Celebrity Endorsements-Scripts, Schema and Roles: Theoretical Framework and Preliminary Tests

Paul Surgi Speck, University of Tennessee-Knoxville
David W. Schumann, University of Tennesses-Knoxville
Craig Thompson, University of Tennesses-Knoxville
ABSTRACT - Subjects were exposed to twelve celebrity type/endorser cue combinations in order to test a script- and schema-based model of celebrity endorsement. Print ads were used. Processing time and product type were also varied. This research differs from previous endorser research in two important ways: (1) it considers the effectiveness of different kinds of celebrity endorsers and (2) it focuses on mechanisms required to integrate celebrity and product information. Information recall measures were significant, but only sometimes in the directions predicted. The researchers conclude that three different effects were observed: endorsement effects, incongruity effects, and distraction effects. All three effects are discussed in terms of script and schema processing.
[ to cite ]:
Paul Surgi Speck, David W. Schumann, and Craig Thompson (1988) ,"Celebrity Endorsements-Scripts, Schema and Roles: Theoretical Framework and Preliminary Tests", in NA - Advances in Consumer Research Volume 15, eds. Micheal J. Houston, Provo, UT : Association for Consumer Research, Pages: 69-76.

Advances in Consumer Research Volume 15, 1988      Pages 69-76


Paul Surgi Speck, University of Tennessee-Knoxville

David W. Schumann, University of Tennesses-Knoxville

Craig Thompson, University of Tennesses-Knoxville

[For further communication: Paul S. Speck, Department of Marketing and Transportation, 324 Stokely Management Center, University of Tennessee, Knoxville, TN 37996-0530. Phone: (615) 974-1647.]


Subjects were exposed to twelve celebrity type/endorser cue combinations in order to test a script- and schema-based model of celebrity endorsement. Print ads were used. Processing time and product type were also varied. This research differs from previous endorser research in two important ways: (1) it considers the effectiveness of different kinds of celebrity endorsers and (2) it focuses on mechanisms required to integrate celebrity and product information. Information recall measures were significant, but only sometimes in the directions predicted. The researchers conclude that three different effects were observed: endorsement effects, incongruity effects, and distraction effects. All three effects are discussed in terms of script and schema processing.


Bill Cosby, Linda Evans, Cliff Robertson and Jimmy Conners-what do they have in common? Aside from being well-known celebrities, they have all endorsed consumer products. Although the use of celebrity endorsers dates back to the 1930's (e.g., Johnny Weissmuller for Wheaties), it is uncertain that such testimonials are always effective. From the multimillion dollar contracts awarded to celebrity endorsers (e.g., James Garner for Polaroid), advertisers apparently assume so. Poloroid credits Garner and Mariette Hartley with selling two million Polaroid cameras (McMahan and Kile, 1982). On the other hand, McCollum/Spielman & Company (Topline, 1985) report that only 41% of the celebrity commercials they have tested have above average brand awareness and influence on consumer attitudes towards the product.

How do consumers view the relationship between a famous person and a product they are considering? What selection criteria for celebrity endorsements should marketers consider? In light of the intense use of celebrities as product spokespeople, answers are needed to determine formulas that maximize celebrity endorsement effectiveness.


To date, few academic researchers have studied celebrity endorsements. Friedman and his colleagues (Friedman and Friedman, 1976, 1979; Firework and Friedman, 1977; Friedman, Santeramo, and Traina, 1979) explored several aspects of this issue. They found a positive relationship between how much a celebrity was liked and how much he/she was trusted. However, in two different tests, they found mixed results across endorser types. Firework and Friedman (1977) found no significant differences in purchase intention across test advertisements that employed experts, celebrity nonexperts, and typical consumers. Conversely, Friedman and Friedman (1979) found an endorser by product interaction. Experts were found to be more effective for household durable goods, whereas celebrities were more effective promoting luxury goods.

In a similar study, Freiden (1982) found that the use of expert endorsers resulted in more favorable product attitudes compared to celebrity and ordinary consumer endorsers. More recently, Swerdlow (1984) found that, between buyers and non-buyers of a product, (1) there were no differences in the consumer's ability to associate celebrities with the products they endorse and (2) no noticeable difference in celebrity influence on buying behavior.

Although there is some evidence that an expert may, in fact, be the best endorser, findings are mixed. In fact, Sternthal and his colleagues (Dholakia and Sternthal, 1977; Sternthal, Dholakia, and Leavitt, 1978) found that situations exist where a low credible source is equal to or greater than a highly credible source in terms of effectiveness. This work reflects the earlier psychological research on source credibility carried on at Yale University in the 1950's (Hovland and Weiss, 1951; Kelman and Hovland, 1953; and Hovland and Mandell, 1952). In addition, endorser effectiveness may be affected by other ad-related or consumer-related variables. Aaker and Brown (1972) suggest that the impact of an advertising variable may be influenced by copy objectives. Friedman and Friedman (1979) point out that spokeperson effectiveness might be a function of communication goals. Pitts, Canty, and Tsalikis (1983) found that personal values increased the influence of a positive message and changed the effect of expertise on attitude toward a social ideal.

Although several studies have compared celebrity endorsers to other types of endorsers, prior research has not explored the relative effectiveness of different kinds of celebrity endorsers. Once advertisers decide to use a celebrity, they must decide which celebrity to use. Further, they must consider how the celebrity should be positioned with respect to the product? The purpose of this study is (1) to explore the effectiveness of different celebrity endorser types and (2) to see whether positioning cues that define the role of the endorser vis a vis the product can influence that effectiveness.

Before taking up these questions, it is important to establish an appropriate conceptual framework: one that allows researchers to consider (1) how several forms of old knowledge interact, (2) how old information interacts with new, and (3) how the structure of the ad mediates these various levels of integration. Script and schema theory offer such a framework (Crocker, 1984; Schurr, 1986; Hunt, Bonfield, and Kernan, 1986).


Schema are higher-order cognitive structures which are thought to guide perception, thought and action (Mandler, 1982). Schema may affect (1) the selection, (2) abstraction, (3) interpretation, and (4) integration of new information (Alba and Hasher, 1983). Schematic theory offers several advantages over other theories of memory and information processing: it posits the existence of higher order memory structures, explains unitary access to entire sets of knowledge, accounts for selective attention, encoding and retrieval, accounts for the purposefulness of processing, and emphasizes relational context as the basis of meaning (Brewer and Nakamura 1984, Alba and Hasher 1983). Schema organize perception by organizing expectations.

Researchers distinguish among several kinds of schema (Hastie 1981). We suggest that celebrity endorsements often involve four types of schema: scripts, role schema, individual person schema, and object schema. Figure 1 suggests how these four schema might be related within a celebrity endorsement. Person, role and object schema are integrated within the context of the script.



Scripts are event-schema that describe the prototypic structure of commonplace actions, such as, "Eating at a restaurant" (Abelson 1976, Schank and Abelson 1977). Like all schema, scripts are propositional and semantically encoded knowledge structures (Lord 1980), but unlike some schema, scripts are highly abstract, temporally organized, and hierarchically structured (Abbott and Black 1980). Since scripts categorically define he principle actions, actors and objects found in a recurring situation (Abelson 1976), they facilitate the study of individuals, objects and roles within organized contexts (Calder and Schurr 1981, Schurr 1986).

Roles are relational-schema used to explain the intention and behavior of people in specific situations (Hastie 1981). They are "relational" in that they define the set of probable relationships among actors and objects within the event. Within the "Eating at a restaurant" schema, people can play many different roles: waiter, patron, or someone entertaining friends. Scripts define both those actors characteristic of a stereotypic event and the roles associated with each actor.

Individual person schema and object schema are trait-based impressions of specific people and things. Their development depends on one's prior experience with the person or object in question. Unlike role schema, person and object schema are not situation specific. For example, one can possess schema for Julia Child and chocolate mousse. Either schema might be activated in the context of various scripts, "Eating at a Restaurant" or"Preparing a Dessert." Furthermore, they may be activated in the context of various roles, "Julia Child as the person eating out" or "Julia Child as someone who recommended this restaurant."

Ads can be viewed as scripts, that is, as prototypic rhetorical structures that present new information within a convention. 1 format. Moreover, different kinds of ads, by having their own traditions of structure and content, involve different kinds of scripts: the before-after comparison, the expert testimonial, the slice of life, and the celebrity endorsement. Through experience, consumers probably develop strong expectations concerning the evens, actors and objects associated with a particular type of ad. For instance, "endorsements" typically involve three actors (seller, endorser, and target), one object (the product), and a protypic sequence of events:

(1) [the seller employs the endorser]

(2) the seller asks the endorser to use or evaluate the product

(3) the endorser tries the product

(4) the endorser communicates his/her findings to the target

(5) the endorser urges the target to consider the Product

(6) [the seller pays the endorser]

In a particular endorsement, various script elements can be missing. Typically, events #] are only hinted at by small print that reminds that the ad is a "compensated endorsement." I instances, events #2 and #3 may be omitted; h once the viewer recognizes that the ad is an endorsement he/she assumes that a product-related experience kind has occurred. The point is this: if most possess a script for "commercial endorsements," recognition of any element from that script can activate the entire script, and events or variables that are not specified by the current stimulus can be supplied by generic default values from memory.

In a general endorsement script, "endorser" serves as a generic category. There are, however, many kinds of endorsers, including humorous, attractive, expert, likable and celebrity endorsers. These endorser traits may be consistent, inconsistent or irrelevant to the role schema implied by a specific script. This study concerns only likable expert and likable nonexpert celebrity endorsers.

When actors serve as endorsers, they are used to activate social and occupational stereotypes (the typical policeman, the typical steelworker, the typical dairy farmer, the typical yuppie). Once the stereotype is activated, a host of cognitive and affective information is brought to bear on the interpretation of the script.

Celebrity endorsements are a special kind of endorsement: By definition, celebrities are people who are well-known -- individuals for whom most people have well-developed individual person schema. Whether or not this preexisting person schema helps should depend on four factors: (1) the relative development of the celebrity schema, (2) the overall affect one attaches to the celebrity schema, (3) the appropriateness of the celebrity schema in regard to the role schema, and (4) the appropriateness of the celebrity schema in regard to the product schema.

If a particular celebrity endorser is well-known, well-liked, positively associated with the product under consideration, and capable of being viewed in an appropriate endorser role, script processing is likely to benefit from the activated celebrity schema. Indeed, given sufficient time, the viewer should be able to integrate material from the celebrity schema into the product schema, so that the entire ad is perceived and remembered as one unit. The endorsement script and the role schema mediate this integration by implying how information in the celebrity schema should relate to information in the product schema. On the other hand, if the celebrity (though well-known and well-liked) cannot be associated with the product and/or cannot be viewed as a typical endorser, then script processing might be impaired, so much perhaps that the celebrity schema and the product schema are never integrated.


This study manipulates two variables within the celebrity endorsement script: (1) the preexisting product relatedness of the celebrity schema and (2) embedded verbal cues that may facilitate activation of a suitable endorser role.

First, we consider two types of celebrities (celebrities who have strong preexisting relatedness to the product and celebrities who have no preexisting relatedness to the product). Strong preexisting product relatedness is operationalized as perceived expert knowledge of and professional connection with the product category [expert celebrity endorsers]. Conversely, no preexisting product relatedness is operationalized as the perceived absence of special knowledge or professional connection with the product category [product irrelevant celebrity endorsers].

H1 A product relevant celebrity schema should produce better recall of product information than a product irrelevant celebrity schema.

The subject's prior knowledge about the expert celebrity should allow him/her to infer an appropriate endorser role schema (expert) and thus facilitate the processing of the ad as an endorsement. On the other hand, the product irrelevant celebrity schema does not contain information required to infer an appropriate role schema and thus should not facilitate the processing of the ad as an endorsement.

Next, we consider the role of advertiser-supplied cues that may facilitate perception of the celebrity as a suitable endorser. Three types of cues were employed: cues which suggest that the celebrity is an expert, cues that suggest that the celebrity is a typical user of the product, and cues that leave the role of the celebrity unspecified.

H2 Cues that imply a specific endorser role (expert or typical user) should lead to greater recall of product information than cues that are neutral.

Again, since script processing requires not merely the successful activation of the product schema and the celebrity schema, but also the activation of a role schema capable of linking the other two, information that helps a subject to activate or construct an acceptable endorser schema will facilitate overall processing of the ad.

How will these three levels of role cue interact with the two levels of celebrity endorser?

H3a Expert celebrities should produce greater recall when combined with expert cues (compared to neutral cues).

H3b Expert celebrities should produce greater recall when combined with typical user cues (compared to neutral cues).

In the first case, the "expert" cues should reinforce the endorser role (expert) implied by the preexisting celebrity schema: from the subject's point of view, an easy job (seeing the connection) should become easier. In the second case, "typical user" cues have the potential for making the expert endorser more likable or attractive. On the other hand, the addition of a dimension that is orthogonal to dimensions in the celebrity schema, could confuse or delay processing. Still, given sufficient time to process the ad, H3b should hold.

H4a Unrelated celebrities should evidence diminished recall when accompanied by expert cues (compared to neutral cues).

H4b Unrelated celebrities should evidence diminished recall when accompanied by typical user cues (compared to neutral cues).

Since the subject has a schema for the unrelated celebrity that is totally unrelated to the product, the addition of expert cues should only heighten the level of incongruity between the celebrity and the endorser role. This heightened inconsistency might cause the subject to focus disproportionate attention on nonproduct-related portions of the message and thus impair processing. In the second case, although subjects are unlikely to view unrelated celebrities as typical users, they may accept the idea if it is suggested by new information ("typical user" cues) and they are allowed sufficient time to process this new information.

Finally, the investigators were uncertain how much time would be required to observe the predicted effects. Therefore two levels of processing time were employed.

H5a In all conditions, information recall should improve with additional processing time.

H5b In several conditions, the predicted results may only appear with longer processing time. This is most likely with interaction effects involving either celebrity schema and the "typical user" cue (H3b and H4b).

In addition to the three variables that are manipulated (preexisting product relatedness of the celebrity schema, cues that facilitate activation of a suitable endorser schema, and time), a number of other script- and schema-related variables are held constant across all treatments: product information, layout, amount of text, number of relational cues, likableness and familiarity of the celebrity. Two products were used to provide replication.


Subjects: One hundred and ninety-two undergraduates at a large state university participated in the experiment. The design allowed for the testing of multiple experimental conditions within the students' own classrooms, with 25-30 students per class participating at any one tim :.

Procedure: The experimenter distributed booklets to all subjects. Each booklet contained two sections, and each section contained advertising stimuli and questions. The first page of the booklet included a brief statement of the study's purpose, a cover story and instructions about the study's procedures. Subjects were informed that the study concerned the evaluation of several print ads and that the marketing department was working with the advertising department on this joint project. After subjects read the instructions, the experimenter verbally clarified the procedure.

Each section of the booklet contained five magazine ads: four filler ads and one target ad. The experimental ad appeared third to control for possible primacy or recency effects. Ads were separated by blank pages. The procedure required subjects to view each ad until instructed to turn the page. Following each 5-ad sequence, subjects answered questions related to the ads in that section. After completing the entire booklet, subjects were asked demographic questions, debriefed and dismissed.

Target ads: Each target ad was structured using a conventional endorser format that is frequently used in print advertising. Each target ad included (1) two cues related to the celebrity schema (a picture and the celebrity's name), (2) two cues related to the product schema (a picture and the brand name), (3) cues (as part of the text) that specified (or failed to specify) the celebrity's role, and (4) four pieces of information about the product.

Manipulations: Booklets were collated to permit optimal balance between combinations of celebrity endorsers, types of cues, products, and processing time. Each combination was repeated an equal number of times across orders with no endorser/message combination repeated twice in the same booklet.

Two products were used to determine whether the results were generalizable across product classes. After pretesting several products for student interest and gender bias, a car was chosen to represent hard durable goods, and a tennis shoe was selected to represent soft consumable goods. In half the booklets, the car ad appeared in the first section and the tennis shoe in the second; in the other half, the car ad was found in the second section and the tennis shoe in the first.

Two celebrity endorsers were chosen for each product on the basis of their relatedness with the product category. Pretests were used to determine the product-relatedness, likability, and familiarity of the endorsers. For the automobile, Richard Petty and Pee Wee Herman were chosen to represent high and low product relatedness, respectively. For the tennis shoe, Chris Evert Lloyd and Bette Midler were selected.

The amount of time allowed for the subjects to process each ad was systematically varied. For one section of the booklet, subjects were allowed only 20 seconds to process each ad, while in the other section, 60 seconds were allotted. For half of the subjects, the 20 second condition came first; for the other half, the 60 second condition came first.

Cues related to different endorser roles were inserted into the ad's text. See Table 1. Expert cues emphasized that an active comparison and judgment had been made between the advertised brand and other competitive products. The words "judges,""evaluate," and "comparing" were used to suggest an expert role. Typical user cues implied that the endorser had used the product as any normal consumer might. The terms "experiences,""try out," and "driving" (or "wearing") were used to establish this testimonial theme. Neutral cues were worded passively and conveyed no reference to either an evaluation- or product use ("and," "Comment on" and "seeing").

Dependent Measures: Each ad contained four pieces of product-related information. See Table 2. After answering questions on one of the non-experimental ads, subjects were given an unaided test of recall for specific information in the target ad. Subjects were given four lines to list all the information they could recall. Three minutes were allowed for this task.


An overall ANOVA was performed on the 2 (celebrity relatedness to product expert vs. non-expert) X 3 (relational cue: expert vs. typical user vs. neutral) X 2 (processing time: 20 vs. 60 seconds) X 2 (product: car vs. shoe) design. Although replication (across products) was sought, the overall analysis yielded effects that varied by product. Consequently, separate ANOVAs were run on each product.

The overall effect of processing time: H5a predicted that recall of product information will generally improve with additional processing time. Subjects were allowed either 20 or 60 seconds to process each target ad. In general, additional time led to more complete recall of the four product-related statements (1.92 vs 2.48, p = .0001).





The differential impact of processing time: H5b suggested that processing time may be more significant whenever subjects are required to process less familiar or less easily related information. Two differential effects are worth noting. First, additional time has a greater impact on the recall of tennis shoe information than on the recall of car information.


Subjects may have been less familiar with tennis shoe information than car information. (See Table 2.) As a result, recall for shoe information at 20 seconds is lower than recall of car information at 20 seconds. On the other hand, after 60 seconds recall for shoe and car information are nearly equal.

Second, whenever expert endorsers appear with neutral processing cues, additional time produces little additional recall ( = .28). However, when product unrelated endorsers appear with neutral cues, extra time leads to significant additional recall ( = .81).


Again, as suggested by H5b, additional time is required whenever subjects are obliged to recall information that is more difficult to associate with other information in the ad. When there is little opportunity to process the ad, inconsistency between the endorser and the message impairs recall. However, when subjects have additional time, inconsistency enhances recall. Apparently, the extra time allows subjects to construct a relationship between the unrelated endorser and the message. This newly constructed relationship appears to be more powerful than the established relationship.

In sum, although additional processing time generally leads to greater recall, this effect is moderated by the familiarity and dimensionality of the target information. When the product schema is more familiar (car) or the endorser's relationship is more obvious, information can be processed in a relatively short span of time. When the product schema is less familiar (tennis shoe) or the endorser's relationship is unclear, more time is required.

The effect of the endorser's preexisting product relevance: H1 predicted that product-relevant celebrities would facilitate greater recall of product information than celebrities who have no prior association with the product category. ID this study, preexisting product-relevance is operationalized as perceived expertise. The data provide very limited evidence for such an effect. Although expert celebrities do produce higher recall of the product information than nonexpert celebrities (2.27 vs 2.15), the main effect for product-related expertise is not significant. When the car data is considered separately, expertise is marginally superior to nonexpertise (2.39 vs 2.15, p = .08). However, this effect must be interpreted in light of a significant expertise x cue interaction (p = .005). Compared to other expertise/cue combinations, the "nonexpert endorser/typical user cue" combination produced significantly lower recall (1.44).

H1 can also be tested by analyzing only those conditions that employ neutral processing cues (since neutral processing cues serve as a control). When the shoe data is considered alone, the product-relevant celebrity facilities marginally higher recall than the nonrelevant celebrity (2.31 vs 1.69, p = .10), but only under the 20 second condition.

In short, there is no evidence of a general facilitory effect related to the preexisting product relevance of a celebrity endorser. To the extent that endorser expertise does enhance recall, the effect appears limited by other conditions (processing time, type of embedded cues, product type and/or the endorser).

The overall effect of processing cues: Each ad included words that specified (or failed to specify) whether the celebrity endorser was acting (1) as an expert, (2) as a typical user, or (3) in an unspecified role. H2 predicted that ads which used specific endorser cues (both expert and typical user) would demonstrate higher recall than ads that did not. The data fail to support this hypothesis. In fact, the nonspecific or neutral cue is associated with higher recall than either specific cue.


Although these differences are not significant, their directional consistency suggests that an endorser's role need not be spelled out. Two explanations appear plausible. On one hand, if the il formation offered by the verbal cue only repeats information implicit in other elements of the ad, then the relational cue may be gratuitous. To the extent that it is redundant, it would serve as a mild distraction. On the other hand, specification could diminish the interpretive role of the subject. That is, when a subject is compelled to infer the endorser's role, processing may be more active and memory could be enhanced. When everything is spelled out by the advertiser, processing may be more passive and recall could be diminished.

In sum, there is no overall effect related to either (1) the endorser's product expertise or (2) the role cues embedded in each ad. Although there was a general effect related to processing time, there was also evidence that "processing time" effects were moderated by other variables. In short, recall appears to depend not on any of these individual variables but on different combinations of product, endorser, cue and time. The remainder of this discussion considers various celebrity cue combinations and attempts to explain why different combinations produced different recall results.

&pert celebrities with expert cues: H3a predicted that expert cues would enhance information recall for ads involving expert celebrity endorsers. As is clear from the following table, expert cues are never significantly better than neutral cues, and in one case (Evert Lloyd/20 seconds) the combination using expert cues is significantly worse.


Why don't expert cues help expert celebrities? Perhaps, the expert cues are unnecessary. Since each expert celebrity's schema contains information about the celebrity's expertise, subjects can infer the celebrity's role (to give expert testimony) as soon as they recognize the celebrity within the context of an ad and long before they encounter additional "expert" role cues provided by the text. If this is the case, the addition of "expert" cues to an expert celebrity is gratuitous. The celebrity's expertise, once recognized, is not enhanced. On the other hand, redundant cues may impede processing. When processing time is limited (20 seconds), redundant information seems to impair recall, but when more time is allowed (60 seconds), this distraction effect is diminished.

This "gratuitous information/distraction" explanation fits the data for Evert Lloyd, but it does not fit the data for Richard Petty. In Petty's ease, expert cues are associated with equivalent or slightly higher recall. Why would Evert Lloyd and Petty evidence different patterns? Perhaps the utility of the role cue depends on the power of the endorser's picture of prompt (1) recall of the celebrity schema and (2) inferences regarding his or her role as an endorser. In a pretest, Chris Evert Lloyd's and Richard Petty's names were found to be equally familiar to the subject population. On the other hand, it is likely that Chris Evert Lloyd enjoys greater face recognition. Facial recognition was not pretested. To the extent that Chris's picture was more recognized, the "expert" role cues would have been less helpful. To the extent that Petty's face was less familiar, the "expert" role cues would remain useful.

Expert celebrities as typical users: H3b predicted that "typical user" cues would enhance information recall for ads involving expert celebrity endorsers. However, it was also suggested that this facilitory effect might only appear in the 60 second condition. As a clear from the following table, "typical user" cues are never significantly better than "neutral" cues, and in one case (Evert Lloyd/20 seconds) the combination using "typical user" cues is significantly worse. This pattern is similar to but even stronger than that evidenced by "expert" cues. Could the same explanations apply? That is, could the "typical user" cue be gratuitous for Chris but marginally useful for Richard Petty.


After much consideration, we suggest that the difference stems from the perceived dimensionality of the two expert celebrities. Richard Petty is well-known, but is thought of almost entirely in terms of auto racing. Chris, on the other hand, was the public's sweetheart. The public followed her through adolescence, courtship and marriage; they know her as a person as well as a tennis star. If this is the case, if the average subject's schema for Petty is strictly ear- and racing-related (expert), while the average subject's schema for Chris is more multidimensional (expert and normal person), then it is clear why "humanizing" cues (1) can make a positive difference for Petty (no matter the time allowed), (2) have no effect for Chris in the 60 second condition, and (3) have a negative effect for Chris in the 20 second. For Chris, the "typical user" cue is just as gratuitous as "expert cue," thus there is a negative distraction effect with limited time and a washout with additional time. For Petty, whose celebrity schema is less likely to include a "normal person" dimension, the "typical user' cue produces the humanizing effect looked for in H3b.

Nonexpert celebrity with expert role cues: H4a predicted that "expert" cues would impair information recall for ads involving product unrelated celebrity endorsers. As is clear from the following table, "expert" cues have no significant effect on recall when combined with product unrelated celebrity endorsers. "Neutral" cues and "expert" cues perform almost the same. Why? The answer may be an "incongruity effect."


First, consider the use of expert cues with Pee Wee or Bette. Subjects are unlikely to accept either one as a product expert. Traits found in each celebrity's schema (unserious and silly) are inconsistent with role traits implied by the expert cues (serious and logical). If this is the case, then subjects perceive a significant mismatch between the celebrity and his/her implied role. In an attempt to resolve this celebrity-role mismatch, the subject might continue to process the ad seeking additional cues to explain how Pee Wee or Bette fit into the ad. When subjects encounter the product information, their incongruity is redoubled. That is, there is also no connection between Pee Wee or Bette and the products they endorse. ln short, incongruities exist on two levels: a celebrity-role mismatch and a celebrity-product mismatch. At this point, Berlyne's (1960) collative principle might take hold. The ad becomes a cognitive challenge, a puzzle, and at least for the moment a subject is motivated to examine various parts of the ad in order to find a solution. Even if no solution were found, the additional search could result in higher recall of product information.

Neutral cues could produce a similar but somewhat simpler "incongruity effect." In this case, since the role cues are unspecified, there would be no celebrity-role mismatch. Nonetheless, subjects would find themselves in a similar situation due to an unresolved celebrity-mismatch. Thus, neutral cues and expert cues would exhibit similar effects when used with Pee Wee and Bette. When no solution is found and there is not sufficient time to construct a solution (20 seconds), recall is very low. When enough time is available to construct a resolution (60 seconds), recall improves significantly.

Nonexpert celebrity with typical user cues. H4b predicted that "typical user" cues would significantly impair recall. As is clear from this table, the "typical user" cue produces radically different effects for Pee Wee and Bette. When the "typical user" cue is used with Bette, recall is unusually high for 20 seconds, then falls off slightly at 60. We suspect that this effect is related to the reasonableness of viewing Bette as a "typical user," that is, as someone who in real life is not too unlike the subject. As with Chris Evert Lloyd, the average subject probably has a multidimensional schema for Bette Midler (singer, actress, comedienne, subject of an interview). If this is he case, then the "user" cues provide the subject with a plausible connection between Bette and the tennis shoe. In this case, since there is no significant unresolved incongruity effect, Bette actually works as an endorser, and since resolution is achieved, no additional processing occurs with additional time.


Finally, consider the use of "typical user" cues with Pee Wee Herman. Since Pee Wee is always in character, subjects only know Pee Wee as "Pee Wee." By contrast, Bette is occasionally seen out of character. As a result, while it may be plausible for subjects to think of Bette Midler as "someone like myself," it is difficult to think of Pee Wee as "someone like myself." Again, there is a mismatch. However, in this case the mismatch is not between the celebrity and the role nor between the celebrity and the product (which would cause the subject to look for the solution in the ad); instead, the mismatch is between the celebrity and the subject. Typical user cues imply that Pee Wee is like the subject and social companion is required. Even if he subject tries to resolve the mismatch (by comparing Pee Wee to him/herself), then he/she will not attend to the product information. The result may be a distraction effect and diminished recall.

In summary, expert cues consistently failed to enhance the effectiveness of celebrities who were already considered to be experts, and typical user cues improved effectiveness only when (1) the expert celebrity was not already considered multidimensional (expert and normal person) or (2) an irrelevant celebrity is capable of being thought of as normal. Furthermore, the use of irrelevant celebrities (when they produce strong incongruity effects) can be as beneficial as the use of product-relevant celebrities (which produce endorser effects). This pattern is consistent with research concerning the schematic processing of consistent and inconsistent data (see reviews in Alba and Hasher, 1983; Hastie, 1981; and Brewer and Nakamura, 1984).

These findings are limited in several ways: only print ads were employed, the imposition of specific amounts of processing time heightened the unnaturalness of the procedure, and delayed recall was not measured. More importantly, the experimenters failed to anticipate the complexity of the celebrity-role-product interaction. Clearly, certain variables that were not controlled for in this experiment (such as the dimensionality of the celebrity and the presence of role information in the celebrity schema) must be included in future research. On the other hand, the overall results support the decision to explain endorsements within a script- or schema-based theoretical framework.

Future research should explore (1) the effects of a celebrity's perceived dimensionality, (2) the conditions required to produce a beneficial incongruity effect, (3) the relative importance of various sources (text, facial cues, or preexisting schema) in allowing subjects to infer acceptable role schema, (4) the importance of processing time, and (5) generalizability of this model to other media (especially television).


Aaker, David A. and Philip K. Brown (1972), "Evaluating Vehicle Source Effects," Journal of Advertising Research, 12 (August), 11-16.

Abbott, Valerie and John B. Black (1980), "The Representation of Scripts in Memory," Cognitive Science Technical Report #. . New Haven, CT: Yale University.

Ableson, Robert P. (1976), "Script Processing in Attitude Formation and Decision Making," in Cognition and Social Behavior, ed. J.S. Carroll and 3.W. Payne, Hillsdale, NJ: Lawrence Erlbaum Assoc., 33-45.

Alba, Joseph W. and Lynn Hasher (1983), "Is Memory Schematic?" Psychological Bulletin, 93, 203-231.

Berlyne, Daniel E. (1960), Conflict, Arousal and Curiosity. New York: McGraw-Hill.

Brewer, William F. and Glenn V. Nakamura (1984), 'The Nature and Functions of Schema," in Handbook of Social Cognition, Vol 1, ed. Robert S. Wyer, Jr. and Thomas K. Srull, Hillsdale, NJ: Lawrence Erlbaum Assoc., 119-160.

Calder, Bobby 1. and Paul H. Schurr (1981), "Attitudinal Processes in Organizations," in Research in Organizational Behavior, Vol 3, eds. L.L. Cummings and Barry M. Staw, Greenwich, CT: JAI Press.

Crocker, Jennifer (1984), "A Schematic Approach to Changing Consumer's Beliefs," in Advances in Consumer Research, Vol 1 ', ed. Thomas C. Kinnear, Provo, UT: Association for Consumer Research, 472-477.

Dholakia, Ruby R. and Brian Sternthal (1977), "Highly Credible Sources: Persuasive Facilitators or Persuasive Liabilities?" Journal of Consumer Research, 3 (March), 223-232.

Firework, Robert B. and Hershey H. Friedman (1977), 'The Effects of Endorsements on Product Evaluation," Decision Sciences, 8 (July), 576-583.

Forkan, James (1980), "Product Matchup Key to Effective Star Presenters," Advertising Age, 6 (October), 42.

Freiden, Jon (1982), "An Evaluation of spokesperson and Vehicle Source Effects in Advertising," Current Issues and Research in Advertising, eds. James H. Leigh and Claude R. Martin, Jr., 77-87.

Friedman, Hershey H. and Linda Friedman (1976), "Whom Do Students Trust" Journal of Communication, 26, 48-49.

Friedman, Hershey H. and Linda Friedman (1979), "Endorser Effectiveness by Product Type," Journal of Advertising Research, 19 (October), 63-71.

Friedman, Hershey, H., Michale J. Santeramo and Anthony Traina (1979), "Correlates of Trustworthiness for Celebrities," Journal of the Academy of Marketing Science, 6 (Fall), 291-299.

Friedman, Hershey, Salvatore Terzuni and Robert Washington (1977), "The Effectiveness of Advertisements Utilizing Four Types of Endorsers," Journal of Advertising, 6, 22-24.

Hastie, Reid (1981), "Schematic Principles in Human Memory," in Social Cognition: The Ontario Symposium, Vol 1, ed. E.Tory Higgins, C.Peter Herman and Mark P. Zanna, Hillsdale, NJ: Lawrence Erlbaum Assoc., 39-88.

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

Havland, Carl I. and W. Weiss (1951), "The Influence of Source Credibility on Communication Effectiveness," Public Opinion Quarterly, 15, 635-650.

Hunt, James M., E.H. Bonfield and Jerome B. Kernan (1986), "The Representation and Recall of Message Arguments in Advertising: Test of a Schema-Based Model," in Advances in Consumer Research, Vol 13, ed. Richard Lutz, Provo, UT: Association for Consumer Research, 562-565.

Kelman, H. C. and Carl I. Hovland (1953), "'Reinstatement' of the Communicator in Delayed Measurement of Opinion Change," Journal of Abnormal and Social Psychology, 48, 327-335.

Lord, Charles G. (1980), "Schemas and Images as Memory Aids: Two Modes of Processing Social Information," Journal of Personality and Social Psychology, 38(2), 257-269.

Mandler, George (1982), "The Structure of Value: Accounting for Taste" in Affect and Cognition, ed. Margaret S. Clark and Susan T. Fiske, Hillsdale, NJ: Lawrence Erlbaum Assoc., 3-36.

McMahan, Harry Wayne and Corwin Mack Kile (1982), "100 Best," Advertising Age, 22 (February), M4.

Pitts, Robert E., Ann L. Canty and John Tsalikis (1985), "Exploring the Impact of Personal Values on Socially Oriented Communications," Psychology and Marketing, 2(4), New York: John Wiley and Sons, 267-278.

Schank, Roger C. and Robert P. Abelson (1977), Scripts, Plans, Goals. and Understanding. Hillsdale, NJ: Lawrence Erlbaum Associates.

Schurr, Paul H. (1986), "Four Script Studies: What Have We Learned," in Advances in Consumer Research, Vol. 13, ed. Richard J. Lutz, Provo, UT: Association for Consumer Research, 498-503.

Sternthal, Brian, Ruby Dholakia and Clark Leavitt (1978), "The Persuasive Effect of Source Credibility: A Test of Cognitive Response Analysis," Journal of Consumer Research, 4, 252-260.

Swerdlow, Robert A. (1984), "Star Studded Advertising: Is It Worth the Effort?" Journal of the Academy of Marketing Science, 1 2 (Summer), 89-102.

Topline (1985),"The Mood/Image Appeal: Giving The Product Good Vibrations!" Vol.4 (September), Great Neck, NY: McCollum/Spielman & Company, Inc.