Carryover Effects of Corrective Advertising

ABSTRACT - A laboratory experiment tested based on communication theory and empirical research, about the delayed or carryover effects of corrective advertising. The results of the tested corrective advertisements present evidence that one exposure of a corrective advertisement may not influence brand beliefs as would be hoped by public policy makers. Problems with the employed longitudinal design are discussed and suggestions for future research are offered.

Citation:

Alan G. Sawyer and Richard J. Semenik (1978) ,"Carryover Effects of Corrective Advertising", in NA - Advances in Consumer Research Volume 05, eds. Kent Hunt, Ann Abor, MI : Association for Consumer Research, Pages: 343-351.

Advances in Consumer Research Volume 5, 1978      Pages 343-351

CARRYOVER EFFECTS OF CORRECTIVE ADVERTISING

Alan G. Sawyer, The Ohio State University

Richard J. Semenik, University of Utah

ABSTRACT -

A laboratory experiment tested based on communication theory and empirical research, about the delayed or carryover effects of corrective advertising. The results of the tested corrective advertisements present evidence that one exposure of a corrective advertisement may not influence brand beliefs as would be hoped by public policy makers. Problems with the employed longitudinal design are discussed and suggestions for future research are offered.

INTRODUCTION

The effects of corrective advertising have been examined in several published laboratory experiments by consumer researchers. The general conclusion that may be gained from this literature is that one exposure of a corrective advertisement can, under certain circumstances, have the desired negative effect on target beliefs and attitudes. This paper examines the persistence of effects of corrective advertising over time. The thesis of this paper is that any obtained immediate effects of a corrective' ad are likely to rapidly dissipate over time. The experiment described in this paper tests this proposition by including measures delayed up to eight weeks after the exposure of a corrective ad. In addition, it includes several factors which provide a communication situation for the tested corrective advertising more externally valid than some past research in this area.

After first discussing the theoretical and empirical rationale for the predictions about the carryover effects of corrective advertising, this paper will review four past experiments about corrective advertising. Following that review, the methodology of this experiment will be described and contrasted with the methodology of the previous research in this area. Although the described results fail to support the predictions about the immediate and delayed effects of corrective advertising, it is hoped that this paper will encourage other researchers to examine delayed as well as immediate effects of advertisements and other marketing communications.

HYPOTHESIZED DECAY OF CORRECTIVE ADVERTISING EFFECTS

[Space prevents a full discussion of the literature discussed in this section. For more details, see Sawyer (1976) and Sawyer and Ward (1976).]

Figure 1 shows the hypothesized results of an experiment which measures beliefs toward an established product both immediately before and after exposure to a corrective advertisement for the product and also several weeks after exposure. Although the effects of the corrective ad (point 2) might immediately neutralize or even reverse the previous cumulative effects of the deceptive advertising campaign (point 1), the belief based on deceptive prior information would reassert itself with time (point 3). Although no previous corrective advertising research has examined carryover effects, several other areas of research about the effects of attitudes over time lend strength to the above prediction. Especially relevant are two areas of investigation of carryover effects: 1) the general nature of attitude decay and 2) the relative persistence effects of discounted or qualified messages.

Attitude Decay

If, as some researchers believe (e.g., Miller and Campbell, 1959), decay of attitude change resembles the negatively accelerated forgetting curve, a prediction of Jost's Second Law of Learning (Jost, 1897) is very relevant to the corrective advertising situation. Figure 2A illustrates the prediction that, for two stimuli equal in strength but of different age at a given point in time (t₂), the effect of the older of the two will decay less rapidly - reaching some asymptotic level higher than that of the newer campaign. Figure 2B shows the same curves as in Figure 2A except that curve D shows the hypothesized "positive decay" of any negative effects of corrective advertising. A major assumption in Figure 2B is that, whereas a brand belief rated highly probable will decay over time "down" toward some neutral point or mean probability for all competing brands; a belief rated highly improbable will decay "up" towards the same point of neutrality or mean probability for all brands. A copy test may misleadingly indicate at time t₁ that a corrective ad (C) is effective at reversing the effects of a deceptive campaign (D). Over time, however, the effects of the much newer and less often exposed corrective ad is likely to decay more rapidly and result in a belief much more positive than the immediate effect.

Discounted Messages

The "discounting cue hypothesis" theorizes that, if there is some aspect of a message that causes the message to be initially discounted or resisted, the passage of time may lead to a dissociation of the inhibiting cue from the original message so that the potential attitude change impact of the persuasive content may gradually take effect (see Gruder, et al., 1974). The combined impact of a positive (deceptive) message followed by a negative (corrective) message can be viewed as a discounted cue situation where the corrective ad discounts the previous positive advertising claim(s). The results of each of the three discounting cue experiments, which examined the delayed effects of messages highly analogous to corrective advertisements, are consistent with the prediction graphed in Figure 1.

FIGURE 1

HYPOTHESIZED IMMEDIATE AND DELAYED EFFECTS OF CORRECTIVE ADVERTISING

Weiss (1953) presented four exposures of a list of eight statements about the effects of smoking on health. Each statement was followed by the words "true" or "false" (e.g., "Studies have shown that smoking is an important factor in producing cancer of the lungs and throat. True"). Immediately after these four learning trials which were designed to produce anti-smoking beliefs, subjects in the discounted message group were also presented with a brief countercommunication in which they were told that more recent smoking experiments conclusively contradicted the findings of the research studies on which the statements were based and that they should be very skeptical about the truth of the original statements. Belief was then measured either immediately, three weeks, or six weeks after the messages.

FIGURE 2A

RELATIVE EFFECTS OF OLDER, MORE REPEATED ADVERTISEMENT (A) COMPARED TO NEWER COMPETITOR AD (B) WITH NEGATIVELY ACCELERATED ATTITUDE DECAY

FIGURE 2B

RELATIVE EFFECT OF DECEPTIVE ADVERTISING CAMPAIGN (D) AND NEW CORRECTIVE AD (C) WITH NEGATIVELY ACCELERATED ATTITUDE DECAY

Figure 3 shows the effects of the discounted message condition which is directly analogous to a situation of a repeatedly exposed, deceptive advertising campaign advocating that smoking is unhealthy followed by a corrective advertisement to the contrary viewpoint. As would be desired by public policy makers, the discounted message produced initially lower beliefs about the unhealthy effects of smoking than the non-discounted message. After six weeks, however, the relative effectiveness of the discounted (corrective) message condition was reversed from the immediate situation and produced significantly higher beliefs that smoking was unhealthy than the non-discounted condition. Ironically, the condition directly analogous to corrective advertising was even less effective in maintaining "correct" beliefs than the "uncorrected" messages when the effects were measured over time.

FIGURE 3

THE EFFECTS OF DISCOUNTED AND NON DISCOUNTED MESSAGES ON BELIEFS: (from Weiss (1953))

A recent experiment by Gruder, et al. (1974) found not only significantly less negative attitude decay in the discounted condition compared to a non-discounted message but also found a significant positive slope of attitude over time for the discounted message. Subjects were exposed to two exposures of either one of two messages -one arguing against a new law allowing right turns on red lights and the other arguing that a four-day work week decreases worker satisfaction. The discounted message condition also included a following message that cautioned the subject that, since the selection of the previous message for the experiment, it had been learned that the message's arguments were not true and had positively been refuted as "inaccurate and, frankly, wrong." Attitude change was measured both immediately and five weeks after the messages.

Not only were the slopes different as predicted, but the attitudes produced by the discounted (corrective) messages actually became more positive over the five week period. Although this increased attitude did not exceed the attitude produced by the message only condition, the large differences in slopes add further support to the prediction that a corrective ad is likely to be ineffective in producing enduring negative effects to eradicate the residual attitude based on deceptive information.

In the corrective advertising used in the Profile and Ocean Spray cases, the employed format was essentially that of a qualified argument where the messages were positive in overall nature but contained corrections of misimpressions that might have been formed by past advertisements. Bartlett (1932) predicted that, for a message arguing for one side of an issue but also presenting some qualifications and reservations against the main arguments, the qualifying details would be forgotten more quickly than the main conclusion. Because of this difference in forgetting, the qualified message would show a delayed action effect on beliefs with more change after a passage of time than immediately following the message.

Papageorgis (1963) tested Bartlett's hypothesis about the relative persistence of belief change of qualified and unqualified messages. Two messages low in ego involvement--one advocating that life expectancy was longer in cities than in rural areas, and the other arguing that there had been few technological advances in oil deposit location practices--were presented. A subject received the unqualified message for one topic and the qualified version for the other. The qualified message conditions contained the same arguments but stated much more equivocably with many reservations about the validity and generality of the contained arguments. Content recall and beliefs were measured immediately, two days, fourteen days, and forty-one days after the message.

The recall results were nearly exactly as predicted. For the qualified messages, recall of the main conclusions barely declined over forty-one days whereas there was substantial decay in the recall of the qualifications. As predicted, the decay in belief produced by the qualified messages was less than that of the unqualified messages. Although this difference averaged across both message topics was not statistically significant, the predicted relative advantage for delayed belief for the qualified message was statistically significant for the life expectancy issue. In other words, the qualified messages, which were quite similar in format to corrective advertisements, produced beliefs that, although initially lower as would be hoped by public policy makers, tended over time to revert toward the position held by subjects who viewed only unqualified messages.

IMMEDIATE EFFECTS OF CORRECTIVE ADVERTISING

Four laboratory experiments on the attitudinal effects of corrective advertising have all indicated that the effects of deceptive ads tended to be at least partially neutralized by one corrective ad. Hunt (1973) found that a print form of a corrective ad about allegedly false claims by Standard Oil of California produced a more negative attitude about Chevron than a group which viewed only the deceptive ad. Similar negative effects of a corrective ad resulted whether the copy was a detailed, research-based refutation of the truthfulness of the original claims or a general statement that the original ad was false and deceptive, and whether the source of the ad was Standard Oil, a government agency, or an independent consumer organization. The explicit, detailed corrective ad appeared to be especially effective.

Kassargian, Carlson, and Rosin (1975) examined the effects on product image of a corrective print ad for a motorcycle safety helmet. This corrective advertisement, which was sponsored by a retailer, was effective in reversing the positive effects of the original deceptive ad on several general dimensions of product image. However, the negative effects of corrective advertising did not generalize to the retailer.

Dyer and Kuehl (1974) reported slightly more equivocal results of fictitious corrective advertisements for nationally advertised products. One corrective ad was a print ad about a diet soft drink and the other was a radio ad about a suntan lotion. Only when the FTC was indicated as the source of the corrective message was the corrective ad for the suntan lotion effective in decreasing intentions to buy the product. When the source was the manufacturer of the product, the suntan lotion corrective ad was ineffective. Neither source was effective in decreasing purchase intention in the radio corrective ad for a diet soft drink.

Mazis and Adkinson (1976) tested four advertising conditions for a well-known mouthwash brand currently negotiating with the FTC about a corrective advertising order. Two corrective ads stressed that the brand was ineffective at preventing colds or sore throats but helped to kill germs. In addition, one of the corrective ads mentioned that the admission about the cold prevention ineffectiveness had been required by the FTC One positive, non-corrective ad stressed the germ-killing effectiveness, and the other positive-only ad emphasized only the breath freshening qualities without making any germicidal claims. The results showed that aided recall of the corrective ads was higher than of the non-corrective ads. Analysis of variance, controlled for brand usage, found that the target belief about the probability that the brand was effective at preventing colds or sore throats was significantly lower after exposure to the corrective ads. A similar decrease was found for the non-target belief about germ-killing effectiveness. Not only was the target belief strength lowered, but the evaluation of the goodness of that product attribute was also lowered by corrective advertising. No effect of corrective advertising on overall attitude toward the brand was found.

METHODOLOGY

Table 1 classifies the four past research studies according to several important methodological dimensions. These areas include the research method, advertised brand, corrective advertising conditions, and measurement details. In addition to including the time factor, this experiment differed from the previous ones in several important respects.

Advertised Brand

The product which was the focus of the corrective advertising in this study was Listerine mouthwash, which is currently under negotiation for corrective advertising with the FTC This product met several important criteria presented by Sawyer (1977). First, the product was a real brand familiar to the subject population. Although research using fictitious or unknown brands as stimuli is "clean", it is also irrelevant to the problem of whether and how corrective advertising can accomplish its goal of erasing the residual of a deceptive campaign. It is highly unlikely to have a strong residual belief about a fictitious or unfamiliar brand. Thus, it is not surprising that, in the studies which showed merely one exposure of a "deceptive" ad to subjects who had never seen that ad before, one exposure of the "corrective" and was able to negate its effects (Hunt, 1973; Kassarjian et al., 1975). [The Chevron F-310 ad in Hunt (1973) was not familiar to the Iowa college student subjects in Hunt (1973). The motorcycle helmet ad which was sponsored by a New York City retailer had not been exposed to the Maryland residents who were subjects in the Kassarjian, et al. (1975) study.] Such a design can be viewed as a primacy-recency experiment which is not appropriate to the research question of whether corrective advertising can overcome and reverse the residual of a familiar, highly repeated deceptive campaign. In the study reported here, the use of a well-established ad campaign as the "deceptive" advertising permitted a natural experimental design (Campbell, 1969) in which the deception was already established or, at least, easily reestablished with a reminder exposure of the deceptive advertisement. Also, the fact that subjects may have been previously exposed to the publicity of the court case involving the mouthwash product and the FTC added to the external validity of the research.

TABLE 1

SUMMARY OF PAST RESEARCH ON CORRECTIVE ADVERTISING

Second, the brand advertised an explicit claim which was believed by the general population. Such believability was ascertained by a pretest--using belief measures described below. Listerine was rated as significantly (p<.05) more probable than Scope, Lavoris, or Cepacol to possess the attributes of "prevents colds", "relieves nasal congestion", and "reduces colds severity"; more likely than Scope or Lavoris to "kill germs"; and less probable than the other competitors to "have appealing taste."

Finally, the advertising claim refuted in the corrective advertising was not telecast after the initial laboratory exposure to the corrective advertising. After early May of 1976 when this experiment was begun, Listerine did not use its campaign which stressed the (allegedly false and deceptive) claim of preventing colds and reducing the severity of their side effects. The former campaign was replaced by a new campaign which stressed an appeal based on germ killing and breath freshening effectiveness.

Ad Copy and Medium

Five 30-second TV commercials comprised the treatment stimuli. These ads, including one deceptive message, three versions of a corrective message, and one non-deceptive positive message, featured a professional actress and were filmed in color by the Photography and Cinema Department of Ohio State University. The advertisements were buried within a half-hour video tape of a television program. The copy for the advertisements employed the recommended corrective copy offered by the Federal Trade Commission for the brand (FTC vs. Warner-Lambert, 1974, p. 102), as well as copy devised by the researchers. It should be noted that the advertisements were of such quality that none of the subjects doubted the authenticity of the advertisements.

Subjects

Unlike past studies, subjects were not confined to a college student population. Of the 142 total subjects (122 experimental and 20 control), 50 were obtained from random mail solicitation and 113 subjects from a commercial marketing research firm. To give a general profile of the laboratory subjects, the average age of the participants was 43.1 years and they had an average household income of between $12,500 and $15,000. The sample included 29% males and 71% females; 32.7% of the subjects had college degrees; and only 15.6% were currently students.

Communication Measures

There have been two major problems with the cited past research on corrective advertising. It has, in our opinion, sometimes employed improper measures of effectiveness, and, as mentioned previously, it has measured only the immediate effectiveness.

If the FTC's goal for corrective advertising is to eliminate the residual of deception, then that residual must be defined. Advertising can attempt to achieve several communication goals including awareness, beliefs, attitudes, and intended or actual behavior. Concerning attitudes, an advertiser can try to influence either consumers' beliefs about some attribute(s) of its brand or competing brands or the evaluation consumers place on various attributes of brands in the product class in question. (Boyd, et al., 1972). It is very important to distinguish among a belief about how a particular brand rates on a particular attribute or dimension, an evaluation of that attribute, and a general affective attitude toward that brand (see Fishbein and Ajzen, 1975).

If an advertiser makes a deceptive claim about a particular attribute, the most direct effect would be to change consumers' beliefs about that attribute (e.g., Lutz, 1977). Of course, the ad may also have secondary effects on other beliefs or, through the changed beliefs, affect attitudes, purchase intention, or behavior. However, in the opinion of the authors, the prime criterion of impact of the deceptive claim is the change in the belief about that dimension and not anything else (see also Olson and Dover, 1978, Wilkie and Gardner, 1974). If belief is the prime residual of deceptive advertising, corrective advertising must focus on the particular dimension in question and change that belief to coincide with the ascertained facts.

In this study, informal interviews and pretests prompted the use of eight attribute dimensions of the mouthwash product category. Thus, the questionnaire measured eight beliefs about the advertised brand as well as three other competing brands and the perceived importance of those attributes. Most critical to the goals of this experiment, these measures were taken not only immediately following exposure of the corrective advertising but also two, four and eight weeks later.

Experimental Procedure and Design

Upon arrival at the laboratory, all participants were informed they would be participating in a study regarding advertising and general social issues and were asked to complete a questionnaire concerning attitudes toward advertising in general, attitudes toward government regulatory activities, product usage, and demographics.

The practice of some past experiments (e.g., Hunt, 1973, Kassarjian, et al., 1975) which included one exposure of the deceptive ad shortly prior to the exposure of the corrective ad seems low on external validity when, in an actual campaign, these ads would be widely separated. On the other hand, inclusion of an exposure of the deceptive ad might be necessary to trigger the deceptively-based belief. In order to test the impact of this factor, exposure to a deceptive ad closely preceding exposure to a corrective ad was experimentally manipulated. Half the experimental participants were exposed to a deceptive advertisement, which argued that the product was effective in reducing colds and sore throats; the other half viewed an ad for an entirely different product. Following this exposure, half of each of these experimental groups was given a questionnaire assessing attitudes toward the firm, attitudes toward several mouthwash brands, and product performance beliefs. The other half was given a "bogus" questionnaire measuring attitudes toward social issues. This procedure was followed to determine whether any learning effects were present for subjects using the same rating scales both before and after the corrective messages. (Subsequent analysis (Semenik, 1976) revealed no effects of this pretest, so all subjects were combined into one analysis).

During the next phase of the procedure, all subjects were informed that they were to view a program aired on network television concerning the solid waste disposal problem. On a video tape monitor and playback unit, subjects watched a seventeen minute program on this topic with the appropriate treatment stimuli contained as part of the program.

Depending upon the treatment group to which they had been randomly assigned, experimental participants viewed one of the four corrective advertisements contained in the program (see Semenik (1976) for the exact copy of each):

Corrective Version 1: One-sided positive appeal only. Since only favorable product attributes were communicated (deleting the deception in the previous advertising for the firm), this was not a corrective ad. This ad emphasized the breath freshening qualities. This version represents a possible outcome of a "cease-and-desist" order issued by the FTC.

Corrective Version 2: One-sided negative only. This ad explicitly stated that the product was not effective at reducing colds and sore throats. A firm might be required to run such an advertisement in which only the previous deception is explained and the firm is not able to present favorable product attributes.

Corrective Version 3: Two-sided, positive first. This communication related positive product attributes first and then revealed the deception of previous advertising.

Corrective Version 4: Two-sided, negative first. The firm first presented unfavorable information (disclosure of deception) and then presented positive product attributes.

Following exposure to one of the four corrective advertisements, subjects rated, as a diversion from the true purpose of the experiment, several open ended questions about the effectiveness of the television program- Subjects were then asked to complete another questionnaire regarding attitudes toward the brand, attitudes toward the firm, and product performance beliefs. Upon completion, they were each paid $2.00 and were told that they might be contacted again as part of their paid participation.

The experimental design was a 2 x 3 x 2 x 4 mixed, factorial design with order of measurement (immediately after ad exposure or delayed) as a within-subjects factor and length of delay (two, four, or eight weeks), prior exposure to original deceptive commercial or not, and corrective ad copy (positive-only; negative-only; two-sided, positive side first; and two-sided, negative side first) as between-subjects factors. Whereas all subjects were measured immediately, subjects in each ad copy-exposure condition were randomly assigned to one of the three delay conditions. Either two, four, or eight weeks later, they received a mail questionnaire which, after reminding them of their earlier participation, asked them to answer a "few last questions." These questions included belief and importance ratings identical to the previous immediate-past measures. Return rates were remarkably (and quite uniformly) high over the three delay conditions with 90.9%, 89.5% and 81.3%. respectively.

In addition to the experimental groups, another group (n=20) acted as an immediate post-measure control group. This latter group saw the same television show and filled out the same questionnaire but did not see any experimental ads. Regrettably, a clerical error caused an omission of delayed questionnaires sent to the control group. Table 2 presents the experimental design and sample sizes.

TABLE 2

EXPERIMENTAL DESIGN AND CELL SAMPLE SIZE

RESULTS

Two experimental hypotheses were tested in an analysis of variance. [For the description and results of several hypotheses about the immediate effects of the three corrective advertisements, see Semenik (1976).]

H1: The corrective ads would have an immediate negative effect on brand beliefs about the prevention/reduction of colds and sore throats. However, this immediate effect would dissipate over the eight week test period in a manner graphed in Figure 1.

H2: The positive-only copy (non-corrective) condition would have an immediate positive effect on brand beliefs. This immediate effect would decrease over the eight week post period (in a fashion directly opposite to the corrective ads).

To support the first hypothesis of initial lowering of beliefs by the corrective ads and a gradual increase over the eight week period, the order of measurement by length of delay interaction term of the ANOVA should have been statistically significant for attributes relating to colds and cold symptoms prevention. A significant main effect of length of delay by itself would not be a sufficient test of hypothesis one since differences among subjects were expected only when measured after a delayed period. There was no reason to expect differences among these groups when they were each measured immediately after exposure to the corrective ad. The second hypothesis concerning the positive (non-corrective) ad would be supported by a significant main effect of the ad term and a significant three-way interaction of ad and length of delay with order of measurement. The ad copy main effect term would test whether the positive and negative (corrective) ads had, as hypothesized, opposite effects, and the third-order interaction term would test whether these opposite effects had opposite decay effects over time.

Table 3 presents the results of the ANOVA for the sixteen measures. In addition, the analysis of the sum of the products of the importance and beliefs of each of the eight dimensions is shown. For convenience of presentation, only the statistical significance levels of .10 or lower are reported. Two separate ANOVA are reported; the first figure in each column is the results of an ANOVA with four levels of ad copy, and the second figure (in parentheses) is the results of a second ANOVA in which only two levels of ad copy are considered - corrective (negative only or two-sided) and non-corrective (positive-only).

A look at the results reveals several interesting findings not directly salient to the research hypotheses. First, the second ANOVA with ad copy collapsed into two levels seems more statistically powerful. This is probably due to the fact that the ad copy factor in both ANOVAs had very little effect and there was a smaller error term in the second analysis. Second, it seems that being exposed to the deceptive ad shortly before one of the four versions of the test Listerine ad consistently had an effect on both the importance and belief perception on the various attributes (see the column entitled "Exposure" in the table). A check of cell means revealed that the effect of such exposure was always to decrease importance or to decrease the perceived probability that the brand possessed the attribute in question. It appeared that this condition artificially increased the impact of the subsequent corrective ad. Perhaps the first ad sensitized subjects to pay more attention to the later exposure.

As stated earlier, the ANOVA term relevant to the hypotheses about the carryover effects of the corrective ad was the length of delay by order of measurement interaction. Table 3 shows that there was no support for this hypothesis; in no instance was this interaction term statistically significant at the .05 level. The main effect of time was significant for several attributes. However, the absence of a significant interaction term testifies to the fact that, in no case, was the predicted increase in importance or belief probability significantly greater in the delayed conditions (where the increases were predicted) than in the immediate conditions (where there was no reason to expect differences among groups who differed only in that, subsequently, they were to be measured after different delays). For most attributes, there were equal increases for those in the immediate measure condition and those in the delayed condition.

The only result that resembled the hypothesized effects was the importance of "long lasting" which was not explicitly addressed in any of the rested Listerine ads. The time of delay by measurement order interacting term was nearly significant in the first ANOVA (F = 2.88, 2 and 98 df, p = .07). Table 4 presents the cell means which show no changes in ratings in the immediate conditions and an increase in the delayed conditions. This was the result that was predicted and not found for beliefs relating to the prevention of colds and cold symptoms addressed in the corrective advertising copy. As an example typical of most of the other ratings, Table 5 presents the mean ratings of the target belief "prevents colds/sore throats". It can be seen that an overall increase was found (F=3.17, 2 and 110 df, p=.05) but very little difference was found between the three immediately measured groups and these same three groups measured, respectively, two, four and eight weeks later.

Given the lack of overall support for the predictions about the decay of beliefs, the question arises as to whether the ads had any effect at all. As previously noted, the ad copy variable showed no significant effects. To test the immediate effects of the ad copy, the positive only (non-corrective) and negative only (corrective) copy appeals were compared to the immediate control group with Dunnett's Comparison test (see Kirk, 1968). These two conditions were chosen because of the conceptual non-ambiguity of the direction of their likely effects.

TABLE 3

SUMMARY OF ANALYSES OF VARIANCE OF IMPORTANCE OF EIGHT DIMENSIONS AND LISTERINE BELIEFS ON THOSE DIMENSIONS

TABLE 4

MEAN IMPORTANCE OF "LONG LASTING" AS A FUNCTION OF ORDER AND TIME OF DELAY OF MEASUREMENT

TABLE 5

MEAN RATINGS OF "PREVENTS COLDS/SORE THROATS" AS A FUNCTION OF ORDER AND TIME OF DELAY OF MEASUREMENT

Table 6 shows only a limited number of immediate effects. There was little difference between the positive and negative copy. The main area of immediate impact was the effect of the negative, corrective ad on the importance of four attributes - "appealing taste" and "long lasting" (decrease) and "relieves nasal congestion" and "reduces coughing/sneezing" (increase). Because, as noted earlier, control groups were unfortunately not remeasured in the delayed conditions, Table 6 presents comparisons of delayed measures with the immediate control group. Although the validity of these comparisons is tenuous at best, it is interesting to note that, in comparison to the immediate control group, both the importance and Listerine belief on the dimension of "prevents colds/ sore throats" went from significantly negative two weeks after the negative copy (corrective ad) exposure to significantly positive at eight weeks. A similar reversal can be noted for the importance of "reduces severity of colds/sore throat" and "prevents bad breath".

SUMMARY AND DISCUSSION

The goals of this experiment were to test, for the first time, the carryover effects of corrective advertising and to add to the external validity of past research about corrective advertising. Unfortunately, the results are accompanied by more questions than answers. Although the tested television commercials were explicitly worded, they showed little immediate effects on target beliefs, This non-effect contrasts with the effects found by others including Mazis and Adkinson (1976) with the same advertised mouthwash product. Moreover, no differences between corrective and non-corrective (positive only) ads were found. Finally, no support for the hypotheses about the carryover effects of the corrective ads was found.

Of course, the limited immediate effects made the tests of delayed effects somewhat moot. The hypotheses of rapidly decaying immediate effects assumed that the ads would, as in past research, have the desired initial impact. It might be concluded that the changes adding to external validity in this study also served to limit the effectiveness of the corrective advertisements. Some of the cited past research which resulted in immediate effects of corrective advertising did not focus on a product with familiar claims as the target of the corrective advertising, and thus made it easier for the corrective advertising to be effective. However, the Mazis and Adkinson (1976) experiment, which used the same product, similar beliefs measures, and somewhat similar advertising copy (although a different medium) as this study, did find significant immediate effects on beliefs. Thus, a conclusion that a "stiffer test" of corrective advertising in this study limited the effects of corrective advertising does not seem to be a valid excuse for the non-effects in this experiment.

TABLE 6

RESULTS OF DONNETT'S COMPARISON OF POSITIVE AND NEGATIVE AD COPY WITH CONTROL GROUP

Although it is conceptually appealing that the proper measures of corrective advertising be belief specific, the few observed effects of the tested corrective ads on rated beliefs and importances were neither intuitively nor theoretically sensible. It seemed that the major impact of the negative corrective copy was to affect importances and not the target ratings. Although the idea that consumers' importances are altered by corrective advertising is not surprising (see Mazis and Adkinson, 1976) it is odd that such effects would be in lieu of effects on beliefs.

If the target beliefs about cold prevention had been immediately lowered as hypothesized, then it would seem that the importance of those beliefs would also decrease since that attribute would be perceived as less of a discriminator among competing brands. Instead, the observed results in Table 6 showed an increase in the importance of the cold and cold symptoms prevention beliefs ("relieves nasal congestion" and "reduces coughing/sneezing") and a decreased importance of (what, conceptually, should be more important) other beliefs ("appealing taste" and "long lasting").

The fact that parallel differences were found in both the immediate and delayed measures is simply baffling. A demand artifact explanation makes no sense since subjects were randomly assigned to a delay condition after they had been immediately measured.

The fairly consistent patterns over two, four and eight weeks suggest the strong possibility that some uncontrolled external force influenced subjects' ratings. It may be that what was thought to be a major strength of the research design - the use of a naturally occurring advertising campaign - proved to be a major problem. Although the mouthwash product did not use the (allegedly deceptive) "colds" campaign during May and June of 1976 when the data were collected, other positive advertising for the product was being telecast. Perhaps this continued advertising increased all ratings over the period of time in question. If so, the observed increases may not represent, as originally hypothesized, a reversion to former beliefs, but may rather represent the positive effects of Listerine's spring advertising campaign. Alternative speculation would suggest a "ceiling effect" to explain the lack of immediate effects. It could be that the FTC publicity and other available information about the Listerine case in combination with the fact that Listerine had recently ceased its "colds campaign" had already, to some degree, lowered the target beliefs of the corrective advertising. If so, merely one exposure of the corrective ads could not be expected to further decrease beliefs very much. Perhaps the observed increase over time was a decay of the effects of the negative publicity surrounding the case and not, as originally hypothesized, the single corrective ad exposure. Of course, the regrettable lack of delayed control groups becomes a crucial void because of the corresponding lack of ability to answer questions about the effects of any brand advertising during the experimental period. It is also possible that the slightly lower return rate at eight weeks delay represented a lower number of negatively oriented subjects.

The authors plan to re-examine the persistence of effects of corrective advertising. Planned improvements include a strengthening of the power of the experimental design by focusing on the negative-only corrective copy condition with no reminder exposure of the deceptive ad and having a greater sample size per cell. Also, repetition of the corrective ad will be experimentally manipulated to (hopefully) increase its initial impact. Inclusion of proper control groups for the delayed measure conditions (which were planned but not implemented in this experiment) should allow a separation of immediate and delayed effects of the experimental ads from ads exposed outside the laboratory in the period following exposure to the test ads. Finally, improved measurement of cognitive structure following the suggestions of Ahtola (1975) will be incorporated. The questions and problems raised in this study deserve answers and renewed attempts at improved solutions. Despite the methodological challenges, attitude research that goes beyond a static perspective is sorely needed in consumer research.

REFERENCES

Ahtola, O. T., "The Vector Model of Preferences: An Alternative to the Fishbein Model." Journal of Marketing Research, 12(1975), p. 52-59.

Bartlett, F. C., Remembering: A Study in Experimental and Social Psychology. New York: MacMillan, 1932.

Boyd, H. W., Ray, M. L., and Strong, E. C., "An Attitudinal Framework for Advertising Strategy," Journal of Marketing, 36(1972), p. 27-33.

Campbell, D. T., "Perspective: Artifact and Control," in Rosenthal, R. and Rosnow, R. (eds.), Artifact in Behavioral Research, (New York: Academic Press), 1969, p. 351-382.

Dyer, R. F. and Kuehl, P. G., "The 'Corrective Advertising' Remedy of the FTC: An Experimental Evaluation," Journal of Marketing, 38(1974), p. 48-54.

Fishbein, M. and Ajzen, I., Belief, Attitude, Intention and Behavior, (Reading, Massachusetts: Addison-Wesley), 1975.

FTC vs. Warner-Lambert, Initial Decision, FTC Docket 8891, November 25, 1974.

Gruder, C.; Hennigan, K.; Halamaj, J.; and Cook, T., "An Empirical Test of the Absolute Sleeper Effect Derived from the Discounting Cue Hypothesis." Unpublished manuscript, University of Illinois at Chicago Circle, 1974.

Hunt, H. K., "Effects of Corrective Advertising," Journal of Advertising Research, 13(1973), p. 15-22.

Jost, A. "Die Assoziationsfestigkeit in ihrer Abhandigkeit von der Verteilung der Wiederholungen." Zeitschrift fur Psychologie und Psysiologie der Sinnesorgane, 14(1897), p. 436-72.

Kassarjian, H. H., Carlson, C. J. and Rosin, P. E., "A Corrective Advertising Study," in M. J. Schlinger (ed.), Advances in Consumer Research, Volume 2 (Chicago: Association for Consumer Research), 1975, p. 631-642.

Kirk, R. G., Experimental Design: Procedure for the Behavioral Sciences, (Belmont, California: Brooks/Cole), 1968.

Lutz, R., "An Experimental Investigation of Causal Relations Among Cognitions, Affect, and Behavioral Intention," Journal of Consumer Research, 3(March 1977), 197-208.

Mazis, M. B. and Adkinson, J. E., "An Experimental Evaluation of a Proposed Corrective Advertising Remedy," Journal of Marketing Research, 12(May, 1976), p. 178-183.

Miller, N. and Campbell, D., "Recency and Primacy in Persuasion as a Function of the Time of Speeches and Measurement," Journal of Abnormal and Social Psychology, 59(1959), 1-9.

Olson, J.C., and Dover, P.A., "Cognitive and Behavioral Effects of Deceptive Advertising," Journal of Marketing Research (1978), forthcoming.

Papageorgis, D., "Bartlett Effect and the Persistence of Induced Opinion Change," Journal of Abnormal and Social Psychology, 67(1963), 61-7.

Sawyer, A. G., "The Need to Measure Attitudes and Beliefs Over Time: The Case of Deceptive and Corrective Advertising,'' in Proceedings, 1976 Educators Meeting, (Chicago: American Marketing Association), 1976 P.380-5.

Sawyer, A. G., "Measurement of the Effects of Corrective Advertising," in Wind, Y. and Greenberg, M. (eds.), Attitude Research Comes to A Head, (Chicago: American Marketing Association), 1977.

Sawyer, A. G. and Ward, S., "Carryover Effects in Advertising Communication: Evidence and Hypotheses from Behavioral Science," Working Paper No. 76-122, Marketing Science Institute, December, 1976.

Semenik, R. J., "Message Forms and Effective Corrective Advertising: An Experiment," unpublished doctoral dissertation, The Ohio State University, 1876.

Weiss, W. "A Sleeper Effect in Opinion Change." Journal of Abnormal and Social Psychology, 48(1953), p. 173-80.

Wilkes, R. E. and Wilcox, J. B., "Recent FTC Actions: Implication for the Advertising Strategist," Journal of Marketing, 38(1974), p. 55-61.

Wilkie, W. and Gardner, D., "The Role of Marketing Research in Public Policy Decision Making." Journal of Marketing, 38(1974), 38-47.

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