Inferential Belief Formation: an Overlooked Concept in Information Processing Research

Philip A. Dover, Amos Tuck School, Dartmouth College
ABSTRACT - A study examined the role of inferential beliefs in the formation and modification of cognitive structure toward a product. After careful observation of existing knowledge structure for the trial product (cold, ready-to-eat cereal), subjects were exposed to limited salient or nonsalient information about a new brand of cereal. Hypotheses were based on the differential impact of these pieces of information on evolving cognitive structure, with emphasis on the extent and nature of the inference process.
[ to cite ]:
Philip A. Dover (1982) ,"Inferential Belief Formation: an Overlooked Concept in Information Processing Research", in NA - Advances in Consumer Research Volume 09, eds. Andrew Mitchell, Ann Abor, MI : Association for Consumer Research, Pages: 187-189.

Advances in Consumer Research Volume 9, 1982      Pages 187-189


Philip A. Dover, Amos Tuck School, Dartmouth College


A study examined the role of inferential beliefs in the formation and modification of cognitive structure toward a product. After careful observation of existing knowledge structure for the trial product (cold, ready-to-eat cereal), subjects were exposed to limited salient or nonsalient information about a new brand of cereal. Hypotheses were based on the differential impact of these pieces of information on evolving cognitive structure, with emphasis on the extent and nature of the inference process.


Fishbein and Ajzen (1975) suggested that three different processes may underlie belief formation. First, a perceived relationship between the object of belief (o) and some object (x) may be actively established on the basis of direct observation (descriptive belief). Second, a link between o and x may be perceived from information obtained from an outside source, and this relationship may be accepted (informational belief). Third, a link between o and x may be established through a process of inference from some other belief about o (inferential belief).

Although evidence from the social psychology literature indicates that inferential beliefs are important in the formation of impressions and attitudes (e.g., Jaccard and Fishbein, 1975), few consumer behaviorists have mentioned these cognitive "spill over effects" in their research (for exceptions, see Lutz, 1975; Mazis and Adkinson, 1976; Olson, 1978; Olson and Dover, 1978). Moreover, no consumer researchers have explicitly examined beliefs formed via the inferential process nor how these inferential beliefs influence other components of cognitive structure (i e., attitudes and behavioral intention). This observation raises serious doubts about the completeness of our treatment of the basic expectancy-value theory and has clear implications for information processing and communications research.

If we accept the importance of inference formation in cue utilization, how good are people as intuitive information processors? Some controversy exists here. Based on mathematical probability theory, various models have been employed to account for relationships between subjective probabilities or beliefs (e.g., McGuire, 1960, Wyer and Goldberg, 1970). These models are normative in that they prescribe what relations should exist between probabilities or how beliefs ought to change in light of new information. Although probability models describe the belief structures of "rational" or "logically consistent" persons rather than those of actual persons, it has been suggested they can be used as first approximations for a psychological theory of inferential belief formation. Here deviations from the normative models indicate that such "nonrational" factors as the person's attitudes or personality characteristics may influence his probabilistic judgments or beliefs. Wyer (1974) reports reasonably strong relationships between his model predictions and independent ratings of attribute associations provided by respondents. Similarly, Fishbein and Ajzen (1975) report a large number of studies that show Bayes's theorem as a reasonably good descriptive model of human information processing. In other words, there is a strong tendency not only for beliefs to be internally consistent but also for people to revise their beliefs in an orderly fashion as a result of new information. On the other hand, a number of judgment researchers (e.g., Slovic and Lichtenstein, 1971; Kahneman and Tversky, 1972) have revealed people to be "quite inept at all but the simplest inferential tasks . . . muttling through a world that seems to let them get through life by gratuitously allowing for a lot of error". Tversky and Kahneman (1974) feel that people rely on a limited number of heuristic principles by which they reduce the complex tasks of assessing likelihoods of uncertain events and predicting their values to simpler judgmental operations. Such heuristics may prove quite useful, but sometimes lead to severe and systematic errors.

It is clearly not possible to answer such an intriguing conundrum in a single paper or research project. Many variables can be suggested (e.g., cognitive complexity: product specific ego-involvement) that may influence the adoption of "rational" information processing. One such variable - information cue saliency -- is examined in this paper. Only a small part of the total research program is described here but it is hoped it provides a suitable stimulus to further study of inferential belief formation.


Devising Measurement Instruments

An explanation for inferential belief formation first requires a consideration of previously learned information upon which one can logically base, or from which one can derive, inferences. Measurement of relevant memory schema has been suggested (Olson, 1978) as a way to identify such knowledge representations. A method was sought that reveals individual consumer's knowledge as it naturally exists, stored in memory. After piloting a number of techniques, Kelly's Repertory Grid (1955) was selected as the test instrument. This was done as the output of the grid, in the form of personally salient product characteristics at a number of discriminable levels, was easily transferred to a structured questionnaire, allowing precise response scoring to create a base for predicting subsequent inference processes. A full description of this and other measures devised for this research are contained in Dover (1980).

The next question was how to expose respondents to information about a new brand of breakfast cereal and best observe the effect on the belief structure towards this new product. Consequently, three linked techniques were used to measure inferential belief formation, each technique being successively more structured. These were as follows:

Cognitive Responses-->Implications Grids-->Cognitive Structure

Immediately after exposure to the single variable information, subjects were asked to write all of their thoughts that occur during exposure. These thought protocols are assumed to indicate the internal, sub-vocal responses that take place during message reception. As such, they should represent the process activated as incoming information reacts with retrieved memory schema (e.g., whether there is a critical or favorable orientation to the external facts) and offer an indication of the mediating factors at work in the formation of new beliefs or changes in existing beliefs. Of particular interest in this study was whether cognitive responses were generated that relate to issues not directly contained within the limited message and thus help explain and predict the inferential belief process.

Following this, implications grids were conducted using the bi-polar approach advocated by Fransella (1972). Respondents were provided with one piece of information (e.g. a new brand of cereal is pre-sweetened) and asked to indicate which other characteristics, from those elicited at the earlier Rep Grid, they would expect to find in the new brand. Observation of the number of implications resulting from exposure to any single information cue are a simple index of the extent of inferential belief formation given knowledge of that cue.

In order to better quantify this inference process, the cognitive structure questionnaire, formulated at the first stage,was readministered. Again the only constructs used were those elicited at the original Rep Test. It was felt that this multiple measure approach to inference formation permits cross validation of research results


Salient attributes were defined as those elicited during the Repertory Grid interview. Pilot tests suggested two breakfast cereal attributes, both related to nutritional considerations, for use in the study. All respondents identified the sugar content of cereal as a product construct while only one mentioned vitamin content or the level of vitamins as a point of similarity or dissimilarity between cereals. Hence, sugar content was used as the salient feature while vitamin level became the non-salient piece of information.

Sixty homemakers, aged 18-49, were recruited to participate in the two-stage study. To be eligible they had to have purchased and used RTE cereal in the home in the past three months. Thirty respondents were pre-sweetened cereal users while the remaining thirty did not currently use pre-sweetened cereal.

Each respondent first undertook the Repertory Grid test in which memory schema components for the generic product category of breakfast cereal were elicited. These idiosyncratic constructs were used for each individual for the remainder of the experiment. About two weeks after this first stage, respondents were exposed to salient and nonsalient information about a new brand of breakfast cereal (i.e., it is pre-sweetened or vitamin fortified) and the successive inferential belief measures were administered. This allowed comparison of the cognitive structure for the generic product category with that emerging for the new brand(s) of breakfast cereal.


The measures of inference progressed from highly qualitative (cognitive responses) to highly quantitative (cognitive structure) in a hopefully fairly nonreactive way. Hypotheses were based mainly on the premise that a greater amount of inferential activity would result following exposure to a limited amount of highly salient information than following a less salient disclosure. This contention was generally well supported. As there is only space to describe one of the measures, findings from the implications grid are shown below.

To recall, subjects were shown cards containing each discriminable product attribute level elicited from the Repertory: Grid application. They were asked to imagine that they knew only one thing about a new brand of breakfast cereal (e.g., "it's pre-sweetened"). Then they indicated, from the cards shown, which characteristics they would expect to find in the new brand, given the limited knowledge they possessed. Replies were recorded on a special scoring sheet. Implications made were presumed to reflect the formation of inferential beliefs. Results are shown in Table 1. Because no difference in response was evident between pre-sweetened and nonpre-sweetened cereal users, findings are shown for the total sample only.



The cue saliency hypothesis was strongly supported. Subjects were able to imply or infer the existence of more associated attributes following the pre-sweetened (highly salient) than the vitamin fortified (less salient) message (4.18 vs. 2.69 for memory schema implications).

Table 1 also presents the proportion of the constructs elicited from the original Repertory Grid test that were implied to exist in the new brands, following exposure to high and less salient information. A higher proportion of the memory schema constructs were inferred following the pre-sweetened message (68%) than after the vitamin fortified message (44%). This suggests that subjects were not merely responding to the implications grid task by simple "yea saying" but perceived a genuine difference in the inferences from each message.

The cognitive response and cognitive structure findings will be discussed in future papers. It is interesting to note, however, that cognitive responses provide a useful indicator of inferential belief formation. Although all responses influence belief formation, it is felt that certain types do so more directly than others. That is, some responses make specific reference to product characteristics involved in belief modification or formulation. As such, counter- and support arguments, and perhaps curiosity statements, suggest strong inferential activity. Similarly, the various cognitive structure measures give an introductory view of the compositional nature of emerging beliefs, adding data on strength and confidence to that on extent of belief formation. These measures of inference are rather simple. But they do not show how orthogonal methods can provide convergent evidence in the study of an important yet little considered area of cognitive research -- inferential belief formation.


A few concluding words should be said about the policy implications resulting from the study of inferential beliefs. The importance of considering the role of inferences in the development of communications should not be doubted, given the strong evidence of inferential activity in the present study.

It is argued that in creating a persuasive communication the following steps should be undertaken: a) identification of the content and structure of topic specific memory schema, b) isolation of the attribute(s) to be used in the message, c) measurement of the information and inferential beliefs formed as a result of such information exposure. Differential impact of such variables as message, vehicle and receiver effects, held-advocated belief discrepancy, cue saliency, and extent of prior product experience should be carefully considered.

In a broad sense the study of inferential belief formation should add considerable insight into the topics of deceptive and corrective advertising. Extensive debate has centered around a definition of deception (cf. Gardner, 1975; Olson and Dover, 1978), with much concern relating to what information to consider. Some have argued that only explicit statements should be judged while others claim that what the consumer implies from incomplete information is just as important. This study provides considerable evidence of inferential workings in information processing and underlines the point that advertisers must be prepared to accept responsibility not only for what they do say but also what they leave unsaid. The application of this knowledge should greatly assist policy makers in measuring and identifying false beliefs and help manufacturers overcome consumer skepticism about advertising by avoiding misleading communications.


Dover, Philip A. (September 1980), "Inferential Belief Formation in Cue Utilization," Proceedings of the American Psychological Association, Montreal.

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

Fransella, Fay (1972), Personal Change and Reconstruction: Research on a Treatment of Stuttering (London: Academic Press).

Gardner, David M. (January 1975), "Deception in Advertising: A Conceptual Approach," Journal of Marketing, 39, 40-46.

Jaccard, James J., and Fishbein, Martin (1975), "Inferential Beliefs and Order Effects in Personality Impression Formation," Journal of Personality and Social Psychology, 31, 1031-1040.

Kahneman, Daniel, and Tversky, Amos (1972), "Subjective Probability: A Judgment of Representativeness," Cognitive Psychology, 3, 430-454.

Kelly, G. A. (1955), The Psychology of Personal Constructs, (New York, Norton).

Lutz, Richard J. (July 1975), "First-Order and Second-Order Cognitive Effects in Attitude Change," Communication Research, 2, 289-299.

Mazis, Michael B., and Adkinson, Janice E. (May 1976), "An Experimental Evaluation of a Proposed Corrective Advertising Remedy," Journal of Marketing Research, 13, 178-183.

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

Olson, Jerry C. (1978a), "Theories of Information Encoding and Storage: Implications for Consumer Research." in A. A. Mitchel (ed.) Effects of Information on Consumer and Marketing Information (Chicago: American Marketing Association).

Olson, Jerry C. (1978b), "Inferential Belief Formation in the Cue Utilization Process," in H. Keith Hunt (ed.) Advances in Consumer Research, Vol 5, (Provo, UT: Association for Consumer Research).

Olson, Jerry C. and Dover, Philip A. (1978a), "Attitude Maturation: Changes in Related Belief Structures Over Time,X' in H. R. Hunt (ed.) Advances in Consumer Vol. 5 (Ann Arbor, MI: Association for Consumer Research).

Olson, Jerry C. (1978b), "Cognitive Effects of Deceptive Advertising," Journal of Marketing Research, 15, pp. 29-38.

Slovic, P., and Lichtenstein, S. (1971), "Comparison of Bayesian and Regression Approaches to the Study of Information Processing in Judgment," Organizational Behavior and Human Performance, 5, pp. 649-744.

Tversky, Amos, and Kahneman, Daniel (1974), "Judgment Under Uncertainty: Heuristics and Biases," Science, 185, pp. 1124-1131.

Wyer, Robert S., and Goldberg, Lee (1970), "A Probabilistic Analysis of the Relationships Among Beliefs and Attitudes," Psychological Review, 77, pp. 100-120.

Wyer, Robert S. (1974), Cognitive Organization and Change: An Information Processing Approach, (Potomac, MD: Lawrence Erlbaum Associates).