ISSA Proceedings 1998 – Evasion In Question-Answer Argumentation: An Empirical Extension

ISSAlogo19981. Introduction
1.1. Rationale
Dialectical theories of argumentation feature question and answer sequences as one basic procedure for building and testing arguments (Hamblin, 1970; Walton, 1989a). In their simplest, ideal forms, questions call on respondents to refine an informational ground presupposed by the question and to commit to the truth of the refined proposition (cf. Bolinger, 1957; Carlson, 1985). Yes/no questions call on the respondent to provide assent (“Yes”) or dissent (“No”) with a questioned proposition. “Are these clothes dirty?” presupposes that either these clothes are dirty or these clothes are not dirty, and the respondent is called upon to commit to one or the other proposition. Alternative questions call on the respondent to select from among a set of exhaustive and mutually exclusive alternatives. “Is this theory a rhetorical, dialectical, or logical approach?” presupposes that this theory is one and only one of the following: this theory is a rhetorical approach, or this theory is a dialectical approach, or this theory is a logical approach. The respondent is called on to commit to one or another of those propositions. WH-questions presuppose some proposition containing a variable (the WH-word/phrase) with an open range of values. For example, “Where is Baluchistan?” presupposes that Baluchistan is somewhere. The respondent is called on to declare a proposition that further specifies the value of the WH-variable (who with someone, where with somewhere, how with somehow, what with something, and why with some reason).

And overlaid on these structural constraints on content are additional pragmatic constraints (see Grice, 1975). What counts as a proper and fitting answer to a question depends upon a mutual understanding of the information space carved out by a question and of the activity for which the information is used. What counts as a relevant, informative, and even truthful or straightforward answer depends upon the purpose of the question, and what is taken to be problematic and what is taken for granted. Thus, “Central Asia” may not be an informative answer to the question “Where is Baluchistan?” if the questioner wants to know where Pakistan exploded its nuclear test bomb but may be informationally sufficient if the questioner wants to know where the extinct giant mammal, the Baluchithere, once lived. Or again, the question “Are these clothes dirty?” may get a quite different truthful answer depending on whether the purpose of the question is to obtain information in deciding whether or not they should be drycleaned, added to the current load of laundry, worn for hanging around the house, or worn to a party.
While there is considerable complexity in the circumstances of their use – as well as complexities and variations in the form of questions themselves – the basic point to see is that questions elicit from respondents pragmatic commitments to propositions. Moreover, questions elicit commitments in ways that pragmatically constrain the kind of propositions the respondent can properly select for commitment. And from these constraints, the informational ground of dialogue is refined and positions may be tested.
Of course, the success of even the simplest idealized question-answer argumentation depends upon clear questions with uncontroversial presuppositions and straightforward, truthful answers to those questions. In practice, questions are often complex, their points opaque, their presuppositions loaded with controversial assumptions. Under less than ideal conditions an appropriate and fitting answer may actually require a reply that is not simple, direct, straightforward, and obviously to the point. Hedging, qualifying, elaborating, and framing answers, and various ways of correcting and pre-empting questions often are cooperative contributions to a complicated situation. Then again, often they are not.

Through the dynamics of questions and answers interlocutors may find themselves faced with defending equally unwelcome choices of position, committed to unanticipated conclusions, forced to abandon positions in which they have a vested interest, or simply compelled to disclose information they would rather not provide. Rather than embrace such consequences, respondents may construct utterances that bend, break, bruise or abandon the principles of cooperative engagement.
Argumentation theorists have long acknowledged that complicated questions of various sorts constitute fallacies that impair argumentative discourse (cf. Walton, 1989a; 1989b; 1991). But they have been less quick to take up systematic problems in answers. This paper examines a type of complicated answer that also constitutes what we think is a fallacy of argument: evasive “answers”. Evasive “answers” are, from our point of view, a subclass of the more general class of answer avoidances. Though they appear to be answers to the question asked, evasive “answers” are not really answers at all (thus the scare quotes).

The distinctive features of evasive “answers” can be highlighted by contrasting them with two other forms of answer avoidance.
First, one can avoid giving an answer by simply “opting out” (Grice, 1975). Here one more or less openly declines to answer a question. Commonly enough, this opting out generates a kind of motivational inference akin to a Gricean implicature. The reason for avoiding an answer is transparently available to the hearer. But it is not the kind of content implicature that Grice was concerned with. The speaker does not implicate information that, once inferred, “saves” Grice’s conversational maxims and would thereby make the message a cooperative contribution. Instead, one at best  only implicates a motive that clashes with the adherence to the Cooperative Principle. So, for example, in addition to a “No comment” a political spokesperson might reply to a question with “I’ve already answered that question several times.” When it is evident that prior replies did not actually answer the question, the spokesperson’s open underinformativeness this time implicates that they are not going to answer the question any further than they already have and that, perhaps, they are growing impatient with the persistence of that line of questioning. Evasive “answers,” like opting out, neither directly nor indirectly supply the called for information. Unlike cases of opting out, however, it is not obvious that a non-answer has been provided.

A second general class of answer avoidance are indirect answers. These utterances do implicate information that answers the question, despite their apparent violation of maxims of relevance, informativeness, truthfulness, and/or manner. They are “avoidances” primarily in these sense that they avoid a direct, open, straightforward, or otherwise perspicuous manner of expression. Often, however, the implicature generated is so nonstraighforward as to be “off-record” (Brown and Levinson, 1987). This is the realm of hint, allusion, innuendo, and insinuation. The speaker can effectively deny commitment to the implicated proposition. Thus, in answer to the question, “Do you think Bill Clinton is guilty of treason?” a radio talk show host might answer, “I’ve never said that, but many people have said that, and I don’t think their concerns can be dismissed lightly.” The host does not come out and commit to the proposition that Bill Clinton is guilty of treason; instead, the proposition is cleverly insinuated and even reinforced. Like indirect answers, evasions appear to indirectly answer the question by appearing to implicate information that would constitute a cooperative answer to the question asked, or at least to what the question is getting at. Unlike indirect answers, evasions don’t really do so, not even off the record.
In a previous study of the answers supplied in political interviews, we illustrated these three general types of answer avoidance (Polcar and Jacobs, 1998). Through detailed textual analysis of excerpts of interviews from various political news shows, we displayed the features summarized above. Analysis of these excerpts, however, was based on our own intuitions about what was and was not being communicated by the avoidances. Skeptics might rightly wonder whether our classification and analysis was not really just an imposition of our own biases, and our textual justifications a series of artfully persuasive interpretations that would not be spontaneously or ordinarily shared by natural language users. To address these sources of doubt, we conducted an empirical study of the impressions ordinary language users have of these three types of utterance.

1.2. Hypotheses
If our typology is correct, then “opt outs,” “indirect answers,” and “evasions” should each display their own distinctive pattern of interpretation when read by ordinary language users. First, since evasions are designed to appear to be efforts to answer the question, ordinary language users should see both indirect answers and evasions as more like actual answers to questions than they do opts outs (which are more or less open refusals to answer). Specifically, we hypothesize the following:
H1a: Respondents will judge opt outs to be less relevant, responsive, and reasonable than indirect answers.
H1b: Respondents will judge opt outs to be less relevant, responsive, and reasonable than evasions.
H2a: Respondents will be more likely to attribute motivational implicatures rather than content implicatures for opt outs than they will for indirect answers.
H2b: Respondents will be more likely to attribute motivational implicatures rather than content implicatures for opt outs than they will for evasions.
Both sets of hypotheses are ways of getting at whether or not evasions, like indirect answers, appear to be answering the question in comparison to opt outs. The outcome of direct comparison of evasions and indirect answers is an open research question. Judgments of relevance, responsiveness, and reasonableness all get at aspects of the quality of answerhood. Motivational implicatures should be preferred where what is inferred is the reason for not answering the question in the first place.

The second comparison distinguishes indirect answers from opt outs and evasions. Ordinary language users should more easily identify information implicated as an answer for indirect answers than for either opt outs or evasions because only indirect answers actually provide information that is an answer to the question. Specifically, we hypothesize:
H3a: Respondents will be more likely to agree on content implicatures for indirect answers than for opt outs.
H3b: Respondents will be more likely to agree on content implicatures for indirect answers than for evasions.

We expect respondents to identify the implicatures we as researchers intended to design into the messages. The crucial point, however, is that whatever implicature respondents recognize they should more readily recognize it in the case of indirect answers than in the cases of opt outs or evasions, which we hypothesize to have no clear content implicature. For this reason, our third set of hypotheses is cast in terms of agreement rather than accurate correspondence with researcher intentions. Once again, the outcome of a direct comparison of opt outs and evasions is an open research question.

2. Method
2.1. Participants and Procedures
Participants consisted of 474 students recruited from undergraduate Communication and Linguistics classes at the University of Arizona. Due to its brevity, the questionnaire was completed during the regularly scheduled class period. Extra-credit was offered at the instructor’s discretion.

2.2. Messages
The independent variable used for this analysis was message type. As proposed in the hypotheses, message type was operationalized as having three levels: opt outs, indirect answers, and evasions. In all, 27 messages were used in this analysis with nine messages typed as evasions, ten as indirect answers, and eight as opt outs. The messages used were generated from the results of a pilot study where 206 participants were placed in hypothetical situations and asked to respond to a “double-bind” question for which no good answer could be directly constructed. Participants’ responses were selected to best exemplify one of the three broad types of avoidance. Since the primary interest in the pilot examination was to determine naturalistic responses and appropriate content for messages, some messages were restructured to enhance readability, brevity, and prototypicality.
Each participant in the main study randomly received one of the 27 messages to evaluate. The final messages evaluated by the participants were single question-answer sequences. Each question-answer sequence was contextualized within one of four scenarios:
Chris and Mary; Sam and Diane; Christine and Jay; and Professor Smith and Jim. Each scenario contained a question that demanded a yes/no answer and each constructed response fell into one of the three general types of avoidance described above. Examples of messages used can be found in Appendix A.

2.3. Dependent Measures
After reading the scenario with its question-answer exchange, participants were asked to evaluate the response for its overall relevance, responsiveness, and reasonableness. Ten Likert-type items were used to assess these variables, mixed with 10 distractor questions to help disguise the purpose of the questionnaire. All 20 items consisted of statements to which the participants responded on a five point scale where 1 indicated strongly agree; 2 indicated agree; 3 indicated neutral; 4 indicated disagree; and 5 indicated strongly disagree.
Relevance was operationalized as the degree to which the response addressed the question, appeared to address the question, provided the information asked for, and was on topic. Responsiveness consisted of measures of participants’ perception of the respondent’s intent to answer the question. This variable was operationalized through assessment of statements that the respondent was trying to answer the question, wanted to answer the question, and did not want to answer the question. Reasonableness was operationalized by two items: an assessment that the respondent’s response was reasonable and that it met the demands of the situation. Type of implicature drawn was measured by one multiplechoice type question. Participants were asked to choose one of three statements that best represented a paraphrase of the speaker’s message in an attempt to identify what participants’ believed the speaker actually communicated. The three answer choices were positive implicature (e.g., “I will move into your apartment”), negative implicature (e.g., “I will not move into your apartment”), or motivational implicature (e.g., “I am not going to answer you now”). For indirect answers, the negative implicature was the target implicature. For opt outs, the motivational implicature was the target implicature. For evasions, none of the options were targeted. (For the latter two message types, the negative implicature could be thought to get at the unexpressed reason for the response). Implicature questions for each scenario are reprinted in Appendix B.

3. Results
In order to protect against alpha inflation, correlations between the dependent variables and a Multivariate Analysis of Variance were conducted before testing hypotheses 1a and 1b (Bray & Maxwell, 1985). Significant correlations (p .000) were discovered among the three dependent measures. As suggested by Keppel (1993), an omnibus F test was performed to protect against Type One error. Results of the MANOVA were significant, F(8,233) = 11.21, p <.000 (Wilks’ A : L = .83), warranting interpretation of tests of the individual hypotheses.

In this analysis, error terms were constructed treating messages as a random factor (Jackson & Brashers, 1994; Keppel, 1993). All F ratios use message type as the numerator with messages nested within message type as the denominator. Overall, the oneway ANOVAs used to test for differences between opt outs, indirect answers, and evasions on the dependent variables were significant.
Message types differed on relevance, F(2,24) = 9.99, p<.001; on responsiveness, F(2,24) = 9.44, p<.001; and on reasonableness, F(2,24) = 4.72, p of planned comparisons were conducted.
Hypothesis 1a predicted a difference between opt outs and indirect answers on each of the three dependent measures. Planned comparisons indicated that indirect answers were rated as more relevant (t(24) = 7.64, p<.0001), more reasonable (t(24) = 5.00, p<.0001), and more responsive (t(24) = 5.75, p<.0001) than opt outs. Likewise, consistent with hypothesis 1b, evasions were rated as more relevant (t(24) = 3.39, p .001), more reasonable (t(24) = 4.39, p .0001), and more responsive (t(24) = 4.97, p<.0001) than opt outs. Further planned contrasts showed that indirect answers were rated as significantly more relevant than evasions (t(24) = -3.76, p<.0001), but no significant differences between the two message types were found for reasonableness or responsiveness (see Table One for reported means).
To test the overall relationship between type of implicature and message type, a chi-square analysis was conducted. The chi-square was found to be significant (x2(4) = 16.80, p<.01). Observed and expected frequencies for the nine cells as well as the proportions of responses evidenced in each category are reported in Table Two. As indexed by Cramér’s statistic, the strength of the relationship between implicature drawn and message type was .13. Specifically, consistent with hypotheses 2a and 2b opt outs were more likely than expected by chance to generate motivational implicatures, but not so for indirect answers and evasions. And consistent with hypotheses 3a and 3b indirect answers are more likely to result in negative implicatures than expected while opt outs are less likely to result in negative implicatures than expected by chance. The remaining cells only nominally contributed to the overall chi-square value and the differences between expected and observed totals were minimal. Interestingly, the observed values for the evasion message type were virtually what we would expect due to chance alone.

4. Discussion
The results of this study provide strong independent corroboration for our intuitions that evasion is a distinctive form of answer avoidance, different in kind from indirect answers or opt outs. The overall pattern of results suggest that participants recognize that all these response types are less than full and straightforward answers. The ratings of relevance, responsiveness, and reasonableness in Table One all tilt to the disagree side of neutral. Likewise, the frequencies in Table Two display a pronounced tendency in all message types to read the motivational implicature (“I am not going to answer you now”) as the best paraphrase of what the respondent means.
Nevertheless, each proposed type of answer avoidance displayed a pattern of interpretation in the expected characteristic fashion. Participants’ ratings of messages suggest that they see evasions as answers to questions like indirect answers and unlike the more or less blatant failure of an answer in opt outs.  This impression was also reflected in their choice of paraphrase: Participants were much more likely to choose “I am not going to answer you now” for opt outs than for either indirect answers or evasions. On the other hand, while evasions look like answers, ordinary language users seem to have a hard time figuring out what the answer is. Subjects in this study were unable to consistently identify exactly what implicature an evasion was expressing – unlike the case of indirect answers where a negative implicature was the predominant choice. Here, the pattern looks more like it does for opt outs. In other words, while participants thought they had an answer with an evasion and did not think they had an answer with an opt out, participants were could no more reliably say what the answer was than if they had been exposed to “I don’t know” as the response.
We think these features of evasions qualify it as a genuine fallacy of answering. While we cannot offer any mechanical procedure for constructing an evasion, nor can we provide clearcut recognition rules independent of the functional characteristics laid out here, we think it is clear that evasions are a recurrent and reproducible phenomenon. Their artfulness makes them no less systematic than say, red herrings or strawmen.

Moreover, evasions are a clearcut violation of dialectical principles. Whatever else they do, evasions covertly violate the maxims of relevance and informativeness which underlie any cooperative communicative exchange (Grice, 1975). Now, it is still something of a puzzle as to how, exactly, evasions are able to get away with these violations (Polcar and Jacobs, 1998). In the case of deception, where the violation involves falsification of content information, this can be done covertly because identifying false information often involves comparison between the message and some independent state of affairs that may not be immediately available to the message recipient. But evasion isn’t like that. Evasion (in question-answer sequences) involves a relationship between the message given and the communicative demands that are established by the question. That relationship should be open, publicly observable, and readily accessible for inspection. Nevertheless, artful respondents do get away with these violations.

Table 1

Table 1

And that is another reason to treat evasions as genuine fallacies. They work, and they work to the detriment of productive dialectical engagement. One might argue that in the case of opting out, no one is tricked or fooled. Such a move may obstruct deliberation, but not in a way that escapes notice or chance for repair. Likewise, with indirect answers, the answer at least does get on the floor – although often in a way that allows the respondent to elude responsibility for defending what has been conveyed. But at least for many indirect answers, the answer is on record and the violation of manner is trivial or only apparent but not substantial.
That is not the case for evasions. They are neither obvious in appearance, nor are they trivial in impact. The natural response of the questioner is to prod no further and to proceed with another line of inquiry because it appears an adequate answer has now been provided. And that is how to stop an investigation cold.

 

Appendix A

Examples of Scenarios and Message Types

1. Opt Outs

1a. Professor Smith and Jim
You are taking a class on Interpersonal Communication. For today, you were supposed to read a book that discussed relationship development.
Jim, another student in class, is sometimes unprepared for class discussion. Dr. Smith, your professor, calls on Jim to discuss the reading.
Professor Smith: Did you do all the reading?
Jim: I don’t know.

1b. Christine and Jay
Christine: I just don’t think the other members of the group like me.
What do you think?
Jay: Come on, let’s talk about something else.

2. Indirect Answers

2a. Chris and Mary
Chris and Mary have been dating for two years. Over dinner, the following conversation occurs.
Chris: Don’t you think you should move into my apartment?
Mary: Well, I am awfully busy with school.

2b. Professor Smith and Jim
You are taking a class on Interpersonal Communication. For today, you were supposed to read a book that discussed relationship development.
Jim, another student in class, is sometimes unprepared for class discussion. Dr. Smith, your professor, calls on Jim to discuss the reading.
Professor Smith: Did you do all the reading?
Jim: I got through a lot of it.

3. Evasions

3a. Diane and Sam
Diane: I know you’ve been working closely with your boss on this recent project. I think he’s been making a lot of mistakes, especially at the last meeting, and I am sure you are worried about how the project will turn out. Are you worried about the project?
Sam: One thing my boss does well is stand behind his work. It is an admirable quality and one that I wish I and more people in this company had.

3b. Christine and Jay
Christine: I just don’t think the other members of the group like me. What do you think?
Jay: You know, I am really sorry you think that. You really shouldn’t be concerning yourself with this and you need to work on developing more self confidence.

Appendix B

Implicature Questions by Scenario Chris and Mary: Which of these best paraphrases what Mary’s answer meant?
a. “I will move into your apartment.”
b. “I will not move into your apartment.”
c. “I am not going to answer you now.”

Professor Smith and Jim:
Which of these best paraphrases what Jim’s answer meant?
a. “I have done the reading.”
b. “I have not done the reading.”
c. “I am not going to answer you now.”

Diane and Sam:
Which of these best paraphrases what Sam’s answer meant?
a. “I am not worried about the project.”
b. “I am worried about the project.”
c. “I am not going to answer you now.”

Christine and Jay:
Which of these best paraphrases what Jay’s answer meant?
a. “The study group likes you.”
b. “The study group does not like you.”
c. “I am not going to answer you now.”

REFERENCES
Bolinger, D. (1957). Interrogative Structures of American English: The Direct Question, Publication no. 28 of the American Dialect Society. Tuscaloosa: University of Alabama Press.
Bray, J. H. & S. E. Maxwell (1985). Multivariate Analysis of Variance, Quantitative Applications in the Social Sciences, series no. 07-054. Newbury Park, CA: Sage.
Brown, P. & S. C. Levinson (1987). Politeness: Some Universals in Language Usage. Cambridge, UK: Cambridge University Press.
Carlson, L. (1985). Dialogue Games. Dordrecht, Holland: D. Reidel.
Grice, H. P. (1975). Logic and conversation. In: P. Cole & J. Morgan (Eds.), Syntax and Semantics, Vol. 3: Speech Acts (pp. 41-51). New York: Academic Press.
Hamblin, C. L. (1970). Fallacies. London: Methuen.
Jackson, S., & D. E. Brashers (1994). Random Factors in ANOVA, Quantitative Applications in the Social Sciences, series no. 07-098. Newbury Park, CA: Sage.
Keppel, G. (1993). Design and Analysis: A Researcher’s Handbook (3rd ed.). Englewood-Cliffs, NJ: Prentice-Hall.
Polcar, L., & S. Jacobs (1998). Evasive answers: Reframing multiple argumentative demands in political interviews. In: J. F. Klumpp (Ed.), Argumentation in a Time of Change: Theories, Definitions and Critiques. Proceedings of the Tenth NCA/AFA
Conference on Argumentation (in press). Annandale, VA: Na
Bolinger, D. (1957). Interrogative Structures of American English: The Direct Question, Publication no. 28 of the American Dialect Society. Tuscaloosa: University of Alabama Press.
Bray, J. H. & S. E. Maxwell (1985). Multivariate Analysis of Variance, Quantitative Applications in the Social Sciences, series no. 07-054. Newbury Park, CA: Sage.
Brown, P. & S. C. Levinson (1987). Politeness: Some Universals in Language Usage. Cambridge, UK: Cambridge University Press.
Carlson, L. (1985). Dialogue Games. Dordrecht, Holland: D. Reidel.
Grice, H. P. (1975). Logic and conversation. In: P. Cole & J. Morgan (Eds.), Syntax and Semantics, Vol. 3: Speech Acts (pp. 41-51). New York: Academic Press.
Hamblin, C. L. (1970). Fallacies. London: Methuen.
Jackson, S., & D. E. Brashers (1994). Random Factors in ANOVA, Quantitative Applications in the Social Sciences, series no. 07-098. Newbury Park, CA: Sage.
Keppel, G. (1993). Design and Analysis: A Researcher’s Handbook (3rd ed.). Englewood-Cliffs, NJ: Prentice-Hall.
Polcar, L., & S. Jacobs (1998). Evasive answers: Reframing multiple argumentative demands in political interviews. In: J. F. Klumpp (Ed.), Argumentation in a Time of Change: Theories, Definitions and Critiques. Proceedings of the Tenth NCA/AFA
Conference on Argumentation (in press). Annandale, VA: National Communication Association.
Walton, D. N. (1989a). Informal Logic. Cambridge, UK: Cambridge University Press.
Walton, D. N. (1989b). Question-Reply Argumentation. New York: Greenwood Press.
Walton, D. N. (1991). Critical faults and fallacies of questioning. Journal of Pragmatics 15, 337-366.