Persuasion As Directed Inference

John Deighton, Dartmouth College
ABSTRACT - The paper argues that persuasion by advertising is, at least in part, the result of directing aa audience's inferences from evidence. Points at which inference is vulnerable to this influence are identified. Studies offering empirical support for the model are reviewed
[ to cite ]:
John Deighton (1986) ,"Persuasion As Directed Inference", in NA - Advances in Consumer Research Volume 13, eds. Richard J. Lutz, Provo, UT : Association for Consumer Research, Pages: 558-561.

Advances in Consumer Research Volume 13, 1986      Pages 558-561

PERSUASION AS DIRECTED INFERENCE

John Deighton, Dartmouth College

[John Deighton is Assistant Professor of Business Administration, Amos Tuck School, Dartmouth College. Hanover, New Hampshire 03755.]

ABSTRACT -

The paper argues that persuasion by advertising is, at least in part, the result of directing aa audience's inferences from evidence. Points at which inference is vulnerable to this influence are identified. Studies offering empirical support for the model are reviewed

INTRODUCTION

There is ample evidence that some messages, under some contingencies, will persuade their audiences more after the lapse of time than at the moment of exposure. The phenomenon is intriguing, because it offers the possibility of insight into at least some of the processes of persuasion When message exposure can be temporally separated from persuasion, we can examine to what extent persuasion depends on the intervening contingencies.

The contingency studied in this paper is evidence. The paper presents a model of a process of persuasion in which the message acts to direct the path of the audience's subsequent inferences from evidence that bears on the message's claim. Phrased differently, the model envisages that one of the ways messages can persuade is by shaping the sense we make of data.

This model of persuasion seems, on the face of it, to have application to persuasion by advertising. It may account for some odd properties of advertising. The contradiction is often noted, for example, that consumers deny that advertising affects them, yet advertisers see value in using it. To the extent that advertising is influencing sensemaking, rather than presenting argumentation, its influence will not be accessible to introspection by its audience and these contradictory reports can be reconciled.

Another curious property of advertising is the fact that it is often used to promote low-cost, frequently purchased products and services to consumers who are already users. The use of advertising to retain the loyalty of existing consumers has been noted by Ehrenberg (1974) and Harper (1976) among others. To the extent that advertising acts by shaping the interpretation of experience, it is plausible that its most responsive audience will be those who are already users of the advertised product or service. Thus the marketing of high-share brands like McDonald's or Coke will be more successful if it can influence existing customer's satisfaction with the next purchase than if it can persuade the small proportion of their market who have never tried the product to do so. The question addressed in the next section of this paper is: how can persuasive communications influence the experience of consuming a product or service?

THE MODEL: HOW INFERENCE IS DIRECTED

The model applies to ambiguous consumption environments: those in which the consumer is not compelled by the event itself to attach to the event one meaning or another. The sense a consumer makes of an ambiguous consumption experience owes more to promotion than to the native event. Consider two kinds of ambiguity to illustrate this claim.

In the first, ambiguity is the result of the bland, uncontroversial nature of the experience. No prominent features of the product intrude themselves on the consumer's attention: there is little information available from the experience. Eating a slice of bread might be such an experience. Unpromoted, the event has little meaning. Promoted, the very blandness of the event allows it to support many assertions: "builds bodies seven ways," or "see how light and airy it is!," or "taste the wheat!"

In the second, the experience is not bland, but its distinctive features may serve as evidence for a variety of interpretations without conclusively eliminating any of them. Here, in contrast to the previous example, the experience offers considerable information, but little diagnosticity. The purchase and wearing of clothing, for example, generates for the wearer a complex set of data. By asserting claims consistent with some of this data and not contradicted by others, a marketer may shape the experience of wearing the clothing in directions not guaranteed by the experience itself.

The purpose of the model is to explain how advertising shapes the meaning consumers ascribe to such ambiguous events. By "meaning" is meant a set of beliefs and evaluations inferred inductively by the consumer from the consumption event. The ascribing of meaning is modeled as the inducing of inferences. The model seeks to identify points in the inductive process at which advertising can be said to direct inference, or influence it to deviate from the canons of classical statistical induction.

To decide what an experience means, the stream of sense data that makes up the experience must be broken into units needing interpretation, and one potential interpretation must win favor over others. In this model of the sensemaking process, advertising plays at least two roles. First, during exposure, it forms an expectation. Second, during experience with the product, it directs inference toward confirmation of the expectation.

In the first step, the audience learns, rather passively, what the advertiser is claiming. It would be wrong to assume that the audience accepts the claim, and probably more accurate to describe the claim as merely entertained. One interpretation of the experience of using the product becomes more salient than others. As McCombs and Show (1977) claimed of the press with respect to politics, advertising cannot determine what we think. But it can influence what we think about.

Because the expectation aroused in the first step is weak, particularly if the audience is sensitive to the partisan nature of advertising, measurement of the expectation is difficult. A subject who fails to report the expectation may do so either because the ad failed to create it, or because the ad succeeded but the subject was reluctant to admit to belief in the claim on such slight evidence. If the subject does report the expectation, it may not exist as a psychological state of expectation but merely as demand-induced verbal compliance. Thus simple pre-post exposure copy testing is likely to be a poor predictor of advertising performance.

The second step, from "thinking about" to "believing," depends on experience. Without wholly accepting the claim, the consumer may find it plausible enough to act on it, at least to the extent of a trial purchase. In this second step, provided the claim is not sharply repudiated by experience or contradicted by prior beliefs, it influences the sense made of experience in several ways. It affects the choice of evidence to be encoded in memory, strategies for interpreting the evidence, the process of combining information to form impressions, and/or the retrieval of evidence from memory. In this manner the claim gradually becomes a belief: external influence feels like learning from experience.

A number of studies have found that the salience of cuse influences their use in judgment. Reyes, Thompson and Bower (1980) found that, in a mock jury decision, evidence made disproportionately available in memory had a correspondingly disproportionate influence on the jury's verdict. In the field of person perception, several studies have found effects of salience on causal attribution (Lepper and Greene 1975, Taylor and Fiske 1978, and see Ross and Anderson 1982 for a survey). Wyer and Srull (1981) induced shifts in impression formation by manipulating the availability of relevant concepts. Schindler and Berbaum (1983) found that salience influenced choice.

Media-induced salience has been shown to influence behavior. Suicide reports in the press were associated with increases in the population suicide rate (Phillips 1974), publicized prize fights affected the indicence of homicide (Phillips 1983), radio reports of prosocial behavior induced subjects in a bargaining experiment to be more co-operative (Blackman and Hornstein 1977), and the rate of violent crime rose after the spate of assassinations in the early and mid-1960's (Berkowitz and Macaulay 1971). More generally, a view of the mass media as agenda-setters for the public's evaluations have been advanced by Shaw and McCombs (1977), and forms the basis of Sutherland and Galloway's (1981) argument that agenda-setting is an important part of the process by which advertising communicates.

Advertising usually makes salient two aspects of a consumption experience: an outcome and an associated cause. The goal may be to make the brand a candidate cause for the outcome, for example when Ritz crackers are given as the reason for a successful party. Or it may be to draw attention to the outcome: the 'quicker picker-upper' advertising for Bounty paper towels tells consumers to notice absorbency among the many other aspects of the consumption experience that might capture their attention. In either event, the advertising invites its audience to test one cause-outcome link among the many they might otherwise consider.

Another relevant research stream has to do with the way such candidate cause-outcome links are tested. A number of studies have found that, in inferring the strength of association between causes and outcomes, lay perception of correlation is disproportionately influenced by confirmatory instances (Beyth-Maron 1982, Hamilton and Rose 1980, Smedslund 1963, Ward and Jenkins 1965. See Crocker 1981 for a comprehensive review of lay covariation assessment). This work suggests that the confidence with which a claimed cause comes to be accepted as a true cause will increase with the number of times the cause and the outcome co-occur, with too little regard for the number of times the outcome occurs without the cause or the cause without the outcome. Thus, for example, belief in the claim that a paper towel was more absorbent because it bore the brand name Bounty would be strengthened whenever a consumer used Bounty and found it absorbed well. Lack of evidence on the absorbency of other brands, or even occasional evidence that other brands pick up quickly too, if not mate salient by advertising, would not interfere with the experiential confirmation of the claim for Bounty.

Combining these two sets of findings, we propose that advertising (1) sets the consumer's diagnostic agenda and (2) benefits from the self-fulfilling nature of consumer inquiry into that agenda. In situations that would otherwise be bland or confusing, its claims become the questions consumers ask. By making one interpretation of an experience more salient or available than others, an advertisement disposes the consumer to confirm it. so the consumer comes to accept it.

EMPIRICAL TESTS

The research reviewed in this paper comprises three studies. The first examined how an advertising claim influenced the inferences drawn from evidence of product performance. The second study examined how a claim affected inferences from product experience. The final study sought to identify characteristics of the consumer's information search and analysis that might explain the confirmatory diagnosis phenomenon.

1. Directing Inference from Evidence. The first study (Deighton 1984) examined how advertising for Ford cars would affect the sense made of evidence of automobile reliability. The advertising stimulus was print advertising from Ford's "Quality is Job 1" campaign which claimed that Ford payed particular attention to quality in its assembly plants. Evidence of reliability took the form of two pages of data from Consumer Reports' Frequency of Repair Records for 1976 to 1981 for sixteen models of car. This data was rather overwhelming, and in the time subjects had to examine it they could do no more than form a rough impression of its meaning.

The experiment can be thought of as a kind of inkblot test: subjects saw a rather ambiguous stimulus (evidence relevant to Ford's claim) and in the value they placed on it they indicated the extent to which the advertising had been effective in prefiguring the sense-making process.

The experiment took the form of a 2x2 factorial design: subjects were assigned to one of four conditions obtained by crossing advertising (present or absent) with evidence (PreSent or absent). Dependent measures were the shifts in ratings of the reliability of six car manufacturers following exposure to the experiential condition.

The result of this experiment is summarized in Figure 1. The Ford advertising had a significant (p=0.03) effect on belief change, but only when subjects had the opportunity to see evidence. Without this 'grist for the mill of inference', the advertising had no measurable persuasive effect. If the no-evidence conditions of the experiment had been run alone, as in a conventional copy test, the test result would have suggested rejecting the campaign. In fact, the campaign appears to have the capacity to shape the sense made of evidence.

FIGURE 1

SHIFT IN ESTTMATE OF RELIABTLITY

By using the evidence contained in Consumer Reports tables, this study was able to control the amount and kind of experience consumers could bring to bear on the advertised claim. This tactic enhanced the internal validity of the study. However, it did little for external validity. If an advertisement can direct inferences from a table of data, does that mean it can alter the experience of using a product? Further research was needed to establish whether inference from experience was similarly responsive to advertising.

2. Directing Inference from ExPerience. The second study (Deighton and Schindler 1985) examined how belief in an advertised claim was affected by experience with the product. The finding was very similar to that in the first study: that although the claim itself was not persuasive, the claim in the presence of experience did induce belief to shift. Thus this study supports the notion that expectancy-confirming inference is not just characteristic of consumers' reading of evidence, but extends to the way they form impressions of their own experiences with advertised products.

In this study, advertising was developed for each of three Boston rock music radio stations, claiming that each played 'the most new music'. Subjects were assigned to one of three groups, and each group heard advertising for a different station. The study measured shifts in belief in the claim before hearing the advertising, after hearing it, and two weeks later. Subjects reported how much exposure they had had to the stations over the two weeks that followed exposure to the advertising.

Figure 2 summarizes the results of this experiment. Again, advertising for a station had no effect on beliefs for subjects who did not subsequently listen to that station. However, subjects who heard the advertising and also listened to the station did respond positively to the claim, and this effect was greater the more experience they had (subject only to a scale ceiling effect). It appears that the audience found confirmation of the claim in their experience with the advertised station, but did not find this confirmation when their interpretation of the experience had not been prefigured bs advertising.

FIGURE 2

SHIFT IN BELIEF THAT STATION PLAYS MORE NEW MUSIC

3. Mechanisms in Directed Inference. The first two studies showed that advertising's persuasive effect is sometimes delayed: that the claims of advertising sometimes become more plausible over time, and not just because they are objectively correct. The third study (Deighton 1983) examined why. How is it that a consumer can acquire a strengthened belief in an advertised claim in the face of inconclusive or even occasionally contradictory evidence?

This study sought the answer in the way consumers test claims about products: in characteristics of consumer inquiry that might discourage or impede accurate learning from experience. This is not to say that we imagine a world of very methodical consumers who set out deliberately to test claims. All we propose is that the learnt claims of advertising become the default opinions that consumers entertain about the state of affairs in markets, and as they operate in these markets they encounter experience which either repudiates or endorses the default opinions. The question then is how experience updates beliefs.

The general answer proposed in this study was that consumers use heuristics (cognitive short cuts) to form the impressions they depend on the make choices among products. These heuristics lighten the burden of thinking, but they systematically reduce the odds that advertised claims will be discredited. Thus impressions are biased toward confirmation of prior expectations.

The first hypothesis tested in this study was that consumers employ positive confirmation as a reliable guide to covariation: that when they form impressions about causation, positive confirmatory evidence is overweighted. What is positive confirmation? When consumers believe that a cue causes an outcome (such as that Crest gives better checkups), they can meet with four kinds of experience. They can use Crest and have better checkups (cue and outcome both present: a positive confirmatory experience) or they can have good checkups while using another brand (outcome present, cue absent), or use Crest but not have good checkups (outcome absent, cue present). Finally they can have poor checkups with another brand (outcome and cue both absent). Normatively, consumers should compare the frequencies of occurrence of all four kinds of experience to infer the strength of the association between Crest and checkups. In practice, we argue, consumers who hold Crest's claim as their default opinion will give too much weight to the occasions when Crest works. They will take too little account of the times when Crest fails to work or of the success rates of other brands.

The second hypothesis dealt with the way consumers gathered information bearing on an advertised claim which they had learnt. Consumers, seeking successful purchase outcomes rather than knowledge for its own sake, will expose themselves to evidence in which the advertised cue is present more than evidence in which it is not. In so doing, however, they shield themselves from evidence that might prove the claim false. The study sought empirical support for this proposal. Einhorn and Hogarth (1978) showed that, as long as there is any correlation above chance between a cue and an outcome, the restriction of evidence to instances in which the cue is present will bias the gathered sample to report higher correlation than is present in the population.

The third hypothesis was for a primacy effect in impression formation: that in the interpretation of a sequence of experiences, early evidence contributes more to the impression than later evidence (Asch 1946, Tversky and Kahneman 1973). If a consumer is persuaded to try a brand and finds support for its advertised claims, this impression will, it is hypothesized, withstand subsequent experience that other brands deliver the same benefit.

The study tested for these sources of confirmatory bias. It examined how a sample of mothers made inferences about the breakfast cereal preferences of a group of children. The mothers saw advertising that built in them expectations about the childrens' cereal preferences, and then were shown data on the childrens' actual cereal choices. Although the data shown to the mothers neither supported nor contradicted the advertised claim, the mothers interpreted them as confirmatory when the hypothesized factors, positive confirmation and primacy, were experimentally manipulated. They also displayed a preference for evidence in which the advertised cue was present over evidence in which it was absent. Because the study dealt with advertising's effect on the inferences subjects drew from the behavior of others, some caution is needed in extrapolating this result to inferences from the performance of products.

CONCLUSION

This paper suggests one reason why persuasion may be delayed beyond the moment at which an audience is exposed to a message. It presents a model in terms of which message exposure merely establishes the salience of a particular interpretation of experience with a product. Actual experience, or at least some dispassionate evidence, must be available to the audience before the persuasive consequences of salience can be observed.

REFERENCES

Asch, Solomon E., (1946), "Forming impressions of personality". Journal of Abnormal and Social Psychology, V. 41.

Berkowitz, Leonard, and Macaulay Jacqueline, (1971), "The contagion of criminal violence." Sociometry, V 34.

Beyth-Marom, Ruth, (1982), "Perception of correlation reexamined." Memory and Cognition, V. 10.

Blackman, Julie A., and Harvey Hornstein, (1977), "Newscasts and the social actuary". Public Opinion Quarterly, V. 41.

Crocker, Jennifer, (1981), "Judgment of covariation by social perceivers". Psychological Bulletin, V. 90.

Deighton, John, (1983), Advertising as influence on consumers' use of evidence: The bias to confirm. Unpublished dissertation, University of Pennsylvania.

Deighton, John, (1984), "The interaction of advertising and evidence", Journal of Consumer Research, V. 11.

Deighton, John and Robert Schindler (1985), "Self-fulfillment of advertised claims", Working Paper, Amos Tuck School of Business Administration, Dartmouth College.

Ehrenberg, Andrew S.C., (1974), "Repetitive advertising and the consumer", Journal of Advertising Research,

Einhorn, Hillel and Robin Hogarth, (1978), "Confidence in Judgment: Persistence of the illusion of validity", Psychological Review, V. 85.

Harper, Paul C., (1976), "What advertising can and cannot do", Paper delivered at the Marketing Conference of the Conference Board, New York, October 20.

Hamilton, Davit L., and Terrence L. Rose, (1980), "Illusory correlation and the maintenance of stereotypic beliefs." Journal of Personality and Social Psychology, V. 39.

Lepper, Mark R., and David Greene, (1975), "Turning play into work: Effects of adult surveillance and extrinsic rewards on childrens' intrinsic motivation". Journal of Personality and Social Psychology, V. 31.

McCombs, Maxwell E. and Donald L. Shaw, (1977), "The agenda setting function of the press", in Donald L. Shaw and Maxwell E. McCombs (Eds.) The Emergence of American Political Issues: The Agenda-Setting Function of the Press. St. Paul: West Publishing Co.

Phillips, David P., (1976), "The influence of suggestion on suicide: Substantive and theoretical implications of the Werther effect," American Sociological Review, V. 39.

Phillips, David P., (1983), "The impact of mass media violence on U.S. homicides," American Sociological Review, V. 48.

Reyes, Robert M., William C. Thompson, and Gordon H. Bower, (1980), "Judgmental biases resulting from differing availability of arguments," Journal of Personality and Social Psychology, V. 39.

Ross, Lee, and Craig A. Anderson, (1982), "Shortcomings in the attribution process: On the origins and maintenance of erroneous social assessments", in Daniel kahneman, Paul Slovic and Amos Tversky, Judgment Under Uncertainty: Heuristics and Biases. Cambridge: Cambridge University Press.

Schindler, Robert H., and Michael Berbaum, (1983), "The influence of salience on choice," Advances in Consumer Research, Volume X, Richard P. Bagozzi and Alice M. Tybout (Eds.), Ann Arbor: Association for Consumer Research.

Smedslund, Jan, (1983), "The concept of correlation," Scandinavian Journal of Psychology, V. 4.

Sutherland, Max, and John Galloway, (1981), "Role of advertising: Persuasion or agenda setting?". Journal of Advertising Research, V. 21.

Taylor, Shelly E., and Susan T. Fiske, "Salience, attention and attribution: Top-of-the-head phenomena", in Leonard Berkowitz (Ed.), Advances in Experimental Social Psychology, V. 11, New York: Academic Press.

Tversky, Amos and Daniel Kahneman, (1973), "Availability: A heuristic for judging frequency and probability", Cognitive Psychology, V. 5.

Ward, William C., and Herbert Y.. Jenkins, (1965), "The display of information and the judgment of contingency," Canadian Journal of Psychology, V. 19.

Wyer, Robert S., Jr., and Thomas R. Srull, (1981), "Category accessibility: Some theoretical and empirical issues concerning the processing of information". In E. Tory Higgins, C. Peter Herman and Hark P. Zanna (Eds.), Social Cognition, Volume 1. Hillsdale, N.J.: Erlbaum.

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