Inferential Belief Formation Through the Use of Non-Information: an Example

David W. Finn, Texas Christian University
ABSTRACT - This paper reports on a surprising source of consumer in-formation--the absence of information about salient attributes. This finding adds another dimension to research into the inference process. Not only do consumers combine information from the message, source, etc., to form beliefs, but they also seem to use information that is conspicuously absent.
[ to cite ]:
David W. Finn (1981) ,"Inferential Belief Formation Through the Use of Non-Information: an Example", in NA - Advances in Consumer Research Volume 08, eds. Kent B. Monroe, Ann Abor, MI : Association for Consumer Research, Pages: 344-348.

Advances in Consumer Research Volume 8, 1981      Pages 344-348

INFERENTIAL BELIEF FORMATION THROUGH THE USE OF NON-INFORMATION: AN EXAMPLE

David W. Finn, Texas Christian University

ABSTRACT -

This paper reports on a surprising source of consumer in-formation--the absence of information about salient attributes. This finding adds another dimension to research into the inference process. Not only do consumers combine information from the message, source, etc., to form beliefs, but they also seem to use information that is conspicuously absent.

INTRODUCTION

Perhaps the strongest single influence on consumers' behavior is the information they have about various choice alternatives--that is, the beliefs they possess about the alternatives on important dimensions. In fact, many attempts to model consumer behavior are structured around consumers' information and experience. For example, the Engel, Kollat, and Blackwell (1978) model discusses how information and experience are updated and changed (Information Processing), how information and beliefs are used to form attitudes and intentions toward alternatives (Product Brand Evaluations), and how existing levels of information influence the purchase process (Decision Process Stages). Similarly, the Howard-Sheth (1969) model places beliefs and brand experiences in crucial positions. Howard and Sheth offer explanations of how Perceptual Constructs mediate the effect of various inputs on beliefs (Brand Comprehension) and discuss how changes in beliefs related to brand knowledge and Choice Criteria influence attitude and intention. In summary, the central influence of beliefs on behavior is widely recognized.

If we accept these views that beliefs influence attitudes and intentions, and that attitudes and intentions, in turn, influence behavior, then the study of consumer belief formation and change becomes important in the study of consumer behavior. Three issues immediately present themselves regarding beliefs: 1) What are they? 2) What kinds are there? and 3) Where do they come from? The first two questions are handled rather easily. If all researchers can agree on a definitional basis for study, the explorational efforts into the third question will be more easily diffused.

Fishbein and Ajzen (1975), in reviewing and condensing much of the literature related to beliefs end attitudes, have offered the definition of beliefs as representing the information a person has about an object. They further instruct that a belief can be expressed as a person's subjective probability that some object (or brand) has some specific attribute. This view of what beliefs are has built into it a method of measuring beliefs--namely via probabilistic scales.

They describe three basic kinds of beliefs, classified according to the informational determinants associated with belief formation. Beliefs formed as a result of direct observation and experience are called descriptive beliefs. A consumer forms descriptive beliefs about different brands mainly through use experience with them. For example, if a person test drives an automobile and feels the comfort and experiences the roominess, s/he will come to form beliefs about the presence or absence of these attributes. Predictably, beliefs formed in this manner are held with a great deal of confidence. Very little research has been conducted on the nature or formation of descriptive beliefs although these beliefs are commonly tapped in various consumer behavior studies. Measures of recall, blind taste tests, and exploratory product concept tests all typically measure levels of consumer experience and preference that can be labeled "descriptive".

A second type of belief, one that is typically held with much less confidence, is an informational belief. Informational beliefs are formed by accepting information provided by an outside source. Most of the opinion formation literature, the literature on credibility and fear appeals, and other examples related to the work of the Yale group (Hovland, et al 1953), is concerned with the study of informational beliefs. In the consumer behavior literature, a study by Holbrook (1978) contains a clear example of informational belief formation. As is typical in message content research, Holbrook's study contained a message composed of belief statements related to the target beliefs that the message was meant to influence. For example, Holbrook listed characteristics of an automobile ventilation system, its acoustical insulation specifications, and its amount of leg room in order to influence receivers' beliefs about the comfort and roominess of the car in question. Clearly, before any change can take place in the target belief (comfort), the listener must first accept or reject the information about the ventilation system, etc. Belief links "A" in Figure A represent the location of this informational belief influence. The degree to which this information is accepted depends, in part, on the source, message, and channel performing the communication.

FIGURE A

ANTICIPATED INFLUENCE CHAIN FROM HOLBROOK (1978)

The third belief type is an inferential belief. Inferential beliefs are formed through the process of utilizing previously learned relationships. In the example above and in Figure A, consumers were expected to form beliefs about the comfort of the car based on their previous experience and knowledge that the ventilation system, insulation, and amount of leg room were all different aspects of the overall comfort of the vehicle. The message itself did not mention comfort, and formation of a belief that the vehicle is comfortable depends both on the level of acceptance of the message statements (informational belief formation) and on the inferences that receivers draw about the relation between comfort and these three attributes. (links "B" in Figure A). A successful change in this target belief will influence a consumer's attitude toward the car if comfort is a salient attribute. Figure A illustrates the influence links between message items and attitude.

Much of the knowledge we have of the inferential belief formation process is related to the literature on trait inferences started by Asch (1946). One of the more promising directions this literature has taken is toward the measurement of probabilistic relationships among beliefs--not only those beliefs related to message content, but also those beliefs residing in memory that can influence further belief formation. An early study of these probabilistic relationships is McGuire's (1960) analysis of logical syllogisms. Further work has been done by Wyer (1970; Wyer and Goldberg 1970), and an excellent positioning of this literature into the realm of consumer research was offered by Olson (1978). The Holbrook study is an important step in measuring these relationships in a consumer behavior setting.

IMPACT EFFECTS

Not all inferential beliefs are formed due to planned influence by a communicator. Sometimes listeners take information from the communication and combine it with previously learned material in a completely unexpected manner. The result is inferential belief formation due to an unexpected impact effect. A good example of an impact effect is found in the corrective advertising study by Mazis and Adkinson (1976).

In their study, Mazis and Adkinson compared the effects of four different advertising messages about a brand of mouthwash. Three of the messages contained a germ killing appeal, while two added a corrective statement that "colds and sore throats could not be prevented, cured or benefited" by the use of the brand. Additionally, the corrective versions varied by apparent source of the corrective statement. The messages were:

1) noncorrective copy with germ killing appeal,

2) FTC required corrective copy with germ killing appeal,

3) corrective copy with germ killing appeal (no source mentioned), and

4) noncorrective copy with no germ killing appeal

All four versions made claims about the brand's usefulness in preventing bad breath. Version #1 differed from version #4 by also emphasizing the brand's germ killing ability. Version #3 differed from #l only in the additional corrective copy that the brand could not prevent colds and sore throats. Finally, advertisement #2 added the phrase, "The previous statement was required by the Federal Trade Commission." This systematic inclusion of additional message statements allows us to measure belief change as influenced by incremental information. Table 1 repeats a portion of the Mazis and Adkinson data. Since Mazis and Adkinson found no difference in beliefs between messages 2 and 3, message #2 is deleted.

TABLE 1

BELIEF MEANS AND STANDARD DEVIATIONS FOR EXPERIMENTAL ADVERTISEMENTS   -  [FROM MAZIS AND ADKINSON (1976)]

Comparing messages 1 and 4, we can isolate the effect of the germ-killing appeal. Because of the large difference in the sample variances, an approximation to the t-test must be used with an adjustment to the degrees of freedom (see Winkler and Hays, 1975 pg. 372). On a seven point scale (1 equals "probable", 7 equals "improbable"), the 23 respondents exposed to massage #4 reported an average rating of 2.39 that the brand kills germs. The average rating for the 19 respondents exposed to message #1 was 1.42. The t-test approximation (28 d.f.) reveals that this difference is significant beyond the .025 level. In other words, the germ killing appeal, alone, had a positive influence. As expected, there is no significant difference on the "cold prevention" belief.

The influence of the corrective statement can be measured by comparing messages 1 and 3. Since they differ on only this piece of information, we should expect to see changes in only this belief. Table 1 reveals that this is not the case! Large and significant changes are present on the "cold prevention" belief (as expected), but large changes are also present in the "kills germs" belief. Introduction of the corrective statement had an unexpected impact on the belief that the brand kills germs.

Even though message #3 emphasized the brand's germ killing properties, an audience exposed to it had significantly weaker beliefs that this was true than an audience exposed to the same message without the corrective statement (message #1). Figure B illustrates what may have happened:

1)  as shown in Figure B, the germ killing emphasis of the message positively influenced receivers' beliefs that the brand kills germs (link "Al" of Figure B and the comparison of messages 1 and 4) while the corrective statement positively influenced their belief that it does not cure colds (link "A2"); and

2)  the belief that the brand does not cure colds was inferentially related to the belief that the brand kills germs and impacted negatively on that belief (link "B" of Figure B).

Depending on the relative sizes of the positive and negative influences, the resulting "kills germs" belief could be stronger, weaker, or the same as the "kills germs" belief of an audience exposed to only message #4 (no mention). However, because of step #2 above, we would expect every audience exposed to message #3 to have weaker beliefs that the brand kills germs than an audience exposed to message #1. The Mazis and Adkinson data support this idea.

FIGURE B

APPARENT INFLUENCE CHAIN FROM MAZIS AND ADKINSON (1976)

This example of an unexpected impact effect emphasizes the need for understanding the experience and belief systems of potential consumers before directing persuasive communications at them. Where the Holbrook study showed that certain inferential processes can be expected, the Mazis and Adkinson paper can be interpreted as warning that inferential relationships can work against the intentions of the communicator. Both of these analyses demonstrate inferential belief formation due to the interaction between previously learned relationships and message-related beliefs--one planned and one unplanned. It is a logical step to expect that inferential beliefs might also result from an interaction between previously learned relationships and perceptions of the communication source or channel (e.g. one learns the characteristics associated with "trust" and forms message-related beliefs that vary in strength depending upon how the source matches those characteristics). We might conclude that the inferential process can distort any massage-related belief or can use message-related beliefs in combination with any other learned material to create new beliefs. This logic, however, does not lead to any predictions concerning the absence of information. This paper presents evidence that impact effects resulting from the inference process can also occur in situations where certain information (particularly about salient product attributes) is totally absent.

PROCEDURE

Two hundred and thirty-one junior and senior business students at a large Northeast university participated in the study. A posttest-only control group design was used (Campbell and Stanley, 1963), with one hundred and one of them serving as a control group. The others were exposed to an advertising-like message about a fictitious brand of toothpaste. The cautions voiced by Ferber (1977) were exercised in choosing the experimental product. The main consideration that guided the choice of the product was that subjects should have use experience with it and be the ones to make purchase decisions for it. Toothpaste was the ideal choice.

The use of an unknown brand was necessary because the study was concerned with belief change. Since beliefs about a brand are a function of information a person has about that brand, using an unknown brand allowed the assumption that the initial, or pre-exposure beliefs of both the control group and the experimental group were based on the same information (little or none). Belief change could then be computed from the control group's position.

Accordingly, a cover story was invented that stated that a local professor (named) was engaged in consulting with a well known market research company (named) that specialized in test-marketing. The brand was said to be manufactured by a large, well known manufacturer (named) and to be in test market in the Mid-West. The members of the control group were told that the purpose of the questionnaire was to find out what their beliefs were concerning the new brand of fluoride toothpaste - - without any other information. They were to answer the questions using their knowledge and expectations about toothpaste in general.

After the cover story, the experimental group was told that the consultant had access to advertising information and that they were about to be shown the voice script of a television advertisement for the toothpaste. They were allowed to read the script (the experimental treatment) as many times as they wanted, but were not allowed to return to it once they started to answer later questions. The belief questions were the same as the control group's. The message is printed as Table 2.

TABLE 2

THE EXPERIMENTAL MESSAGE

To measure subjects' beliefs about the brand, questions were asked about the probability of the brand being associated with various attributes. A seven-interval scale anchored by the phrases "very improbable" (1) and "very probable" (7) was used to record the response. Changes due to the measure were measured from the control group mean.

TABLE 3

MEAN BELIEF CHANGES FROM CONTROL GROUP MEANS

An earlier questionnaire had revealed that the salient attributes of toothpaste desired by members of the subject pool included; fluoride content, cavity prevention, low abrasiveness, whitening ability, good taste, the seal of approval of the American Dental Association, and a "reasonable" price. Many of these attributes ware specifically claimed by the advertisement (Table 2) but no mention was made about the price or the seal of approval. It was expected that any message effects would be only on beliefs about those attributes specifically mentioned in the message (the "A" links) or on beliefs that could be inferentially linked to message content (for example, the "clean mouth" and "good taste" claims in the advertisement could result in a belief about "mouthwash effect"). Table 3 lists the beliefs measured. The beliefs above the dashed line are directly related to message elements; beliefs below that line were not contained in the message, but were measured to study possible impact effects.

RESULTS

The results of the analysis are contained in Table 3. The first belief (made with a new type of fluoride) was the belief most dramatically affected by statements in the message. The control group (with no information) had an average belief of 3.01 on the seven interval scale. Members of the treatment group increased their belief to 4.31 on this (P < .001). Other changes in message related beliefs were less dramatic with only the third and fourth beliefs changing significantly.

Although changes in these early beliefs in the change chain from message to attitude would be necessary, if targets lower in the chain (e.g., attitudes) were expected to change, the focus of this paper is on changes in beliefs not related to the message. The belief that the brand "works like a mouthwash" probably is inferentially related to message beliefs, and changes in it can be expected through the same processes as in the Holbrook paper. The last two beliefs however, appear to be unrelated to any message elements, yet change in them was among the largest of any belief changes. The message said nothing about the presence or absence of the seal of approval of the American Dental Association, yet the treatment group had a significantly weaker belief in its presence than the control group. A similar result is evident in the belief that the brand sells at about the same price as other brands.

Observation of the control group means revealed that, based simply on their past experiences with the toothpaste category, respondents had a fairly high expectation that any new fluoride toothpaste would possess the seal of approval and be priced in a competitive range. Unexpected impact effects of the message or the message environment changed these beliefs in a negative direction. Respondents exposed to the message did not report these strong beliefs.

SUMMARY AND CONCLUSIONS

The intent of this paper is to report on an interesting observation of impact effects of an advertisement. Other impact effects may be present, but to discover all of them, a complete inventory of beliefs would have to be measured - -an impossible task.

Where did the large negative changes in beliefs about the seal of approval and price come from? Earlier in the paper it was demonstrated that impact effects occur on beliefs related to informational items in a message. It is very doubtful that "seal of approval" and "price" are related to informational items in the toothpaste message. However, it can be argued that they are related to information contained in the complete message environment. That is, the information processed by the treatment group included the observation that nothing was said about these two salient attributes. If this is the case, then the schema proposed by Olson (1978) is applicable to non-information as well as to information. That is, the absence of any mention of "seal of approval'' or "price" may have been perceived as a "cue" in the task environment, leading to the formation of a descriptive belief ("nothing was mentioned about ____"). This belief could then have combined with the expectation that a company would mention the attribute if the brand had it, resulting in an inference that the brand might or might not possess the attribute. Unfortunately, the results of this study were neither predicted nor expected, due to its exploratory nature, and the data were not collected in a manner that allows quantifying the possible relationships among beliefs. A tentative hypothesis is that "top of the mind" salient attributes are noticed in their absence as well as when mentioned.

The general findings of this study suggest the need for future research in at least two areas. First, what are the dynamics of inferential belief formation due to advertising? Holbrook (1978) attempted to explore this issue and has provided a meaningful framework for further study of the problem; the work of Mazis and Adkinson (1976) highlights the existence of unexpected impact effects related to inferential belief formation; and this study suggests that impact effects may result not only from the interaction of message elements, but also from the entire message environment.

The second area for future research evolves from the question, do consumers expect an advertiser to state product claims for those salient attributes for which he can make legitimate claims? This question has implications for identifying the more salient attributes of different product classes. That is, it may be that belief change as noted above is a good index of relative salience of various attributes. For example, if consumers make stronger inferences about salient than non-salient attributes, we could systematically leave out mention of various attributes and note the effect on beliefs. Those beliefs changing the most when not mentioned in a advertisement would then be identified as most salient.

Together, these two questions raise the issue, "what is the information content of non-information?"

REFERENCES

Asch, S. E.(1946), "Forming Impressions of Personality," Journal of Abnormal and Social Psychology, 41, 258-90.

Campbell, Donald T., Stanley, Julian C.(1963), Experimental and Quasi-Experimental Designs for Research, Chicago: Rand McNally.

Engel, James F., Blackwell, Roger D. and Kollat, David T. (1978), Consumer Behavior (3rd edition), Hinsdale, Illinois: The Dryden Press.

Ferber, Robert (1977), "Research by Convenience," Journal of Consumer Research, 4, 57-58.

Fishbein, Martin and Ajzen, Icek (1975), Belief, Attitude, Intention, and Behavior: An Introduction to Theory and Research, Reading, Massachusetts: Addison-Wesley.

Holbrook, Morris B.(1978), "Beyond Attitude Structure: Toward the Informational Determinants of Attitude," Journal of Marketing Research, 15, 545-56.

Hovland, C. I., Janis I. L. and Kelley, H. H. (1953), Communications and Persuasion, New Haven: Yale University Press.

Howard, John A. and Sheth, Jagdish N.(1969), The Theory of Buyer Behavior, New York: Wiley.

Mazis, Michael B. and Adkinson, Janice E.(1976). "An Experimental Evaluation of a Proposed Corrective Advertising Remedy," Journal of Marketing Research, 13, 178-83.

McGuire, William J.(1960), "A Syllogistic Analysis of Cognitive Relationships," in Hovland, C. I. and Rosenberg, M. (eds.), Attitude Organization and Change, New Haven: Yale University Press, 65-111.

Olson, Jerry C.(1978), "Inferential Belief Formation in the Cue Utilization Process," in Hunt, H. K. (ed.), Advances in Consumer Research, Vol. 5, Association for Consumer Research, 706-13.

Winkler, Robert L. and Hays, William L.(1975), Statistics (2nd edition), New York: Holt, Rinehart and Winston.

Wyer, Robert S., Jr.(1970), "Quantitative Prediction of Belief and Opinion Change: A Further Test of a Subjective Probability Model," Journal of Personality and Social Psychology, 16, 559-70.

Wyer, Robert S., Jr. and Goldgerg, Lee (1970), "A Probabilistic Analysis of the Relationships Among Beliefs and Attitudes," Psychological Review, 77, (no. 2), 100-120.

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