The Generation Effect in Advertising Appeals

David J. Moore, University of Oklahoma
Richard Reardon, University of Oklahoma
Francis T. Durso, University of Oklahoma
ABSTRACT - When subjects are allowed to generate for themselves semantic responses to experimental stimuli (internally-generated information), memory performance is often better than if they are allowed to be mere passive recipients of the information (externally-presented information). This paper reviews the current research related to this generation effect", and presents the findings of two experiments. The first demonstrates the role of the degree of elaboration of semantic networks in facilitating a generation effect. The second, using radio broadcast materials, confirms that the generation effect can also be produced in an advertising context. Implications for marketing researchers and directions for future research are discussed.
[ to cite ]:
David J. Moore, Richard Reardon, and Francis T. Durso (1986) ,"The Generation Effect in Advertising Appeals", in NA - Advances in Consumer Research Volume 13, eds. Richard J. Lutz, Provo, UT : Association for Consumer Research, Pages: 117-120.

Advances in Consumer Research Volume 13, 1986      Pages 117-120


David J. Moore, University of Oklahoma

Richard Reardon, University of Oklahoma

Francis T. Durso, University of Oklahoma


When subjects are allowed to generate for themselves semantic responses to experimental stimuli (internally-generated information), memory performance is often better than if they are allowed to be mere passive recipients of the information (externally-presented information). This paper reviews the current research related to this generation effect", and presents the findings of two experiments. The first demonstrates the role of the degree of elaboration of semantic networks in facilitating a generation effect. The second, using radio broadcast materials, confirms that the generation effect can also be produced in an advertising context. Implications for marketing researchers and directions for future research are discussed.

A large body of recent research in cognitive psychology has demonstrated that retention of experimental verbal material may be more significantly enhanced when subjects are allowed to be active cognitive participants in the learning process rather than mere passive recipients of the experimental stimuli (e.g., Jacoby 1978; Slamecka & Graf 1978; Tyler, Hertel, McCallum, & Ellis 1979). In other words, memory tends to be better when subjects are forced to engage in a greater degree of thinking in order to generate a response to an experimental stimulus or problem (internally-generated information) than when those same subjects are presented all the answers/solutions by the experimenter (externally-presented information). We believe that this phenomenon has significant implications for research involving memory for advertising materials.

Most of the current literature in advertising has focused on memory for perceptual events derived from externally-presented stimuli, ignoring the potential influence of internally-generated information (Edell & Staelin 1983; Kisielius & Sternthal 1984; Saegert 1978; Saegert & Young 1982). There have been a few attempts to examine the role of internal generation on the memorability of advertising stimuli. For example, Reimbach and Jacoby (1969) found better recall of incomplete versions of TV ads compared to complete versions. MacLachlan and Jalan (1984) reported better memory for ads that were incomplete or preceded by question prompts. However, there has been an obvious lack of a unified theoretical framework which could give guidance to this area of research. Information processing interpretations (e.g., Craik & Lockhart 1972) are intermingled with drive reduction explanations (e.g., Berlyne 1965; Zeigarnik 1927).

More attention should perhaps be paid to the growing body of research in cognitive psychology that relates to specific, testable issues concerning cognitive generation and memory. A consideration of this literature may provide a basic-process oriented theoretical framework which is more useful in directing the applied research in advertising and marketing communications. For example, a recent line of research in cognitive psychology on the "generation effect" (Slamecka & Graf 1978) may have something very compelling to say about the basic processes underlying cognitive generation.

This paper will review the current research related to the generation effect and then present the results of two experiments. The first demonstrates some of the conditions under which the generation effect is likely to occur. The second demonstrates the generalizability of the generation effect to a broadcast advertising context. The paper concludes with a discussion of the implications for advertising researchers and practitioners, and directions for future research.


The generation effect refers to the greater memorability of internally-generated versus externally-presented stimuli. Slamecka and Graf (1978) conducted a series of experiments with the following basic paradigm. Subjects were presented with a number of complete and incomplete word pairs. Their task was to read the complete pairs, and to read the first word and covertly generate the second word of the incomplete pairs. The complete pairings constituted the externally-presented conditions; the incomplete pairings, the internally-generated conditions. In an external-presentation condition, for example, subjects would read both the stimulus word "rapid," and the response word "fast." In an internal-generation condition, subjects would be shown the word rapid," but would have to generate the response word given only a first letter cue, f___." The generation effect was found with several different encoding rules (e.g., the pairs were synonyms, antonyms, homonyms, and so forth), and, of course, with appropriate counterbalancing of materials across conditions. Free recall and recognition confidence performance were better for the internally-generated words than for the externally-presented words.

Slamecka and colleagues turned to a consideration of possible theoretical explanations for the phenomenon. Several experiments implicated the involvement of semantic memory. McElroy and Slamecka (1982) noted that the effect was not found when subjects generated nonwords. Graf (1982) found the effect for completions of meaningful sentences but not for completions of anomalous sentences. Slamecka and Fevreiski (1983) demonstrated that the activation of the correct semantic network that presumably occurs when subjects attempt to generate the correct word, even if they fail to actually generate Lt, is sufficient for subjects to better remember the correct word compared to a word they read. Additionally, Slamecka and Fevreiski looked at incorrect generations and found that the more semantically related the error to the desired correct response, the larger the generation effect. Gardiner and Rowley (1984) have neatly concluded that Zany type of item may give rise to the generation effect provided that it activates some existing representation in some knowledge system in semantic memory" (p. 445).

We turn now to a presentation of our experiments. For the first, a more complete account was presented in a recent professional symposium (Reardon & McGahan 1985). However, we felt it important to give a summary here because it confirms the importance of semantics, specifically from the perspective of degree of elaboration of semantic networks, in facilitating a generation effect. In the second, we attempted to extend the generation effect to an advertising message retention problem.


If, as Slemacka and colleagues have argued, the generation effect rests on the reliance on semantic features compared with surface features, it should have important implications for memory differences between experts and novices. This, in turn, may be an important consideration in advertising communications because expertise defines a dimension on which consumers vary with respect to products (Conover 1981; Marks & Olson 1981). Some consumers are familiar and involved with a particular class of products, or with a societal problem or need which the class of products proposes to meet. Others are not so familiar. Research on expertise has shown that such differences can be considered in terms of the degree of elaboration of semantic networks in memory

Experts differ from novices in having a larger knowledge base in the area of expertise that benefits from an efficient chunking of smaller units into larger units based on the meaningful relations among the smaller units (Charness 1976; Chase & Simon 1973; Egan & Schwartz 1979; Newell & Simon 1972). Additionally, the units for experts are connected extensively, and often redundantly, with other units in the domain (Anderson 1976; Hayes-Roth 1977). Experts organize domain-relevant stimuli conceptually, at a deeper more abstract level, whereas novices organize domain-relevant stimuli using surface or syntactic features (Adelson 1981; Chase & Ericsson 1981; Chi, Feltovich, & Glaser 1981). The rich knowledge base of the experts together with the reliance of novices on surface features leads to the expectation that the generation effect should be attenuated for novices in general, and for experts outside of their area of expertise.

One of the primary purposes of Experiment 1 (Reardon & McGahan 1985) was to examine the proposed relationship between the generation effect and expertise. Experts (22 psychology faculty and advanced graduate students) and novices (22 junior-senior level psychology majors; the male-female ratio was the same for both experts and novices) engaged in a task requiring them to generate some words and read others (e.g., Durso, Reardon, & Jolly 1985; Johnson, Raye, & Durso 1980). They saw or generated words in a domain for which only one group was expert (psychology: domain-relevant) and words in a domain for which both groups were comparable (sports/ leisure: domain-irrelevant). Recognition memory was then tested by asking subjects to identify words as ones they had read, ones they had generated, or as ones new to the experiment (Johnson & Raye 1981). We expected differences in the magnitude of the generation affect only with materials for which the groups differed in expertise.

Our results showed that superior memory for internally-generated events relative to externally-presented events was greatest for experts dealing with events from their area of expertise (i.e., the critical test, the internal/ external source x domain x expert/novice interaction, was significant, F(1.42)=4.47,p < .05. Thus, the expectation that experts would show a large generation effect, but only in their area of expertise was confirmed.

Consistent with Slamecka and Fevreiski's (1982) contention, experts, who tend to process domain-relevant information semantically, showed a strong generation effect relative to novices, who tend to rely more heavily on surface features. The mechanism underlying the generation effect may be that richly interconnected networks provide multiple access routes to the to-be-retrieved concept. These access routes are needed in order to generate the word initially, but they are not important when the subject simply reads the word presented by the experimenter. The fact that experts and novices differed little in their recognition of externally-presented psychology terms supports the idea that the expertise advantage in memory lies in the experts' ability to utilize the rich knowledge structure. When this knowledge structure is not brought to bear, as when subjects read a terminal word the expertise advantage is not found.

What do these results suggest for advertising communications? Simply that the generation effect is most likely to occur when the stimulus materials (e.g., an advertising appeal) activate a rich semantic network/knowledge base. That is, a generation effect may occur only when consumers are involved/familiar with a product area and the class of products designed to address that area. Only then can cognitive generation tap, and thus take advantage of an extensive, redundantly connected network. In the experiment below, we attempted to extend the above findings and theoretical discussion to a commercial message retention problem.



Subjects. Subjects were 62 undergraduate business administration students who participated for course extra-credit.

Procedure. A sugar-free high-nutritional soft drink was selected as the target product. Informal discussions conducted with other groups of subjects from the same population had revealed that soft drinks were frequently consumed by our population, and that the ingredients of these drinks were often of concern. The product was positioned as a new brand of soft drink just being introduced to the market. A fictitious brand name, Prime Time, was chosen in order to minimize any experimental complications caused by varying degrees of prior familiarity with well-known brands.

In order to minimize possible sources of reactivity within the design and prevent any artificial set to respond unnaturally to the stimulus, the target ad was embedded within a 15 minute excerpt of radio broadcast material containing 10 other advertisements and several songs. The broadcast material (which actually could have been presented on the air) was taped and appropriately edited in the studio of a local radio station, and the target ad was read by a professional radio announcer. Subjects were led to believe that the material was taken directly from a radio broadcast. The target ad itself (almost 70 seconds long) was designed so that the copy was delivered with the same cadence and inflection as many radio ads that use mock news bulletin- formats. It stressed the following points: (1) Recent findings of medical science have shown that a high intake of sugar (usually found in soft drinks) reduces the body's ability to fight disease. (2) In response to the health needs of consumers, General Foods has developed Prime Time, a new, delicious, sugarless soft drink sweetened with aspertane. (3) Prime Time is fortified with vitamin B complex and vitamin C to help build resistance to disease and infections.

Scripts for both versions of the ad were essentially identical, with the product name being mentioned four times during the course of the ad. The manipulation of the independent variable occurred at the end of the ad. In the external-generation condition, the ad ended with an overt mention of the target product brand name. For the internal-generation condition, the end of the ad was structured so that subjects would covertly generate the brand name. This version ended with the words "Hey, what was the name of that new soft drink again....?"

Two days after the experimental exposure, subjects were surprised with a delayed recognition test. They were presented with a list of 13 brand names. Included in this list, and randomly ordered, were 7 of the brand names heard on the tape, 5 new brand names (foils) that were similar to the names heard, but which were not in fact on the tape, and the target brand name, Prime Time. Using a scale from 1 (not confident at all) to 7 (extremely confident), subjects were asked to indicate the degree to which they were confident that a given brand name was one that was mentioned on the tape they heard. Finally, to confirm the degree of semantic memory engagement, and to ensure that this engagement was comparable for both conditions, subjects were asked to respond to 8 items that assessed their interest and knowledge of health, nutrition, and concern with product ingredients (e.g., frequency of exercise, regularity of vitamin consumption, "junk" food habits, frequency of keeping up with nutritional news, etc.). Subjects responded to these items on 7-point scales, where higher numbers indicated greater familiarity/involvement.

Results and Discussion

Mean recognition confidence for the taret product for the external-presentation group was 3.83; for the internal-generation group it was 6.25. These means were significantly different t(60)=4.42, 2 < .0001. In contrast, mean confidence responses of the two groups to other products that actually appeared in ads on the tape did not significantly differ (external-presentation: 4.17; internal-generation: 4.74). An examination of false positive responding, i.e., recognition confidence that the foils appeared on the tape, was undertaken to test for a possible recognition response bias on the part of either group. If anything, there was a slightly greater tendency on the part of the external-presentation group to indicate greater confidence that a foil was on the tape, but the difference between groups was not significant (external-presentation: 2.45; internal-generation: 1.96). The recognition confidence results must be considered in light of our postexperimental checks on subject familiarity/involvement. These measures confirmed that our subjects were, in fact, reasonably well-informed and concerned about health and the nutritional value of the foods they consume (both groups' mean scores on these items were above the midpoint of the 7-point scale in the direction of greater familiarity/involvement; external-presentation group: 4.21; internal-generation group: 4.02). Importantly, there were no differences in involvement/familiarity between the two groups. As a precaution, the measures of involvement/familiarity were covaried (individually, and as a sum) from the target recognition scores. The difference between groups remained significant (all Fs > 15.00 with 1 and 59 df), and none of the covariates reached significance (all Fs < 1.00).

One concern applied researchers have had with basic research in cognitive psychology is that it is conducted in artificial settings with socially impoverished stimuli (cf, Taylor 1981). Therefore, the generalizability of effects produced there are sometimes suspect until they are demonstrated in more ecologically valid settings. In the case of our study, we were able to replicate a finding in basic cognitive psychological research with "real world" materials in an applied context. We took two groups of relatively involved and aware subjects, and with broadcast advertising stimuli, produced a generation effect in the condition in which one was expected with no effect in the other condition. Experiment 2 is, of course, preliminary. By choosing a stimulus domain with which all of our subjects were fairly familiar, we could not test adequately the familiarity/involvement dimension. One should also be cautious about a result that is based on a single ad. We are currently planning more comprehensive tests of the generation effect in the marketing-advertising domain that involve a more systematic manipulation of generation and involvement/familiarity across a number of different types and modes of advertising presentations. Nonetheless, experiment 2 opened the door on this research direction for us, and hopefully for others in marketing and advertising research who may share our interests.


What does the generation effect literature offer to marketing researchers and practitioners? At a concrete level, it suggests a technique for getting people to better remember certain aspects of an advertising appeal (e.g., brand name). There are boundary conditions on this, however. The audience should have a reasonably extensive knowledge base/semantic network which can be activated by the generation process.

The generation effect literature, at a more general level, offers something important that goes beyond a simple, concrete advertising technique. For example, the theoretical underpinnings associated with the generation effect may contribute something unique to recent studies in persuasive communications based on cognitive response/ cognitive structure models (e.g., Olson, & Dover 1982;

Wright 1973; among others). These models have been used to explain how the message recipient's cognitive responses (defined as subvocal spontaneous thoughts generated during exposure to a persuasive message) mediate subsequent attitudinal and behavioral responses. The generation effect posits that elaborate knowledge bases/semantic networks are activated by the generation process. These networks are made up of meanings, relations, rules, images, and so on, but also include evaluative/emotional states and reactions. The generation effect is simply a manifestation of an activation of a semantic network. The cognitive response models may involve the same basic process.

The most extensive model of persuasion that includes cognitive response components is Petty and Cacioppo's (1983) Elaboration Likelihood Model (ELM). Nevertheless, Petty and Cacioppo (1985) admit that the ELM is a particularly simple and general framework, which does not -ultimately indicate why certain variables affect information processing" (p. 86). The generation effect phenomenon can conceivably provide a basic process explanation for that aspect of the ELM that deals with self-generated responses to persuasive appeals. To a limited extent, both the ELM and the generation effect make similar predictions concerning the message recipient's responses to various experimental stimuli. For example, according to the generation effect literature, internal generation leads to better memory performance when the person making the generation is equipped with a rich knowledge base for which the generation is relevant. Similarly, it may be inferred from the ELM that when the motivation and ability to respond to a message is high (as, for example, when a subject has high prior knowledge of or familiarity with the attitude issue) message elements may become more salient in memory, leading to better memory performance. Indeed, Cacioppo, Harkins & Petty (1981) have demonstrated that the advantage of self-generation over passive reception in persuasion settings is more pronounced when message recipients possess some prior knowledge base concerning a given topic, or are equipped with some relevant schema.

Future research will have to be directed toward a confirmation of the linkage of the generation effect to other cognitive response/cognitive activity models. Certain conceptual issues and consistency of terminology will have to be addressed if the two research areas are to be linked successfully. For instance, cognitive response researchers typically make a distinction between memory (defined as recall or recognition of message elements) and cognitive responses (Petty, Ostrom, & Brock 1981). Cognitive psychologists, as we have seen, make no such distinction. For them, message elements and generations are both memorial, but they have different sources (external versus internal; Johnson & Raye 1981; Slamecka & Graf 1978). We think the latter formulation is the more useful, particularly in helping provide a link between marketing research and current thinking in cognitive psychology.

Finally, it might be interesting to establish a linkage with the reality monitoring literature (Johnson & Raye 1981). Reality monitoring deals with the confusions people tend to make between internal and external sources of memories. Johnson and Raye (1981) have demonstrated that individuals make surprisingly many of these confusions. That people often confuse what they think with what they hear and see may have important implications for research on persuasive appeals.


Adelson, B. (1981), "Problem Solving and the Development of Abstract Categories in Programming Languages, Memory & Cognition, 9, 422-433.

Anderson, J. R. (1976), Language, Memory, and Thought, Hillsdale, NJ: Erlbaum.

Berlyne, D. E. (1965), Structure and Direction in Thinking, New York: Wiley.

Cacioppo, J. T., Harkins, S. G., and Petty, R. E. (1981), "The Nature of Attitudes and Cognitive Responses and Their Relationship to Behavior," in Cognitive Responses in Persuasion, R. Petty, T. Ostrom, and T. Brock, eds. Hillsdale, NJ: Erlbaum.

Charness, N. (1976), "Memory for Chess Positions: Resistance to Interference," Journal of Experimental Psychology: Human Learning and Memory, 2, 641-653.

Chase, W. G., and Ericcson, K. A. (1981), "Skilled Memory,- in Cognitive Skills and Their Acquisition, J. R. Anderson, ed. Hillsdale,-NJ: Erlbaum.

Chase, W. G., and Simon, H.A. (1973), "Perception in Chess," Cognitive Psychology, 5, 55-81.

Chi, M. T. H., Feltovich, P. J., and Glaser, R. (1981), "Categorization and Representation of Physics Problems by Experts and Novices," Cognitive Science, 5, 121-152.

Conover, J. N. (1981), "Familiarity and the Structure of Product Knowledge,- Advances in Consumer Research, Vol. 9, A. A. Mitchell, ed. Ann Arbor, MI: Association for Consumer Research, 494-498.

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

Durso, F. T., Reardon, R., and Jolly, E. (1985), "Self-nonself Segregation and Reality Monitoring, Journal of Personality and Social Psychology, 48, 447-455.

Edell, J. A., and Staelin, R. (1983), "The Information Processing of Pictures in Print Advertisements," Journal of Consumer Research, 10, 45-61.

Egan, D. E., and Schwartz, B. J. (1979), "Chunking in Recall of Symbolic Drawings, Memory & Cognition, 4, 541-547.

Gardiner, J. M., and Rowley, J. M. C. (1984), -A Generation Effect with Numbers Rather than Words," Memory & Cognition, 12, (5), 443-445.

Graf, P. (1982), "The Memorial Consequences of Generation and Transformation," Journal of Verbal Learning and Verbal Behavior, 21, 539-548.

Rayes-Roth, B. (1977), "Evolution of Cognitive Structures and Processes,- Psychological Review, 84, 260-278.

Heimbach, J. T., and Jacoby, J. (1972), "The Zeigarnik Effect in Advertising. In M. Venkatesan (ed.), Proceedings of the Third Annual Conference of the Association for Consumer Research. College Park. MD. 3. 746-757.

Jacoby, L. L. (1978), "On Interpreting the Effects of Repetition: Solving a Problem Versus Remembering a Solution," Journal of Verbal Learning and Verbal Behavior, 17. 649-667.

Johnson, M. K., and Raye, C. L. (1981), "Reality Monitoring," Psychological Review, 88, 67-85.

Johnson, M. K., Raye, C. L., and Durso, F. T. (1980), "Reality Monitoring: Second Perceptions and Thoughts," Bulletin of the Psychonomic Society, 15, 402-404.

Kisielius, J., and Sternthal, B. (1984), "Detecting and Explaining Vividness Effects in Attitudinal Judgments," Journal of Marketing Research, 22, 54-64.

Marks, L. J., and Olson, J. C. (1981), "Toward a Cognitive Structure Conceptualization of Product Familiarity," Advances in Consumer Research, Vol. 8, K. B. Monroe, ed. Ann Arbor, MI: Association for Consumer Research, 145-150.

McElroy, L. A., and Slamecka, N. J. (1982), Memorial Consequences of Generating Nonwords: Implications for Semantic-Memory Interpretations of the Generation Effect, Journal of Verbal Learning and Verbal Behavior, 21, 249-259.

MacLachlan, J. (1984), making a Message Memorable and Persuasive, Journal of Advertising Research, 23, 51-59.

MacLachlan, J., and Jalan, P. (1984), "The Effect of Questions and Ambiguity on Brand Recall of Advertised Products. Unpublished manuscript, Rensselaer Polytechnic Institute, Troy, NY.

Newell, A., and Simon, H. A. (1972), Human Problem Solving, Englewood Cliffs, NJ: Prentice-Hall.

Olson, J. C., Toy, D. R., and Dover, P. A. (1982), "Do Cognitive Responses Mediate the Effects of Advertising Content on Cognitive Structure?," Journal of Consumer Research, Vol. 9, 245-262.

Petty, Richard E. and John T. Cacioppo (1985), "The Elaboration Likelihood Model of Persuasion, in Berkowitz, L. (ed.) (1985), Advances in Experimental Social Psychology, 19, New York: Academic Press (Forthcoming).

Petty, Richard E., John T. Cacioppo and David Schumann (1983), Central and Peripheral Routes to Advertising Effectiveness: The Moderating Role of Involvement, Journal of Consumer Research, 10, 134-148.

Petty, R. E., Ostrom, T. M., and Brock, T. C. (1981), Cognitive Responses in Persuasion, Hillsdale, NJ: Erlbaum.

Reardon, R., and McGahan, J. R. (1985, April), Expertise and Internal versus External Sources of Memories, in Human Expertise: Papers in Honor of William Chase, F. T. Durso (Chair), Symposium Conducted at the Meeting of the Southwestern Psychological Association, Austin, Texas.

Saegert, J. (1978), "A Demonstration of Levels of Processing Theory in Memory for Advertisements," Advances in Consumer Research, Vol. 6, W. L. Wilkie, ed. Ann Arbor, MI: Association for Consumer Research. 82-84.

Saegert, J., and Young, R. K (1981), "A Comparison of Effects of Repetition and Levels of Processing in Memory for Advertisements," Advances in Consumer Research, 8, K. B. Monroe, ed., Ann Arbor, MI: Association for Consumer Research, 431-434.

Slamecka, N. J., and Fevreiski, J., (1983), "The Generation Effect When Generation Fails," Journal of Verbal Learning and Verbal Behavior, 22, 153-163.

Slamecka, N. J., and Graf, P. (1978), "The Generation Effect: Delineation of a Phenomenon," Journal of Experimental Psychology: Human Learning and Memory, 4, 592-604.

Taylor, S. E. (1981), "The Interface of Cognitive and Social Psychology," in Cognition Social Behavior, and the Environment, J. H. Harvey,ed. Hillsdale, NJ: Erlbaum.

Tyler, S. W., Hertel, P. T., McCallum, M. C., and Ellis, H. C. (1979), Cognitive Effort and Memory, Journal of Experimental Psychology: Human Learning and Memory, 5, 607-617.

Wright, P. (1973), "The Cognitive Processes Mediating Acceptance of Advertising," Journal of Marketing Re- search, Vol. 10, pp. 53-62.

Zeigarnik, B. (1927), "Uber Das Leehalten von Erledigten und Unerleigten Hand Bengen," Psychol. Forsch., 9, 1-85.