Advances in Consumer Research Volume 6, 1979 Pages 82-84
A DEMONSTRATION OF LEVELS-OF-PROCESSING THEORY IN MEMORY FOR ADVERTISEMENTS
Joel Saegert, The University of Texas at San Antonio
The levels-of-processing theory of memory maintains that stimulus material will be remembered as a function of the amount of personal elaboration to which it is subjected by the observer. An experiment was conducted which varied level of processing of brand names by directing attention to either physical or semantic aspects. Memory was superior for items presumed to be processed at a deep level.
Olson (1977) has recently discussed the potential of the "levels-of-processing" theory of memory (Craik and Lockhart, 1972) for understanding the processes which underlie advertising recall. This theory postulates that memory is a function of the type of cognitive processes which occur during exposure to a to-be-remembered stimulus and has elements in common with a number of recent theories of advertising effects (e.g., Leavitt, 1974; Wright, 1974).
Craik and Lockhart argued that memory is a function of the "level" to which material is processed and that this is independent of the amount of repetition or rehearsal of the stimulus. While a precise definition of "level" has not been specified, the variable involves the degree to which material is subjected to elaboration by the learner in terms of relating it to his prior experience and knowledge. Items processed to a deeper level (that is, subjected to greater elaboration) are hypothesized to be more readily remembered.
An example of a typical levels-of-processing experiment will serve to illustrate how the concept is operationally defined. Hyde and Jenkins (1973) manipulated the orienting task of a number of groups of experimental subjects. Each group was presented with a long list of words that were to be processed in one of several ways. In one condition, the subjects were asked to look at the words and rate how pleasant or unpleasant each one was on a five-point scale. Another group was simply asked to count and write down the number of times the letter E appeared in each word they were shown. Following this, the subjects in both groups were given a surprise recall task in which they were to write down all the words they could remember. This procedure corresponds to the usual "incidental learning" paradigm used in many memory experiments.
Differences in recall were observed among the groups. The letter counting group was only able to recall about 28% of the words while the pleasantness rating group got 48% correct. Remarkably, this was even better recall than that of a group of subjects who had been specifically instructed to try to learn the words.
It is presumed that the groups differed only in the level of processing of the material since the word list and exposure time were identical across conditions. In terms of Craik and Lockhart's theory, it is argued that when a person is required to ascertain whether a word is pleasant or unpleasant, it must be considered (thought about) in terms of all of its available attributes and associations; such deep processing is sufficient to maintain a high level of memorability. The letter-counting activity, on the other hand, directs the attention to superficial aspects of the material to be remembered, resulting in a shallow level of processing.
The levels theory has been subjected to a great deal of theoretical and empirical elaboration in the past several years. For example, Craik and Tulving (1975) have argued that the effects are more accurately characterized as "spread" of processing rather than depth. That is, the successful recall of an item depends upon its encoding in additional contexts (e.g., structural, phonemic or semantic) of the existing cognitive make-up of the individual.
In Craik and Tulving's experiments, subjects were asked to respond to questions about words presented to them. Again, no mention was made that the words were to be remembered. An example of a deep processing question would be "Is the word a part of the body?"; in this case, the attention is directed to the meaning (category membership) of the word. A shallow processing question example would be "Is the word in upper case letters?", while an intermediate processing question was "Does the word rhyme with weight?". A series of ten experiments demonstrated consistently better memory for deeper levels of processing. The experiments also ruled out the possibility that longer processing time for deep questions could account for the superior performance; this was done by asking questions which took a long time to answer but still only required shallow processing. Finally, Craik and Tulving varied the degree of elaboration required to respond to questions and found better recall for more elaborate processing.
The concept of memory maintenance as a function of the quality rather than the quantity of processing would seem to be of profound significance for information processing theories of consumer behavior. The present experiment was designed to extend Craik and Tulving's operational definition of depth of processing to an advertising context. The task required subjects to attend to brand names in advertisements; for half the ads, attention was directed to physical features while for the other half, semantic features were stressed. In this way, an attempt was made to manipulate the level at which each ad was processed.
Forty magazine advertisements were used as presentation stimuli. The ads were chosen to represent a wide variety of products which would vary considerably on such features as familiarity, distinctiveness or personal relevance. Examples included Pioneer Stereo, Dole Pineapple, Arrow Shirts and American Airlines. Subjects were asked to respond to a single question about each ad. The questions used to elicit deep processing concerned personal experience with the brands advertised; examples include "Do you have this brand in your home?" and "Have you ever bought this brand?". Shallow processing questions directed the attention of the observer to the formal aspects of the brand name; "Is the brand name in script letters?" and "Are there seven letters in the brand name?".
To insure that any observed difference in memory for the ads was not specific to the ads chosen for each condition, a control procedure was used. This involved shifting questions for each ad from the shallow do the deep condition or vice versa after half of the participants in the study had been tested. Thus, for half the subjects, one half of the ads (arbitrarily chosen) had shallow questions while for the other half of the subjects, that same half of the ads had deep processing questions. The reverse was true for the other half of the ads.
Brand names were randomly dispersed throughout the presentation list with the restriction that no more than three ads with questions of the same level of processing appeared contiguously.
Subjects were seated at a table in a small room. The experiment was explained as an attempt to study perceptions of brand names in ads. The subjects were to look briefly at a number of ads and quickly answer a question about each one. An example was given of what was meant by "brand name" and a sample question was asked. The instructions indicated that a "yes" or "no" response was to be checked on a numbered answer sheet as soon as each question was asked.
The experimenter presented the ads by turning the pages of a notebook to the beat of an electronic timer set at a five-second rate. As each page was turned, the appropriate question was read. Subjects were given no prior indication that they would later be asked to remember the ads; post-experimental interviews indicated that the memory tests were indeed a surprise. "Buffer items", indistinguishable from other items but not scored, were included at the first and last of the list of ads to control for "serial position effects".
Two dependent variables, recall and recognition, were measured. In the recall test, the subjects were told to write down as many of the brand names as they could remember. This immediately followed presentation and two minutes were allowed for recall. For the recognition test, the subjects were given a list containing 80 brand names and were told to check "yes" or "no" to indicate whether they had seen each brand name in the stimulus notebook. Half of these items had actually been presented in the experiment while the other half were distractor items. The distractors were brand names in the same product class as the originally presented ads, but they had not been seen in the experiment. Thus, in the recognition test, each "old" brand name was "matched'' with a "new" brand of the same product. Previously presented items and distractors were randomly dispersed through the list with the restriction that no more than three items of each type appear contiguously.
The recognition scores were corrected for guessing by subtracting the number of false alarms (brand names in correctly checked "yes") from the number of hits (brand names correctly checked "yes"). The logic for this standard correction procedure is that each subject might be expected to guess "yes" by chance about as of ten for correct items as they would for incorrect ones.
The subjects in the experiment were 30 adults (members of the secretarial staff) at a university in a south western metropolis. Fifteen subjects served in each group and approximately half were males.
The mean number of brand names correctly recalled and the mean number of brand names recognized (corrected for guessing) for the two types of questions for both subject groups are presented in Table 1. As can be seen, about twice as many brand names for which deep processing questions were asked were recalled compared to those for which shallow processing questions were asked. (The "groups" variable refers to the control procedure of reversing deep and shallow questions for the first and second half of the subjects.) The deep processing question ads were also recognized substantially better.
MEAN BRAND NAME RECALL AND RECOGNITION (STANDARD DEVIATIONS IN PARENTHESES)
ANALYSIS OF VARIANCE SUMMARY TABLES
Table 2 presents the summary tables for analyses of variance of the two dependent variables. The design was a 2 x 2 factorial with type of question as a repeated measures variable. In both recall and recognition, the differences between the deep and shallow question items were significant (p < .001); the small differences between groups and the interactions between groups and type of question were not significant.
Thus, for both recall and recognition, retention was superior for ads presented in the context of deep processing questions.
While it cannot be claimed that the situation studied in the present experiment is precisely analogous to the experience of consumers viewing ads in the "real world", it can be said that there are elements in common. When a researcher calls to ask what ads were seen last night on TV, it is likely that the level to which the interviewee related ads to personal experience will influence recallability. In this context, it has been demonstrated in the present study that the recall and recognition of ads varies as a function of the type of questions asked and hence, the level of processing to which the ad was subjected.
Of course, the advertiser who wishes to elicit deep processing by the consumer cannot ask questions which will cause brand names to be related to a person's personal experience. However, ads can be designed in such a way that they elicit active participation by the viewer. That is, any ad which invites and effects the participation of the viewer by requiring him to supply some omitted aspect or to compare a brand with his personal experience or to predict the outcome of a hypothetical situation might be expected to require comparison of the brand name with previous knowledge. The result may be that the brand name becomes an integral part of the viewer's knowledge and hence is available for future retrieval.
Levels-of-processing theory places emphasis in the memory situation on the kind of processing that material receives rather than the number of exposures. While such a distinction between quality and quantity of processing may come as no surprise to advertising strategists, there has been an emphasis on the use of associative theory to explain repetition effects in advertising (e.g., Sawyer, 1974). Although it cannot be disputed that repetition is related to the likelihood that an ad will be remembered, it may be that repetition chiefly serves to provide multiple opportunities for observers to apply cognitive elaboration to the ad material, rather than to build up "habit strength" in accordance with classical associative theory. It seems safe to assume that such cognitive theories as Craik and Lockhart's levels of processing will receive a great deal of attention in future attempts to explain the memory of advertisements.
F. I. M. Craik, and R. S. Lockhart, "Levels of Processing: A Framework for Memory Research," Journal of Verbal Learning and Verbal Behavior, 11 (1972), 671-684.
F. I. M. Craik, and E. Tulving, "Depth of Processing and the Retention of Words in Episodic Memory," Journal of Experimental Psychology: General, 104 (1975), 268-294.
T. S. Hyde, and J. J. Jenkins, "Recall for Words as a Function of Semantic, Graphic and Syntactic Orienting Tasks," Journal of Verbal Learning and Verbal Behavior, 12 (1973), 471-480.
C. Leavitt, "Strong versus Weak Effects of Mass Communication: Two Alternative Hypotheses," in G. D. Hughes and M. L. Ray, (Eds.) Buyer/Consumer Information Processing. (Chapel Hill: University of North Carolina Press, 1974, 256-269).
J.C. Olson, "Theories of Information Encoding and Storage: Implications for Consumer Research," Working Series in Marketing Research, No. 65 (August, 1977), College of Business Administration, The Pennsylvania State University.
A. G. Sawyer, "The Effects of Repetition: Conclusions and Suggestions about Experimental Laboratory Research," in G. D. Hughes and M. L. Ray, (Eds.) Buyer/Consumer Information Processing. (Chapel Hill: University of North Carolina Press, 1974, 190-219.)
P. L. Wright, "On the Direct Monitoring of Cognitive Response to Advertising," in G. D. Hughes and M. L. Ray (Eds.) Buyer/Consumer Information Processing. (Chapel Hill: University of North Carolina Press, 1974, 220-248.)