Format Effects on and In-Ad Disclosure

Edward T. Popper, Bryant College
Keith B. Murray, Northeastern University
ABSTRACT - This paper presents the results of an experimental portfolio test of print advertising that manipulated two components (the size of the type used and the background color) of an in-ad health warning for a smokeless tobacco product. The study found that neither increased warning size nor a contrasting background significantly increased disclosure communication. The study also found that the disclosure failed to communicate the health warning to nearly half of all subjects.
[ to cite ]:
Edward T. Popper and Keith B. Murray (1989) ,"Format Effects on and In-Ad Disclosure", in NA - Advances in Consumer Research Volume 16, eds. Thomas K. Srull, Provo, UT : Association for Consumer Research, Pages: 221-230.

Advances in Consumer Research Volume 16, 1989      Pages 221-230


Edward T. Popper, Bryant College

Keith B. Murray, Northeastern University

[The authors wish to thank Dr. Greg Connelly of the Commonwealth of Massachusetts' Department of Public Health without whose support and advice this project could not have been conducted.]


This paper presents the results of an experimental portfolio test of print advertising that manipulated two components (the size of the type used and the background color) of an in-ad health warning for a smokeless tobacco product. The study found that neither increased warning size nor a contrasting background significantly increased disclosure communication. The study also found that the disclosure failed to communicate the health warning to nearly half of all subjects.

In late 1986 the Federal Trade Commission (FTC) was considering the format that would be required for the newly mandated health warnings to appear in advertisements for smokeless tobacco products. Although the FTC staff had decided to recommend a circle and arrow shape (which had been proven effective in other countries, Daube, 1989), the specifics of how the warning would appear within that shape had not been determined. This study was conducted to provide input regarding two format elements: the size of the lettering in the warning and its background color.


The Comprehensive Smokeless Tobacco Health Education of 1986 required that the FTC specify the manner in which three different health warnings would be presented in the advertising for smokeless tobacco products (dry snuff, semi-moist snuff and chewing tobacco). 1 he goal of the disclosure was to clearly and conspicuously disclose and, hopefully, communicate the health risks associated with smokeless tobacco.

Previous tobacco warning disclosures (cigarettes) required by the FTC had met with less than complete success. A 1981 FTC staff report on cigarette warnings concluded that the warning then appearing on cigarette labels was "not effective" (Myers, et al. 1981). A principal reason for this conclusion was the finding that a significant percentage of smokers were unaware of the basic message of the warning (that smoking was hazardous to health) even though the warning had been appearing in advertising and on packages for over ten years since. In an attempt to improve warning communications the FTC staff recommended adoption of a series of rotational health warnings that would appear in a specially shaped label (a circle with an arrow pointed into it). These recommendations were based, in part, on the success of similarly shaped, rotational warnings that had been a component of an effective Swedish health education program (Daube, 1982).

Although the new warning shape was not part of the revised cigarette rotational warnings adopted by the US Congress and the FTC in 1985, that shape (circle and arrow) was selected by the FTC staff as the basic format for the in-advertising warnings that Congress mandated for smokeless tobacco. Beyond the basic shape of the warning there were other format elements that were likely to increase exposure.

Studies of consumer product safety warnings (Viscusi & Magat, 1986; Karnes & Leonard, 1986) indicated that increasing the type size of the warning, would increase communications. Further, Bettman, Payne and Staelin (1985) suggest that increased size and contrasting color would both add to the communications effectiveness of health warning disclosures.

This leads to the two hypotheses of the study:

H1. A larger warning (both type size and overall warning size) will increase warning communication in print advertisements for a smokeless tobacco product.

H2. A contrasting background for the warning (as compared to the background of the advertisement) will increase warning communication in print advertisements for a smokeless tobacco product.


Subjects were drawn from a cross section of undergraduates attending a major metropolitan university. While student samples, generally, are inappropriate for tests of advertising communication, in this instance here are two reasons that they are, in fact, a preferable sample.

First, students and young adults (both male and female) are principal target markets for smokeless tobacco products (Warder & Popper, 1985). Thus, a student sample is drawn from the population of the product's potential users. Second, the use of a student sample provides a conservative test of advertising disclosure communication. Students can be expected to be more literate and more analytic in their reading than the public at large. Thus, if a print disclosure can be communicated at all, it will be communicated to a student sample. On the other hand, a disclosure that is ineffective in communicating to students (when they are, in fact, the targets of the advertising) is almost certain to be ineffective in communicating to a non-student audience.

A total of 270 subjects (52% male, 48% female; median age of 21) were randomly assigned to one of the following experimental groups (which varied with regard to the format of the in-ad health warning):

Control Group Smokeless Ad Without Warning

Test Group 1 Small black type on white background.

Test Group 2 Small white type on grey background.

Test Group 3 Large black type on white background.

Test Group 4 Large white type on grey background.

Subjects in the groups each received a packet of three advertisements. These three advertisements included two distractor ads (for student relevant products) and one test ad. The order of the ads was rotated between subjects. All ads were high quality, 8" X 10", color photocopies. The test ads were all modifications of a single advertisement for Skoal Bandits (semi-most snuff packets) that was taken from the current Sports Illustrated. The warning was positioned in the advertisement according to FTC placement guidelines and appeared against a (generally) grey background portion of the ad. The only difference between the test ads was the size and color of the health warning. In the small type version of the warning, the type size used was the same size (10 point) as the tape specified for use in FTC mandated cigarette warnings for a similarly sized print ad. The large version s type size was 40% larger (14 point). In the contrasting background version, the background was white and the type was black. In the non-contrast version the background was grey (similar but not identical to the background of the ad) and the type appeared in white. The control group received the same smokeless tobacco advertisement without any warning.

After exposure, the ads were removed from the subject's view. Subjects were than asked an unaided recall question about the messages in the ad for the smokeless tobacco product. Next they were asked an aided recall question on the presence of a warning message. Next the subjects were asked recognition question on both the warning message the warning symbol. The message recognition question included the following options: the warning that had appeared in the ad, two other health messages regarding smokeless tobacco, a paraphrase (to remove the reference to cigarettes) of the 1970-1985 Surgeon General's warning that had appeared in cigarette advertising, and the response that the ad did not contain a warning message. The choices for the symbol recognition question included the circle and arrow symbol (that had appeared in the ad), the rectangular shape used for warnings in cigarette advertising, and an octagon (stop sign shape), warning the response that the warning was not contained in a symbol or border. Finally the subjects were asked a series of classification questions that included general demographics, cigarette smoking history and smokeless tobacco usage.

The dependant measures of the study were unaided recall (whether the subjects recalled a warning as part of the open ended question about the ad's principal messages); aided recall (whether the subjects responded yes to the question "was there a warning message in the advertisement"); message recognition (recognition of the ad's warning message) and symbol recognition (recognition of the circle and arrow warning symbol). Proven unaided and aided recall measures were created by adjusting the unaided end aided measures removing those who claimed recall but were unable to correctly recognize either the warning message or symbol.


The recall of the warnings was at a moderately low level (see Table 1) with about one fifth (20.7%) of the subjects exposed to a version of the warning recalling it unaided (compared to a 4% unaided recall level for the control group). This unaided recall level changed only minimally when verified in the proven recall measure. Aided recall was substantially higher with almost two thirds (62.3%) of the subjects who were exposed to some form of the warning recalling the presence of a warning when asked (compared to just over a quarter of the subjects in the control group who said they saw a warning). The high level of aided recall in the control group appears to be yea saying, in that it dropped to a negligible level when verified in the proven aided measure. Similar yea saying responses may have occurred within the test groups as well since proven recall declined to a average level of just over one half (53.2%).

Chi square statistics for each of the different measures (based on a table composed of recall versus non-recall) were statistically significant (at a p <.35 level) for all measures when the control group was included in the analysis. However, when the control group was excluded and just the warning test groups were compared, all chi square statistics dropped to a non-significant level.

Both message and symbol recognition levels (see Table 2) showed that about half (47.3% for message recognition and 52.7% for symbol recognition) of the subjects exposed to warningS were able to recognize that warning. Only a negligible portion of the subjects in the control group "recognized" the warning (fewer, in fact, than would be expected by chance). Like the recall analysis, chi square statistics for recognition measures were statistically significant (p<.001) when the control group was included and dropped to non significance when the control group was excluded.

Looking at the effects of the experimental factors, type size showed very little effect on either unaided recall (see Figure 1) aided recall (see Figure 2) or recognition (see Figure 3). One way analyses of variance on these four dependent measures for the typesize factor all produced statistically significant F statistics (p <.01). However, a variety of multiple comparison tests (c.g., Duncan's multiple range test; Tukey's HSD) all indicated that the significant variation occurred between the control group and the treatment groups and there were no statistically significant differences (p <.05) between the two different type size groups.

















A similar examination of the effect of background on the four dependent measures shows that the contrasting (white) background appears to have some effect on unaided recall (see Figure 4) and a less pronounced effect on aided recall (see Figure 5) and on the two recognition measures (Figure 6). One way analyses of variance again showed statistically significant F statistics (p <.05). However, multiple comparison tests showed the significant differences to be between the control groups and the two different background treatments. No statistically significant (p <.05) differences were found between the two different background treatments.

Two way analyses of variance were conducted on each of the dependent measures for the two manipulated factors (type size and background) and no statistically significant (p <.05) F statistics were found for the factors or their interaction. There were no statistically significant (p. <.05) correlations (Pearson's r) between any of the dependent measures and the subjects use (either current or historical) of either cigarettes or smokeless tobacco. Thus, not surprisingly, using these variables as covariates in the two way ANOVA had no substantive effect on the outcome of those analyses.


These data do not support the hypotheses. Neither varying the size of the type used in the warnings nor varying the background of the warnings increased the disclosure's communication. This somewhat unexpected finding can, perhaps, be explained by the low level of warning communication. The low level of communications might suggest that those subjects who did recall or recognize the disclosure were somehow different than those who did not retrieve the communication. The analysis demonstrated that the differences were not in product use. Perhaps the difference was in the subject's sensitivity to warning information. If that were so, than this disclosure was only effective for those subjects with a predisposition for disclosure.

As discussed above, a limitation of this study's sample was that it was likely to be more prone to disclosure communication than the population at large. In this way, the measures of communication were likely to be conservative in that if there was any likelihood of the disclosure being communicated under natural conditions, it would certainly be communicated here. Thus, the communications results shown in Tables 1 and 2 reflect relatively poor disclosure communication. If, with a communication prone population, less than quarter of the subjects car. prove unaided recall and only half can recognize the warning (seen only moments before), effective disclosure has nol occurred.

A second factor that could have been expected to increase the level of communications (perhaps unnaturally) was the novelty of this warning. Health warnings had never appeared in advertising for smokeless tobacco products. Further, the circle and arrow format was new to tobacco warnings (and, indeed, to all advertised warnings) in the United States. These two factors should, together, have resulted in communication levels that were dramatically higher than would occur with a familiar warning in a familiar product environment (eg., cigarettes).

Assuming that both these factors (more communications conscious sample and novel warning) did, in fact, elevate communication scores suggests that in a natural environment with a less novel warning little or no disclosure would occur.

These low levels of the communications measures are relatively constant (showing low standard deviations) across treatment levels. Even for those communications measures that would be expected to have been most sensitive to the manipulations no significant differences within the factors were apparent.

The message recognition measure would be -most likely to have been impacted by the type size manipulation. If the larger type size were to have any impact it should appear in the measure that specifically tapped the subjects' ability to recognize the words that appeared in the larger type face. There was, however, no significant difference between message recognition scores for the two type size conditions. Indeed, the percentage of subjects who correctly recognized the message was slightly lower for the large type size (52.1%) than for the small type size (53.3%).

The symbol recognition measure would, similarly, be most likely to be impacted by the background color manipulation. This would merely be reflecting that the communication of a visual image would be most likely affected by differences of the visual content and environment of that image. While the contrasting background resulted in higher correct recognition of the symbol (50.8%) than the noncontrasting background (43.3%), there was no significant difference in the symbol recognition measure. This non-significant difference is, none the less, in the hypothesized direction.


These findings may suggest that perhaps the manipulations were not sufficiently large. While, from a research perspective, this can be readily remedied, the pragmatic implications are more substantial. The type-size manipulation used in this experiment increased the size of the type and, therefore, the overall warning, by 40%. It would be easy, in an experimental setting, to increase warning size continually, until an effect is observed. It is unlikely, that an advertiser would be quite so willing to allow a warning (particularly a mandated health warning) to consume his expensive media space.

This study would suggest that if the FTC were to mandate that the disclosure for smokeless tobacco was to use either the sizes or the backgrounds studied here, the disclosure would not be effective in warning the target market of the product's health risks. In fact, the FTC did, November, 1986, mandate that the warnings would appear in a circle and arrow format using the smaller size tested here. However, the required warning would have to appear in a contrasting color (such as the white background, black type tested here).

Future research can track the communications effects of those warnings directly in the effected population. Other research can determine what size is necessary before an effect is noted and whether the effect of contrasting background will increase as the size of the warning increases. Additional factors to be studied would include whether communications could be enhanced by increasing the size of the warning space without increasing the size of the typeface (in this study warning space increased directly with type size)-and whether the effects would be different for a less novel warning (both in terms of novelty of format as well as the novel presence of any warning).

Totally aside from the effects of format. This study suggests that in-ad disclosures are not particularly effective in communicating health warnings. If that finding is confirmed by other researchers it would suggest the need for serious reconsideration of the use of in-ad disclosures. If they are used by companies as a protection from liability (i.e., by including the in-ad disclosure the company has dispatched its obligation to inform consumers of its product's risks), a disclosure that doesn't communicate would provide little protection. If they are used by public policy makers to inform the public, a disclosure that doesn't communicate can provide little information.


Bettman, J.R., J.W. Payne and R. Staelin, "Cognitive Considerations in Designing Effective Labels for Presenting Risk Information", Journal of Public Policy and Marketing, 5:1-28, 1986.

Daube, M., "In Support of Health Warnings." World Smoking And Health, (7)3:2,42-46,1982.

Karnes,E.D. & D.L.Leonard, "Consumer Product Warnings: Reception and Understanding of Warning Information by Final Users", in Karwowhis, W. Ed., Trends in Ergonomics/Human Factors 111, Amsterdam:Elsevier Science Publishers, 1986.

Mazis, M. B. & W.K. Viscusi, Review of the Research Literature on the Effects of Health Warning Labels; Rockville, MD:US Department of Health & Human Services, 1987.

Myers, M. L., C. Iscoe, C. Jennings, W. Lennox, E. Minstry & A. Sacks, Staff Report on the Cigarette Advertising Investigation. Washington, DC: Federal Trade Commission, 1981.

Viscusi, W.K. & W.A. Magat, Analysis of Economic Benefits of Improved Information: Project Period 2 Report, Durham, NC: Center for the study of Business Regulation, Duke University, 19&6.

Warder, N. & E.Popper, Staff Report on the Smokeless Tobacco Investigation, Washington,DC: Federal Trade Commission, 1985.