Attitude Maturation: Changes in Related Belief Structures Over Time

ABSTRACT - This paper presents some initial ideas about the dynamic changes which may occur within the belief structure associated with a stimulus concept. In particular, we focus on a phenomenon termed attitude maturation -- the observed tendency for the belief structure and overall attitude associated with a product to become increasingly related and consistent over time as knowledge about the product accumulates. An initial conceptualization of the attitude maturation phenomenon is offered along with several alternative explanations and these ideas are illustrated with data from a five-wave longitudinal experiment. The paper concludes with suggested theoretical perspectives derived from cognitive psychology.


Jerry C. Olson and Philip A. Dover (1978) ,"Attitude Maturation: Changes in Related Belief Structures Over Time", in NA - Advances in Consumer Research Volume 05, eds. Kent Hunt, Ann Abor, MI : Association for Consumer Research, Pages: 333-342.

Advances in Consumer Research Volume 5, 1978      Pages 333-342


Jerry C. Olson, Pennsylvania State University

Philip A. Dover, Dartmouth College

[This research was funded by a Research Initiation grant to the first author from the Pennsylvania State University. We are grateful to Rajesh Kanwar for his assistance with the data analysis and to James Bettman, Joel Cohen and Richard Lutz for helpful comments on a draft version of the paper.]

[Associate and Assistant Professor of Marketing, respectively. This study and paper were completed while the second author was a doctoral candidate in Marketing at the Pennsylvania State University.]


This paper presents some initial ideas about the dynamic changes which may occur within the belief structure associated with a stimulus concept. In particular, we focus on a phenomenon termed attitude maturation -- the observed tendency for the belief structure and overall attitude associated with a product to become increasingly related and consistent over time as knowledge about the product accumulates. An initial conceptualization of the attitude maturation phenomenon is offered along with several alternative explanations and these ideas are illustrated with data from a five-wave longitudinal experiment. The paper concludes with suggested theoretical perspectives derived from cognitive psychology.


How are attitudes acquired? Are crystallized, stable attitudes formed "instantly" upon learning something of an object's existence and its characteristics? Or, do our attitudes develop (evolve or mature) more slowly as we accumulate information and cognitively consider that information? How long (how many experiences, how much information) does it take to form a stable, mature attitude -- and an associated stable, mature belief structure? Does the formation of beliefs through the acquisition of information directly and automatically lead to the formation of a stable (reliable) attitude that possesses clearly defined, stable relationships with associated salient beliefs and with other attitudes? Or, does this process take time and some degree of conscious, cognitive processing effort? This paper presents our initial thinking, accompanied by some illustrative data, regarding these and other questions involved with attitude formation, attitude change, and the dynamics of the relationship between attitude and belief structure.

The theoretical literature on attitude formation typically does not address such issues. Fishbein's attitude theory presents perhaps the clearest account of the attitude acquisition process (cf. Fishbein, 1963, 1967; Fishbein and Ajzen, 1975) and has been adopted or advocated by many consumer researchers (Ahtola, 1975; Cohen, Fishbein and Ahtola, 1972; Lutz, 1975; Olson and Mitchell, 1975). In this view, attitude, defined as uni-dimensional evaluation or affect, is a direct causal function of one's beliefs about associated concepts and one's evaluations of those perceived associations. Beliefs and their evaluative aspects are developed through learning principles of classical and instrumental conditioning and mediated generalization (Fishbein, 1967; Olson and Mitchell, 1975). Fishbein quantified the relationship between attitude (A) and belief structure (indicated by belief strength, bi and evaluation, ei) via the now familiar additive compensatory model:


Implicit in most attitude research which takes this perspective is the assumption that a completely formed and internally consistent cognitive structure (here consisting of an attitude and its related belief structure) is formed essentially immediately upon processing whatever information is involved (cf. Lutz, 1975, 1977). However, the questions posed above challenge this assumption. To provide answers to these questions a dynamic, longitudinal research approach is required in which changes in attitude, belief structure, and their interrelationships are monitored over time for a given group of subjects. Unfortunately, the vast majority of consumer attitude studies are static and cross-sectional and thus do not address these issues. And, the relatively few longitudinal experiments that do exist (cf. Ginter, 1974; Winter, 1974) typically do not report the types of data analyses required for a study of cognitive dynamics and thus neglect the issues identified above.

Our purposes in writing this paper are several. First, we wish to present and briefly discuss some interesting conceptual issues regarding the dynamics of cognitive structure--from an attitude perspective. Where appropriate, specific issues will be illustrated with data from our own research. It should be noted that the data reported here are intended to be only heuristic and illustrative--not conclusive--as they are derived from a longitudinal experiment originally designed for another purpose. Second, we wish to identify critical theoretical issues that seem to warrant future research attention. Broadly speaking, our goals are to arouse interest in issues of attitude structure dynamics and to encourage research specifically directed at such issues.



The common tendency in attitude/cognitive structure research to use concepts and terms somewhat loosely suggests that specific, explicit definitions of the major concepts used in this paper would minimize semantic misunderstandings.

Attitude. Consumer researchers appear to be gradually reaching a consensus that the construct of attitude is best considered as the unidimensional evaluation or affect (favorable-unfavorable, good-bad, like-dislike) associated with an object, event, action, or concept. This simple definition was adopted for the present study as the most easily justified theoretically and the most useful definition from a measurement standpoint.

Cognitive Elements. The term cognitive elements as used in this paper refers to several major cognitive variables, including beliefs (bi), the evaluative aspects of each belief (ei), attitude toward the concept of interest (attitude toward the object, Ao, or act, Aact), and behavioral intention (BI, in consumer research usually operationalized as purchase intention).

Cognitive Structure. An implicit assumption in most cognitive research is that all of the cognitive elements associated with a stimulus object or event are organized together in a more or less consistent (in some sense) way and this organized set of cognitions may be viewed as a structure. The most basic cognitive elements in a structure are the encoded representations of the stimulus object and its characteristics. These learned concepts are represented cognitively by symbolic codes assigned to the concept during the encoding process. Beliefs refer to the cognitive link between two concepts and belief strength is the perceived degree of probabilistic association between any two concept symbols (e.g., brand X and attribute A). Taken as a whole, the associations between these concepts may be considered to represent a belief structure as discussed below. The belief structure that is salient for a particular object/concept is thought to be related in a causal way to the overall attitude toward the concept of interest (see equation 1; Fishbein, 1967). In turn, attitudes (and perhaps other factors such as social norms) influence intentions to behave, which in turn are thought to exert a causal effect on overt behavior (see Ryan and Bonfield, 1975). Since concepts or codes, linking beliefs, evaluation/ attitudes, and intentions are all cognitive elements, the overall pattern of their interrelationships may be considered as partially representing the cognitive structure for a particular concept such as a brand. However, because this paper is primarily concerned with attitudes, we shall deal mainly with belief structure interrelationships and belief-attitude relationships.

Belief Structure. The various salient and non-salient beliefs associated with an attitude concept can be considered as a structure of interrelated beliefs. That is, each belief concept is associated with the attitude concept and may also be "interconnected" with other belief concepts. The overall evaluation associated with a belief structure may be quantified by summing the cross products of belief strength and belief evaluation, Sbiei (cf. Lutz, 1975; Olson and Dover, in press). This belief structure evaluation should influence (perhaps cause) the evaluation or attitude associated with the overall concept.


The study which produced the data used here for illustrative purposes was designed to investigate another phenomenon--disconfirmation of expectations. The relevant methodology has been reported in sufficient detail elsewhere (Dover and Olson, 1977; Olson and Dover, 1976) and therefore is only briefly reviewed below.

Two treatment groups were randomly created from among 38 adult housewives. Group A (n = 20) received five exposures to information about a new, unfamiliar (actual) brand of ground coffee. The first three informational stimuli were three written, ad-like messages each of which stated (in different ways) that the coffee had absolutely no bitterness. No information about other coffee attributes were contained in any of the ads. Following the third message, consumers actually tasted a coffee sample intentionally prepared to be bitter and thus to disconfirm pretrial expectations. Then, to provide an extended period of product experience (and extensive product information) consumers received a free one-pound can of the coffee to try in their homes for about 10 days. The first four stages were each separated by approximately four days. Following each of the five stages, the major cognitive elements (associated with this brand of coffee) were measured. In sum, the five experimental stages for Group A can be seen as five opportunities for consumers to acquire information about a previously unknown product, with at least some of the information likely to be inconsistent with the remainder.

Consumers in Group B (n = 18) received no pretrial information regarding the coffee or its level of bitterness. These consumers merely tasted the purposefully bitter coffee and then took a one-pound sample home for an extended trial period. Thus, the cognitive structures of group B subjects were measured twice. Consumers in this group had two opportunities to acquire information about the unfamiliar brand--both involving direct trial experiences with the brand.

For purposes of the present paper, this longitudinal experiment allows a close examination of (a) the dynamics of the belief structure-attitude relationship and (b) changes in the interrelationships within belief structure, as a function of multiple, sometimes inconsistent product information. The relationship between beliefs and attitude was modeled by the Ahtola (1975) adaptation of the Fishbeinian expectancy-value model (see Dover and Olson, 1977). [This study adopted Ahtola's vector model, which separates belief strength from belief content, in order to more precisely monitor belief structure changes. In this model, a specific product attribute is reduced to a vector of discriminable levels or amounts of the attribute. Measures of belief strength and evaluation are taken for each attribute level. Summing the belief strength-evaluation cross products (Sbiei) within a vector yields an index of evaluation for that attribute. Summing across vector scores (SSbiei) yields the composite evaluative index for the entire belief structure.]


Magnitude of Experimental Effects

The experimental treatments, as expected, had significant effects on several cognitive elements, particularly on beliefs about the degree of coffee bitterness (see Olson & Dover, 1976, 1977). A substantial amount of "movement" over time was produced in the various cognitive variables by the exposures to the experimentally controlled information (see Figure 1, Dover and Olson, 1977). Our present interest, however, is with the dynamics (trends, shifts, or fluctuations) involved in the interrelationships between cognitive variables.

Major Interrelationships Among Cognitive Elements

One relationship of major interest to an attitude theorist involves the degree of correspondence between beliefs and overall attitude. As discussed previously, belief structure evaluation can be operationalized by the simple model, SSbiei. Column 1 in Table 1 contains the results pertaining to the validity of the basic Fishbein/Ahtola notion that belief structure is strongly and causally related to attitude (SSbiei = Ao). [Attitude toward the object (the coffee brand) was measured by the mean of three measures of evaluation --good-bad, like-dislike, and high quality-low quality--each measured on 5-point bi-polar scales.] A steady increase [We were unable to find a procedure for statistically testing differences between correlations of the same variables measured at different points in time for the same subjects.] in the strength of relationship between the belief structure index and attitude was obtained for Group A, increasing from an r2 of .14 upon first learning something about the product to .61 after three weeks of varying experiences with the product. It is this effect that we have termed the attitude maturation phenomenon. The change in the attitude-intention relationship was even more dramatic, steadily increasing to a correlation of .95 (r2 = .90) by the study's end. Group B also evidenced increases (although less dramatic) in the strengths of both relationships over the extended trial period (see Dover and Olson, 1977, for a more complete analysis and discussion of these results from an Expectancy value model perspective).



Interpretation. Despite the relatively small samples involved, at least compared to most correlational studies, increases of this magnitude and consistency in theoretically predicted relationships aroused our interest. A number of rather simple explanations for these dynamic effects were considered and tentatively judged unable to account for the effect (see Dover and Olson, 1977). For instance, increasing model correlations were not caused by a steady monotonic increase in the magnitude of the individual cognitive elements which might in turn produce increasing consistency between the cognitions. In fact, the variables fluctuated in magnitude across the experimental stages as would be expected given the conflicting information at stage 4 (see figures in Dover and Olson, 1977; Olson and Dover, 1976). However, the observed variations of the manipulated bitterness belief, belief structure, attitudes, and intentions were relatively consistently related (see figure 1, Dover and Olson, 1977). That is, although cognitive changes were not monotonic over time, the change patterns for individual cognitive elements were similar.

An alternative explanation for the attitude maturation effect is that the observed increases in correlations were due to the repeated measurement of consumer subjects. However, this simple statement does not explicate precisely how the multiple measurements could have caused the correlation increases. For instance, it seems unlikely that subjects remembered their previous responses and/or consciously modified their responses to create a more consistent cognitive structure over time. The number and complexity of the measures and the relatively long intertrial time (at least four days) argue against this explanation. However, repeated, indeed any, measurements of cognitive variables certainly affect the measured cognitive structure and presumably also influence the relationships between cognitive elements. Thus, questions remain as to exactly what these measurement effects are and how they operate.


A more conceptually intriguing explanation for the observed attitude maturation effect concerns the dynamic changes that may have occurred within the belief structure underlying the product attitude. Using terms somewhat loosely at this point, the belief structure may become "clearer" (to the subject), more stable and reliable, more internally consistent--in other words, increasingly mature--and, as a consequence, more strongly associated with overall attitude. Such changes would not necessarily be evidenced in the absolute magnitudes or levels of specific cognitive variables, such as increasing belief strength, but rather might be more subtly represented by changes in the interrelationships between cognitive variables. In the following sections we discuss these possibilities and propose specific, theoretically based explanations regarding the belief structure changes, and illustrate their viability with data from the experiment described above.

How is Information Structured and Stored?

It has been common in cognitive psychology and to a lesser extent in consumer research to discuss cognitive structure in terms of specific memory stores; e.g., sensory, short-term, intermediate, and long-term memories (see Chestnut and Jacoby, 1977 or Olson, 1977, for overviews). This approach led to rather mechanistic theories of information transfer between these structural "warehouses" via the "forklift trucks" of rehearsal or various retrieval processes. Currently these notions are being rejected for a less rigid concept of memory structure. In this view, memory or cognitive structure takes form as information is encoded and organized (cf. Estes, 1975; Hayes-Roth, 1977; Restle, 1974). That is, information enters memory and becomes a part of cognitive structure simply by having been encoded during initial processing activity.

From this view, cognitive structure involves the content and organization of knowledge as stored in memory. Because of the focus on the meaning assigned to processed informational stimuli and the relations between that meaning and other, previously acquired knowledge, this aspect of cognitive structure or memory has been termed semantic memory (see Tulving, 1972). In sum, semantic memory concerns not the structural aspects of the storage system, but the structural characteristics of the knowledge that is represented in memory.

The traditional theoretical paradigm to explain knowledge representation in cognitive structure has been associationist theory (e.g., Anderson and Bower, 1973). Associative models of memory typically assume that knowledge is represented as networks of nodes and connecting arcs, where the nodes encode specific discriminable concepts and the arcs between nodes encode the perceived relationships between concepts (see Figure l-a). Thus, cognitive structures can be modified by changing arcs (modifying the perceived association between two concepts) or by adding new concepts (and, of course, connecting arcs) to the cognitive structure.

Consumer attitude researchers will note the distinct similarity of these notions to their typical conceptualization of belief structures (Figure l-b). Lutz (1975), for example, discussed strategies for modifying belief structures (and subsequently, attitude) in analogous terms. In their ubiquitous measurement of consumers' belief structures for products and brands, consumer researchers have in fact been investigating aspects of consumers' knowledge structure in semantic memory. However, the usual focus in consumer behavior/marketing has been on predicting attitudes from such belief structures rather than on describing the structural characteristics of the knowledge representations associated with a product concept.



Given this perspective, this paper can be seen as an initial, exploratory attempt to examine consumers' semantic structure per se, particularly focusing on the belief structure and the dynamics of its relationship with attitude. Following are several possible (and probably only partial) explanations for the attitude maturation effect. These ideas focus on changes within the structure of salient beliefs regarding the attitude object. Where possible, each notion is illustrated by data analyses from the study described above.

Belief Stability

One explanation for the observed maturation effects is that beliefs may not be well-formed or stable or "held with conviction" until several "experiences" with the belief object have occurred. There may be some problem in developing measures of this "stability" or "conviction'' construct, since belief strength scores may not adequately assess such changes. That is, it seems possible that one could have a stable, clear belief that a particular concept is only weakly (or occasionally) related to another concept. In fact, belief strength scores for Group A did not increase (or decrease) consistently over the multiple stages (see Figures 1, 3, and 4 in Olson and Dover, 1976).

In the social and cognitive psychology literatures, researchers have taken independent measures of subjects' confidence in their belief ratings (cf. Wyer, 1973, 1974). Such confidence scores could be treated as direct evidence for the "clarity," stability, or degree of conviction of one's beliefs. In the present study, subjects were asked to rate their degree of confidence in the belief strength allocations for each of the five attribute vectors. [Confidence in the 10-point belief strength allocation to the levels along each belief vector was measured on a 5-point scale with bi-polar labels not-at-all confident and very confident.] Table 2 presents the resulting five sets of confidence scores for each experimental stage. Confidence in beliefs about each attribute increased over the five stages for group A (all p's < .10). [Due to the relatively small sample sizes (which decrease the power of statistical tests) and our greater tolerance for Type 1 errors and reluctance to commit Type 2 errors at this early state in our research, an alpha level of .10 was adopted for these analyses.] Similarly, confidence increased over the two stages for group B for all belief vectors except bitterness (p's< .10). However, group A's confidence after five exposures to product information was stronger than group B's confidence after two such exposures, for all attributes except the bitterness vector (p's < .10). Thus, it would seem that a relatively steady increase in belief clarity, stability, or certainty occurred with successive exposures to information about the belief object. The confidence data, therefore, support the notion that an increasingly well-developed, stable belief structure develops over time as one becomes more knowledgeable or familiar with the relevant object.

Structural Relationships Between Beliefs

Another perspective to belief structure dynamics is provided by the notion that the interrelationships between the beliefs associated with an object/concept develop rather slowly and require time and experience before they become clearly defined and stable. This idea reflects the commonly-accepted tendency toward consistency among cognitive elements, but suggests that consistency may evolve over time with successive experiences which "activate" the belief structure. It seems logically reasonable that a well-formed, internally-consistent belief structure is more strongly related to overall attitude toward the belief object than a less well-developed structure.

Note, however, that the basis for consistency among beliefs has not been specified. Broadly speaking, two types of cognitive consistency have been investigated in past research -- consistency in terms of evaluative reactions and consistency based on perceived logical, meaningful, or semantic linkages between cognitive concepts (a review of this area is provided by Fishbein and Ajzen, 1975, pp. 143-156). For example, we may develop belief structures composed of beliefs that are increasingly similar in evaluative content--e.g., all or most of a consumer's beliefs about this coffee become more positive (or more negative). Alternatively, beliefs may be consistent in a meaningful, semantic sense; for example, one belief connotatively implies another belief. Stated differently, beliefs may be consistently related because they "fit together" on a logical basis--their meanings are consistent with one's previously learned expectations. For example, if one learns that a winter jacket is filled with goose down, one might also logically develop beliefs that the jacket is lightweight, expensive, and casual in style. In this case the consistency within the belief structure would be measured by the probabilistic relationships between belief concepts (belief strength) and not necessarily by increasing evaluative consistency. In summary, one might postulate a "need" or tendency to acquire internally-consistent cognitive structures, the evolution of which could occur over time as a function of processes "driven" by goals of evaluative or semantic consistency, or both.



Evaluative Consistency

The following analyses concentrate on the relative changes in evaluative consistency among the belief elements in cognitive structure. Two indices of the evaluative consistency among the beliefs about the five product attributes were derived from the belief vector scores (Sbiei). These scores reflect the overall direction and degree of evaluation associated with each attribute. Estimates of internal consistency among these product attribute evaluation scores are provided by the coefficient alpha (a) index (see Nunnally, 1967, for a thorough discussion) and the average zero-order correlation between the five vector scores. Table 3 presents these two indices of evaluative consistency for both groups over the experimental stages. On both measures, group A evidenced a fairly dramatic increase in evaluative consistency after the second informational exposure, with relatively little increase in evaluative consistency thereafter. The evaluative consistency of belief structure for group B increased to a relatively low level over the two trial experiences.



Yet another perspective on the evaluative consistency within a belief structure is provided by examining the relationship of each attribute vector (Sbiei) with the independent, direct measure of product attitude. Changes in these relations over time provide additional evidence for dynamic shifts in evaluative consistency within the belief structure. Table 4 presents the correlations between each belief vector evaluation score (Sbiei) and the direct rating of overall product attitude (Ao) for group A (recall that "no bitterness" was the product attribute belief most directly manipulated by the experimental treatments). By the study's end for group A, the evaluation associated with bitterness beliefs was the most strongly related to Ao, r = .78(r2 = .61), followed by that associated with the amount-of-caffeine beliefs, r = .50(r2 = .25), and beliefs about the coffee's strength--of flavor, r = .47(r2 = .22). The expensiveness and consistency vectors were only weakly related to overall attitude (r2 = .08). A basically similar pattern of relationships was found for group B.

In sum, changes in evaluative consistency among salient beliefs seem to be somewhat related to the attitude maturation effect. It should be noted, however, that the present study probably attenuated the degree of evaluative consistency because the product information was not consistent over the course of the experiment. However, it does seem reasonable that as one's attitude structure "matures," one tends to acquire beliefs relatively consistent in terms of evaluation. A major problem with this explanation is in identifying sufficient and/or necessary amounts or degrees of evaluative consistency for a strong belief structure-attitude relationship. Threshold effects seem a distinct possibility. Such issues could be relatively easily explored in longitudinal correlational studies with larger samples.

Semantic Consistency. It is of interest to note that the three belief attributes most closely related to overall product attitude also seem closely related in a connotative or semantic sense as well. That is, an experienced coffee drinker might be expected to draw logical inferences about caffeine content and flavor strength after learning about a coffee's level of bitterness. There are probably relatively clear, semantically-based, expected linkages between these three attributes for experienced, knowledgeable consumers. In numerous studies, Wyer (cf. 1974, 1976) has quantified such semantic "connections'' by asking subjects to rate the probability of one concept (or attribute level) given knowledge about the likelihood of another concept (or attribute level).



Although such data was not collected in the present research, similar types of analyses could be done in consumer product judgments research, in effect yielding predictions of inferential belief formation, based upon conditional probability theory and analyses (see Olson, in press).

Summary. The confidence data (Table 2) and evaluative consistency data (Tables 3 and 4) support the inference that changes occurred within the product belief structure over the experimental stages. It seems logical that such changes in belief structure may cause (or at least partially influence) the increasing predictive validity for the Fishbein/Ahtola attitude model (see Table 1). Thus, the present analyses provide post-hoc support for the notion that the attitude maturation effect is due at least partially to changes occurring within the associated structure of salient beliefs. Clearly, however, before much confidence can be placed in these relationships and interpretations, similar measures and analyses must be included in subsequent research so that these initial findings can be replicated, or not.

Mediators of the Attitude Maturation Effect

To provide additional insight into the dynamics of the belief structure-attitude relationship as a function of multiple exposures to information, a final set of analyses was conducted to investigate possible mediators of such changes. This approach coincides with Chronbach's (1957, especially 1975) contentions that individual difference factors (variable characteristics of research subjects) tend to interact with experimental treatment variables in determining their effect on the dependent variable of interest, and thus should be explicitly examined. Similarly, Underwood (1975) has argued convincingly that researchers should make and test specific predictions of the mediating effects of individual difference variables in order to provide more rigorous tests of the theory of interest.

In anticipation of such use, several characteristics of the consumer subjects were measured in the study reported here (e.g., perceived product risk, coffee purchase volume). Because of their intuitive implications for belief structure dynamics, two of these variables were selected for presentation in this paper: (a) ego-involvement with the ground coffee product class, and (b) product specific self-confidence in evaluating and choosing alternative brands of ground coffee.



Ego-involvement. Consumers in group A were categorized either "higher" or "lower" in product ego-involvment based upon the relative number of coffee brands they sorted into the acceptable and neutral (noncommitment) regions for possible purchase. [This procedure is an adaptation of Sherif's ordered alternatives procedure for establishing involvement with an attitude issue. See Jacoby and Olson (1976) or Olson and Dover (1977) for detailed descriptions of the task and analysis.] The 10 subjects who placed one-third or less of the familiar brands into the acceptance and noncommitment latitudes (or, alternatively, who rejected two-thirds or more of the brands) were deemed higher in product class involvement than the other 10 subjects.

Table 5 presents the relevant cognitive states and cognitive interrelationships for these "high" and "low" involvement subjects. Of greatest interest are the correlations between belief structure and attitude. Although the "high" and "low" involvement subjects yielded essentially the same belief structure-attitude relationship at the study's conclusion (r's = .81 and .79, respectively), the more highly involved subjects reached that level of evaluative consistency more slowly. In fact, only after trial did the belief structures of the higher involvement group begin to become strongly related to overall attitude.

The explanation advocated for the lack of consistency between belief structure and attitude is that belief structures which are not well-formed, stable, or internally consistent may not be highly related to overall attitude. General support for this notion may be found by comparing the pattern of attitude model correlations for the "high" and "low" involvement subjects with the respective indices of evaluative consistency (see Table 5). Except for the data for the second stage (ad B), the evaluative consistency of belief structures is relatively low for those subjects with lower attitude-belief structure relationships, and vice versa for subjects with higher model correlations.

Product Specific Self-Confidence. A similar internal analysis was conducted by dividing group A subjects into two categories (n = 10 each) based on their relative self-confidence in making quality judgments of ground coffee. [Self-confidence was assessed by a simple question which asked subjects to rate on a 5-point bipolar scale the degree to which they were "confident" in their ability to evaluate the quality of alternative brands of ground coffee.] The product self-confidence variable also mediated dramatic differences in belief structure-attitude relationships (see Table 5). "Higher" confidence consumers had much greater consistency of belief structure and attitude than did subjects lower in confidence (r = .73 versus .25, respectively). However, the "internal evaluative consistency explanation" for this pattern of relationships was supported for only three of the five stages (2, 3 and 5).


The basic phenomenon of interest in this paper involves the cognitive structure changes termed attitude maturation--namely, the tendency for the salient belief structure associated with a stimulus concept to become increasingly predictive of the overall attitude towards that concept. Unfortunately, the present research design does not allow analysis of the unique effects of multiple measurement of cognitive structure (measurement and exposure to information were confounded), nor is it possible to completely eliminate the alternative explanation that the effect is caused by increasingly strong halo effects. [Note, however, that the two internal analyses yielded data that do not support the explanation of a steadily increasing halo effect. For example, the "high" involvement group produced strong belief structure-attitude relationships only after initial trial. And, the "high" confidence group produced relatively strong model correlations immediately upon the first informational exposure.] Rather, this paper focuses on the heuristic and more interesting (to us) idea that cognitive structures, including the interrelationships among cognitive elements, develop and evolve toward evaluative and/or semantic consistency as one accumulates knowledge about the stimulus concept of interest. The illustrative data were generally consistent with this explanation, although there were exceptions. However, these analyses provide no explanations of the processes that cause such changes to occur.

The concluding section identifies three theoretical ideas derived from the cognitive psychology literature that may provide useful perspectives regarding the causal processes underlying cognitive structure dynamics. It is suggested that these theories can be combined to form an initial conceptual framework for explaining changes in cognitive structure, and can generate a number of interesting hypotheses to guide future research.

Evolution of Cognitive Structure

Very little has been written in the basic cognitive psychology literature regarding dynamic changes in cognitive structure. Recently, however, Hayes-Roth (1977) presented a clear theoretical description of the evolution or maturation of cognitive structure from initial formation to a well-formed state. Her theory is essentially consistent with our interpretations above. Briefly, Hayes-Roth stated that an organized cognitive structure begins with the establishment of encoded representations of elementary knowledge (cogits), which become stronger (more stable?) as experience accumulates. Basically this is the concept formation process (see Olson and Mitchell, 1975). As experience increases, these simple representations become cognitively linked to other representations and these associations become stronger. This is analogous to the belief formation process described by Fishbein (1967) and an attitude theorist might operationalize such changes as increases in either belief strength or confidence in belief ratings. Through repeated activations of these cognitions (caused by multiple exposures to the controlling stimulus), configurations of associated representations develop and become stronger. At some point a configuration may become so well-learned that it becomes "unitized," after which the entire cognitive structure may be treated as a single discrete memory representation or cogit (or perhaps a chunk, cf. Miller, 1956; Simon, 1974). Hayes-Roth emphasized that the development process for a cognitive structure proceeds "slowly," requiring repeated "activations" of the evolving structure.

Memory Schemata

The ideas regarding the evolution of highly organized cognitive structures are basically consistent with the heuristic concept of memory schemata (Olson, 1977). Norman and Bobrow (1975; Bobrow and Norman, 1975) based their conceptualization of memory schemata on the idea that the central goal of cognitive processing is the formation of a meaningful interpretation of the world; therefore, the sensory information available to a person must be encoded and organized relative to and in terms of some coherent framework of previously acquired knowledge (Norman and Bobrow, 1975, p. 119). Schemata are the large number of such structural frameworks of knowledge that a person has learned (developed or evolved) as his/her experiences have accumulated. A schema consists of the knowledge or encoded information about a concept as organized into a meaningful set of structural relationships. Thus, schemata not only contain representations of knowledge, but also "rules" of relationship between cognitive elements and even, for well-developed schemata, evaluation or decision rules for overtly responding to the schema concept. From a memory schemata perspective, conscious perception of a stimulus first involves the activation of the appropriate schema and then the interpretation or encoding of the stimulus "in light of" that schema. Thus, schemata may be considered to provide the framework of basic knowledge regarding a concept within which (and from which) initial encoding of incoming information (from ads or trial) relevant to that concept take place.

A number of interesting ideas and hypotheses regarding the attitude maturation effect can be derived from the schemata concept (see Olson, 1977). For instance, one might postulate that the development of a stable, internally consistent belief structure with a strong relationship to overall attitude is essentially equivalent to the formation of a well-developed schema. Following this reasoning, the belief structure encodes the interrelationships between knowledge concepts related to the attitude object. As the schema "matures," the "rule" for relating the knowledge structure to overall attitude toward the object concept develops and presumably gets stronger and more stable. One would expect a general expectancy-value model (if it is roughly similar to the true "rule") to show increasing correlations over time in such a situation. The present results are consistent with this line of reasoning. As one more example, one might expect consumers who are high in product self-confidence to possess more stable and well-developed schemata for that product. Such consumers, upon learning something about a new brand, should be able to quickly form a stable, well-developed belief structure for that brand, which in turn should be more strongly related to brand attitude. The internal analysis for self-confidence presented in Table 5 produced data consistent with this scenario. The belief structures for "higher" self-confidence consumers were much more strongly related to attitude than were those of the "lower" self-confidence consumers.

Depth or Level of Processing. A final notion from cognitive theory that appears relevant to the attitude maturation issue is "depth of processing." Craik and Lockhart (1974) introduced this concept to account for the common finding that memory performance (typically recall ability) is substantially enhanced when subjects perform cognitive "analyses" or tasks on the incoming stimuli that require their consideration of the meaning of the stimuli (cf. Jenkins, 1973). These results suggest that the encoding processes--that is, the specific cognitive operations carried out during initial processing of incoming information--are critical determinants of later memory performance. In particular, it seems that the degree of conscious attention focused on the meaning or semantic aspects of the stimulus is the critical type of encoding operation for improved memory performance.

Craik and Lockhart's basic notion, later modified somewhat (see Craik and Tulving, 1975), is that certain types or domains of encoding operations exist that can be characterized in terms of "depth," from less semantic to "deeper" levels of semantic processing. "Shallow" processing might concentrate on encoding physical characteristics of the stimulus, while semantic processing would focus on various aspects of the meaning of the stimulus. The encoding operations applied to a stimulus do not necessarily proceed from sensory, shallow levels to deeper, more semantic domains. Rather, encoding processes may be elaborated at a particular level or domain, as indicated by the term "spread of processing."

Several implications for the attitude maturation phenomenon can be derived from the levels-of-processing concept. Perhaps most broadly, it would seem that the development of a well-developed, internally consistent belief structure (or schema) would require relatively "deep" conscious attention to the semantic meaning of incoming product information, over a series of exposures. That is, consumers would have to consciously consider the meaning of product information (e.g., the function of an attribute) and determine for themselves the value of that characteristic. Perhaps in the real world this degree of "semantic involvement" with product information, derived either from advertising or direct experience, is not common, especially at the time of exposure. But, in the present study, this level of conscious processing was essentially "forced" upon the subjects via the process of completing the extensive questionnaire regarding their beliefs, attitudes, and intentions about the product. That is, a subject had to engage in rather "deep" semantic encoding processes in order to answer the detailed belief questions regarding the various product attributes. By this reasoning, one would expect, ceteris paribus, that successive exposures to information and forced semantic processing should lead to the development of relatively consistent, stable, mature attitudes and associated belief structures.

Note that this predicted relationship between "depth of processing" and the formation of well-developed belief structures or schemata provides a theoretically-based and more satisfying account of the measurement effect explanation discussed previously. In one sense, the extensive postinformation measurements can be seen as manipulations of relatively deep levels of processing which, over time should create well-developed, stable schemata (belief structures). Future research could test this explanation (a) by including treatment conditions which receive the information of interest in the experiment, but receive no measurement of cognitive structure until the final stage, or (b) by forcing semantic analysis of the product information in ways other than responding to a questionnaire (e.g., tell someone else about the product).


The theoretical ideas presented here suggest that attitudes and associated belief structures evolve or mature over time as a function of repeated activations involving "deep" processing. The illustrative data cited were generally consistent with this hypothesis. The concept of attitude/cognitive structure maturation is theoretically interesting and, if verified, has clear practical implications for marketing research. For example, one should not be dismayed to find weak correlations for an expectancy-value attitude model applied to a new or unfamiliar product. Second, these ideas suggest that communication research might be focused on developing methods of enhancing cognitive structure "maturity." Finally, the maturation phenomenon is highly relevant to major marketing issues such as repetition effects of advertising (Mitchell and Olson, 1977) and the development of brand loyalty (Jacoby and Olson, 1976), among many others.

Many problems, of course, remain to be solved. Foremost among these are the clear conceptualization and development of valid measures of structure "stability/consistency/maturity.'' This paper presents several measure candidates, but future work must carefully examine the convergent and divergent validities of these measures in a variety of situations. In contrast to the empirical problems, theoretical accounts of the maturation phenomenon are more easily provided. Several concepts from the cognitive theory seem useful in explaining these dynamic phenomena and in formulating future research questions. These include the Hayes-Roth (1977) theory of evolving cognitive structures, the Norman and Bobrow (1975) concept of memory as composed of numerous structural frameworks of organized knowledge called schemata, and Craik and Lockhart's (1972) notions of various types of encoding operations and their effects on information storage and memory.

As mentioned at the outset, our major goal was to stimulate interest in issues regarding changes in cognitive structure over time and as a function of specific informational manipulations. We hope this paper is an additional stimulus to the emerging interest among consumer researchers in cognitive structure dynamics.


Ahtola, Olli T. "The Vector Model of Preferences: An Alternative to the Fishbein Model," Journal of Marketing Research, 12(February 1975), 52-9.

Anderson, J. R. and Gordon H. Bower, Human Associative Memory (Washington, D.C.: Winston, 1973)

Bobrow, Daniel G. and Donald A. Norman, "Some Principles of Memory Schemata," in D. G. Bobrow and A. M. Collins, eds., Representation and Understanding: Studies in Cognitive Science (New York: Academic Press, 1975)

Chestnut, Robert W. and Jacob Jacoby, "Consumer Information Processing: Emerging Theory and Findings," in Arch G. Woodside, Jagdish N. Sheth and Peter D. Bennett, eds., Consumer and Industrial Buying Behavior (New York: Elsevier North-Holland, 1977)

Chronbach, Lee J., "Beyond the Two Disciplines of Scientific Psychology," American Psychologist, 30(February 1975), 116-27.

Chronbach, Lee J., "The Two Disciplines of Scientific Psychology," American Psychologist, 12(1957), 671-84.

Cohen, Joel B., Martin Fishbein, and Olli T. Ahtola, "The Nature and Uses of Expectancy-Value Models in Consumer Attitude Research," Journal of Marketing Research, 9(November 1972), 456-60.

Craik, Fergus I. M. and Robert S. Lockhart, "Levels of Processing: A Framework for Memory Research," Journal of Verbal Learning and Verbal Behavior, 11 (December 1972), 671-684.

Craik, Fergus I. M. and Endel Tulving. "Depth of Processing and the Retention of Words in Episodic Memory," Journal of Experimental Psychology: General, 104 (September 1975), 268-94.

Dover, Philip A. and Jerry C. Olson, "Dynamic Changes in an Expectancy-Value Attitude Model as a Function of Multiple Exposures to Product Information," in Barnett A. Greenberg and Danny N. Bellenger, eds., Contemporary Marketing Thought, 1977 Educators Proceedings (Chicago: American Marketing Association, 1977), 455-60.

Estes, William K., "Structural Aspects of Associative Models for Memory," in Charles N. Cofer, ed., The Structure of Human Memory (San Francisco: W. H. Freeman, 1975).

Fishbein, Martin, "An Investigation of the Relationships Between Belief About an Object and the Attitude Toward that Object," Human Relations, 16(1963), 233-40.

Fishbein, Martin, "A Consideration of Beliefs and Their Role in Attitude Measurement," in Martin Fishbein, ed., Readings in Attitude Theory and Measurement (New York: Wiley, 1967)

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

Ginter, James L, "An Experimental Investigation of Attitude Change and Choice of a New Brand," Journal of Marketing Research, 11(February, 1974), 30-40.

Hayes-Roth, Barbara, "Evolution of Structures and Processes,'' Psychological Review, 94(May 1977), 260-78.

Jacoby, Jacob and Jerry C. Olson. "An Attitudinal Model of Brand Loyalty: Conceptual Underpinnings and Instrumentation Research." Paper No. 159, Purdue Papers in Consumer Psychology, Department of Psychological Sciences, Purdue University, 1976.

Jenkins, James J. "Remember that Old Theory of Memory? Well, Forget It!" American Psychologist (November 1974), 785-95.

Lutz, Richard J, "Changing Brand Attitudes Through Modification of Cognitive Structure," Journal of Consumer Research, 1(March 1975), 49-59.

Lutz, Richard J, "An Experimental Investigation of Causal Relations Among Cognitions, Affect and Behavioral Intention,'' Journal of Consumer Research, 3(March 1977), 197-208.

Miller, George A, "The Magic Number Seven, Plus or Minus Two: Some Limits in Our Capacity for Processing Information,'' Psychological Review, 63(March 1956), 81-97.

Mitchell, Andrew A. and Jerry C. Olson, "Cognitive Effects of Advertising Repetition," in William D. Perrault, Jr., ed., Advances in Consumer Research: Vol. IV (Atlanta: Association for Consumer Research, 1977), 213-20.

Norman, Donald A. and Daniel G. Bobrow, "On the Role of Active Memory Processes in Perception and Cognition," in Charles N. Cofer, ed., The Structure of Human Memory (San Francisco: W. H. Freeman, 1975)

Nunnally, Jum, Psychometric Theory (New York: McGraw-Hill, 1967).

Olson, Jerry C., "Theories of Information Encoding and Storage: Implications for Consumer Research." Paper presented at the Conference, "Effects of Information on Consumer and Market Behavior," Carnegie-Mellon University, May 20 and 21, 1977.

Olson, Jerry C., "Inferential Belief Formation in the Cue Utilization Process," in H. Keith Hunt, ed., Advances in Consumer Research: Vol. 5. Association for Consumer Research, in press.

Olson, Jerry C. and Philip A. Dover, "The Effects of Expectation Creation and Disconfirmation on Belief Elements of Cognitive Structure," in B. B. Anderson, ed., Advances in Consumer Research: Vol. III (Cincinnati, Ohio: The Association for Consumer Research, 1976), 168-175.

Olson, Jerry C. and Philip A. Dover, "Cognitive Effects of Deceptive Advertising," Journal of Marketing Research, in press.

Olson, Jerry C. and Andrew A. Mitchell, "The Process of Attitude Acquisition: The Value of a Developmental Approach to Consumer Attitude Research," in Mary J. Schlinger, ed., Advances in Consumer Research: Vol. II (Chicago: Association for Consumer Research, 1975), 249-264.

Restle, Frank, "Critique of Pure Memory," in Robert L. Solso, ed., Theories in Cognitive Psychology: The Loyola Symposium (Potomac, Md.: Lawrence Erlbaum, 1974)

Ryan, Michael J. and E. H. Bonfield, "The Fishbein Extended Model and Consumer Behavior," Journal of Consumer Research, 2(September 1976), 118-37.

Simon, Herbert A., "How Big is a Chunk?" Science, 183 (February, 1974), 482-8.

Tulving, Endel, "Episodic and Semantic Memory," in Endel Tulving and W. Donaldson, eds., Organization of Memory (New York: Academic Press, 1972)

Underwood, Benton J, "Individual Differences as a Crucible in Theory Construction," American Psychologist, 30 (February 1976), 128-34.

Winter, Frederick W, "The Effect of Purchase Characteristics on Postdecision Product Reevaluation," Journal of Marketing Research, 11(May 1974), 164-171.

Wyer, Robert S, "Category Ratings as 'Subjective Expected Values'; Implications for Attitude Formation and Change," Psychological Review, 80(1973), 446-67.

Wyer, Robert S., Jr., Cognitive Organization and Change: An Information-Processing Approach (Potomac, Md.: Lawrence Erlbaum, 1974.

Wyer, Robert S., Jr., "Effects of Previously Formed Beliefs on Syllogistic Inference Processes," Journal of Personality and Social Psychology, 33 (March 1976), 307-16.



Jerry C. Olson, Pennsylvania State University
Philip A. Dover, Dartmouth College


NA - Advances in Consumer Research Volume 05 | 1978

Share Proceeding

Featured papers

See More


Surprise! The Positive Impact of Uncertainty on the Evaluation of Experiential Purchases

Iñigo Gallo, IESE Business School
LILY JAMPOL, Queen Mary University of London
Alberto Rampullo, IESE Business School
Thomas Gilovich, Cornell University, USA

Read More


Consumer Attachment Styles and Preference for Sharing Products

Kivilcim Dogerlioglu Demir, Sabanci University
Ezgi Akpinar, Koc University, Turkey
mehmet okan, ITU

Read More


When Taking Action Means Accepting Responsibility: Omission Bias Predicts Reluctance to Vaccinate Due to Greater Anticipated Culpability for Negative Side Effects

Gary Sherman, Stony Brook University
Stacey R Finkelstein, Stony Brook University
Beth Vallen, Vilanova University, USA
Paul M Connell, Stony Brook University
Kristen Feemster, Vaccine Education Center at Children’s Hospital of Philadelphia and University of Pennsylvania, Perelman School of Medicine, USA

Read More

Engage with Us

Becoming an Association for Consumer Research member is simple. Membership in ACR is relatively inexpensive, but brings significant benefits to its members.