Proceedings of the Third Annual Conference of the Association for Consumer Research, 1972 Pages 746-758
THE ZEIGARNIK EFFECT IN ADVERTISING
James T. Heimbach, Nationwide Research Center
Jacob Jacoby, Purdue University
If advertising is to be effective, its impact should persist over time. It is not enough that an advertisement have some effect at the moment of presentation; rather, it must continue to exert its effect throughout the decision-making process.
In general, there are two broad approaches to enhancing the persistence of an advertisement's effect. First, the strength of the initial impact may be raised, thus raising the residual impact at any point along the decay curve. Second, the slope of the decay curve (i.e., the rate of loss of effectiveness) may be decreased. Either of these processes will lead to an advertisement presented at time t retaining greater impact at time t + x, the time of decision-making or purchase. The primary purpose of this investigation was to examine a technique for increasing the initial impact of an advertisement and the slope of the subsequent decay curve.
Numerous operational definitions of "advertising impact" have been developed. In general, they fall into three categories:
1. Memorial effect (e.g., awareness, recognition, recall, knowledge of product)
2. Attitudinal effect (e.g., opinion, liking, belief, preference, intention, conviction)
3. Behavioral effect (e.g., inquiry, purchase).
The pair of studies to be discussed in this paper focus on two types of memorial effects--awareness and recall. The assumption which underlies this investigation is: "That version of an advertisement which produces the greater memorial impact is the more successful version of the advertisement." Immediate measures (e.g., awareness, immediate recall) reveal changes in the initial impact of the advertisement, while delayed measures (e.g., awareness change, delayed recall) reveal the compound effect of changes in the initial impact and changes in the rate of decay of effectiveness. Both types of measures were employed in this investigation.
Since Zeigarnik's (1927) classic study, it has repeatedly been demonstrated that incomplete tasks are better remembered than complete tasks (cf. Butterfield, 1964). Zeigarnik's explanation for this effect (now named for her) was that the subject beginning a task develops a need to complete it. If he is prevented from doing so, he is left in a state of tension. which manifests itself in improved memory for the uncompleted task.
Extending this explanation from tasks to messages (i.e., advertisements), one could suggest that hearing the beginning of a message leads to the development of a need to hear the rest of it--rather like waiting for the second shoe to drop. The resulting tension leads to improvement in memory for that part of the message which has already been heard.
As might be supposed, this is a rather short-term phenomenon--one does not stay up all night waiting for the second shoe to drop. So what is its significance for advertisers? The key lies in the fact that, in the situation discussed, the subject is unable to complete the task on his own--he does not know the rest of the message. This is not the case with an advertisement. Few people hearing "Winston tastes good like a " would be unable to complete the message. Most probably, a good proportion of readers have completed it just now. This illustrates two further forces leading to improved recall for the message. The first, suggested by Heller (1956), is that completing the message is positively reinforcing, thereby improving learning and memory. A second force is provided by the fact that the audience will have actively participated in the message, and it is a well-known learning principle that active participation, as opposed to passive reception, improves learning and memory. These effects would be expected to persist over extended periods of time.
In addition to testing the general hypothesis (i.e., that incomplete ads would be better remembered than complete ads), the two experiments also examined other factors which could be hypothesized to strengthen or diminish the magnitude of the Zeigarnik effect.
1. Subject's need for achievement (nAch). Several studies (e.g., Atkinson, 1953; Weiner, 1966) have found that high n-Achievers show a stronger Zeigarnik effect than low n-Achievers.
2. Subject's prior familiarity with the advertisement. It is expected that greater familiarity will lead to better ability to complete an interrupted message, thus strengthening the Zeigarnik effect.
3. Subject's prior involvement with the product and brand advertised. It is predicted that subjects more involved and interested in the subject of the commercial will pay closer attention to it and will make more effort to complete an interrupted message.
4. Presence or absence of a jingle in the advertisement. The presence of a jingle should strengthen the Zeigarnik effect in three ways: (a) by adding strain for completion of the musical phrase to that strain already present for the verbal material; (b) by increasing the awareness of the interruption; and (c) by providing a well-known slogan which should be highly familiar and therefore easy to complete when interrupted.
5. Presence or absence of a post-commercial pause. Providing a brief pause between the termination of the commercial and the resumption of the program should enhance the Zeigarnik effect by giving subjects a chance to complete the commercial without interference.
Two video tapes were prepared of a 30-minute television program, each containing four test and five filler commercials. These commercials were inserted into existing program breaks (i.e., replaced the original commercials appearing in the program as broadcast). Two test commercials (for cigarettes and soft drinks) were for highly familiar products, were highly familiar themselves (as established in a separate pre-test with different subjects), and contained jingles. The other two test commercials (for rug shampoo and automotive services) were relatively unfamiliar and did not contain jingles.
One commercial of each type was presented in complete form in Tape 1 and incomplete--the last 5 to 6 seconds deleted--in Tape 2. In a crossover control design, the other two were presented complete in Tape 2 and incomplete in Tape 1. This design allows analysis utilizing a replication factor to determine the reproducibility of the results.
Subjects were selected by administering the nAch scale from the Personality Research Form (Jackson, 1967) to 172 Introductory Psychology students at Purdue University during the Spring of 1970. The 30 highest and 30 lowest scoring students were selected for this experiment.
The two video tapes were shown simultaneously on 24-inch monitors in different rooms, with 30 subjects--15 high and 15 low in nAch--assigned to each room. Immediately following presentation of the program, the subjects completed an advertisement-recall measure. This measure consisted of asking the subjects to identify the product class (e.g., toothpaste) and the brand name and to provide a detailed description of the contents of each of the nine commercials. Upon completion, this measure was collected and an evaluation measure was distributed. This consisted of a listing of all the commercials, each accompanied by a seven-point, good-bad semantic differential scale.
The recall measure was scored on a five-point scale. A score of zero represented total non-recall for a commercial, one point was assigned for recall of the product class, one point for the brand name, and zero, one, or two points were given for recall of specific copy points. The questionnaires were scored by two independent raters, who achieved a conspect reliability of .96.
The hypotheses tested in Experiment I were:
1. Incomplete versions of commercials will produce greater recall than complete versions.
2. High nAch subjects will manifest a stronger Zeigarnik effect than low nAch subjects.
3. Familiar commercials with jingles will produce a stronger Zeigarnik effect than unfamiliar commercials without jingles.
Profiles of the recall scores are shown in Figure 1. It can be seen that the high-nAch subjects exhibited a Zeigarnik effect for both high-familiarity commercials with jingles (soft drinks and cigarettes) while the low nAch subjects showed the effect only for the cigarette commercial. The unfamiliar commercials without jingles consistently produced a small anti-Zeigarnik effect, in that they tended to be recalled less well when interrupted. The analysis of variance, however, revealed that neither the main effect for completeness nor the interaction between completeness and nAch were significant. The predicted interaction between completeness and "familiarity/jingle" was marginally significant at p < .10.
PROFILES OF RECALL SCORES, EXPERIMENT I
While it was not expected that the technique of interrupting a commercial would have any effect upon subjects' liking for it, this measure was included because it was felt that the possibility of either a positive or adverse effect should be explored. One significant effect of interruption upon evaluation did appear: a marginally significant (p < .10) triple interaction between completeness, nAch, and familiarity/jingle. This interaction may be due to an unplanned confounding factor. The two high-familiarity, jingle-containing commercials were for pleasure-oriented products (cigarettes and soft drinks), while both low-familiarity, non-jingle commercials were for functional products (rug shampoo and automotive services). It appears, then, that the high-nAch subjects preferred the complete versions of functional-product commercials and the interrupted versions of commercials for pleasure-oriented products, while the low-nAch subjects exhibited the opposite pattern, preferring complete pleasure-oriented and incomplete functional product commercials.
Although this experiment did not provide strong support for the application of the Zeigarnik effect to broadcast advertising, it was felt that since all results had been in the expected directions, the question did merit further study. Therefore, a second, somewhat more elaborate and better controlled experiment was designed to clarify some of the questions raised, as well as to explore new issues.
Specifically, the following questions were raised:
1. What is the effect of different levels of subjects' involvement with the brand and product advertised? Will those who use the product or brand pay closer attention to the commercial or make more effort to complete an interrupted message than those who do not use the product, thus showing a stronger Zeigarnik effect?
2. What is the effect of different levels of prior awareness of the brand advertised? Will subjects more highly aware of the brand pay closer attention to the commercial, thereby exhibiting a stronger Zeigarnik effect?
3. Will allowing a brief pause after the commercial strengthen the Zeigarnik effect by allowing the subjects a chance to complete the interrupted commercials without distraction?
4. Does the presence of a jingle enhance the Zeigarnik effect in the absence of large differences in the familiarity of the commercials?
5. What would be the effect of truncating the beginning of a commercial rather than deleting the end?
6. Can a Zeigarnik effect be shown for other memorial criteria han unaided recall of copy points? Specifically, can basic awareness of the brand name be stimulated via the Zeigarnik effect?
7. Finally--and most importantly--does the Zeigarnik effect persist over time?
The second experiment was performed to answer these questions.
Six audio tapes were prepared of the sound portion of a TV show. Again, each contained four test commercials, two with and two without a jingle. In this experiment, however, all test commercials were of roughly equal familiarity. A pre test, in which 24 subjects viewed the four commercials (counterbalanced for order effects via a Latin square design) and assigned familiarity ratings from 1 to 7, established that the range of familiarity among the commercials was from 5.38 to 5.70; no differences were significant at the p <.25 level. The products advertised were cigarettes, chewing gum, mouthwash, and a headache remedy.
Each commercial was presented in one of six modes on each of the six tapes: it could be presented complete, interrupted at the end, or truncated at the beginning; in addition, it could be followed by immediate resumption of the program or by a four-second delay preceding program resumption.
One week prior to the presentation of the program, students in three classes of Introductory Psychology at Purdue University (Spring 1971 semester) were tested on their usage patterns and brand awareness for a number of low-price consumer products. This testing was performed by a graduate student having no ostensible connection with the present study. The awareness measure consisted of asking the subjects to list five brands of each product class. The product/brand usage measure asked if the subject used each product, whether he had a regular brand, and what that brand was. This pretest was used to select brands for use in the experiments as well as to determine which tape each subject would hear.
Each tape was played to a group of 15 subjects which included three subjects in each of the following five categories for each advertised brand: (1) regularly uses advertised brand; (2) uses product but not advertised brand and is highly aware of brand; (3) uses product but not advertised brand and is not highly aware of brand; (4) does not use product and is highly aware of brand; and (5) does not use product and is not highly aware of brand.
The subjects were told that the experiment was concerned with the relative amount of information carried by the audio and video portions of TV broadcasts. Immediate recall for the commercials was tested directly following the presentation, using the same recall measure as was used in Experiment I. The page containing this measure was the second of three pages of the questionnaire; the other pages asked questions about the program itself in order to continue the deception.
Two days later all subjects were retested on brand awareness; this testing was performed by the same graduate student who had administered the pretest under the pretext that he was checking its reliability. One week after the presentation, the advertisement recall test was again administered to measure delayed recall. At this time the deception was dropped and the subjects were debriefed.
The recall measures were scored by two raters, with an overall conspect reliability of .97. The awareness measure was scored in two ways. The first provided a measure of "change in awareness." This was achieved by subtracting the awareness score on the pretest from awareness on the posttest, counting five for listing the advertised brand first, four for listing it second, and so on, down to zero for failing to list it at all. The second scoring method provided a measure of "first-brand awareness"--listing the advertised brand first on the posttest gave a score of "one," while anything else gave a score of "zero."
Thus, this experiment contained five independent variables: (l) Completeness (complete; cut-at-beginning; and cut-at-end); (2) Subject's Product/ Brand Involvement (does not use product; uses product but not brand; uses brand); (3) Subject's Prior Awareness of Brand (high; low);-(4) Jingle (absent; present); and (5) Post-Commercial Pause (none; four-seconds).
There were also four dependent variables: (l) Immediate Recall (measured within 10 minutes of the presentation); (2) Delayed Recall (measured one week after the presentation); (3) Change in Awareness (posttest measure two days after the presentation); (4) First-Brand Awareness (measured two days after the presentation).
It was anticipated that the dependent variables would react similarly to the manipulations, although first-brand awareness was expected to be the least sensitive measure.
The major hypotheses were:
1. Interrupting the end of a commercial is superior to cutting the beginning, which in turn is superior to presenting it in complete form.
2. Increased involvement with the product and brand advertised leads to a stronger Zeigarnik effect.
3. Increased brand awareness produces a stronger Zeigarnik effect.
4. The presence of a jingle strengthens the Zeigarnik effect.
5. The presence of a post-commercial pause strengthens the Zeigarnik effect.
The primary hypothesis--that interrupting commercials produces greater memorial impact than presenting them complete--was strongly supported. This effect was significant beyond the p < .01 level for immediate recall, delayed recall, and change in awareness; it was in the expected direction, although not significant, for the first-brand awareness. In addition, the hypothesis that cutting the end of the commercial would be best, cutting the beginning second best, and leaving the commercial uncut worst was also strongly supported. This pattern was exhibited for all four dependent variables, and each interval was significant at the p < .05 level on the Tukey (a) test for immediate and delayed recall and for change in awareness. The profiles of these data appear in Figure 2.
As was noted earlier, it is important that the memorial impact of a commercial persist over time. Therefore, the relative performance of complete, cut-at-beginning, and cut-at-end commercials was examined over time (i.e., from immediate to delayed recall--a period of one week). For the sake of simplicity, and since two points are not enough to establish the nature of the curve, a linear decay has been assumed. This plot appears in Figure 3.
PROFILES OF RESULTS FOR ALL DEPENDENT VARIABLES
PROFILES OF RECALL ACROSS TIME
It can readily be seen that cutting the beginning of a commercial produces an effect which decays over time, although it still produces recall superior to that for complete commercials after a period of a week. But the effect of cutting the end of the commercial appears to be a lasting one--its superiority over cut-at-beginning ant complete versions in producing recall is even slightly greater after one week.
The hypotheses concerning the effects of product/brand involvement, brand awareness, ant the presence of a jingle were not supported. None of these factors exerted any effect upon the Zeigarnik effect on any of the dependent variables.
The presence of a four-second pause following the commercial tit have a significant effect, although this effect was considerably more' complex than had been predicted. It appears that the presence of a post-commercial pause has different effects upon subjects with low ant high initial brand awareness. As is shown by the profile of the triple interaction between completeness, post-commercial pause, ant brand awareness (Figure 4), post commercial pause improved immediate recall of the cut-at-beginning commercials ant decreased recall of complete commercials ant those cut at the end for the subjects high in brand awareness. For those low in brand awareness, on the other hand, the post-commercial pause improved immediate recall of commercials cut at the end and decreased it for those presented complete or cut at the beginning. This same pattern also occurred at statistically significant (p < .05) levels on delayed recall ant on first-brand awareness
It thus appears that a number of tentative conclusions can be drawn concerning the effects of the incomplete-commercial technique; however, a considerable number of questions remain.
First, it is apparent that the technique can have a substantial ant significant effect upon recall for commercials, both immediately following the presentation of the commercial ant one week later. In Experiment II, commercials cut at the end generated 33.8% more immediate recall than did complete versions of these commercials; one week later the recall advantage of the cut-at-end version was 52.4%. A similar effect was also observed for brand awareness.
It appears that truncating the beginning of a commercial is also an effective technique for increasing its memorial impact, although this technique is not as effective as deleting the end of the message.
While the data on familiarity of the commercial are not as complete as one would desire, it appears that the Zeigarnik effect will not occur with unfamiliar commercials.
There seems to be no support for the suggestion that the presence of a jingle will enhance the Zeigarnik effect, nor for the hypothesis that the viewer's involvement with the product ant brand advertised will affect the strength of the effect.
INTERACTION BETWEEN COMPLETENESS, POST-COMMERCIAL PAUSE, AND BRAND AWARENESS
The question of providing a brief pause after the commercial obviously requires further research--the effects of this procedure are rather nebulous at this point.
Several basic questions remain. For example, to what degree is the improved recall for interrupted commercials due to a uniqueness or "shock" effect? Will a single interrupted commercial lose its effect with repetition, or would the production of a number of different interrupted commercials for different products cause the effect to disappear?
The effect of interruption of a commercial upon other than memorial criteria is also unknown. The single evaluation scale employed in Experiment I revealed the interaction which was interpreted--on a post hoc basis--in terms of functional vs. pleasure-oriented products, but did not reveal any main effect due to interruption. It must be emphasized, however, that this scale was probably not especially sensitive; the absence of significant effects suggests only that attitudinal effects are probably not very large. Further research on attitude, preference, and intention to buy as a function of interruption of commercials might well prove fruitful.
Finally, it should be noted that none of the advertisements utilized in these experiments were originally designed to be interrupted. It is probable that commercials could be specifically tailored to maximize the Zeigarnik effect; in addition, experiments could be performed with only the audio channel interrupted ant the video channel left uncut. This technique also might prove effective.
In summary, it appears that the interrupted-commercial technique shows considerable promise for enhancing an advertisement's impact, but further research is necessary if the full potential of the technique is to be revealed.
Atkinson, J. W. The achievement motive and recall of interrupted and completed tasks. Journal of Experimental Psychology, 1953, 46, 381-390.
Butterfield, E. C. The interruption of tasks: methodological, factual, and theoretical issues. Psychological Bulletin, 1964, 62, 309-322.
Heller, N. An application of psychological learning theory to advertising. Jouranl of Marketing, 1956, 20, 249-254.
Jackson, D. N. Personality Research Form Manual. Goshen, New York: Research Psychologists Press, 1967.
Weiner, B. Achievement motivation and task recall in competitive situations. Journal of Personality and Social Psychology, 1966, 3, 693-696.
Zeigarnik, B. Uber das behalten von erledigten und unerledigten Handlungon. Psychologische Forschungen, 1927, 9, 1-85.