Advances in Consumer Research Volume 9, 1982 Pages 431-434
COMPARISON OF EFFECTS OF REPETITION AND LEVELS OF PROCESSING IN MEMORY FOR ADVERTISEMENTS
Joel Saegert, The University of Texas at San Antonio
Robert K. Young, The University of Texas at Austin
Level of processing orientation (shallow vs deep) and number of repetitions (one vs two) were varied simultaneously in an advertising memory experiment. One deep processing presentation of advertisements was found to produce greater recall than two shallow processing presentations. An interaction was also found such that the increase in percent recall from one to two repetitions was substantially greater for deep processing, compared to shallow. Thus, the type of processing and the number of repetitions combined to yield the best advertisement memory in the study.
Despite recent controversy concerning its validity (e.g., Krugman, 1977), day-after recall remains one of the most widely used indicators of advertising effectiveness (Wall Street Journal, 1980). While this technique can make few claims of effectiveness in terms of attitude change, purchase intentions, or actual consumer behavior, demonstration of day-after recall of an ad at least gives an indication of potential brand name awareness, presumably a component of ad effectiveness. Thus, a major advertising goal can be said to be the creation of a brand name memory trace in the cognitive structure of the target viewer which will be maintained over time, at least until a purchase opportunity becomes available.
A number of consumer researchers (e.g., Mitchell, 1978; Olson, 1978; Saegert, 1978) have discussed advertising memory from the standpoint of Craik and Lockhart's (1972) "levels-of-processing theory." This memory research paradigm directs the attention of subjects (in this case, viewers of ads) to the semantic and/or personal features of stimuli (brand names) as opposed to physical features. The procedure used requires research subjects to answer a question about each of a number of (brand name) stimuli. Although no indication is given that memory for the stimuli is involved, a surprise recall and/or recognition task is then given to ascertain effects of different types of questions on memory. The levels-of-processing variable is manipulated by asking questions concerning meaning (Have you bought this brand before?) or form (Is the brand name in script letters?). An experiment by Saegert (1978) found recall for items about which semantic (deep processing) questions were asked to be approximately two times that for form (shallow processing) questions. Subsequently, Reid and Soley (1980) replicated this effect for retention of brand names in television commercials, demonstrating that the levels-of-processing principle can be applied to the day-after recall concept.
Although the levels-of-processing proposal as a theory of memory has been greeted with some reservation by psychologists (e.g., Nelson, 1977; Baddeley, 1978), as well as consumer behaviorists (Bettman, 1979), there is little cause for doubting that substantial differences in retention follow operations designed to manipulate the levels-of-processing variable. Some of the criticisms have been answered recently in the psychology literature. For example, the argument that a non-circular operational definition had not been developed was answered by Seamon and Virostek (1978) in a study which obtained an independent establishment of a hierarchy of processing levels, and which subsequently demonstrated differential memory performance as a function of the levels variable. Even if the theoretical basis for the "levels" effect is not, as yet, completely worked out, it seems clear that material is better remembered when questions about the stimulus refer to some meaningful aspect, such as past personal experience with the stimulus object, compared to questions about formal c t is us features.
The introduction of the levels-of-processing notion into advertising theory focuses attention on the character or quality of advertising design, as opposed to the quantity or frequency of advertising presentation. The frequency or repetition variable has been investigated extensively, with an eye to determining the number of presentations needed for maximum effectiveness (Sawyer, 1974), and from the standpoint of determining when advertising "wear-out" begins to occur following frequent exposure (Craig, Sternthal ant Leavitt, 1976; see Engle, Blackwell and Kollat, 1978, for a comprehensive review). In the present study, both the levels-of-processing variable and the frequency of presentation were manipulated simultaneously to determine the relative effects of the two, as well as to look for a possible interaction between them. It has been suggested (Saegert, 1978; Reid and Soley, 1980) that subsequent repetitions of ads may simply serve as opportunities for consumers to perform more of the type of elaborative activities hypothesized to occur in the "deep processing" operationalization. If this is so, it should be found that a second repetition of a brand name will increase memory following deep processing more than following shallow processing. Thus, the present investigation was specifically designed to compare relative effects of levels of processing ant repetition as well as to look for an interaction effect between the two variables
Subjects. The subjects of the experiment were 117 volunteer undergraduate students at The University of Texas at Austin who were fulfilling a research participation requirement for an introductory psychology course. There were between 12-16 subjects in each of eight experimental groups.
Design. The subjects' memory for advertising was studied by presenting them a series of slides containing magazine ads for a variety of well-known national brands (e.g., Arrow, Toyota, Wesson Oil, etc.). The variables of interest in the study were the presentation frequency (once or twice), the levels of processing (shallow vs deep) and the retention interval (immediate memory test vs 24-hour delay). Finally, a counterbalancing variable, by which each stimulus ad was presented with both deep and shallow level questions, was included to rule out the possibility that any levels-of-processing effects obtained in the experiment could be list-specific.
Manipulation of the presentation frequency was carried out simply by presenting each stimulus slide to half of the subjects for one presentation while the other half of the subjects viewed two presentations of the same slides. In this latter case, all of the slides were shown before any slide was shown a second time, with the second order of presentation being different from the first.
The levels-of-processing variable was manipulated in the manner used in previous studies (e.g., Saegert, 1978; Reid & Soley, 1980). Briefly, for half the slides, tape recorded questions were asked such that the brand names would be considered as simple aggregates of letters (e.g., "Is the brand name in blue letters?"). For the other half of the slides, the taped questions were designed to have the brand names treated as if they had meaning in the context of past experiences (e.g., "Rave you heard of this brand name before?"). Following earlier convention, these conditions are referred to as shallow and deep processing, even though a number of authors (e.g., Nelson, 1977; Baddeley, 1978) have questioned whether attributing the observed effects to such theoretical processes as "levels" is warranted (other labels which have been used include "non-semantic" vs "semantic" processing).
For the retention interval variable, subjects were either given retention tests (recall followed by recognition) immediately following presentation of the slides or they were asked to return 24-hours later. In neither case were they told that they would be asked to remember the brand names depicted in the slides.
The counterbalancing variable was designed such that two sets of questions were recorded, Tape A and Tape B. For Tape A, half the presented advertisements had deep processing questions while the other half had shallow, with the order of the questions varied randomly. For Tape B, ads having a deep processing question on Tape A received a shallow question and vice versa. Thus, in the analysis of the results, the true levels-of-processing effect was contained in a Tape (A vs B) by Ad Set (l vs 2) interaction. All references to the levels-of-processing effect in the discussion of the results refer to this counterbalancing interaction. The design of the experiment was thus a 2 x 2 x 2 x 2 factorial with Presentation Frequency (once or twice), Retention Interval (O or 24 hours) and Tape (A or B), serving as between -subject variables and Ad Set (l or 2) serving as a within-subject variable; again, the levels-of-processing effect is represented by the Tape by Ad Set interaction.
Procedure. Groups of 5 to 8 subjects were tested in a small semi-darkened amphitheater. Taped instructions were played to the subjects and any questions about the procedure were answered. As each slide was presented, a question was heard from the tape recorder followed by a 4-sec. silence to allow the subjects to check YES or NO on their answer sheets in response to the question asked on the tape. For each set of 50 slides, 25 shallow processing and 25 deep processing questions, similar to the examples above, were asked. In all, six different questions were used for shallow and six for deep processing. The order of the questions was random and the total time allotted for asking each type of question was the same for deep and shallow processing. After the 50 slides had been presented, the first YES/NO answer sheet was collected. In the case of those groups getting two presentations a second answer sheet was then distributed and a new tray with the same 50 slides in a different random order, along with a second set of taped questions, was played. In the second presentation, a different question of the same type as that asked during the first presentation was asked for each slide.
During the retention phase of the experiment, the subjects were either measured immediately after their answer sheets were collected or after they returned the next day. They were asked to write down as many of the brand names as they could on a blank sheet of paper. Two minutes were allocated for recall after which these papers were collected. Next, a check list with 80 brand names was distributed and the subjects were then asked to indicate, by checking the appropriate column, which brands had been previously presented and which had not. To reduce serial position effects, only the middle 40 brand names (i.e., slides 6 through 45) were included in the check list. In addition, 40 new brand names, not previously presented, were added to the recognition check list in random order. Each original brand name was paired with a new brand name (e.g., an Exxon ad was one of the slides presented and both the Exxon and Texaco brand names were included on the check list).
Recall. In analysis of the mean number of brand names recalled (out of the 40 middle names), recall was found to be poorer after a 24-hour retention interval (mean = 5.70 out of 40) than after an immediate test (8.84), with F (df = 1,101) = 36.89, p <.001. [All F's reported were calculated via unweighted means analysis of variance.] In addition, more brand names were recalled after two presentations (8.26) than after one presentation (6.28), with F (df = 1,101) = 14.69, p <.001. A third main effect, Ad set, was also significant, with F (df = 1,101) = 13.15, p < .001. Comparison of the two sets of advertisements indicated that Michelob and Smirnoff ads were in the ad set which was easier to recall. It is hypothesized that these brands may have been differentially memorable because they were more salient to the student subjects of the study or because of an inherently greater simplicity of Liquor Ads (it has been reported that Starch Adnorms tend to be higher for beer ant liquor ads than for other product categories. [The authors thank an anonymous reviewer for this observation.])
Of central interest in the experiment was the relative effect of the levels-of-processing orientation given by the taped questions. This effect was significant, with F (df = 1,101) = 142.32, p <.001. When a subject was asked a question like, "Is the brand name in blue letters?", the mean number of brand names remembered was equal to 2.30 out of 20; when the subject was asked a question like, "Have you ever used this brand?"; the mean number of brand names recalled was 4.98-a greater than two-to-one advantage. Thus, the levels-of-processing variable was significant in the direction predicted, and is consistent with the previous demonstrations of the effect by Saegert (1978), and Reid and Soley (1980).
Even though a substantial recall advantage for deep brand names was observed, the difference between the two levels of processing may actually be underestimated because certain ads with a great deal of salience to a college audience (e.g., brand names such as Michelob and Smirnoff) would be likely to be well remembered even if the question they were asked was a shallow processing one. Some supportive evidence for this view is found in the interaction between levels of processing and retention interval. This interaction was significant, with F (df = 1,101) = 17.569 P < .001, and resulted from greater forgetting of the brand names in the deep processing condition than for those brand names in the shallow processing condition. Greater resistance to forgetting for shallow brand names would be expected if those few brand names which were remembered under the shallow processing condition were remembered because the product shown had a greater individual impact than would be expected on the basis of the question asked.
Further evidence supportive of this position is found in the interaction between number of presentations ant level of processing. These data are presented in the left hand portion of Table l (see next page). The interaction shown is significant, with F (df = 1,101) - 9.72, p < .01. As can be seen, a second presentation increased retention of deeply processed brand names by 41 percent, while a second repetition increased retention of shallowly processed brand names by only 14 percent. Using Newman-Keuls procedures suggested by Winer (1971), it was fount that the increase Ln retention from one to two presentations for the deeply processed brand names was significant, p < .01, while the same increase for the shallowly processed brand names was not significant, p > .05. Similarly, more brand names were recalled following a single presentation of deep processing (4.13) than were recalled following two presentations of shallow processing (2.45), p <.01.
RECALL AND RECOGNITION AS A FUNCTION OF NUMBER OF PRESENTATIONS AND DEEP VS. SHALLOW PROCESSING QUESTIONS
Recognition. The number of correct recognitions of presented ads was the basis of the analysis of the recognition data. In this analysis, retention was better after two presentations than after one (27.16 of 40 brand names vs 31.78), with F (df = 1,10.) = 17.25, p <.001. [No correction for guessing was computed. However, analysis of previous studies using identical procedures indicated equivalent results with or without a correction for guessing.] As in the analysis of the recall data, the brand names in the deep processing condition were remembered better than were those in the shallow processing condition (17.05 vs 12.73), with F (df = 1,101) = 194.78, p < .001. Also, as before, two presentations with shallow processing resulted in lower performance than one presentation with deep processing. However, in this case (right-hand portion of Table l), performance increased from one to two presentations for both shallow and deep processing conditions (p's <.01); moreover, in contrast to the recall data, the percent increase was greater for shallowly processed ads than for deep (217. vs 10%). This almost certainly results from a ceiling effect for recognition performance; that is, since 16 of 20 ads were correctly recognized following one deep presentation, the room for improvement following two presentations was relatively small.
A second analysis of the recognition data added an addition al within-subject variable to the design by including correct rejections of brand names which were not presented. The results of this analysis were essentially the same as in the analysis of presented items except that the ANOVA was based on correct responses to 80 ads (40 presented and 40 non-presented rather than only the 40 presented ads). The only new significant effect of interest was that of retention interval, which, like the recall data showed a decrease in performance over the 24-hour period, with F (df = 1,101) = 12.25, p <.001. A mean of 67.04 brand names (out of 80) were correctly classified immediately after presentation, but after 24 hours this fell to a mean of 62.56. Thus, when the correct rejection of brand names not presented are combined with correct recognitions of presented items, performance declined over the 24-hour period.
Although the levels-of-processing paradigm is only a rough analogy to the processes involved in advertising effectiveness, the distinction between deep ant shallow processing draws attention to the importance of "semantic" or "elaborative" encoding of material if memory is to be well maintained. The present study implies that repetition per se is not very effective in maintaining recall for an ad; what seems to be important is the kind of processing that occur during or immediately following the presentation. The experiment supports the idea that a repetition provides an opportunity for deep processing operations to be carried out if appropriate orientation (semantic vs formal) is given.
The interaction between levels of processing and number of presentations fount in the present experiment has been shown earlier with word stimuli by Craik and Tulving (1975, Experiment 3). Jacoby, Bartz and Evans (1978) have taken this result as evidence for a multiplicative theory of the relationship between levels of processing and repetition. That is, the combination of deep processing and several presentations results in greater performance than would be expected on the basis of either variable alone. Jacoby, et al. refer to the underlying process as being a hierarchical one involving l) organization of items in memory and 2) the ease of recall of individual items. They further suggest that level of processing enhances "between-unit organization" while repetition enhances integration of items within units. In the present case, this argument would specify that thinking about a brand name under deep processing conditions promotes associative organization between organizational units containing the to-be-remembered brand name while successive presentations strengthens the integration of the brand name within those units. Thus, deeper levels of processing would increase the probability of accessing a given memory unit containing the co-be-recalled brand name while extended presentations would determine the recall of the brand name, once the unit is accessed. Therefore, in answer to the question of whether quality (deep processing) or quantity (repetition) of advertising presentation is more important, the evidence supports the notion that both of these variables act in concert to enhance memory for the brand name presented in an ad.
Another issue which has elicited considerable discussion is that of whether recall or recognition is a more appropriate test of ad memory (e.g., Krugman, 1977). Krugman maintains that recall does not adequately measure what viewers actually remember since they can recognize ads in checklist with a much higher rate than they can verbalize ads in unaided recall. The present data certainly give some credibility to this argument. Not only was recognition performance characteristically higher, but the superior increase in performance across presentations for deep compared to shallow ad questions found on recall was not observed (indeed, was reversed) for recognition. Thus, even ads for which the recall test had shown minimal learning were found to have a high level of retention when measured by the recognition test.
On the other hand, it seems reasonable to question the quality or strength of an advertisement memory which can only be measured through recognition but not recall. Craik (1979) has recently attempted to shed some light on the "similarities and differences between recall and recognition" in the psychology literature. He suggests that recall may be more indicative of retrieval of stored meanings while recognition provides the meaning and asks only for retrieval of the original episodic context (Tulving, 1972). From an ad memory standpoint, it is important to consider the issue on the basis of the goals an advertiser is trying to accomplish. It is presumed that a goal of most ads is to enhance the memory of the brand name in the "semantic" memory of the viewer, as opposed to the "episodic" memory. That is, it is more important that the consumer become aware of the brand name as an alternative in his/her problem solving set than just that brand name X was mentioned at a given time and place (i.e., on last night's TV offerings). Thus, in defense of the traditional "day after" procedure, recall may be more consistent with an advertiser's goal of establishing an available and retrievable memory trace, especially when the consumer's problem solving behavior is dependent on his/her "stored knowledge" as opposed to the "task environment" (Mitchell, 1978).
One question which reasonably follows from the present study is what strategies are available to advertising practitioners as a function of awareness of the levels-of-processing effect. Unfortunately, the small amount of research to date has not been particularly encouraging in this regard. Reid and Soley (1980) have made suggestions to enhance the memorability of ads based on the levels-of-processing theory. These include "prompting audience members to relate information...to personal experience and knowledge." One way, they argue, that this could be accomplished is by asking viewers to imagine themselves in the context of the product or brand portrayed in the ad. However, recent study by Mowen (1980) has failed to demonstrate that such a procedure enhances (semantic differential) brand evaluations or memory for brand attributes. Whether this single demonstration of the null hypothesis is attributable to ineffective experimental manipulations in the laboratory, or constitutes a genuine disconfirmation of the implications of levels-of-processing theory, is a subject for future investigation. Of course, other operationalizations of the levels concept in an advertising context are possible. It seems safe to say that much further work remains to be done if the processes underlying advertising effectiveness are to be fully understood. The levels-of-processing paradigm seems to provide a useful framework to investigate semantic and frequency effects and should ultimately yield applications for advertising strategists.
Baddeley, Alan D. (1978), "The Trouble with Levels: A Reexamination of Craik and Lockhart's Framework for Memory Research." Psychological Review, 85, 139-152.
Bettman, James R. (1979), "Memory Factors in Consumer Choice: A Review." Journal of Marketing, 43, 37-53.
Craig, Samuel C., Sternthal, Brian and Leavitt, Clark. (1976), "Advertising Wearout: An Experimental Analysis." Journal of Marketing Research, 13, 365-372.
Craik, Fergus I.M. (1979), "Human Memory." In Annual Review of Psychology, Vol. 30, Mark A. Rosenzweig and Lyman W. Porter, eds. Palo Alto, Calif: Annual Reviews Inc.
Craik, Fergus I.M. and Lockhart, Robert S. (1972), "Levels-of-Processing: A Framework for Memory Research." Journal of Verbal Learning and Verbal Behavior, 11, 671-84.
Craik, Fergus I.M., ant Tulving, Endel. (1975) "Depth of Processing and the Retention of Words in Episodic Memory." Journal of Experimental Psychology: General, 104, 268-294.
Engel, James F., Blackwell, Roger D. and Kollat, David T. (1978), Consumer Behavior, Hinsdale, Illinois: The Dryden Press.
Jacoby, Larry L. Bartz, Wayne H., and Evans, James D.(1978), "A Functional Approach to Levels of Processing," Journal of Experimental PsychologY: Human Learning and Memory, 4, 331-346.
Krugman, Herbert E. (1977), "Memory without Recall, Exposure without Perception," Journal of Advertising Research, 17, 7-12.
Mitchell, Andrew A. (1978), "An Information Processing View of Consumer Behavior, " Proceedings, Marketing Educator's Conference, American Marketing Association, 188-197.
Mowen, John C. (1980), "The Availability Heuristic: The Effect of Imagining the Use of Product on Product Perceptions." Proceedings of the American Marketing Association.
Nelson, Thomas D. (1977), "Repetition and Depth of Processing." Journal of Verbal Learning and Verbal Behavior, 16, 151-71.
Olson, J. C. (1978), "Theories of Information Encoding and Storage: Implications for Consumer Behavior." In The Effect of Information on Consumer and Market Behavior, Andrew Mitchell, ed. Chicago: The American Marketing Association, 49-60.
Reid, Leonard N. and Soley, Lawrence C. (1980), "Levels-of-Processing in Memory and the Recall and Recognition of Television Commercials." In Current Issues and Research in Advertising, J. B. Leigh and C. R. Martin, eds. Ann Arbor, Mich., Division of Research, Graduate School of Business Administration, University of Michigan, 135-146.
Saegert, Joel (1978), "A Demonstration of Levels-of-Processing Theory in Memory for Advertisements." Advances in Consumer Research, Vol. 6, William L. Wilkie, ed. Ann Arbor, Mich: Association for Consumer Research, 82-84.
Sawyer, A. G. (1974), "The Effects of Repetition." In Buyer/Consumer Information Processing, G. David Hughes and Michael L. Ray, eds. Chapel Hill: University of North Carolina Press.
Seamon, John G., and Virostek, Susan (1978), "Memory Performance and Subject-defined Depth of Processing," Memory and Cognition, 6, 283-287.
Tulving, Endel. (1972), "Episodic and Semantic Memory," in E. Tulving and W. Donaldson, eds., Organization of Memory, New York: Academic Press, 385-387.
Weiner, B. J. (1971), Statistical Principles in Experimental Design, 2nd Edition, New York: McGraw-Hill Book Co.
Wall Street Journal (July 24, 1980), "Should 200 Viewers Memories Decide Whether Ads Live or Die?"