Immediate Effects of the Arizona Alcohol Warning Poster

Andrea M. Fenaughty and David P. MacKinnon

Exposure to, awareness of, beliefs about, and memory for the warning on the recentlyrequired Arizona alcohol warning poster were measured in a sample of college studentsbefore fN = 362) and after fN = 332) the legislation became effective on January I, 1992.The introduction of the warning poster was associated with greater exposure, awareness,and memory. The authors found inconsistent evidence for an association between theintroduction of the warning poster and beliefs. Secular trends did not appear to beresponsible for these changes.

F etal alcohol syndrome (FAS), a pattern of malforma-tions found in children whose mothers drank alco-hol excessively during pregnancy, includes mental re-

tardation, reduced birth weight and length, and microceph-aly [Jones and Smith 1976; Jones et al. 1973; Rosett, Ouell-ette, and Weiner 1976]. The term "fetal alcohol effect"(FAE) denotes less pronounced symptomatology whenonly a few of the criteria for FAS are met [U.S. Departmentof Health and Human Services 1990].

Available estimates suggest a rate of 2.6 FAS births per1000 live births for black and American Indian mothers oflow socioeconomic status and .6 FAS births per 1000 livebirths for white middle class mothers [U.S. Department ofHealth and Human Services 1990]. The prevalence of FAEis estimated to be even higher. The cost of caring for FASbabies is estimated to be nearly a third of a billion dollars an-nually in the United States alone [U.S. Department ofHealth and Human Services 1990]. The estimated cost of car-ing for all of the FAS children bom in Arizona during 1989may be as high as $160 million over their lifetimes [Roth1991].

Prevention of Pregnant Women's AlcoholConsumptionFetal alcohol syndrome now outranks Down's syndromeand spina bifida as the leading cause of mental retardationin the United States [Abel and Sokol 1986]. Because FASis the only one of these causes that is preventable, programsto reduce its occurrence have received high priority [War-ren and Bast 1988]. Treatment programs have been devel-oped to reduce alcohol consumption among pregnantwomen [Halmeski 1988; Rosett, Weiner, and Edelin 1983].Other programs have been designed to increase knowledgeand awareness among health providers so that they can iden-

ANDREA M . FENAUGHTY is a doctoral candidate and DAVID P.MACKINNON is Assistant Professor Department of Psychology, Ar-izona State University. The research was supported in part by aPublic Health Service grant (AA8547) to Dr. MacKinnon (Princi-pal Investigator) from the National Institute on Alcohol Abuse andAlcoholism. Parts of this article were presented at the 1992 West-em Psychological Association Conference, Portland, OR. The au-thors thank the anonymous JPP&M reviewers for several helpfulcomments.

tify women at risk of having FAS and FAE babies [Little etal. 1983; Russell, Kang, and Uhteg 1983].

The objective of a third type of prevention activity is toincrease the general public's awareness of the risks of a preg-nant woman's alcohol consumption [MacKinnon et al.1991; May et al. 1983; Waterson and Murray-Lyon 1990].The rationale of this approach is that (1) FAS awarenessshould occur before pregnancy because alcohol consump-tion is problematic even in the earliest stages of pregnancyand (2) the maximum number of people should be exposedto FAS infonnation so that an informed general public willdisseminate this awareness to the pregnant women whomay actually engage in the alcohol consumption that placesthe fetus at risk. Two major legislative strategies that havebeen enacted to increase public awareness of FAS are warn-ing labels on containers of alcohol and warning posters in es-tablishments that sell alcohol.

Alcohol Warning LabelsWarning labels have been required on all alcohol beveragecontainers since November 18, 1989. The fetal risk associ-ated with pregnant women's alcohol consumption is one ofthe most prominent warnings on the label. The risk mes-sage was written to inform and remind the public about therisks of alcohol use and to reduce uncertainty and mispercep-tions about alcohol use consequences. The content of thewarning label is: "GOVERNMENT WARNING: (1) Ac-cording to the Surgeon General, women should not drink al-coholic beverages during pregnancy because of the risk ofbirth defects. (2) Consumption of alcoholic beverages im-pairs your ability to drive a car or operate machinery, andmay cause health problems" [H.R. 5409, 1988; H.R. 5210,1988].

Though evaluations are not yet complete, there is alreadysome positive evidence about the effects of the alcohol warn-ing label. Hilton [1991] found that 27.5% of respondents toa phone survey were aware of the label after it appeared,whereas 7.8% reported being aware of the label before it ac-tually appeared. Mazis, Morris, and Swasy [1990] also re-ported gains in awareness of the warning label and greaterincreases in awareness among heavier drinkers and youngerpeople. Kaskutas and Greenfield [1991] similarly foundlarger increases in alcohol warning label awareness amongthe heaviest drinkers in their sample. MacKinnon, Pentz,

Vol. 12 (I)Spring 1993, 69-77 Journal of Public Policy & Marketing 69

70 Warning Poster Effects

and Stacy [1993] found substantial changes in exposure to,awareness of, and memory for the warning label, but onlynegligible changes in perceived harmfulness of alcohol use.A multiple-wave study by Scammon, Mayer, and Smith[1991] compared devout Mormon (nondrinkers) with non-Mormon (drinkers) residents of Utah on measures of aware-ness and waming-label-relevant risk perception before andafter the alcohol warning label legislation became effective.Results indicated that warning labels were noticed moreafter they had appeared, but no change was found in the per-ception of risk associated with alcohol consumption.

Alcohol Warning PostersThe second legislative strategy to increase public awarenessof FAS, warning posters, has not received much research at-tention. Beginning in 1983, New York City required post-ers warning about the risks associated with drinking whilepregnant. Georgia, South Dakota, and California mandatedthat, starting in 1986, such warnings appear in all establish-ments in their states where alcoholic beverages are sold.These posters primarily warn of the fetal injury associatedwith a pregnant woman's alcohol consumption, thoughother risks also are mentioned (e.g., the risk of cancer is men-tioned on the California poster). The Arizona alcohol warn-ing poster legislation, which targets both the general publicand pregnant women in particular, requires a poster describ-ing solely the risk of fetal injury. The Center for Science inthe Public Interest now identifies nine states and 17 citiesthat require such posters [Hayes 1991].

Apparently only one study has been done on the effectsof these posters. In that study of the New York City warn-ing poster, 54% of the persons interviewed before the warn-ing poster appeared mentioned birth defects as a possibleconsequence of a pregnant woman's alcohol consumption.A year later, after the poster had appeared, 68% mentionedbirth defects [Prugh 1989]. The results of this study suggestthat warning posters are capable of increasing awareness ofthe possible consequences of a pregnant woman's alcoholconsumption.

The focus of our research is the Arizona alcohol warningposter legislation. The following warning message was re-quired to appear in all establishments that sell alcoholic bev-erages as of January 1, 1992: "WARNING: DRINKINGDISTILLED SPIRITS, BEER, COOLERS, WINE, ANDOTHER ALCOHOLIC BEVERAGES DURING PREG-NANCY CAN CAUSE BIRTH DEFECTS" [Arizona Re-vised Statutes § 4-261, 1991]. Failure to comply with thestate law after March 1992 may result in suspension and rev-ocation of the license to sell alcohol or in refusal to renewa license.' Our study was designed to assess the effect ofthis legislation on college students' exposure to and aware-ness of the poster, beliefs about the risk of drinking alcoholwhile pregnant, and memory for the warning.

Study Method

SubjectsSubjects were 694 Arizona State University undergraduatestudents who participated in the research in partial fulfill-

Table 1. Sample Description Before and After WarningPoster Legislation, Including Demographic andOther Covariates

VariableFemale gender

Age in years (mean)

White

Drank at least onealcoholic drink last month

Worked past year in anestablishment that sellsalcohol

Lived last year in a statethat had an alcohol warningposter

Saw the alcoholwarning label"

Knew the labelexists"

Knew the law forthe label exists^

Before (%)49.4

20.581.4

74.0

11.3

11.6

57.6

58.2

57.9

After (%)52.0

20.3

16.1

IIA

9.9

12.1

63.0

63.4

62.3

"The exposure to the alcohol warning label measure described in the text is an indexof these three items (coefficient alpha=.88).

ment of a psychology course requirement. Three hundredsixty-two subjects completed a questionnaire in the fall of1991, before the Arizona warning poster was required to ap-pear; 332 completed the questionnaire in the spring of1992, after the warning poster was required to appear. Col-lege student samples have been used in previous studies ofthe alcohol warning label [Andrews, Netemeyer, and Dur-vasula 1990]. Issues of external validity are discussed inmore detail in the Discussion section.

As shown in Table 1, the fall sample was 49.4% female,had a mean age of 20.5 years, and was 81.4% white. Thespring sample was 52.0% female, had a mean age of 20.3years, and was 16.1% white. The proportion of subjectswho had consumed at least one alcoholic drink in the lastmonth was 74.0% and 77.4% for the fall and spring sam-ples, respectively.

DesignBecause natural exposure to the warning poster was notunder experimental control, randomization of subjects to dif-ferent levels of exposure to the poster was not possible. Insuch a nonexperimental design, comparisons were neededto address alternative explanations for observed changes[Cook and Campbell 1979]. We evaluated the effect of thealcohol warning poster by examining pre- and post-legisla-tion measures that should be affected by an alcohol warningposter, adjusting for possible confounding variables, includ-ing measures that are similar to the alcohol dependent meas-ures but should not be affected by the poster.

Because different subjects were measured before andafter the legislation, an alternative explanation of changesin the dependent measures is that the sample taken after the

Journal of Public Policy & Marketing 71

legislation is different from the sample taken before the leg-islation. Measures of gender, age, race, alcohol use, expo-sure to the poster through employment at an establishmentthat sells alcohol, and having lived in a state (in the pastyear) that has an alcohol warning poster were used toequate the samples taken before and after the legislation. Be-cause increases in awareness of the risk of drinking whilepregnant may be due to alcohol container label legislation,exposure to the alcohol warning label was also included asa covariate.

One additional set of covariates, cigarette measures thatcorrespond to the alcohol dependent measures, were also in-cluded. For example, we asked respondents whether theyhad seen a poster warning of the harmful effects of mater-nal cigarette smoking on the fetus (a poster that does notexist in Arizona) and whether they were aware of the risk ofsmoking cigarettes while pregnant. The paired cigarette andalcohol measures were positively correlated. For example,people who believed that cigarettes are harmful during preg-nancy were likely to believe that alcohol is harmful duringpregnancy. If there is a general trend in beliefs about the ef-fects of mother's drug use on the fetus and awareness ofposters, a variable that codes the appearance of the alcoholwarning poster (hereinafter called the law variable) will notaccount for significant variance in the alcohol dependentmeasure when we control for the corresponding cigarettemeasure. If the alcohol warning poster is responsible for in-creases in exposure, awareness, beliefs, and memory, thelaw variable should account for significant variance in the al-cohol measure above and beyond the variance accountedfor by the corresponding cigarette measure.

MeasuresAs shown in the Appendix, the following constructs weremeasured for both alcohol and cigarettes: exposure to thewarning poster, awareness of the warning poster and thelaw requiring it, beliefs about the risk to the fetus of mater-nal alcohol consumption (and cigarette smoking), and be-liefs about the benefits of avoiding alcohol (and cigarettes)if pregnant. Free recall and recognition memory for the mes-sage on the alcohol warning poster also were measured.The alcohol measures were used as dependent variables, thecorresponding cigarette measures were used as covariates.The scale used for the exposure, awareness, and the first be-liefs measures was: 1 (no), 2 (no, I don't think so), 3 (yes,probably), 4 (yes, definitely). The Likert-type scale used tomeasure beliefs about the benefits of avoiding alcohol (andcigarettes) ranged from 1 (not at all beneficial) to 6 (ex-tremely beneficial). Both free recall and recognition mem-ory were scored as either correct (1) or incorrect (0). Half ofthe subjects randomly received the free-recall memory test,for which a response describing the FAS/birth defects riskof alcohol use during pregnancy (and no mention of anyother risk) was coded 1 and any other response (or lack ofa response) was coded 0. The other half of the subjects ran-domly received a forced choice recognition test, for whichthe correct response was coded 1, and the incorrect distrac-tor response and the "I don't know" response were coded0. The distractor warning on the recognition test was identi-cal to the alcohol warning label. For all measures, higher

numbers indicate positive effects of the warning poster (i.e.,more exposure and awareness, more accurate beliefs, bettermemory).

The primary independent variable was whether the stu-dent completed the questionnaire before (0) or after (1) thewarning poster law went into effect (law variable). Addi-tional predictor variables were male (1) or female (0), age(in years), white (1) or nonwhite (0), alcohol use in the lastmonth (no use to heavy use), employment in an establish-ment where alcohol is sold (1) or not (0), having lived thepast year in a state that had an alcohol warning poster (1) ornot (0), and previous exposure to the alcohol warning label.The warning label exposure variable is an index created bytaking the mean of three standardized items: (1) havingseen the alcohol warning label, (2) being aware of the alco-hol warning label, and (3) being aware of the law requiringthe alcohol warning label. These three warning label itemswere measured on the 1 to 4 scale described previously,with higher numbers indicating greater warning label expo-sure in each case. Coefficient alpha for this index is .88, in-dicating acceptable reliability.

ProcedureSubjects arrived at a designated room in groups of 10 to 20and were seated at separate desks. Each subject received aquestionnaire and was instructed to complete it without talk-ing to other subjects. Once finished (approximately 15 min-utes), each subject turned in his or her completed question-naire and was thanked for participating in the study.

Data AnalysisAnalysis of covariance (ANCOVA) was used to estimatethe relationship between the introduction of the alcoholwarning poster (i.e., law) and changes in the alcohol meas-ures, with adjustment for the following covariates: gender,age, race, alcohol use, the exposure variables^, and the cor-responding cigarette measure. For example, in the analysisof each alcohol dependent variable (e.g., awareness of the al-cohol warning poster), the matching cigarette measure (e.g.,awareness of the cigarette warning poster) was included inthe ANCOVA. We predicted that the law variable would ac-count for significant variance in each of the dependent meas-ures, even when all covariates were included.

The analyses were repeated with difference scores as thedependent variable. The difference scores were created bysubtracting each cigarette covariate from its correspondingalcohol dependent variable. Ordinal logistic regression anal-yses were also conducted to establish that the same conclu-sions were obtained under the assumptions of ordinal and in-terval-level data. The ordinal logistic model assumes that dif-ferences between response options are ordinal rather than in-terval as assumed in ordinary regression models [Muthen1987; Muthen and Kaplan 1985]. These analyses led to con-clusions identical to those of the ANCOVA analyses.

An additional analysis was conducted to test for the pres-ence of a law by alcohol use interaction effect on the aware-ness dependent measure. Previous research has demon-strated an interaction such that after the alcohol warninglabel appeared, heavy users of alcohol had a greater in-crease in awareness of the label than did light users of alco-

72 Warning Poster Effects

hoi [Kaskutas and Greenfield 1991; MacKinnon, Pentz, andStacy 1993; Mazis, Morris, and Swasy 1990]. The inter-action term was created from the product of centered alco-hol use and waming poster introduction variables. Wamingposter introduction, alcohol use, and the interaction termwere then included in the ANCOVA.

Test for Parallel SlopesOne of the assumptions underlying analysis of covarianceis homogeneity of regression slopes. This assumptionmeans that for ANCOVA results to be interpretable, the re-lationship between the covariate(s) and the dependent vari-able should be the same at all values of the independent var-iable. Hence, we should be able to demonstrate the same re-lationship between each covariate and the dependent varia-ble with the wave 1 pre-waming-poster sample as with thewave 2 post-waming-poster sample.

The appropriate model for testing and adjusting for heter-ogeneity of regression slopes is one that contains the inter-action terms of each covariate by the independent variable[Huitema 1980]. These models are created by first center-ing each variable in the model, then including both themain effect and interaction terms in the model. A signifi-cant interaction term indicates heterogeneity of regressionslopes. We tested for homogeneity of regression slopes forthe three most consistently significant covariates in ourstudy—each corresponding cigarette measure, work expo-sure, and waming label exposure.

Results

Descriptive AnalysisTable 2 gives the means (unadjusted and adjusted for all co-variates-*) for the dependent variables before and after the Ar-izona alcohol warning poster was required to appear. Expo-sure, awareness, beliefs, and memory for the alcohol wam-ing poster increased after the waming poster law was en-acted.

Analysis of CovarianceTable 3 reports the standardized betas for the law variableand each covariate across all six dependent variables. Wediscuss these results according to each of the dependent var-iables: exposure, awareness, beliefs, and memory. All re-ported p-values are two-tailed.

Poster ExposureThe law variable accounts for variance in subjects' reportsof having seen the poster (P = .15, p < .001). Exposure tothe cigarette poster also accounts for significant variance (P= .35, p < .001). Gender is marginally significant (P = -.07,p < .10); women reported more exposure. Age accounts forsignificant variance in this dependent measure (p = . 13, p <.001). Exposure to the poster at work (P = .15, /7 < .001)and exposure to the alcohol warning label (P = .09, p < .05)account for significant variance in exposure to the alcoholwaming poster.

Table 2. Unadjusted (and Adjusted") Means for Exposure,Awareness, Beliefs, and Memory Before andAfter the Warning Poster Law

[item

ExposureHave you seen the alcoholwaming poster?

Poster AwarenessIs there an alcohol waming poster?

Law AwarenessIs there a law requiringalcohol waming posters?

BeUefs 1Can drinking alcohol duringpregnancy cause birth defects?

Beliefs 2How beneficial is it to avoiddrinking if pregnant?

MemoryAccurate recall for wamingAccurate recognition of the waming

1 I Before \

2.02(1.81)

2.27(1.83)

2.17(1.67)

3.75 (3.21)

5.85(1.71)

.17 (.28)

.18 (.22)

! After

2.43(2.12)

2.59 (2.08)

2.38(1.87)

3.84 (3.30)

5.85(1.72)

.30 (.43)

.30 (.33)

' Means adjusted for gender, age, race, aleohol use, Ihe exposure variables (work,waming label, stale with a poster), and the cigarette measure are in parentheses.

Poster AwarenessThe introduction of the alcohol waming poster accounts forsignificant variance in awareness of the Arizona alcoholwaming poster (P = .14, p < .001), as does awareness of acigarette warning poster (P = .35, p < .001). Of the demo-graphic variables, gender (P = -.07, p < .10) and age (P =.08, p < .05) account for some of the poster awareness vari-ance; women and older subjects were more aware of the al-cohol warning poster. Other significant covariates are expo-sure to the waming poster at work (P = .I2,p< .001) and ex-posure to the alcohol waming label (P = .13, p < .001).

Law AwarenessThe law variable similarly accounts for significant variancein awareness of the law requiring the Arizona alcohol wam-ing poster (P = .12, p < .001). Awareness of a law requiringan Arizona cigarette waming poster accounts for significantvariance in this dependent variable as well (P = .42, p <.001). Age accounts for marginal variance in law awareness(P = .06, p < .10). Work exposure to the warning poster (P= .16,/7 < .001) and exposure to the waming label (P = .12,p < .001) also accounts for significant variance in wamingposter law awareness.

BeUefs 1The introduction of the poster accounts for significant vari-ance in the belief that matemal consumption of alcohol canlead to birth defects (P = .11, p < .01), as does the belief

Journal of Public Policy & Marketing 73

Table 3. Standardized ANCOVA Estimates for Law and Significant Covariates

DependentVariables

Posterexposure

Posterawareness

Lawawareness

Birthdefectsbeliefs

Benefitsbeliefs

Recall''

Recognition''

Law

.15***

.14***

.12***

.11**

.01

.25**

.17*

Cig«

.35***

.35***

.42***

.17***

74***

—

—

' Cigarette poster exposure, awareness, and beliefs.•> Logistic regression estimates.+ Two-tailed p < . 10.• Two-tailed p < .05.

** Two-tailed p< .01 .•** Two-tailedp<.001.

Gender

-.07+

-.07-t-

-.03

-.15***

-.02-.02

-.07

Independent Variables

Age

.13***

.08*

.06+

-.05

-.04

.19**-.13

Race

.05

.04

.04

.03

.02

.06

.05

AlcoholUse

.05

.05

.00

.00

-.00

.09-.00

WorkExposure

.15***

.12***

16***

.01

-.04

.25***

.15*

LabelExposure

.09*

.13***

.12***

.07+

.02

.05

.15+

PosterState

.03

.00

.04

.05

-.00

.01

.09

that maternal cigarette smoking is harmful to the fetus ((3 =.11, p < .001). Significantly more women believed that ma-temal alcohol consumption is associated with birth defects(P = -.15, p < .001). Subjects exposed to the alcohol wam-ing label were marginally more likely to believe that alco-hol consumption can lead to fetal injury (p = .07, p < .10).

Beliefs 2The only variable that accounts for significant variance inthe belief that it is beneficial to avoid alcohol during preg-nancy is the corresponding cigarette measure (P = .74, p <.001).

Warning Poster MemoryThe law variable accounts for significant variance in accu-racy of both free-recall (P = .25, p < .01) and recognition (P= .17, p < .05) memory for the waming on the Arizona al-cohol warning poster. No comparable cigarette memoryquestion was available because there is no Arizona cigarettewaming poster. Age significantly predicts recall (P = .19, p< .01), but not recognition memory. Exposure to the wam-ing poster at work is a significant covariate for both recalland recognition memory (P = .25, p < .001), and (P = .15,p < .05), respectively. Exposure to the alcohol warninglabel is marginally significant for recognition memory (P =.15, p < .10), but not for recall.

Homogeneity of Regression SlopesThe test for homogeneity of regression slopes was carriedout for two sets of models: (1) a set of equations containingjust the independent variable (LAW), the cigarette covariate

(CIG), and their interaction term (LAWXCIG) for each de-pendent variable and (2) a set of equations containing theseterms (LAW, CIG, and LAWXCIG) as well as the main ef-fect and interaction terms of the two most consistently sig-nificant covariates—work exposure (WORKALC and theLAWXWRK interaction term) and warning label exposure(LABEL and the LAWXLBL interaction term).

Model Containing Cigarette Covariate OnlyThe introduction of the alcohol warning poster is signifi-cantly related to reports of exposure to the waming poster(P = .14, p < .001). Reports of having seen a cigarette wam-ing poster are also related to alcohol warning poster expo-sure (P = .36, p < .001). The interaction term is not signifi-cant (P = .01, n.s.). Similarly, both waming poster introduc-tion and the cigarette covariates are related significantly toalcohol waming poster awareness (Ps = .15, p <.OO1, and P= .37, p < .001, respectively), waming poster law awareness(Ps = .11, p <.O1, and P = .42, p < .001, respectively), andthe beliefs item asking about the relationship between sub-stance use during pregnancy and birth defects (P = . 11, p <.01, and P = .19, p < .001, respectively). Thus, none ofthese models indicate heterogeneity of regression slopes.

The second beliefs item, however, does reveal a signifi-cant interaction term (P = .09, p < .001). As in previous anal-yses, the regression coefficient for the cigarette covariate ishighly significant (P = .74, p < .001) and the coefficient forthe introduction of the alcohol warning poster is nonsignifi-cant. No interaction term was tested in the models for therecognition and recall dependent measures because therewas no cigarette covariate to include in these models.

74 Warning Poster Effects

Model Containing Two Additional CovariatesThe same pattem of results reported in the preceding AN-COVA section is found for the exposure, poster awareness,and law awareness variables: the appearance of the wamingposter accounts for significant variance in each dependentvariable (P = .15,p < .01, P = .18,p < .01, and P = .12,p <.05, respectively), as do each cigarette covariate (P = .35, p<. 001, P = .35, p < .001, and P = .41, p < .001, respec-tively), the work exposure variable (P = .14, p < .001, P =.12, p < .001, and P = .15, p < .001, respectively), and thewaming label exposure variable (P = .10, p < .01, P = .14,p < .001, and P = .13, p < .001, respectively). None of theinteraction terms are significant.

The model including the beliefs about birth defects de-pendent variable also has no significant interaction terms,however, their inclusion in the model reduces the law regres-sion coefficient and this term is statistically nonsignificant(P = .04, n.s.). The cigarette covariate is the only term toreach significance in this model (P = .19, p < .001). Themodel containing the beneficial-to-avoid beliefs dependentvariable also yields a significant cigarette covariate (p =.75, p < .001). In contrast to the results of previous modelsusing this dependent variable, the appearance of the warn-ing poster is marginally related to beliefs about whether itis beneficial to avoid drinking during pregnancy (P = .07, p< .10). Though the two remaining covariates do not accountfor significant variance in this beliefs item, the law by workexposure interaction term is significant, indicating heteroge-neity of regression slopes. Work exposure and beliefs weremore highly related after the poster appeared, which may ac-count for the now significant law coefficient.

For both recall and recognition memory items, the intro-duction of the alcohol waming poster remains significantlyrelated to more accurate memory (P = .19, p = .06, and P =.23, p <.O5, respectively). In the recall model, waming labelexposure does not account for significant variance in mem-ory, but exposure to the poster at work does (p = .24, p <.001). A significant law by waming label exposure interac-tion term indicates heterogeneity of regression slopes forthe warning label exposure covariate (P = -.19, p < .05).Both exposure covariates are related significantly to recog-nition memory (work exposure P = .14,p < .05 and label ex-posure P = . 15, p < .05), and neither of the interaction termsis significant.

DiscussionWe used several different outcome measures (e.g., exposureto the poster, beliefs about birth defects) as potential indica-tors of the short-term effectiveness of the Arizona alcoholwaming poster. The pattem of results with these indicatorssuggests that the waming poster had an immediate effect ona college sample's exposure to the poster, a.wareness of theposter and the law requiring it, and memory for the wamingmessage on the poster. We find inconsistent evidence for aneffect on beliefs about the association between matemal al-cohol consumption and birth defects.

ExposureThough establishing a link between the introduction of andhaving seen the waming poster may seem trivial, some re-

searchers believe health warnings are rarely even noticed[Myers 1981, cited by McCarthy et al. 1984]. More sub-jects reported seeing the alcohol waming poster after the leg-islation. More subjects also reported seeing a cigarette wam-ing poster, which is surprising because a cigarette versionof the alcohol warning poster does not exist in Arizona.This finding may reflect subjects' exposure to the cigarettewaming poster in the bordering state of Califomia. Anotherexplanation is that what subjects are reporting as exposureto the Arizona cigarette waming poster is actually exposureto the Surgeon General's cigarette waming, which they seeon billboards and in various other sources throughout thestate. Finally, there may be a general trend to report expo-sure to all health-related wamings and messages describingdrug effects on the fetus. The most parsimonious explana-tion for these results is that the introduction of the alcoholwarning poster led to a change in the exposure dependentmeasure because significant variance in the exposure depend-ent variable is accounted for by the law variable, even whenthe cigarette exposure variable is included as a covariate.

AwarenessAfter the law was enacted, more subjects were aware thatthe alcohol warning poster existed and that it was requiredin Arizona. As with the exposure dependent variable, aware-ness of a fictitious cigarette waming poster also increasedfrom wave 1 to wave 2. Despite the variance accounted forby awareness of a fictitious cigarette waming poster, the in-troduction of the alcohol waming poster is a statistically sig-nificant predictor of awareness.

Other researchers have found a significant interaction be-tween introduction of a waming label and alcohol use suchthat heavier drinkers have the highest awareness of thelabel after it appears [Kaskutas and Greenfield 1991; MacK-innon, Pentz, and Stacy 1993; Mazis, Morris, and Swasy1991]. The interaction of alcohol use and alcohol warningmessage introduction is not statistically significant in ourstudy. The effects of a warning poster may differ fromthose of a warning label. The waming poster may have amore subtle effect because it is not, for example, seen di-rectly on the alcohol container during drinking behavior. Per-haps both heavy and light drinkers look at the warningposter briefly upon entering a drinking establishment. If so,the interaction may be present only when heavy drinkershave much greater exposure to the waming poster than lightdrinkers.

BeliefsWe find inconsistent evidence for an association betweenthe introduction of the alcohol warning poster and beliefsabout the relationship between matemal consumption of al-cohol and birth defects. The results from the main effect AN-COVA model suggest an effect of the poster on more accu-rate beliefs, but the model in which interaction terms are in-cluded to test for parallel regression slopes indicates thatthe effect on beliefs is nonsignificant. Interestingly, none ofthe interaction terms in that model are significant, indicat-ing homogeneity of regression slopes. This finding sug-gests that the initial ANCOVA model was an appropriatemodel to use.

Journal of Public Policy & Marketing 75

The lack of association between the introduction of theposter and beliefs about the benefits of avoiding alcohol dur-ing pregnancy may be due to a ceiling effect, as the wave 1mean response to this item was very high.

MemorySubjects were better at both recalling the content and recog-nizing the exact wording of the alcohol waming poster afterit appeared.

SummaryThe overall pattem of results suggests immediate effects ofthe waming poster. The introduction of the Arizona alcoholwarning poster is associated with increased exposure to theposter, increased awareness of the poster and the law requir-ing it, and more accurate memory for the waming messageon the poster. Some evidence suggests the introduction ofthe waming poster is associated with more accurate beliefsabout maternal alcohol consumption and birth defects,though analyses with different underlying assumptionsyield inconsistent results.

The obtained pattem of results is very similar to findingsfrom alcohol waming label evaluations. Alcohol waming la-bels, like the alcohol waming poster, are being noticed bythe public [MacKinnon, Pentz, and Stacy 1993; Mazis, Mor-ris, and Swasy 1991; Scammon, Mayer, and Smith 1991].However, as in the case of the alcohol waming poster, therelationship between waming message introduction and per-ception of risk [Scammon, Mayer, and Smith 1991] and be-liefs about birth defects [MacKinnon, Pentz, and Stacy1993; Mazis, Morris, and Swasy 1991] has not been clearlydemonstrated.

Our study has several limitations. First, the effects of thewaming poster legislation may be over or underestimated be-cause of the short followup period. The effects of the postermay increase as more people see it and are exposed repeat-edly to it. Alternatively, the warning poster may have onlyimmediate effects that dissipate. Furthermore, the media cov-erage of the alcohol waming poster legislation, rather thanthe poster itself, may be responsible for some portion of theimmediate effects we report. To address this possibility, weanalyzed responses to the following questionnaire item:' 'Where have you seen or heard about the risks of drinkingalcohol while pregnant?" Response choices included "news-papers," "magazines," "television," "radio," and "bill-boards," in addition to "a poster." Significantly more sub-jects chose the "poster" response after the alcohol warningposter legislation became effective. We find no significantchange in any of the other responses. As this question wasasked near the beginning of the questionnaire, the wordingof other items is not likely to have influenced responding.

A second issue, hinted at in the Exposure subsection, isthe high level of reported awareness of and exposure to anonexistent cigarette waming poster. Though interpretationof results would have been clearer without this finding, itdoes not invalidate the conclusion that the alcohol wamingposter was associated with changes in awareness of and ex-posure to the Arizona alcohol waming poster. There are plau-sible explanations for subjects' reports of seeing a cigaretteposter (e.g., seeing the poster in Califomia), and for the re-

lationships between exposure to and awareness of the alco-hol and cigarette waming posters (e.g., health conscious peo-ple might report both). However, the introduction of the al-cohol warning poster accounts for significant variance inthe exposure and awareness dependent measures when ad-justed for other variables, including the cigarette covariates.

A third issue is that some subjects in wave 1, or the ba-seline measurement point, are likely to have had previous ex-posure to a similar alcohol waming poster or similar wam-ing message. For example, subjects may have been exposedto a poster in Phoenix or Califomia (or another city or statethat already had the poster), or they may have been exposedto the message on the alcohol warning label. Exposure ineach of these conditions was measured and included as a co-variate in the ANCOVA. If exposure to these other mes-sages had been less, perhaps stronger effects would havebeen found.

Fourth, because of the questionnaire format, subjectsread items that may have influenced how they responded tosubsequent items. For example, after responding to the ques-tion, "Have you seen posters in Arizona that warn of theharm of a pregnant woman's alcohol consumption to her un-bom baby?,'' subjects were given either a recognition or afree recall memory test for the contents of the poster.Though this ordering may have affected subjects' re-sponses, this problem is probably minimal for two reasons.First, half of the subjects were given a recognition memorytest, which had the correct waming and a very similar dis-tractor waming. Previous questions could not have aided re-sponses to the recognition memory test unless they re-flected the exact wording of the warning poster, which theydid not. Second, both recognition and free-recall memoryitems became more accurate from wave 1 to wave 2. As thequestion ordering was constant across waves, response setis an unlikely explanation for the increase in accuracy ofmemory.

A final potential criticism of the study is that the best sam-ple for this type of study is a sample of pregnant women.Pregnant women are the group whose behavior ultimatelymust change if FAS is to be prevented. However, the Ari-zona alcohol waming poster legislation was created to "ed-ucate(s) the general public, and pregnant women in particu-lar" of the risks associated with matemal alcohol consump-tion [Arizona State Senate Fact Sheet for S.B. 1132 1991,1]. Thus, the legislation targets both the general public andpregnant women. Our analysis of the immediate, overall ef-fects of the alcohol waming poster is important because it ad-dresses a primary'' objective of the Arizona alcohol wamingposter legislation.

ImplicationsThe health, social, and economic costs associated with ma-temal alcohol consumption make it a topic of immediate na-tional concem [U.S. Department of Health and Human Ser-vices 1990]. Alcohol warning posters, if effective, are onerelatively low cost strategy for informing the public aboutthe risk. Many researchers believe warnings in general areineffective—many asserting that this ineffectiveness ismore detrimental than the absence of a warning [Driver1987; Engs 1989]. One rationale for this assertion is that

76 Warning Poster Effects

the public is so inundated by wamings that truly importantwamings may be missed [Driver 1987]. Others believe thatan ineffective waming merely creates a protective loopholethrough which businesses can avoid legal suits [Scammon,Mayer, and Smith 1991]. This concem may be warranted asthe distinction between a waming and a disclaimer may beunclear [Weinstein et al. 1978].

The arguments for warnings are also compelling. Wam-ings provide information that may be used to avert injury ata low cost. Furthermore, preliminary analysis reveals thatthe purely informational format such as that of the currentArizona alcohol warning poster evokes an almost univer-sally positive response. Other researchers have found thatwamings directing behavior (e.g., "You should....") in ad-dition to or instead of providing information may be metwith reactance [see Brehm 1966], thus having the oppositeof the intended effect on behavior [Bensley and Wu 1991].

Whether wamings are a good or bad idea, evaluations ofwamings are needed to determine what, if any, effects theyhave and how closely these effects are related to the initialgoals of the legislation. Our fmdings suggest the Arizona al-cohol waming poster is having immediate and positive ef-fects on exposure, awareness, and memory. Future researchtargeting pregnant women, and perhaps covering a longerfollowup period, is needed to address whether the posterwill reduce the number of FAS and FAE babies bom.

Notes1. Though a similar law went into effect for just Phoenix several

months earlier (July 1991), our research is on the effects of thestate law. Subject reports of residence are included to adjust forpersons who lived in Phoenix during the summer of 1991.

2. The exposure variables are (1) working in an establishment thatserves alcohol, (2) having lived in a state that has the alcoholwaming poster, and (3) exposure to the alcohol waming label.

3. The means are adjusted for gender, age, race, alcohol use, em-ployment in an establishment that serves alcohol, residence thepast year in a state that has an alcohol waming poster, and ex-posure to the alcohol waming label.

4. Though increasing the general public's awareness is mentionedfirst, it does not indicate priority, but rather the numerical orderin which the target populations are mentioned in the legisla-tion.

ReferencesAbel, Emest L. and Robert J. Sokol (1986), "Fetal Alcohol Syn-

drome Is Now Leading Cause of Mental Retardation," iMncet(November 22), 1222.

Andrews, J. Craig, Richard G. Netemeyer, and Srinivas Durvasula(1990), "Believability and Attitudes Toward Alcohol WamingLabel Information: The Role of Persuasive CommunicationsTheory," Journal of Public Policy & Marketing, 9, 1-15.

Arizona Revised Statutes § 4-261 (1991).

Bensley, Lillian Southwick and Rui Wu (1991), "The Role of Psy-chological Reactance in Drinking Following Alcohol Preven-tion Messages," Journal of Applied Social Psychology, 21,1111-1124.

Brehm, J.W. (1966), A Theory of Psychological Reactance. NewYork: Academic Press, Inc.

Cook, Thomas D. and Donald T. Campbell (1979), Quasi-Experi-mentation Design & Analysis Issues for Field Settings. Chi-cago: Rand McNally College Publishing Company.

Driver, Russell W. (1987), "A Communication Model for Deter-mining the Appropriateness of On-Product Wamings," IEEETransactions on Professional Communication, PC-30 (3), 157-163.

Engs, R.C. (1989), "Do Waming Labels on Alcoholic BeveragesDeter Alcohol Abuse?" Journal of School Health, 59 (3), 116-18.

Halmeski, Erja (1988), "Alcohol Counselling of 85 Pregnant Prob-lem Drinkers: Effect on Drinking and Fetal Outcome," BritishJournal of Obstetrics and Gynecology, 95, 243-247.

Hayes, Janice W. (1991), personal communication from SubstanceAbuse Program Coordinator, Office of the City Manager, Cityof Phoenix, AZ.

Hilton, Michael (1991), "Preliminary Findings of Research on Al-cohol Waming Label," report presented to Bureau of Alcohol,Tobacco and Firearms.

H.R. 5409 (1988), lOOth Congress 2D Session.

H.R. 5210 or S. 2852 Omnibus Anti-Drug Abuse Act (1988),100th Congress 2D Session.

Huitema, Bradley E. (1980), The Analysis of Covariance and Al-ternatives. New York: John Wiley & Sons.

Jones, Kenneth L. and David W. Smith (1976), "The Fetal Alco-hol Syndrome: Clinical Delineation," Annals of the New YorkAcademy of Science, 273, 130-137.

, , Christy N. Ulleland, and Ann Pytkowitcz Streissguth(1973), "Pattern Malformation in Offspring of Chronic Alco-holic Mothers," Lancet, 1, 1267-1271.

Kaskutas, Lee and Tom Greenfield (1991), "Knowledge of Wam-ing Labels on Alcoholic Beverage Containers," Proceedings ofthe Human Factors Society 35th Annual Meeting, Vol. 1. SantaMonica, CA: Human Factors Society, 441-445.

Little, Ruth E., Ann Pytkowicz Streissguth, Gay M. Guzinski, Har-rison L. Gratwohl, Judy M. Blumhagen, and Catherine E. Mcln-tryre (1983), "Change in Obstetrician Advice Following a Two-Year Community Educational Program on Alcohol Use and Preg-nancy," American Journal of Obstetrics & Gynecology, 146(1), 23-28.

MacKinnon, David P., Mary Ann Pentz, and Alan Stacy (1993),"The Alcohol Warning Label and Adolescents: The FirstYear," American Journal of Public Health, 83 (4), 585-587.

, Katharine Taft, Laurie Bizzell, Liva Nohre, Alan Stacy, andMary Ann Pentz (1991), "Adolescent Beliefs About DrinkingWhile Pregnant," paper presented at the American PublicHealth Association 119th Annual Meeting, Atlanta, GA (Novem-ber).

May, Philip A., Karen J. Hymbaugh, Jon M. Aase, and JonathanM. Samet (1983), "Epidemiology of Fetal Alcohol SyndromeAmong American Indians of the Southwest," Social Biology,30 (4), 374-387.

Mazis, Micheal B., Louis A. Morris, and John L. Swasy (1991),"An Evaluation of the Alcohol Waming Label: Initial SurveyResults," Journal of Public Policy & Marketing, 10(1), 229-241.

Journal of Public Policy & Marketing 77

McCarthy, Roger L., James P. Finnegan, Susan Krumm-Scott, andGail E. McCarthy (1984), "Product Information Presentation,User Behavior, and Society," Proceedings of the Human Fac-tors Society 28th Annual Meeting, Vol. 1. Santa Monica, CA:Human Factors Society, 81-85.

Muthen, Bengt O. (1987), LISCOMP. Analysis of Linear Struc-tural Equations With a Comprehensive Measurement Model.User's Guide. Mooresville, IN: Scientific Software, Inc.

and David Kaplan (1985),' 'A Comparison of Some Method-ologies for the Factor Analysis of Non-Normal Likert Varia-bles," British Journal of Mathematical and Statistical Psychol-ogy,7,^, 171-189.

Prugh, Thomas (1989), Point-of-Purchase Health Warning No-tices," Alcohol Health and Research World, 36.

Rosett, Henry L., Eileen M. Ouellette, and Lyn Weiner (1976), "APilot Prospective Study of the Fetal Alcohol Syndrome at theBoston City Hospital, Part 1: Matemal Drinking," Annals of theNew York Academy of Science, 213, 118-122.

, Lyn Weiner, and Kenneth C. Edelin (1983), "Treatment Ex-perience With Pregnant Problem Drinkers," Journal of Ameri-can Medical Association, 249 (15), 2029-2033.

Roth, Robert A. (1991), "Report to the Legislature," report pre-pared as public service document for St. Joseph's Hospital andMedical Center, Mesa, AZ.

Russell, Marcia, Gay E. Kang, and Lois Uhteg (1983), "Evalua-tion of an Educational Program on the Fetal Alcohol Syndromefor Health Professionals," Journal of Alcohol and Drug Educa-tion, 29 (1), 48-61.

Scammon, Debra L., Robert N. Mayer, and Ken R. Smith (1991),"Alcohol Wamings: How Do You Know When You Have HadOne Too Many?" Journal of Public Policy & Marketing, 10(1), 214-228.

U.S. Department of Health and Human Services (1990), SeventhSpecial Report to the U.S. Congress on Alcohol and Health.Rockville, MD: National Institute on Alcohol Abuse and Alco-holism.

Warren, Kenneth R. and Richard J. Bast (1988), "Alcohol-RelatedBirth Defects: An Update," Public Health Reports, 103 (6),638-642.

Waterson, E. J. and Iain M. Murray-Lyon (1990), "Preventing Al-cohol Related Birth Damage: A Review," Social Science & Med-icine, 30 (3), 349-364.

Weinstein, Alvin S., Aaron D. Twerski, Henry R. Piehler, and Wil-liam A. Donaher (1978), Products Liability and the ReasonablySafe Product: A Guide for Management, Design, and Market-ing. New York: John Wiley & Sons, Inc.

Appendix Questionnaire Items

Demographic Information—Gender—Age—Ethnic/racial category

Covariates and Dependent Variables—Where do you work now?—How many alcohol drinks have you had in the last month?—Where have you seen or heard about the risks of drinking

alcohol while pregnant?—Are there posters in Arizona restaurants, bars and liquor stores

that wam of the harm of a pregnant woman's cigarette smokingon her unbom baby?

—Have you seen posters in Arizona restaurants, bars and liquorstores that wam of the harm of a pregnant woman's cigarettesmoking on her unbom baby?

—To the best of your knowledge is there an Arizona lawrequiring posters in Arizona restaurants, bars and liquor storesthat wam of the harm of a pregnant woman's cigarette smokingon her unbom baby?

—Are there posters in Arizona restaurants, bars and liquor storesthat wam of the harm of a pregnant woman's alcoholconsumption on her unbom baby?

—Have you seen posters in Arizona restaurants, bars and liquorstores that wam of the harm of a pregnant woman's alcoholconsumption on her unbom baby?

— T̂o the best of your knowledge is there an Arizona lawrequiring posters in Arizona restaurants, bars and liquor storesthat wam of the harm of a pregnant woman's alcoholconsumption on her unbom baby?

—To the best of your knowledge, is there a waming on alcoholbeverage containers?

—Have you seen waming labels on alcohol beverage cans orbottles?

—To the best of your knowledge, is there a law requiring waminglabels on cans and bottles of beer, wine, and liquor?

—Please write the contents of the Arizona Alcohol Wamingposter. If you do not believe there are waming posters inArizona restaurants, bars, and liquor stores, check here Q anddo not complete this question.

—Which of the following wamings is written on the ArizonaAlcohol Waming Poster?1. Drinking alcoholic beverages during pregnancy can cause

birth defects2. Drinking distilled spirits, beer, coolers, wine, and other

alcoholic beverages during pregnancy can cause birth defects3. I don't know what waming is written on the Arizona

Waming Poster—How beneficial would it be for a pregnant woman to entirely

avoid drinking alcohol during her pregnancy?—How beneficial would it be for a pregnant woman to entirely

avoid smoking cigarettes during her pregnancy?—Can drinking alcohol during pregnancy cause birth defects?—If a pregnant woman smokes cigarettes, it may cause injury to

the baby?