Using Conversation Theory to Investigate Conclusion-Drawing: Implications For Persuasion

ABSTRACT - Kardes' (1988) recent article examining spontaneous inference processes in advertising is one of the most significant studies of conclusion-drawing in persuasion in two decades. Consistent with earlier conclusion-drawing studies, Kardes' approach relies heavily on the use of the logical syllogism as a theoretical basis for exploring persuasion. In this article we explore the value of reconsidering conclusion-drawing from a conversation theory point-of-view. The three target arguments used by Kardes for conclusion-drawing are analyzed and evaluated to highlight important differences between message-based and receiver-based meanings in advertising processing.

Citation:

Mark Toncar, James M. Munch, and Michael Mayo (1994) ,"Using Conversation Theory to Investigate Conclusion-Drawing: Implications For Persuasion", in NA - Advances in Consumer Research Volume 21, eds. Chris T. Allen and Deborah Roedder John, Provo, UT : Association for Consumer Research, Pages: 343-347.

Advances in Consumer Research Volume 21, 1994      Pages 343-347

USING CONVERSATION THEORY TO INVESTIGATE CONCLUSION-DRAWING: IMPLICATIONS FOR PERSUASION

Mark Toncar, Kent State University

James M. Munch, University of Texas at Arlington

Michael Mayo, Kent State University

ABSTRACT -

Kardes' (1988) recent article examining spontaneous inference processes in advertising is one of the most significant studies of conclusion-drawing in persuasion in two decades. Consistent with earlier conclusion-drawing studies, Kardes' approach relies heavily on the use of the logical syllogism as a theoretical basis for exploring persuasion. In this article we explore the value of reconsidering conclusion-drawing from a conversation theory point-of-view. The three target arguments used by Kardes for conclusion-drawing are analyzed and evaluated to highlight important differences between message-based and receiver-based meanings in advertising processing.

INTRODUCTION

In her recent article, Thomas (1992) suggests utilizing principles of everyday conversation to investigate a variety of marketing topics, including issues in both personal selling and advertising. One topic which may potentially benefit from a conversation theory approach is the effects of conclusion-drawing in persuasion. The purpose of this paper is to illustrate how considering a conversation theory approach to conclusion-drawing may help improve our understanding of how conclusion-drawing affects persuasion.

THE PRINCIPLES OF CONVERSATION

Conversation contains a system of implicit and explicit rules (Grice 1975,1978). These rules govern when and how long we talk, the words we use, and expectations of our conversational partners during our talk (Nofsinger 1991). Although many of us are unaware of these principles, socialization processes lead us to follow them nonetheless. We use these rules not only to guide us in our own talk, but to interpret and assign meaning to the talk of others. This system of rules, termed the Cooperative Principle, allows us to say one thing, and be confident that our message will be interpreted as something more, or something different than the literal meaning of the words that we use.

One domain where conversational principles may apply is advertising. Upon hearing an advertising claim that appears to be irrelevant, consumers may change the meaning to make it relevant. For example, to understand and make relevant General Motors' "Heartbeat of America" advertising slogan, consumers are expected to infer meanings that go beyond the literal meaning of the words used. Inferences that go beyond the explicitly stated and logically implied meanings of a claim are termed pragmatic inferences. Claims which require a pragmatic inference to arrive at the intended meaning of the claim are termed pragmatic implications. Decoding of the pragmatic implication using the principles of conversation results in a pragmatic inference by the receiver.

Advertising studies have shown that consumers will pragmatically infer a subjective meaning if it is not explicitly stated or logically implied (Harris 1977; Gaeth and Heath 1987; Bruno and Harris 1980). Further, prior research has demonstrated that consumers confuse pragmatically implied product claims with claims that are directly asserted (Harris 1983; Grunert and Dedler 1985). However, previous research has not applied the principles of conversation to the investigation of advertising effectiveness.

A related area which has received significant research attention concerns the effects of conclusion-drawing on persuasion. A pragmatic inference is one type of inference, and an inference can be thought of as a conclusion.

PERSUASION AND CONCLUSION-DRAWING

The logical syllogism is the most popular argument structure used in marketing to examine conclusion-drawing effects. The syllogistic framework in conclusion-drawing research was first explored in Hovland and Mandell's (1952) classic study. The authors exposed two independent groups of subjects to a communication advocating devaluation of the American dollar. For both groups the communicator outlined the general economic conditions that would make devaluation of a country's currency desirable and then, using a syllogistic argument, demonstrated that these conditions existed in the United States. For the explicit conclusion group, the conclusion to the logical syllogism was clearly stated while for the other group it was not. As it turned out, an explicit statement of the conclusion was a great deal more effective in eliciting opinion change than leaving the conclusion to be drawn by the audience.

To equate level of ability between groups, Thistlethwaite, de Haan, and Kamenetzky (1955) administered a comprehension test to groups so the effects of stating the conclusion or leaving it unstated could be determined with understanding held constant. Thistlethwaite et al. found no significant differences in attitude change between groups.

Linder and Worchel (1970) argued that differences in subjects' levels of motivation to effortfully draw the conclusion may account for Thistlethwaite et al.'s data. To consider this explanation, their procedure required subjects to view a series of seven logical syllogisms projected on a screen in front of them. Syllogisms were sequentially arranged, (i.e., the conclusion of one syllogism served as the first premise of the next one). The final syllogism was:

(First premise): Smoking filtered and unfiltered cigarettes introduces large quantities of tar into the lungs.

(Second premise): Tar released from burning tobacco causes lung cancer.

(Conclusion): Smoking filtered and unfiltered cigarettes causes lung cancer.

Subjects were assigned randomly to one of three conditions. In the high effort conditions, the communication explicitly presented the "correct" conclusion of the first syllogism and subjects were asked to infer the conclusion of each of the remaining six syllogisms. In moderate effort conditions, subjects were given three conclusions and were asked to draw four conclusions themselves. In low effort, five conclusions were generated for subjects and they were asked to generate two. After viewing the syllogisms subjects responded to attitude items including the linkage between smoking and lung cancer. Results indicated acceptance of the target conclusion increased as motivation increased.

Spontaneous Inference Formation

Extending Linder and Worchel's research, Kardes (1988) argued that we know subjects can be induced to draw conclusions by explicitly asking them to do so, but can marketers use advertisements to motivate subjects to spontaneously infer conclusions without explicit prompting from the experimenter?

TABLE 1

CONCLUSION LATENCIES AS A FUNCTION OF CONCLUSION EXPLICITNESS AND INVOLVEMENT

To test this general proposition Kardes randomly assigned subjects to conditions in a 2 X 2 X 2 between-subjects design with two levels of conclusion explicitness (explicit or implicit), two levels of involvement (high or low), and two levels of measurement order (measurement of response latency first or measurement of evaluation latency first). Subjects were told that they would be asked to evaluate several ads and were given a folder containing four print ads. The target ad contained three sets of arguments and was presented last.

Kardes reasoned that explicit conclusion subjects should respond to inquiries about the relevant conclusions quickly (irrespective of involvement) because they can simply retrieve from memory conclusions provided to them earlier. Implicit conclusion subjects on the other hand should respond to inquiries more slowly when involvement is low, because they should be insufficiently motivated to generate missing conclusions on their own. Kardes' critical test then involves implicit conclusion-high involvement subjects. If these subjects respond as quickly as the explicit conclusion subjects, we can infer that the involvement manipulation was effective in eliciting spontaneous inference formation.

Consistent with his theorizing, Kardes' conclusion latency data (provided above in Table 1), when averaged across the three conclusions ((4520+4413+4503)/3 vs. 4722, F(1,188)= 5.28, p<0.02), provides support for the hypothesis that when conclusions are omitted, spontaneous inference formation is more likely in high than in low involvement conditions.

Conclusion-Drawing and Attitudes

Consistent with prior research findings of the persuasiveness of explicit conclusions, Kardes reasoned that if the message arguments are compelling, any variable that enhances comprehension should increase persuasion. Since involvement should enhance motivation, Kardes hypothesized and found that brand attitudes are significantly more accessible and favorable (see Kardes' Table 3, and page 230) in the implicit conclusion-high involvement and explicit conclusion conditions than in the implicit conclusion-low involvement conditions.

RECONSIDERING CONCLUSION-DRAWING VIA CONVERSATION THEORY

Interestingly, the data in Table 1 highlight large response time differences between conclusions 1, 2, and 3. This suggests that subjects may have expended different levels of effort or may have assigned differing levels of subjective meaning to Kardes' words in trying to reach "appropriate" conclusions (cf. Mick 1992). Also, because response times are so different between conclusions, Kardes' footnote 3 becomes especially interesting:

"Errors were operationalized as the failure to respond to a target conclusion item within the allotted time period of seven seconds, or pressing the button labeled 'No' (the target questions were always worded in the affirmative direction). Response latencies to questions upon which errors were committed were deleted from subsequent analyses." (Kardes 1988 p., 229).

Given these issues, we decided to reconsider Kardes' conclusion types and their potential meanings from a conversation theory point-of-view.

Conclusion Types

As a first step in reconsidering subjects' inferences we evaluated the structure of the three arguments used in Kardes' research. Then we collected data to improve our understanding of how subjects may have processed these structures. We begin by presenting Kardes' three argument structures.

Conclusion 1: The CT-2000 also features a horizontal disc load, a current track display, and a motorized drawer. Other CD players lack a motorized drawer. Inserting a disc is difficult without one.

(Explicit conclusion) Inserting a disc is easy with the CT-2000.

Consistent with past persuasion research on conclusion-drawing, conclusion 1 may best be considered a logical syllogism:

First premiseC CT 2000 has a motorized drawer.

Second premiseC Inserting a disc is difficult without one.

ConclusionC Inserting a disc is easy with the CT 2000.

Kardes' planned contrast for conclusion 1 revealed that, as predicted, conclusion latencies were faster in the implicit conclusion-high involvement condition and the explicit conclusion conditions than in the implicit conclusion-low involvement condition (F(1,157) = 4.73, p < 0.04). Thus, the personal relevance of the message and form of the syllogistic argument influenced subjects' inference processes.

Conclusion 2: All CD players require digital filters, because the decoding of digital sound creates sampling frequency distortions that must be filtered out. Digital filters are expensive and each filter accounts for a large portion of the total price. One advanced filter is sufficient for filtering out sampling frequency distortions and two less advanced filters are no better than one advanced filter. Most CD players have two less advanced filters. The CT-2000 has one advanced filter.

(Explicit conclusion) The CT-2000 filters out sampling frequency distortions at less cost.

Conclusion 2 does not conform to the hierarchical steps in a logical syllogism, but could be labeled a compound or "double-barreled" syllogism. The sentences comprising conclusion 2 actually contain two sets of syllogistic arguments. The first sentence contains two major premises: (1) All CD players require digital filters; and (1A) Decoding of digital sound creates frequency distortions. The first portion of sentence two is a second premise to major premise 1: Digital filters are expensive. Sentence three is a syllogistic second premise for major premise 1A: One advanced filter is sufficient for filtering distortions. The last clause in sentence three (i.e., the notion that two less advanced filters are no better than one advanced filter) is a syllogistic conclusion about the ability to filter distortions (i.e., a response to second premise following major premise 1A). Finally, sentence four notes that most CD players have 2 less advanced filters, while sentence five states that the CT 2000 has one advanced filter. These sentences add strength to the issue of number of filters, but offer no undisputed conclusion about filter cost. Therefore, subjects may logically arrive only at the portion of Kardes' conclusion about filtering out distortions.

As is evident in Kardes' explicit conclusion, the author's intent is a dual conclusion about filtering distortions at less cost. Consistent with his results for conclusion 1, the planned contrast for conclusion latencies for conclusion 2 were also significant. Subjects' spontaneous inferences in the implicit conclusion-low involvement conditions were significantly slower than the other three conditions (F(1,150) = 4.41, p < 0.04). Thus, even though conclusion 2 is quite obtuse, subjects' conclusion latencies were significantly affected by the treatment variables.

Conclusion 3: Best of all, the CT-2000 brings you a sophisticated laser technology. The purpose of lasers is to reduce distortion from dust and scratches. Most CD players have one laser. The CT-2000 has three.

(Explicit conclusion) The CT-2000 reduces more distortion from surface irregularities than most CD players.

Conclusion 3 is not syllogistic. To arrive at a conclusion, subjects must infer the meaning of the argument by employing the maxims of conversational theory. Subjects must pragmatically infer that the greater the number of lasers a CD player has, the greater the reduction in distortion.

The planned contrast for conclusion 3 was not significant. As rationale for his findings, Kardes suggests that because conclusion 3 is presented at the end of the text subjects may have processed these arguments less extensively than earlier arguments. One alternative interpretation for these data resides in conversation theory. If subjects were "required" to pragmatically infer the meaning for conclusion 3 in order to comprehend it, the formation of receiver-based meanings may account for the dramatically slower response latencies shown for conclusion 3. In addition, this deeper level of subjective comprehension processing in response to this conclusion may have rendered Kardes' treatment variables ineffective (Mick 1992).

Pragmatic Inferences

To examine the alternative, subjective meanings that subjects may assign to the above three conclusions, forty-two students were provided a copy of Kardes' three arguments (implicit conclusion version) and were told: "Advertisers and marketers use product claims to sell products. We have provided you with three claims that might be used in advertising. Please write down the meaning of each of these claims." After writing meanings for each of the three arguments subjects were provided with Kardes' explicit conclusions and asked to judge whether their answers matched Kardes' intended meanings or not. Finally, subjects were asked if they did or did not understand each of the three arguments. Subjects were then dismissed. Two researchers coded subjects' responses for the incidence of counterargumentation. Counterarguments were defined as any disagreement with Kardes' intended, syllogistic conclusion. Interjudge reliability was 90% and differences were resolved by discussion. The results are presented in Table 2.

Interestingly, Conclusion 1, the most "logical" argument from a syllogism framework, yields the largest number of incorrect conclusions. This finding suggests that subjects may not typically process advertising claims in a syllogistic fashion. Consistent with Mick's (1992) point-of-view, although the syllogistic information-processing model remains the most widely adopted theoretical basis for persuasion research, the model may over-emphasize message-based meanings while disregarding receiver-based meanings.

Subjects' responses to Conclusion 2 suggest that processing this compound structure results in a good deal of inaccuracy, some problems in understanding and significant counterarguing. Perhaps advertisers desiring to highlight multiple product attributes should exercise caution in developing complex message structures.

Conclusion 3, the most conversational argument, and the argument which "requires" the greatest leap or pragmatic inference, is most accurately identified. However, Kardes' data (see Table 1) suggest claims that necessitate a pragmatic inference are significantly deeper levels of processing than syllogistic reasoning (4877 vs. 3832 milliseconds).

Conclusion-Drawing and Attitudes

Differences in the number of counterarguments reported for each conclusion (see Table 2) provides weak evidence that subjects' attitudes may significantly differ in response to the various types of conclusions. The pragmatic inference "required" for conclusion 3 creates twice the level of counterargumentation compared to the logical syllogism form in conclusion 1. Interestingly, this counterargumentation occurred even though no explicit instructions to counterargue were given to subjects. Apparently, the heightened processing involvement required to conversationally infer conclusion 3 activated subjects' need to affectively respond. (Complexity of the compound syllogism contained in conclusion 2 may have required deeper processing as well).

TABLE 2

FREQUENCY OF SUBJECTS AGREEMENT WITH KARDES' CONCLUSIONS

In his recent article examining subjects' levels of subjective comprehension in response to an advertisement for a CD player, Mick (1992) found that deep comprehension levels (receiver-based meanings such as pragmatic inferences) were positively related to both ad credibility and ad attitude when the valence of those meanings were taken into consideration. Also, surface level comprehension (message meanings related to the explicit or asserted message content) was shown to be negatively related to these dependent variables. Given these findings, it would be interesting to reassess Kardes' conclusions to explore whether significant differences exist between conclusion 3 (i.e., a receiver-based conversational implication) and conclusion 1 ( i.e., an explicit message-based syllogistic argument).

While it is plausible that Kardes' conclusion 3 did not "work," and his experiment was successful in spite of this, a reexamination of the data from a conversation theory viewpoint offers an alternative, and equally plausible explanation. Because conclusion 3 is non-syllogistic and requires a pragmatic inference, it may have had unmeasured and unintended effects. These effects may be one basis for Kardes' results. He eliminated those subjects who either counterargued or took over seven seconds to respond to a target conclusion item. The mean response latency for conclusion 3 (which elicited significant counterargumentation in our sample) was nearly 5 seconds! Those subjects who processed the message most deeply, and took longer to do so, may have been eliminated from Kardes' study. Similarly, subjects who recognized that conclusion 3 was not substantiated, and counterargued with the conclusion, were also deleted from analysis. In light of the relationship between level of comprehension and attitude discussed above, it appears that Kardes' results may not have been obtained in spite of conclusion 3, but because of conclusion 3!

SUMMARY

Past studies examining persuasion have tended to adopt an objective orientation toward message comprehension (cf. Mick 1992) and studies of conclusion-drawing have relied on use of the logical syllogism as a theoretical basis for persuasion research (Kardes 1988; Linder and Worchel 1970). The syllogistic model and the basic idea that humans have a need to maintain consistency among feelings, thoughts, and actions, will no doubt remain a powerful model for future research (cf. McGuire 1978). Kardes' research clearly illustrates the strength of the syllogistic approach. Both the simple syllogistic structure and the compound syllogism are shown to be consistent with the notion that when subjects possess sufficient ability and motivation, they should be left to draw their own inferences.

A great many advertisements do not conform to the syllogistic structure, however. "Have you driven a Ford Lately?"; "Get Met. It pays."; "We drove our competition into the copier business." These are just a few examples of current advertising messages that have meanings that go beyond the words used, and require the receiver to subjectively assign meaning to the message. This paper suggests that when arguments do not clearly conform to a syllogistic structure, marketers may consider receiver-based models of spontaneous inference processes. A Conversation Theory Framework which considers the role of pragmatic implication may be one promising approach for exploring these structures and their relationships to receivers' advertising processing strategies.

Conversation theory, with its emphasis on the generation of receiver-based meanings for messages, offers a useful method for investigating the subjective interpretations of persuasive communications, and the effects of these self-generated meanings on subsequent attitude structure. The Conversation Theory framework need not be limited to advertising issues, however. Research in the areas of buyer-seller interaction, relationship marketing, and globalization of advertising issues are a few of the areas that could benefit from a conversation theory perspective.

This paper offers evidence that people may process information in a manner that is not syllogistic. Further, through a reexamination of an important paper in the conclusion-drawing literature, we demonstrate that using principles derived from the investigation of everyday conversation offers insights and suggests issues that have not been addressed in the marketing literature. Conversation theory offers a receiver-based approach to understanding the process of social influence and persuasion.

REFERENCES

Bruno, Kristin J., and Richard J. Harris (1980),"The Effects of Repetitions on the Discrimination of Asserted and Implied Claims in Advertising," in Applied Psycholinguistics, 1 (August), 307-321.

Gaeth, Gary J. and Timothy B. Heath (1987),"The Cognitive Processing of Misleading Advertising in Young and Old Adults: Assessment and Training," in Journal of Consumer Research, 14 (June), 43-54.

Grice, H.P. (1975),"Logic and Conversation," in P. Cole and J.L. Morgan (eds) Syntax and Semantics: Volume 3. Speech Acts, pages 41-58, Academic Press, New York, N.Y.

Grice, H.P. (1978),"Further Notes on Logic and Conversation," in P. Cole (ed) Syntax and Semantics: Volume 9. Pragmatics, pages 113-128, Academic Press, New York, N.Y.

Grunert, Klaus G. and Konrad Dedler (1985),"Misleading Advertising: In Search of a Measurement Methodology," in Journal of Public Policy & Marketing, 4, 153-165.

Harris, Richard J. (1977), "Comprehension of Pragmatic Implications in Advertising," in Journal of Applied Psychology 62, Number 5, pages 603-608.

Harris, Richard J., ed. (1983), Information Processing Research in Advertising, Hillsdale, NJ: Lawrence Erlbaum.

Hovland, Carl I. and Wallace Mandell (1952), "An Experimental Comparison of Conclusion-Drawing by the Communicator and the Audience," in Journal of Abnormal and Social Psychology, 47 (July), 581-588.

Kardes, Frank R. (1988), "Spontaneous Inference Processes in Advertising," in Journal of Consumer Research, 15 (September), pages 225-233.

Linder, Darwyn E. and Stephen Worchel (1970), "Opinion Change as a Result of Effortfully Drawing a Counterattitudinal Conclusion," Journal of Experimental Social Psychology, 6 (October), 432-448.

McGuire, William J. (1978), "An Information-Processing Model of Advertising Effectiveness," in Behavioral and Management Science in Marketing, ed. Harry L. Davis and Alvin J. Silk, New York: Ronald, 156-180.

Mick, David Glen (1992), "Levels of Subjective Comprehension in Advertising Processing and Their Relations to Ad Perceptions, Attitudes, and Memory," Journal of Consumer Research, 18 (March), 411-424.

Nofsinger, Robert E. (1991) Everyday Conversation, Sage Publications, Newbury Park, CA.

Thistlethwaite, Donald L. Henry de Haan, and Joseph Kamenetzky (1955), "The Effects of `Directive' and `Non-directive' Communication Procedures on Attitudes," Journal of Abnormal and Social Psychology, 51 (July), 107-113.

Thomas, Gloria Penn (1992), "The Influence of Processing Conversational Information on Inference, Argument Elaboration, and Memory," in Journal of Consumer Research, 19 (June), pages 83-92.

----------------------------------------