
Authors and Affiliations
Corresponding author
Rights and permissions
Copyright information
© 2009 Springer Science+Business Media B.V.
About this chapter
Cite this chapter
van Eemeren, F., Garssen, B., Meuffels, B. (2009). Ad Hominem Fallacies: An Exemplary Study. In: Fallacies and Judgments of Reasonableness. Argumentation Library, vol 16. Springer, Dordrecht. https://doi.org/10.1007/978-90-481-2614-9_3
Download citation
DOI: https://doi.org/10.1007/978-90-481-2614-9_3
Publisher Name: Springer, Dordrecht
Print ISBN: 978-90-481-2613-2
Online ISBN: 978-90-481-2614-9
eBook Packages: Humanities, Social Sciences and LawPhilosophy and Religion (R0)