The Relationship Between Distractor Similarity and the Recognition of Print Advertisements

James W. Peltier, University of Wisconsin-Whitewater
John A. Schibrowsky, University of Nevada-Las Vegas
ABSTRACT - This paper reports the findings of an investigation of the impact that distractor similarity has on the recognition memory for print advertisements. Two issues are investigated. First, the direct impact of distractor similarity on brand and claim recognition memory scores is investigated. Second, the attenuating impact of distractor similarity on the study of other variables posited to impact memory is investigated. A method for assessing distractor similarity is discussed, along with implications for advertising research.
[ to cite ]:
James W. Peltier and John A. Schibrowsky (1992) ,"The Relationship Between Distractor Similarity and the Recognition of Print Advertisements", in NA - Advances in Consumer Research Volume 19, eds. John F. Sherry, Jr. and Brian Sternthal, Provo, UT : Association for Consumer Research, Pages: 94-100.

Advances in Consumer Research Volume 19, 1992      Pages 94-100

THE RELATIONSHIP BETWEEN DISTRACTOR SIMILARITY AND THE RECOGNITION OF PRINT ADVERTISEMENTS

James W. Peltier, University of Wisconsin-Whitewater

John A. Schibrowsky, University of Nevada-Las Vegas

ABSTRACT -

This paper reports the findings of an investigation of the impact that distractor similarity has on the recognition memory for print advertisements. Two issues are investigated. First, the direct impact of distractor similarity on brand and claim recognition memory scores is investigated. Second, the attenuating impact of distractor similarity on the study of other variables posited to impact memory is investigated. A method for assessing distractor similarity is discussed, along with implications for advertising research.

INTRODUCTION

The use of recognition as a measure of memory has increasingly become a prominent topic in both academic and applied research (Leckenby and Plummer 1983; Stewart et al. 1985; Singh et al. 1988). Much of this interest stems from the growing concern that recognition may be more appropriate than recall for assessing certain types of advertising effects (Bettman 1979; Krugman 1972, 1985; Singh et al. 1988; Zielske 1982). The fundamental difference between these two measures is that in a recall task an individual must describe a non-present stimulus, while in a recognition task a stimulus is judged on the basis of whether it had been previously seen or heard (Bettman 1979).

Critics of recognition contend that it is a less sensitive measure for detecting differential memory effects than is recall. This belief is based on the substantially higher memory scores and resulting ceiling effects that are often associated with recognition (Haber 1970; Krugman 1979; Shepard and Chang 1967). However, the ease with which a stimulus is recognized is partially a function of the similarity between the stimulus and relevant distractor items. Specifically, researchers in cognitive psychology have shown that stimulus recognizability decreases as distractor similarity increases (Jorg and Hormann 1978; McNulty 1965; Postman et al. 1948).

Despite the potential inverse relationship between distractor similarity and the recognition of an advertisement, this issue has been largely ignored in the marketing literature. The typical approach used by advertising researchers has been to arbitrarily assess the similarity between a target stimulus and its distractors. This subjective procedure for selecting distractors creates two major problems for advertising researchers. First, the level of difficulty for a particular recognition task can not be quantified. Second, and potentially more important, a similarity bias may exist due to the inability to equate distractor similarity across advertisements and/or experimental treatments. These are critical issues that must be addressed before recognition tests can be effectively employed in advertising research (Singh and Cole 1985).

We report the findings from an experiment that examines the effect that distractor similarity has on the recognition of brands and claims presented in print advertisements. In addition, an attempt is made to determine which level(s) of similarity is (are) most sensitive to memory differences. A simple method for equalizing distractors across advertisements and treatments is also discussed.

LITERATURE REVIEW

This section is organized into two separate parts. First, the different types of recognition tests are described. Second, the extant literature that has investigated distractor similarity is examined.

Types of Recognition Tests

In general, recognition tests involve exposing subjects to target stimuli, then, after a specific length of time, the subjects are asked to select those stimuli from a list of items, some of which are distractor items (e.g., they have not been previously seen). There are essentially three different ways to administer a recognition test:

'Yes'/'No' Test. Subjects are shown a set of stimuli one at a time. For each item, subjects are asked to respond 'yes' if they have previously seen that item and 'no' if they have not. Typically, there is one distractor for every item in the original list of stimuli.

Batch-Testing. In a batch-testing recognition test all of the original stimuli and all of the distractor items are presented at the same time. Subjects are asked to select those items that they had previously seen.

Forced-Choice. In this method, subjects are asked to identify a previously seen stimulus from a list containing one or more distractor items. This task is repeated once for every stimulus on the original list. Unlike batch-testing, there is only one 'correct' response per set. When one distractor is used this is called a two-item test, when two distractors are used it is a three-item test, and so on.

An additional measure that is often obtained along with recognition is to have subjects indicate the degree to which they are confident with their responses. Confidence ratings are typically obtained via a three- to five-point scale ranging from very confident that the response is correct to not confident at all. The confidence ratings are then used to transform correct and incorrect responses into a more sensitive multi-point scale. It has been argued that this procedure contributes considerable information for data analysis. In particular, statistical analyses can be conducted on the transformed data that are free of distortions (Singh and Rothschild 1983b).

Distractor Similarity Literature

While the issue of distractor similarity is of extreme importance for measuring advertising effectiveness, it has received surprising little attention in the marketing literature. The work that has been done in this area has been of a conceptual nature. Specifically, Singh and Rothschild (1983a) suggested that recognition scores could be lowered, and thus made more difficult, by employing very similar distractors. More recently, Singh and Cole (1985) stressed the importance of equalizing distractor similarity across test ads and treatments.

There is a body of knowledge within the cognitive psychology literature that has examined how distractor difficulty affects recognition memory. Furthermore, each type of recognition test has been used as a dependent measure. Early work by Postman et al. (1948) using a four-item forced choice recognition test showed that subjects incorrectly selected similar distractors more frequently than dissimilar distractors. McNulty (1965) found analogous results via a batch recognition test--subjects performed more poorly for similar, rather than dissimilar distractors. Using a yes/no test, Dale and Baddeley (1962) showed that stimulus recognizability decreased as distractor similarity increased.

The robustness of the distractor similarity findings is evidenced by the variety of stimulus devices that have been used to manipulate this construct. Specifically, distractor similarity has been manipulated via nonsense syllables/words (McNulty 1965; Postman et al. 1948), numbers (Dale and Baddeley 1962; Shepard and Chang 1963), line drawings (Bahrick et al. 1967; Bower and Glass 1976; Jorg and Hormann 1978), and other pictorial stimuli (Bahrick and Bahrick 1964; Nagae 1980). In all cases, distractor similarity had a significant negative impact on stimulus recognizability.

Distractor Similarity and Advertisements

The degree to which the distractor similarity findings in the cognitive psychology literature can be generalized to an advertising context is not well documented. Part of this dilemma is in measuring the similarity between distractors and advertising stimuli. In particular, measuring the similarity between distractors and specific brand names and advertising claims is a much different process than for nonsense syllables, numbers and line drawings.

In general, the measurement of target-distractor similarity in previous research studies focused on the structural relationship (i.e., what they looked like) between a target stimulus and its distractors. Conceptually, when measuring distractor similarity for words (e.g., brand names) and phrases (e.g., claims), both the structure and the meaning of these items must be compared to the target stimuli. In principle, distractors and stimuli that share comparable meanings may be judged similar even if structurally they are not, and vice-versa.

HYPOTHESES

The research project reported here had two major objectives. One of the objectives was to explore the effect that brand and claim distractor similarity have on recognition memory. The cognitive psychology literature provides insight into this phenomenon. The following prediction is tested:

H1: An inverse relationship will exist between brand and claim distractor similarity and recognition.

The contention that recognition tests attenuate memory differences due to task simplicity (difficulty) and resulting ceiling (floor effects), suggests that an experimental manipulation may be judged unsuccessful when this not the case. Exposure time is manipulated in this study for the purpose of creating experimental conditions where recognition differences are expected to occur (Rothschild et al. 1989). If Singh and Rothschild (1983a) are correct, making recognition tests more or less difficult by manipulating distractor similarity should affect recognition within and across exposure time treatments.

This suggests the following hypothesis:

H2: A significant distractor similarity X exposure time interaction will exist, with recognition memory differences being eliminated as distractor similarity approaches the extremes.

Each of the above two hypotheses are investigated in the study discussed below.

METHOD

Subjects and Design

Ninety-one undergraduate students were randomly assigned to conditions in a 4 (degree of distractor similarity: very similar, similar, somewhat similar and very dissimilar) X 2 (exposure time: 15 seconds and 10 seconds) within and between factorial design. Degree of distractor similarity was the within subject factor and exposure time was the between subject factor.

Procedure

The cover story given to subjects was that a local advertising agency wanted feedback on a number of print ads that they had developed. Eight test ads were then viewed via a slide presentation in small groups of approximately 15 subjects. Order effects were controlled for by developing four different ad presentation sequences. It should be noted that a significant order effect was found for the various memory scores.

After the slides were viewed, subjects completed a three minute distractor test. The dependent measures were then administered. In a response booklet, subjects were given a verbal description of each of the products promoted in the ads that they had viewed. They were asked to identify the correct brand and claim associated with each product and to rate how confident that they were with their responses. In the response booklets, each of the four distractor similarity conditions was assigned two different ads. Order effects were controlled through the use four different response booklets that varied distractor similarity across the test ads. Finally, subjects were debriefed and were asked not to disclose the purpose of the study to others.

TABLE 1A

MEAN BRAND SIMILARITY SCORE BY CONDITION

TABLE 1B

MEAN CLAIM SIMILARITY SCORE BY CONDITION

Stimulus Materials

Slides of color illustrated print ads were used in the study. Each of these ads had a character, a product illustration, a brand name and a claim. These ads generated differential recognition responses across exposure time in a different study.

Distractor Similarity Pretest

A relatively simple pretest was employed to develop recognition measures for each of the distractor similarity conditions. Several graduate students were given the task of generating distractor brands and claims for each of those to be used in the actual study. They were told that these distractors should fall along a similarity continuum ranging from very similar to very dissimilar. When developing these distractors, they were instructed to consider both the structural nature of the targets and their meanings. From this list 25-30 thirty brand and claim distractors were selected for each target stimulus. Using a seven-point scale (very similar to very dissimilar), a second set of pretest subjects rated these distractors with respect to specified target brands and claims. No conditions for judging target-distractor similarity were stated.

After analyzing these judgments, brand and claim distractors for each similarity condition were selected on the following basis (1=very similar and 7=very dissimilar):

1. The mean similarity score for each brand and claim had to fall in the range of 1.6-1.9 for the very similar condition, 2.6-2.9 for the similar condition, 3.6-3.9 for the somewhat similar condition, and greater than 6.0 for the very dissimilar condition.

2. Distractors with lower standard deviations were preferred.

3. For each ad, brand and claim distractors were equally similar.

Table 1A and Table 1B report mean similarity scores for brands and claims for each of the distractor conditions. There were no significant differences in similarity scores between ads within each condition. Differences in mean scores between conditions were significant.

Dependent Measures

As indicated above, each subject was exposed to one of two exposure time conditions and received all four distractor similarity manipulations for two ads each.

TABLE 2

F VALUES FOR DIFFERENCES OF RECOGNITION MEASURES ACROSS FOUR LEVELS OF DISTRACTOR SIMILARITY

Brand and Claim Recognition: Using a five-item forced choice test, subjects were asked to identify the correct brands and claims that were in the ads that they had viewed.

Confidence Rating: Confidence ratings for brand and claim responses were assessed via a five-point scale ranging from very confident to not confident at all.

A forced choice recognition test was employed in the study to limit response biases that are often associated with yes/no and batch-testing procedures (Shepard and Chang 1963). A five-item test was selected as a compromise between a two-item and a nine-item test.

RESULTS

To investigate the hypotheses, the brand recognition, brand recognition confidence, claim recognition, and claim recognition confidence scores obtained from the subjects were used to compute three summary scores. A brand summary score for each ad (BSUM) was obtained by multiplying the brand recognition score (0 or 1) by the brand recognition confidence score (1 to 5). A claim summary score (CSUM) was computed by multiplying the claim recognition score (0 or 1) by the claim recognition confidence score (1 to 5). A total ad recognition score (TSUM) was computed by adding together the brand summary score and claim summary score.

Results for Hypothesis 1

Recall that H1 hypothesized that a significant inverse relationship exists between distractor similarity and brand/claim recognition. Table 2 shows the results of the investigation of this hypothesis. It contains F values for differences in the computed measures of recognition across the four levels of distractor similarity. In a total of 22 of the 27 conditions (81%) there was a significant difference in recognition scores across different levels of distractor similarity. In each case, memory scores improved as distractor similarity decreased. In summary, these results provide strong support for H1.

Results for hypothesis 2

Recall that H2 hypothesized that recognition memory differences would be attenuated as distractor similarity approached the extremes. This prediction suggests that floor and ceiling effects would eliminate differences that would otherwise exist. To test this hypothesis, exposure time was manipulated as a between subjects variable. Exposure time was selected because it has a robust effect on recognition memory scores. As exposure time increases so will recognition scores.

To test H2, the impact of differences in exposure time (10 and 15 seconds) on recognition scores was investigated at four different levels of distractor similarity: very similar, similar, somewhat similar, and very dissimilar. The three computed measures of recognition, brand recognition (BSUM), claim recognition (CSUM), and total brand and claim recognition (TSUM) summed across all ads were used as the dependent variables. The results were comparable for all three measures. Figures A, B, and C show the effects of exposure time on the TSUM, BSUM, and CSUM measures of recognition in the four distractor conditions.

In each case the figures support H2. The differences in recognition scores for the different exposure times are reduced in both extreme conditions. Table 3 summarizes the results for each of the three different measures of recognition.

As table 3 indicates, the effect of exposure time on all three measures of recognition were attenuated at the extremes. In the extreme conditions only 2 of the 6 cases were significant, while in the middle conditions 5 of the 6 conditions were significant. In summary, the findings provide strong support for H2.

FIGURE A

MEAN RECOGNITION SCORES (TSUM) FOR DIFFERENT EXPOSURE TIMES AND LEVELS OF DISTRACTOR SIMILARITY

FIGURE B

MEAN RECOGNITION SCORES (BSUM) FOR DIFFERENT EXPOSURE TIMES AND LEVELS OF DISTRACTOR SIMILARITY

DISCUSSION AND CONCLUSIONS

The results clearly show that distractor similarity affects brand and claim recognition memory. Brand and claim recognition scores for the within subject analysis across distractor similarity conditions varied considerably. In addition, the between subject analysis across exposure time showed that distractor similarity attenuated recognition memory at the extremes.

These findings suggest that when distractors are very dissimilar with respect to target stimuli, the recognition task is too easy and that ceiling effects may occur. When this is true, the criticism that recognition tests are insensitive for detecting memory differences may be valid. In contrast, when distractors are too similar to targets, the task becomes very difficult and memory differences that otherwise may have existed may also be eliminated. It should be noted that the extreme conditions did not attenuate the results in every case; however, the likelihood of detecting memory differences is definitely reduced.

FIGURE C

MEAN RECOGNITION SCORES (BSUM) FOR DIFFERENT EXPOSURE TIMES AND LEVELS OF DISTRACTOR SIMILARITY

TABLE 3

T-VALUES OF THE EFFECT OF EXPOSURE TIMES ON MEAN RECOGNITION MEASURES FOR EACH LEVEL OF DISTRACTOR SIMILARITY

Moderate levels of distractor similarity worked best for detecting recognition differences. While absolute recognition scores varied across distractors categorized as being similar or somewhat similar, differences across exposure time were relatively equal. Thus, either one of these two categories could be used as long as similarity is held relatively constant.

The procedure used here for equating distractor similarity is relatively simple and not very time consuming. While all levels of distractor similarity had to be developed for the current study, subsequent research need only focus on moderately similar distractors. Furthermore, over the course of the pretest the authors became quite proficient at predicting how similar distractors were to target stimuli. As such, a distractor similarity judgment learning curve for researchers should result.

Future research may wish to develop different ways to measure and control for distractor similarity biases. In addition, the most efficient level of distractor difficulty across different types of conditions and task environments need also be tested.

REFERENCES

Bahrick, Harry P. and Phyllis O. Bahrick (1964), "A reexamination of the Interrelations Among Measures of Retention," Quarterly Journal of Experimental Psychology, 16, 310-324.

Bahrick, Harry P., Phyllis O. Bahrick, Sandra Clark and Phyllis O. Bahrick (1967), "Generalization Gradients as Indicants of Learning and Retention of a Recognition Task," Journal of Experimental Psychology, 75, 464-471.

Bettman, James R. (1979), An Information Processing Theory of Consumer Choice, Reading, MA: Addison Wesley Publishing Company.

Bower, Gordon H. and Arnold L. Glass (1976), "Structural Units and the Reintegrative Power of Picture Fragments," Journal of Experimental Psychology: Human Learning and Memory, 2 (4), 456-466.

Dale, H.C.A. and Alan D. Baddeley (1962), "Alternatives in Testing Recognition Memory," Nature, 196 (October 6), 93-94.

Haber, Ralph N. (1970), "How We Remember What We See," Scientific American, 222 (May), 104-12.

Jorg, Sabine and Hans Hormann (1978), "The Influence of General and Specific Verbal Labels on the Recognition of Labeled and Unlabeled Parts of Pictures," Journal of Verbal Learning and Verbal Behavior, 17, 445-454.

McNulty, John A. (1965), "An Analysis of Recall and Recognition Processes in Verbal Behavior," Journal of Verbal Learning and Verbal Memory, 4, 430-6.

Krugman, Herbert E. (1972), "Why Three Exposures May Be Enough," Journal of Advertising Research, 12 (December), 11-14.

Krugman, Herbert E. (1979), "Low Involvement Theory in the Light of New Brain Research," in Attitude Research Plays for High Stakes, J. C. Maloney and B. S. Silverman, (eds), Chicago: American Marketing Association.

Krugman, Herbert E. (1985), "Measuring Memory--An Industry Dilemma," Journal of Advertising Research, 25 (4), 49-51.

Leckenby, John D. and Joseph T. Plummer (1983), "Advertising Stimulus Measurement and Assessment Research: A Review of Advertising Testing Methods," Current Issues and Research in Advertising, 2, 135-65.

Nagae, Seiji (1980), "Nature of Discriminating and Categorizing Functions of Verbal Labels on Recognition Memory for Shapes," Journal of Experimental Psychology: Human Learning and Memory, 6 (4) 421-429.

Postman, Leo, William O. Jenkins, and Dorthy L. Postman (1948), "An Experimental Comparison of Active Recall and Recognition," American Journal of Psychology, 61 (February), 156-63.

Rothschild, Michael L., Lisa Qualheim, Brian Deith, and Yong J. Hyun (1989), "Constructing a More Difficult Recognition Test for Television Commercial Scenes," in Advances in Consumer Research, Vol 17, Marvin E. Goldberg, Gerald Gorn and Richard W. Pollay (eds), Provo, UT: Association for Consumer Research, 785-791.

Singh, Surendra N. and Catherine A. Cole (1985), "Forced-Choice Recognition Tests: A Critical Review," Journal of Advertising, 14 (3), 52-8.

Singh, Surendra N. and Michael L. Rothschild (1983a), "Recognition as a Measure of Learning from Television Commercials," Journal of Marketing Research, 20 (August), 235-48.

Singh, Surendra N. and Michael L. Rothschild (1983b), "The Effect of Recall on Recognition: An Empirical Investigation of Consecutive Learning Measures," in Advances in Consumer Research, Vol. 10, Richard P. Bagozzi and Alice M. Tybout, (eds), Ann Arbor, MI: Association of Consumer Research, 271-6.

Singh, Surendra N., Michael L. Rothschild and Gilbert A. Churchill Jr. (1988), "Recognition Versus Recall as Measures of Television Commercial Forgetting," Journal of Marketing Research, 25 (February), 72-80.

Stewart, David W., Connie Peckman, Srinivasan Ratneshwar, Jon Stroud, and Beverly Bryant (1985), "Methodological and Theoretical Foundations of Advertising Copytesting: A Review," Current Issues and Research in Advertising, 2, 1-74.

Shepard, Roger N. and Jih-Jie Chang (1967), "Forced-Choice Tests of Recognition Memory Under Steady-State Conditions," Journal of Verbal Learning and Verbal Behavior, 2, 93-101.

Zielske, Hubert A. (1982), Does Day-After Recall Penalize 'Feeling' Ads?" Journal of Advertising Research, 20 (April), 7-13.

----------------------------------------