The Impact of Changing Cigarette Warning Message Content and Format

Gaurav Bhalla, University of Kansas
John L. Lastovicka, University of Kansas
ABSTRACT - As regulatory agencies become more proactive, the need for impact evaluation studies, prior to enactment of legislation, increases. Laboratory experimentation is useful for such studies. The paper reports a study which evaluated the impact of a potentially proposed change in the statutory warnings in cigarette print advertisements.
[ to cite ]:
Gaurav Bhalla and John L. Lastovicka (1984) ,"The Impact of Changing Cigarette Warning Message Content and Format", in NA - Advances in Consumer Research Volume 11, eds. Thomas C. Kinnear, Provo, UT : Association for Consumer Research, Pages: 305-310.

Advances in Consumer Research Volume 11, 1984      Pages 305-310

THE IMPACT OF CHANGING CIGARETTE WARNING MESSAGE CONTENT AND FORMAT

Gaurav Bhalla, University of Kansas

John L. Lastovicka, University of Kansas

[The authors thank Jack W. Brehm, George Coan, Jr. and Surendra N. Singh for their helpful comments on this research. This research was supported by a grant (project #8227) from the University of Kansas Business Research fund, provided by the Fourth National Bank of Wichita, Kansas.]

[Gaurav Bhalla is Doctoral Candidate and John L. Lastovicka is Associate Professor at the School of Business, University of Kansas, Lawrence, Kansas 66045.]

ABSTRACT -

As regulatory agencies become more proactive, the need for impact evaluation studies, prior to enactment of legislation, increases. Laboratory experimentation is useful for such studies. The paper reports a study which evaluated the impact of a potentially proposed change in the statutory warnings in cigarette print advertisements.

INTRODUCTION

In a major report, the Surgeon General isolated fifteen priority areas essential for achieving national health goals (U. S. Department of Health 1980). Smoking is one of the priority areas. One of the specific recommendations, made to enhance consumer knowledge of smoking hazards, suggests strengthening both the visibility and content of the statutory warning statement required in cigarette advertising

The FTC's evaluation of the current warning has led to these conclusions: a) the current warning describes the hazards of smoking in a generalized and non-personal way, thus making it inadequate as an educational device; and b) the current message is overexposed and worn out due to its invariant shape and content (FTC Staff Report 1981). Consequently, there is considerable support for improving consumer education of smoking hazards via changing the content and shape of the warning statement. However, before consumers learn, consumers must notice and read. It is the purpose of this study to determine how and why factors like size and content impede and/or facilitate noticeability and effect of current and potentially proposed warning messages.

THEORY

Before changing warning message policy, it is appropriate to speculate why the current warning is judged ineffective in educating the consumer about the hazards of smoking.

Due to the dominance of cognitive psychology in consumer research (Kassarjian 1982), there are no lack of theories explaining how consumers process information and make choices (Hansen 1972, Bettman 1979, Sternthal and Craig 1982). Yet, there is sufficient evidence in both marketing (Olshavsky and Granbois 1979) and social-psychology (Langer et al. 1978), suggesting that much behavior occurs without paying attention to the substantive details of the "informative" environment.

Abelson (1976) explains this type of behavior by postulating that when confronted with incoming information stimuli, prior "scripts" written when similar information really was once new, may be stereotypically enacted. A person engaging in scripted behavior would systematically process a part of the incoming information and systematically ignore a part of it. This is done not because the information is irrelevant, but because it is already known. This knowledge has been obtained because the incoming information has been seen many times in the past and aspects of its structure are known. Consequently, knowledge of message structure yields inference of message content. Thus, knowledge that it is mandatory for all cigarette advertisements to carry a warning message, the regular monotony with which the warning appears in the corners of print advertisements, the invariant sans-serif and condensed type face enclosed in a small rectangular box with black border lines, etc., may trigger scripted behavior and lead to non-comprehension of any new warning message presented in the same format as the old warning message.

A complementary explanation for the above is provided by using "schemas" as units of explanation. Schemas are generic knowledge structures that guide the comprehender's interpretations, inferences, expectations and attention. Thus, a schema guides the comprehension of incoming information stimuli through filling the various variables that it is comprised of. As these variables form a tightly knit structure, filling of one variable may have repercussions on other variables, sometimes causing a specific variable to be filled by default (Graesser and Nakamura 1982). Berlyne (1971) addresses the same phenomenon, though with a different approach. By invoking the interdependence between space and its constituent elements, he explains that once a particular element has been found to occupy a particular location, certain other elements are more likely, in comparison with others, to occupy other locations. He calls this syntactic information transmission, a concept similar to the information theoretic concept of redundancy.

In a series of experiments conducted by Langer and Abelson (1972) and Langer et al. (1978), it was demonstrated that for thoughtful and non-scripted action to occur in the context of written and oral communication, it is necessary for the transmitted message to be structurally novel, rather than merely semantically difference. Novelty of structure stimulates exploratory behavior and, hence, provokes thoughtful action (Berlyne 1971). Structure or format of a message refers to the manner in which a message is presented; the terms structure and format will be used interchangeably. In contrast to how a message is presented, message semantics refer to message content or what is transmitted; the terms message semantics and content will also be used interchangeably.

Given these theory-based explanations, the issue becomes one of evaluating the impact of potential action proposed by the FTC in preventing the occurrence of scripted behavior or the filling in of the current warning message content, despite a change in cigarette warning message content. This leads to the formulation of the first hypothesis:

H1: Message format impacts message content effect. More specifically:

H1A: Not changing the format of the warning message will inhibit message content effect, even if the message content is altered.

H1B: Incrementally changing the format of the warning message should increasingly facilitate message content effect.

A quick look at some current cigarette advertising demonstrates that much cigarette advertising can be classified in at least one of two broad categories. The first group consists of advertisements for brands like "Carlton." These advertisements almost exclusively contain verbal information on tar and nicotine content and may make inter-brand comparisons. These advertisements will be referred to as "textual" in the rest of the paper. The other group, e.g., "Marlboro," contain very little information and rely on visuals for communication. These advertisements will be referred to as "pictorial." Clearly as the warning message must operate in one of these two contexts, their differential impact must be accounted for. This is true for a number of reasons. To begin with, advertising media vehicle context plays an important role in advertising message recall (Clancy and Ostlund 1976). Further, category size plays an important role in cued recall. For any retrieval cue to be effective at test, it must have been encoded during the study trial. It is reasonable to assume that the likelihood of encoding decreases as the category size increases (Nelson 1981). Textual advertisements present a larger set size and may thus inhibit encoding and subsequent retrieval as compared to say a pictorial advertisement. Also, consumers may have different schemas to guide comprehension of different types of advertisements. To the extent that the warning message is essentially verbal, it represents an atypical element in a visual processing schema; atypical elements are remembered better when the stimulus material involves pictorial scenes (Friedman 1979, Goodman 1980). This leads us to the second hypothesis:

H2: Changed content and format of the warning message should facilitate warning content effect to a greater degree in a pictorial advertising context than in a textual advertising context.

METHOD

The hypotheses were tested with a 3 (format types) x 2 (message content statements) x 2 (advertisement types) experimental design in a laboratory setting. The three format levels used are shown in the Figure and are labelled "current," "mild" change and "severe" change. In a manipulation check carried out prior to running the experiment, subjects rated the "severe" change above the "mild" change format level in terms of degree of difference from current format.

The two warning statements used were the current warning: "The Surgeon General Has Determined That Cigarette Smoking Is Dangerous to Your Health" and a new, more specific warning: "The Surgeon General Has Determined That Cigarette Smoking Produces Cancer and Asthma." Both the current and the new warning messages used have exactly 73 letters each. This prevented the alteration of the format by default as the new message, like the old, could be reproduced in exactly two lines in the current format and in exactly one line in the mild format manipulation. The two single page advertisements used were a textual "Carlton" and a pictorial "BelAir" advertisement. Both were obtained from past issues of Newsweek.

The manipulations (i.e., the warning messages and their formats) were professionally prepared and inserted in the advertisements to minimize altering the original advertisements. The advertising manipulations and seven other advertisements and news stories, consisting of four single-page advertisements (Myers's Rum, Alitalia, Cannon Camera and Oldsmobile) and three single-page self-contained news stories (a photo feature on Princess Grace, Howard Baker's candid camera and a plane crash in Spain) were color xeroxed.

The experimental material for the study consisted of identical red ring folders, each with eight clear poly sheet protectors. Into these protectors were inserted one cigarette advertisement, representing one of the twelve experimental cells, and the seven additional items described above. The order of presentation of materials was randomized across the various cells.

Subjects for the study were recruited using a signup sheet from a pool of college students enrolled for an introductory psychology course. The relevance of student subjects for this research can best be appreciated with the following: "Indeed more teenagers are smoking today than ever and they are starting to smoke at a younger age." (FTC Staff Report 1981). Further, the FTC considers teenagers, along with pregnant women and the elderly, to he especially vulnerable consumer groups across product categories.

The experiment simulated a living room scenario, a likely setting for reading magazines such as Newsweek. Subjects were told the experiment's purpose was to understand the distracting effects of music on reading. Subjects were randomly assigned to one of the experimental conditions. Equally important, the experimenter was blind to the condition, also. Once subjects had relaxed, the folder was handed over and the tape recorder turned on. After the first three bars of Simon and Garfunkel's "Mrs. Robinson" played, subjects were signaled to open to the binder's first page. Thereafter, every 15 seconds subjects were asked to turn the page. This was repeated till such time as subjects had finished leafing through the entire folder. It is important to note that throughout the experimental situation, neither the experimenter, nor any experimental material was in view of the subjects. The purpose behind the 15 second limit was twofold: a) to simulate as closely as possible real life viewing of print advertisements, and b) Krugman (1966) reported that subjects spent an average of 19.4 seconds in unforced viewing of print advertisements. To counter the likely excessive cognitive activity triggered by the laboratory setting, it was thought safer to err on the lower side, 15 seconds, than on the higher side, 20 seconds.

After stimuli exposure, four types of measures were used to assess learning of the advertising and editorial material in the simulated magazine. Learning of the non-cigarette advertising stimuli was measured to minimize subjects' suspicions that the true focus of the study was cigarette advertising. The four types of measures were: non-specific unaided recall, specific aided recall, recognition and belief. For the cigarette advertisement

measures, the non-specific unaided recall assessed if subjects claimed to have seen a warning message when asked: "What do you recall having seen in the folder?" Again for the cigarette ads, the specific aided recall was measured by responses to this open-ended request: "Describe what the warning statement in the cigarette advertising said." In short, unaided recall measured attention (i.e., noting the presence of a warning message) while aided recall measured comprehension of the content of the warning message. Recognition was measured by asking subjects to choose between a pair of warning content statements, presented on two separate index cards. The cigarette warning recognition task was one of a set of other recognition tasks based on other ads and news stories read by the subjects. Both the order of presentation and the placement of the recognition alternatives either to left or right was varied between subjects. Finally, subjects were asked to indicate belief on a seven-point scale ranging from "extremely likely" to "extremely unlikely" or from "extremely possible" to "extremely impossible" regarding the presence/absence in the folder of a set of warning message content statements. Again, the actual warning statement belief question was only one of a set of belief questions based on the folder's editorial and advertising content. Here again, the order of presentation of the various statements/scenarios was varied between subjects.

FIGURE

FORMAT MANIPULATIONS

TABLE

LEARNING OF THE WARNING MESSAGE

Finally, subjects were checked for plausibility of the cover story, debriefed as to the true purpose of the experiment, awarded credit and allowed to leave. The entire experiment took, on an average, 25 minutes to complete. The post-experimental session strongly suggested that the cover story had been maintained.

Two further points: first, the entire experimental procedure was pretested; second, all subjects were run by one experimenter.

RESULTS

The table summarizes the proportion of correct responses (i.e., proportion of comprehension), for each of the four response variables, in each of the twelve experimental conditions.

The recall and recognition measures are dichotomous (i,e., a subject either correctly comprehends the warning message or does not). Also, the Fishbein-type belief measure, which was recorded on a seven-point scale, was dichotomized by using the scale's mid-point as a split. Consequently, given this categorical :lata and sparse cell values, it was decided to conduct a log-linear analysis of the 2 response levels (comprehension and miscomprehension) x 2 advertisement types (textual, pictorial) x 2 warnings (current, new) x 3 formats (current, wild, and severe) table, to determine the model which best fit the cell frequencies.

Briefly, the analysis is based on fitting a hierarchical log-linear model to the cell frequencies; that is, the logarithm of the expected cell frequency is written as an additive function of main effects and the interactions in a manner similar to the more common analysis of variance model. (For a more detailed discussion of the log-linear model, see Bishop et al. (1975) or Fienberg (1977).) Of special interest would be to verify whether those components of a fully specified log-linear model, as indicated by the hypotheses, are able to explain the data in the table. It is pertinent to point out that unlike the more common analysis of variance case wherein the hypotheses are largely independent, the log-linear testing of hypotheses in this study will necessarily be interdependent. The sequence in which analysis and results are presented will, thus, not necessarily follow the order in which the hypotheses were presented.

Taken individually, the hypotheses have implications for the significance of these interaction effects: two-factor interaction effects between response and format (RF) to explain response variations across levels of format, and three-factor interaction effects between response, content and format (RCF) as postulated by H1B. Hypothesis H2 specifies that the two factor interaction effect between response and type of advertisement (RA), reflecting response variation across advertisement type, should be significant. Tn addition, H1A stipulates that changing the content of the message in the current Format node should be equivalent to no change at all. Since comprehension should increase due to structural manipulation, the two-factor interaction effect between response and content (RC) should be non-significant.

One further point before the analysis and results are presented is in order. A look at the table indicates that the proportion of responses for both recognition and belief are the same. Consequently, as the analysis for both will be identical, results for only recognition will be presented.

H1B related tests

Analysis yielded only partial support for H1B. The two-factor interaction effect of RF, indicating response variations across levels of format, is significant only for unaided recall (X = 13.3, p < 0.01). For the recognition measure, though insignificant at the criterion probability of 0.05, it is significant at p = 0.10 (X2 _ 4.9, p < 0.08). For the aided recall measure it is insignificant. These results are not altogether unexpected. In terms of complexity of the learning task, the dependent measures form a hierarchy, with the non-specific unaided recall task being the least demanding, aided recall the most demanding and recognition falling in between. In fact, non-specific unaided recall may essentially be labeled an attention or a noticeability measure, while the other two may be called comprehension measures. Format manipulations were successful in stimulating noticeability and, hence, scored significantly on unaided recall. But as the learning task became progressively more complex, the interaction effect of RF impacted only weakly on message content, and declined from marginal significance for the recognition measure to nonsignificance for the aided recall measure. The three-factor interaction effect of RAF, between response, advertisement type and level of format is not significant for any of the response measures.

H2 related tests

Analysts yielded complete support for H2. The two-factor interaction effect, RA, reflecting response variation across type of advertisement is significant for unaided recall (X2 = 6.13, p < 0.01), for recognition (X2 = 8.26. p < 0.004) and for aided recall (X2 = 7.68, p < 0.005).

H1A related tests

Support for H1A was also partial. A look across the rows labeled "Current Format" in the table for the dependent measures of aided recall and recognition, suggests that changing the content of the warning message in the current format is equivalent to no change at all. Either none of the subjects comprehended the new content, or at best, an insignificant few. The second prediction of H1A, namely, that the two-factor interaction effect, RC, should be non-significant met with only limited success. While it is insignificant for the unaided recall measure, it is significant for both the aided recall measure (X2 = 48.38, p < 0.001) and the recognition measure (X2 = 66.66, p < 0.001).

While the above results are contrary to those hypothesized, it is easy to see why they occurred. It is a direct outcome of failure of variation in format to boost comprehension for the new content. An inability expressed earlier as the failure of the two-factor interaction effect of RF to achieve significance under H1B. The last two columns in the table clearly indicate that either none of the subjects were able to comprehend the new content or, at best, a few.

In summary then, the results indicate only partial support for both H1A and H1B, while fully supporting the predictions of H2. Clearly, varying semantics alone, given the current format mode, is insufficient in aiding learning. Variations in format and advertising context both seem to be aiding learning though the effects of the former are not as strong as predicted. Deductive reasoning to explain this will be presented in the next sect ion

DISCUSSION

Using experimental research to aid public policy decision making by assessing potentiaL impact of proposed regulation is a recent phenomenon (Houston and Rothschild 1980). As agencies become increasingly proactive, it becomes imperative to empirically pre-test proposed policy before it is enacted. The above study is cast in this mold.

To help put this study's results in perspective, the study's limitations should be addressed. The first limitation deals with the number of exposures . In this study, each subject was exposed to the cigarette advertisement only once. To the extent that several exposures facilitate advertising effects beyond attention, the findings are limited. Second, it is highly likely that if any new warning message was implemented, it would be accompanied by announcements over TV and other mass media. This would alter consumer expectations, thus disrupting scripted behavior. No such extraneous information was available to subjects in the study. Third, the sample size (n=84) for this study is low. This is really not a limitation as obtaining statisticaLly significant results with a small sample suggests the detection of substantial, non-trivial effects (Sawyer and Peter 1983).

Clearly, the results support a part of H1A, i.e., changing warning message content was as effective as no content change at all, in the absence of format change. Though it was not, perhaps, the recurring aspect of structure that triggered scripted behavior. Further, the results clearly show that variations in format levels and advertisement type impacted warning message comprehension . The implications of this finding should not be overlooked. First, to the cigarette advertiser operating under current regulations, this suggests that the warning message's effectiveness can be either decreased or enhanced, depending upon whether, say, a "Reason Why" textual advertising format or an "Imagery" pictorial format is used. Second, to public policy makers, this suggests that regulations ought not to be "means"-oriented by mandating one particular warning message tactic. Rather, these results suggest the necessity for more "ends"-oriented regulations specifying that some targeted level of comprehension be achieved without specifying the "means" or particular warning message format for accomplishing some comprehension "end." Advertising managers have long ago realized the desirability of defining clear-cut advertising campaign goals and then deveLoping the creative message strategy and tactics to help achieve the campaign goals. It seems that attention to some basic advertising management ideas could help the management of mandatory warning message campaigns. The dangers of being overly message tactics- or "means"-oriented are well known in the advertising community.

Speculations as to why hypotheses H1A and H1B only met with partial support wilt now be presented. Ironically, perhaps, the attempt to maintain internal validity prevented further thoughtful In action from occurring. Recall that to prevent variation of format by default, the new warning message was set at exactly seventy-three letters and began, "The Surgeon General Has Determined That Cigarette Smoking ...." It is likely that this leading set of words triggered scripted behavior and led to filling in of the rest, with the old message, even when the rest of the warning read ". . . Causes Cancer and Asthma." This led to not comprehending the new message and resulted in the RC two-factor interaction effect, overpowering the hypothesized RF two-factor interaction effects.

Further, there remains the nagging suspicion of stored knowledge surreptitiously influencing the aided recall, recognition and belief measures. It is impossible to determine whether a subject correctly responding to the recognition question is doing so, because the subject actually recognizes the statement as similar to the one seen, or the statement was recognized as the one expected to be in the advertisement. Altogether, four subjects voluntarily admitted not having seen the message, and yet responding as if they had seen it, in the post-experiment interview. Clearly, there were others, too, in the same category. Because the analysis was conducted with raw, unadjusted data (i.e., data unadjusted for guessing), an understandable dilution of effects postulated in H1A and H1B occurred.

The results and speculation discussed above point to the following conclusions: first, variation in levels of format and advertisement types prominently influence advertising effect; second, the more severe the departure of the format from the existing format and the less textual the advertising context, the greater the potential effect. It is speculated, however, that subsequent elaboration will be facilitated, only if the warning message has minimum overlap with existing scripts used to process cigarette advertisements and their warning message.

It is generally agreed (Jacoby 1974, Wilkie 1975) that an a priori definition of program goals is crucial for effective evaluation of mandated consumer information programs. In the current context, the final objective is improving consumer knowledge of smoking hazards. However, before consumers learn they must notice and read. The current study demonstrates that manipulating both content and format simultaneously enhances noticeability and attention. Further, research should seek to determine in a field-setting whether sufficient repeat exposure of the new message content in its new format does in fact promote significantly higher levels of comprehension and recall of smoking hazards.

REFERENCES

Abelson, Robert P. (1976), "Script Processing in Attitude Formation and Decision Making" in Cognition and_Social Behavior, eds. John S. Carroll and John W. Payne, Hillsdale, N.J.: Lawrence Erlbaum, 33-46.

Berlyne, David E. (1971), Aesthetics and Psychobiology, New York: Appleton-Century-Crofts.

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

Bishop, Y. M. M., Fienberg, S. E, and Holland, P. W. (1975), Discrete Multivariate Analysis, Cambridge, Mass.: HIT Press.

Clancy, Kevin J., and Ostlund, Lyman E. (1976), "Commercial Effective Measures," Journal of Advertising Research, 16, 29-34.

Federal Trade Commission (1981), "Staff Report on the Cigarette Advertising Investigation: Public Version."

Fienberg, S. E. (1977), The Analysis of Cross-Classified Categorical Data, Cambridge, Mass.: MIT Press.

Friedman, A. (1979), "Framing Pictures: The Role of Knowledge in Automatic Encoding and Memory for Gist," Journal of Experimental Psychology: General, 108, 316-355

Goodman, G. S. (1980), "Picture Memory: How the Action Schema Affects Retention," Cognitive Psychology, 12, 473-495.

Graesser, Arthur C., and Nakamura, Glenn V. (1982), "The Impact of a Schema on Comprehension and Memory," in The Psychology of Learning and Motivation, Vol. 16, ed. Gordon H. Bower, N.Y.: Academic Press, 60-110.

Hansen, F. (1972), Consumer Choice Behavior: A Cognitive Theory, New York: Free Press.

Houston, Michael J. and Rothschild, Michael L. (1980), "Policy-Related Experiments on Information Provision: A Normative Model and Explication," Journal of Marketing Research, 17 432-449.

Jacoby, Jacob (1974), "Consumer Reaction to Information Displays: Packaging and Advertising," in Advertising and the Public Interest, ed. S.F. Divita, Chicago: American Marketing Association. 101-118.

Kassarjian, Harold H. (1982), "Consumer Psychology," Annual Review of Psychology, 33, 619-649.

Krugman, Herbert E. (1966), "The Measurement of Advertising Involvement," Public Opinion Quarterly, 30, 583-596.

Langer, E. and Abelson, R. P. (1972), "The Semantics of Asking a Favor: How to Succeed in Getting Help without Really Dying," Journal of Personality and Social Psychology, 24, 26-32.

Langer, E., Blank, A., and Chanowitz, B. (1978), "The Mindlessness of Ostensibly Thoughtful Action: The Role of 'Placebic' Information in Interpersonal Interaction," Journal of Personality and Social Psychology, 36, 635-642.

Nelson, Douglas L. (1981), "Many Are Called but Few Are Chosen: The Influence of Context on the Effects of Category Size," The Psychology of Learning and Motivation, Vol. 15, ed Gordon H. Bower, N.Y.: Academic Press, 129-163.

Olshavsky, R. W. and Granbois, D. M. (1979), "Consumer Decision Making - Fact or Fiction?" Journal of Consumer Research, 6, 93-100.

Sawyer, A. G. and Peter, J. P. (1983), "The Significance of Statistical Significance Tests in Marketing Research," Journal of Marketing Research, 20 122-33.

Sternthal, B., and Craig, C. S. (1982), Consumer Behavior An Information Processing Approach, Englewood Cliffs, N.J.: Prentice Hall.

U.S. Department of Health and Human Services (1980), "Promoting Health/Preventing Disease: Objectives for the Nation."

Wilkie, William L. (1975), "How Consumers use Product Information: An Assessment of Research in Relation to Public Policy Needs," Washington D.C., U.S. Government Printing Office.

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