FIFTEEN YEARS OF FEAR AROUSAL: RESEARCH ON THREAT

Psycholofical Bulletin1969, Vol. 72, No. 6, 426-444

FIFTEEN YEARS OF FEAR AROUSAL:

RESEARCH ON THREAT APPEALS: 1953-1968

KENNETH L. HIGBEE 1

Purdue University

Research on threat appeals has yielded conflicting findings concerning therelative effectiveness of high threat versus low threat in persuasion. Studiesreviewed here have investigated the effects on persuasion of variables in eachof five areas: (a) the nature of the recommendations, (6) personality char-acteristics of the recipients, (c) source credibility, (d) learning of the messagecontent, and (e) the interest value of fear. Few variables have been foundwhich consistently interact with fear. One conceptual consideration and fourmethodological considerations are suggested as possible sources of the incon-sistency in the findings. A postulated curvilinear relationship between fear leveland persuasion may help reconcile some of the conflicting findings on high-versus low-threat appeals.

The use of threat, or fear arousal, is quitecommon in many types of persuasive at-tempts. The procedure generally used is eitherto associate an undesirable practice (e.g.,smoking) with negative consequences (e.g.,lung cancer), or to associate a desirable prac-tice (e.g., brushing teeth) with the avoidanceof negative consequences (e.g., cavities).Then, after making the association, recom-mendations are offered for attitude change orfor action to take to avoid the consequences.Typically the projected consequence of notconforming to the communication's recom-mendations is spelled out in vivid, threateningdetail. The assumption is that the higher thelevel of fear aroused, the greater will be thepersuasiveness of the communication.

Areas of persuasion which use such a"scare" technique might include public-opinion campaigns (e.g., safe-driving, health,and anti-smoking campaigns), propaganda ef-forts (e.g., atrocities committed by the enemyin war), advertising (e.g., mouthwash, tooth-paste, and deodorant advertisements), andpreaching (e.g., hellfire and damnation forsinners).

However, despite the widespread use ofthreat appeals in applied settings, the tech-

1 The author is indebted to Richard Heslin andW. C. Redding for their valuable comments onearlier versions of this manuscript.

Requests for reprints should be addressed to Ken-neth L. Higbee, Department of Psychology, PurdueUniversity, W. Lafayette, Indiana 47907.

426

nique has only recently come under muchrigorous experimental investigation. Table 1shows 27 studies of fear-arousing appeals,published since 1953, which have involvedexperimental manipulation of the level ofthreat in a persuasive communication. Ofthese 27 studies, 17 (63% of the publishedexperimental research in the area in the last15 years) have been published within thelast 4 years. Table 1 also indicates the levelsof threat used in each study, which will bediscussed later.

Reviews of the threat-appeal literature(see Janis, 1967, 1968a; Janis & Leventhal,1968; Leventhal, 1965, 1967; McGuire, 1966,1968; Miller, 1963) have revealed consider-able inconsistency among the findings regard-ing the relative effectiveness of high threatversus low threat in persuasive messages. Thefirst section of this paper indicates some ofthe variables which research since 1953 hasindicated may affect the persuasiveness ofthreat appeals (see Hovland, Janis, & Kelley,1953, for a review of the relevant researchprior to 1953). In the second section aresuggested some possible sources of the in-consistency among the findings on high threatversus low threat. Finally, the present stateof theorizing and knowledge concerning fear-arousing appeals is summarized.

High Threat versus Low Threat

The "common-sense" approach used inmany applied settings of persuasion suggests

FEAR AROUSAL AND THREAT APPEALS 427

that the more you scare someone the morelikely he will be to accept your recommenda-tions for avoiding the portrayed threat.However, Berelson and Steiner (1964) stated,as one of the scientifically substantiated factsof human behavior, that "strong appeals tofear, by arousing too much tension in theaudience, are less effective in persuasion thanminimal appeals [p. 552]." This conclusionmay also be found in many introductorypsychology textbooks (e.g., Hilgard, 1962;Morgan & King, 1966) and in some socialpsychology textbooks (e.g., Kretch, Crutch-field, & Ballachey, 1962; Secord & Backrnan,1964).

The conclusion that low threat is superiorto high threat in persuasion is based mainlyon a study by Janis and Feshbach (1953)in which three groups of high school studentswere presented with three versions of anillustrated lecture on dental hygiene. Eachversion stated the dangers of dental neglectand recommended specific procedures for

tooth care, but the three versions differed intheir vividness and in the degree to whichthey emphasized the possible dire conse-quences of such neglect. The results of thestudy indicated that the minimal-threat ap-peals were more effective than the high-threatappeals in eliciting reported conformity to therecommended procedures for tooth care.Subsequent studies which have been widelycited as supporting this negative relationshipbetween fear arousal and persuasive effec-tiveness (although such an interpretation oftheir results is not completely justified insome cases, cf. Leventhal & Niles, 1965,Footnote 3) include Goldstein (1959), Haef-ner (1956), Janis & Feshbach (1954), andJanis and Terwilliger (1962).

Some studies, however, have found no rela-tionship between fear arousal and persuasion(Frandsen, 1963; Millman, 1968; Moore,1965; Payne, 1963). In some of these studies,the lack of a clear-cut relationship betweenfear level and persuasion may have been due

TABLE 1

THREAT-APPEAL STUDIES EXPERIMENTALLY MANIPULATING THREAT LEVEL

Studies Fear, intensity, anxiety, or threat levels

Janis & Feshbach (1953)Janis & Feshbach (1954)Moltz & Thistlethwaite (1955)Goldstein (1959)Nunnally & Bobren (1959)Berkowitz & Cottingham (1960)Janis & Terwilliger (1962)Leventhal & Perloe (1962)Frandsen (1963)Leventhal & Niles (1964)Gollob & Dittes (1965)Hewgill & Miller (1965)Insko, Arkoff, & Insko (1965)Leventhal, Singer, & Jones (1965)Powell (1965)Stern, Lana, & Pauling (1965)Chu (1966)Dabbs & Leventhal (1966)Leventhal, Jones, & Trembly (1966)Leventhal & Singer (1966)Leventhal & Watts (1966)Miller & Hewgill (1966)Duke (1967)Fischer, Cohen, Schlesinger, & Bloomer (1967)Leventhal, Watts, & Pagano (1967)Leventhal & Trembly (1968)Millman (1968)

Minimal, moderate, strongMild, strongNone, weak, strongMinimal, strongLow, highMinimal, strongMild, strongOptimistic, threateningMinimal, moderateLow, mild, highNo-threat, threatLow, highLow, highLow, highMild, highNeutral, fear-arousingMild, moderate, strongNone, low, highLow, highNone, low, highLow, medium, highMild, strongNeutral, fearLow, moderate, highModerate, highLow, highNeutral, anxiety provoking

428 KENNETH L. HIGBEE

at least in part to methodological considera-tions, such as failure to arouse differentialfear levels in the subjects (see Hewgill &Miller, 1965; Moltz & Thistlethwaite, 19SS;Sastrohamidjojo, 1968).

Some research has yielded mixed findingsconcerning threat level and persuasion. Forexample, high fear has been found to affectattitude change without affecting behaviorchange (Leventhal, Singer, & Jones, 196S;Radelfinger, 1963); and Leventhal and Watts(1966) found that high fear produced morecompliance in adopting the recommendationsto decrease smoking, but less compliance fortaking an X ray.

Thus, some studies have indicated a nega-tive relationship between fear level and per-suasion, others have found no relationship,and others have yielded mixed findings.However, most of the recent studies in thearea have found a positive relationship be-tween fear and persuasion. Studies findingthat strong fear is more effective than weakfear have involved such diverse topics asdental hygiene practices (Haefner, 196S;Leventhal & Singer, 1966; Singer, 1965),smoking (Insko, Arkoff, & Insko, 1965;Leventhal & Niles, 1964; Leventhal & Watts,1966; Leventhal, Watts, & Pagano, 1967;Niles, 1964; Snider, 1962), tetanus inocula-tions (Dabbs & Leventhal, 1966; Kornzweig,1968; Leventhal, Jones, & Trembly, 1966;Leventhal et al., 1965), safe-driving practices(Berkowitz & Cottingham, 1960; Leventhal& Niles, 1965), fallout shelters (Hewgill &Miller, 1965; Miller & Hewgill, 1966; Powell,1965), tuberculosis (DeWolfe & Governdale,1964), roundworms (Chu, 1966), properviewing of the sun during an eclipse (Kraus,El-Assal, & DeFleur, 1966), and even theuse of stairway handrails for safety (Piccolino,1966).

Not only have numerous studies found thata persuasive communication containing strongfear-arousing appeals can be effective, butthere is evidence that even the arousal offear which is irrelevant to the communicationcan facilitate persuasiveness (Lundy, Simon-son, & Landers, 1967; McNulty & Walters,1962; Simonson & Lundy, 1966).

VARIABLES WHICH MAY AFFECT THEPERSUASIVENESS OF THREAT

There is thus some inconsistency amongthe studies on fear-arousing communications,some indicating a negative relationship be-tween threat level and persuasion, some indi-cating no relationship, and most studies indi-cating a positive relationship. Considerableresearch has been aimed at attempting todetermine the variables which may affect thepersuasiveness of threat (i.e., interact withthe fear level). Although the review in thissection is largely objective, at least two pointsare brought out in the studies summarized:First, few of the variables which have beenstudied have been found to interact with fearlevel consistently enough to account for in-consistencies in past studies. Second, some ofthe findings contradict the fear-drive model,which views fear as a force which motivatesthe recipients to reduce the fear.

Recommendations

Katz (1960) has stated that the findingsof Janis and Feshbach (1953) may be dueto a lack of a clear-cut relationship in theminds of the children between the failure tobrush their teeth properly and the gangrenejaws of the aged (which were portrayed asone of the undesirable consequences in thehigh-fear group). He suggested that the useof fear arousal depends for its effectivenessupon the presence of well-defined paths foravoiding the punishment, that is, upon therecommendations which are given for avoid-ing the undesirable consequences portrayedin the communication. The nature of therecommendations has been the subject of sev-eral subsequent studies.

Dabbs and Leventhal (1966) found thatmanipulations of the effectiveness of therecommendations did not have any significanteffect either on intentions to take actionor on action taken (obtaining an inoculationfor tetanus). Similarly, Moltz and Thistle-thwaite (1955) found that although explicitassurance as to the efficacy of the recom-mended procedures was effective in producinganxiety reduction, the anxiety reduction wasnot associated with greater reported conform-ity to the recommendations. Thus, although


one study (Chu, 1966) did find a slight tend-ency (p < .07, n = 1024) for subjects tominimize the threat when the recommendedsolution was not seen as efficacious, researchhas not sufficiently indicated that perceivedefficacy of the recommendations offered in athreat appeal significantly interacts with fearlevel in affecting persuasion.

Leventhal (1965), after his review of somerelevant findings, concluded that when theactions recommended are clearly effective,attitude and behavior change are more likelyto take place than if doubts exist about re-sponse effectiveness. However, this conclusiondescribes efficacy as a main effect; that is,regardless of the fear level, persuasion is morelikely if the recommendations are perceivedas efficacious. Therefore, the efficacy variablewill not aid in reconciling inconsistencies inpast studies.

Another variable which appears to be im-portant as a main effect is the specificity ofthe recommendations. In communications ontetanus, Leventhal et al. (1966) and Leven-thal et al. (196S) varied the fear levels andthe specificity of instructions for obtainingshots. More of these subjects who receivedspecific instructions actually took shots thanthose who did not receive specific instructions,regardless of the fear level of the communica-tion. Similarly, Leventhal and Niles (1964)noted that the ease of carrying out therecommended action facilitated the effective-ness of a fear-arousing message. Thus, specificinstructions are more effective in persuadingsubjects to change their behavior than aregeneral instructions.

The lack of an interaction between fear leveland efficacy or specificity of recommendationsis not consistent with the fear-drive model.If fear becomes increasingly motivating asfear level increases, then specificity and ef-ficacy of recommendations should have moreeffect for a high-fear message than for alow-fear message, since the recipient shouldbe more motivated to take action to reducethe fear and to avoid the threat.

There is some suggestion that althoughefficacy and specificity may not interact withfear level, the mere presentation of recom-mendations (as compared with no recommen-dations on how to avoid the threat) may

interact with fear level. Leventhal et al.(1967) found a tendency (p < .10) towardan interaction between fear and instructionson how to stop smoking; the presentation ofinstructions made more of a difference in thehigh-fear condition than in the moderate-fearcondition.

The positioning of the recommendationsrelative to the threat in the communicationwas suggested by Cohen (1957) as an im-portant variable. Cohen found that a fear-arousing communication in which informa-tion assumed to satisfy the aroused needs isplaced after need arousal brought more ac-ceptance of that information than a situationin which information was placed before theneed arousal.

In Cohen's study, however, fear level wasnot experimentally manipulated. Later studiesdo not indicate that positioning of the recom-mendations affects persuasion. Leventhal andSinger (1966) found that although fear re-actions were reduced by moving the recom-mendations on dental hygiene from before, tointermixed with, to after the fear stimuli,this positioning had no effect on acceptance.Singer (1965) also found that the order ofpresentation of the recommendations did notsignificantly affect attitude change or behav-ior change. It may be noted that, as was trueof the findings on efficacy and specificity (seeprevious discussion), the finding of no inter-action between positioning and fear does notsupport the fear-drive model, which wouldsuggest that the positioning of recommenda-tions after the threat should be more effectivethan before the threat.

Although positioning of the recommenda-tions apparently does not affect persuasion,one study indicates that positioning may af-fect the learning of the content of a threatappeal. Fischer, Cohen, Schlesinger, andBloomer (1967) interspersed threatening ma-terial throughout safe-driving manuals andfound that threatening stories after instruc-tions were superior to stories preceding theinstructions, in the retention of the messagecontent by the subjects.

Thus, research on the recommendations ina fear appeal indicates that the specificityand efficacy of the recommendations mayfacilitate the persuasiveness of the appeal,


regardless of the fear level; that the presenta-tion of recommendations may be more impor-tant in a high-fear appeal than in a low-fearappeal; and that the positioning of the recom-mendations relative to the threat is not animportant factor in persuasion, although itmay affect the retention of the message.

Personality Characteristics

In their review of some of the fear-arousalliterature, Miller and Hewgill (1966) sug-gested that individuals differ in the kinds ofstatements they perceive as strong fear-arousing appeals. Similarly, Leibler (1967)suggested that the type of person who re-sponds to a threatening mailed questionnaireis different from the type of individual whoresponds to a nonthreatening questionnaire.Thus, the characteristics and predispositionsof the threat-appeal recipients may contributeto the relative effectiveness of high threatversus low threat.

Self-esteem of the recipients contributes todifferential effects of fear on persuasion, ac-cording to a number of studies. In their in-vestigation of the effects of threatening versusoptimistic messages on the perception of armylife, Leventhal and Perloe (1962) found thatself-esteem was negatively related to accept-ance of the threat appeals. Subjects high inself-esteem were influenced more by the opti-mistic communications, while subjects low inself-esteem were influenced more by thethreatening communications. These results,however, occurred only among subjects whoreceived communications from sources dis-similar to themselves with respect to person-ality characteristics.

The above study by Leventhal and Perloedid not vary fear level but rather had one pessi-mistic (threat) appeal and one optimistic ap-peal. In addition, the results held only for acertain kind of subjects. In contrast with thefinding by Leventhal and Perloe of a nega-tive relationship between self-esteem and per-suasiveness of threat, subsequent studies haveconsistently found a positive relationship;that is, high-esteem subjects tend to be morepersuaded by high-threat appeals, and low-esteem subjects tend to be more per-suaded by low-threat appeals. Leventhal andTrembly (1968) found that increasing the

intensity of threat strengthened the coping(see following discussion) efforts of middle-and high-esteem subjects but decreased theefforts of low-esteem subjects. Going frommild- to high-threat messages, Dabbs andLeventhal (1966) found increases for high-esteem subjects and decreases for low-esteemsubjects on stated intentions to take tetanusshots; and Kornzweig (1968) recorded asimilar and stronger effect for actual shottaking. Consistent with these findings-,Zemach (1966) observed that low self-esteemsubjects were more persuaded by a low-guilt-arousing appeal than by a high-guilt-arousingappeal.

Thus, the interaction of self-esteem withfear level in persuasion is quite well sup-ported. People with high self-esteem are morepersuaded by a high-threat appeal than arepeople with low self-esteem. One reason forthis may be that high-esteem subjects areless personally threatened by a high-threatappeal and thus can react to such an appealby taking realistic action rather than by at-tempting to avoid thoughts about the threat(see subsequent discussion on these two reac-tions to fear), which may be the reaction ofthe low-esteem subjects.

Reference has already been made to copingbehavior; and the "coping style" of a person(his characteristic means of dealing withtension-producing material) has been foundby Goldstein (1959) to affect how the personwill respond to a high-threat appeal. Gold-stein's results indicate that "copers" receivestrong-fear appeals better than "avoiders,"whereas among avoiders the minimal-fear ap-peals receive greater acceptance. Coping stylehas also been found to be related to psycho-physiological responses to tension arousal(Goldstein, Jones, Clemens, Flagg, & Alex-ander, 1965). Ability to cope with tension-producing material may be a facet of a highself-esteem (cf. Leventhal & Trembly, 1968).The directions of the results concerning self-esteem and coping style are consistent withthis suggestion.

In addition to self-esteem and coping style,the subject's perceived vulnerability to dangerappears to interact with threat level. Niles(1964), in a study on smoking and lungcancer, divided subjects on the basis of their


initial feeling of vulnerability to lung cancer.Only those people expressing low susceptibil-ity were increasingly persuaded to stopsmoking and to take X rays as threat in-creased; more vulnerable people showed nodifference in their expressed desire to stopsmoking or to take X rays as messages be-came more frightening. This indicates thathigh threat is more effective than low threatfor people who do not feel especially vulner-able to the threat, but that threat level makesno difference for those who do feel vulnerable.An even stronger interaction between fear andvulnerability is indicated by the findings ofWatts (1967) and Leventhal and Watts(1966) that higher feelings of vulnerabilityactually tended to reduce the desire to takeaction as the fear level of the messageincreased.

Leventhal et al. (1966) and Leventhal andSinger (1966) did not find significant effectsof perceived vulnerability on persuasivenessof fear arousal. However, the consistencyamong those studies which did find such ef-fects indicates that perceived vulnerabilitymay be an important source of individual dif-ferences in responses to threat appeals. Theindication is that perceived vulnerabilityinteracts negatively with fear level, so thathigh fear is more effective for low-vulnerablesubjects and low fear is more effective forhigh-vulnerable subjects.

Janis and Feshbach (1954) suggest thatanxiety level may be an important variablefor threat appeals. They found that subjectshigh in anxiety were less influenced by astrong fear appeal than were low-anxiety sub-jects. Consistent with this finding is a studyon frustration and language intensity by Car-michal and Cronkhite (1965), which suggeststhat a subject at a high activation level tendsto reject a persuasive speech that wouldfurther increase his activation. This negativerelationship between chronic anxiety level andpersuasiveness of increasing threat is anassumption underlying the defense-avoidancehypothesis of Janis and Feshbach (1953),who stated that high fear may be less effectivethan low fear because it arouses a high levelof anxiety in the subjects and causes themto reject the anxiety-provoking message. Suchan effect would presumably be greatest for

subjects who are high in anxiety before themessage is presented.

However, the effect of chronic anxiety levelreported by Janis and Feshbach (1954) hasnot been replicated in a number of subse-quent studies which examined anxiety (e.g.,Goldstein, 1959; Niles, 1964; Singer, 1965).Thus, one cannot say that an interaction be-tween anxiety and fear level is substantiatedby the research evidence.

Note that, again, the fear-drive model doesnot predict the results which have been ob-tained for some of the personality character-istics, such as the interactions of esteem andvulnerability with fear level and the lack ofan interaction for anxiety and fear level. Oneresponse to such findings has been to abandonthe concept of fear as a drive and to developa model that explains the responses to fearappeals as being mediated by variables otherthan fear (see Leventhal, 1967, discussedsubsequently).

Of the personality characteristics whichhave been suggested as interacting with fearlevel, self-esteem and perceived vulnerabilityappear to be well supported, coping stylehas been suggested in one study, and chronicanxiety level does not appear to be an impor-tant variable. Two other suggested variablesare the recipient's concern with the topic(Singer, 1965) and his need for cognition(Cohen, 1957).

Source Credibility

The results of a study by Hewgill andMiller (1965), on attempts to persuade PTAmembers of the importance of fallout shelters,indicate that strong fear appeals bring aboutgreater attitude change than do mild fearappeals when the source has high credibilitywith the recipients. Powell and Miller (1967)also found a positive relationship betweensource credibility and persuasiveness of in-creasing fear levels.

Incidental observations by the authors oftwo studies lend further support to the im-portance of source credibility in fear-arousingappeals. Insko et al. (1965), in their studyon smoking, noted that their source washighly credible (source was a teacher, sub-jects were junior high students) and that thisfact probably influenced the results of their

432 KENNETH L. H1GBEE

study, which yielded a positive relationshipbetween fear level and attitude change.Rosenblatt (1962) found an interaction be-tween threat and discrepancy (strong threatbecame less effective than weak threat thegreater the amount of change advocated) andsuggested that this was due to the effects ofboth variables on the intervening variable ofcredibility.

If the source is not highly credible thenthe response of the subjects may be merelyto discredit the source further rather thanto accept the recommendations of a high-threat communication. Miller and Hewgill(1966) presented evidence of a lack of atti-tude change by subjects exposed to strong-fear messages which were attributed to low-credibility sources, and they attributed thisresult to the possibility that the subjects mayhave further discredited the source in reject-ing the message. Related to this suggestionis one made by Katz (1960), that when thereis no clear relation between the punishmentand the desired behavior in a threateningmessage, then the effect of high threat maybe merely to create negative attitudes towardthe person (s) associated with the negativesanctions (i.e., the source of the communica-tion). Such a result appears to be morelikely if the source were low in credibilitythan if he were highly credible.

In conclusion, there is considerable evi-dence that source credibility does interactwith fear and thus is an important variablein determining whether high fear or low fearis more effective. High fear may be moreeffective than low fear when the sourcecredibility is high but not when the sourcecredibility is low.

Learning of Message Content

It may be that subjects exposed to differentlevels of threat learn different amounts of thefactual materials presented in the communica-tions and that this differential learning affectstheir attitude change. Duke (1967) sug-gested that this is so. Duke felt that thevariable of communication length may havebeen confounded with level of fear in theJanis and Feshbach (1953) study, in whichthe high-fear messages were longer than thelow-fear messages. In fact, the high-fear

messages are longer in many studies, becausethey often consist of the same material as thatin the low-fear messages plus additional fear-arousing material. For example, such a pro-cedure resulted in the high-fear messagesbeing longer than the low-fear messages (270words as compared with 190 words—42 %longer) in the study by Janis and Milholland(1954) described subsequently. Duke's resultsindicate that there were differential learningeffects as a function of communication length;subjects presented with the shorter communi-cation (low fear) learned more.

In contrast with the finding by Duke(1967) of differential learning due to dif-ferential message lengths, several studies havenot found any differences between the dif-ferent levels of threat in the subjects' com-prehension of the factual content of thecommunications (Fischer et al., 1967; Gold-stein, 1959; Gollob & Dittes, 1965; Janis& Feshbach, 1953; Millman, 1968; Singer,1965).

Although there is considerable evidencethat threat level does not influence how muchis learned, one study suggests that threatlevel may influence what is learned from thecontent of a persuasive communication.Janis and Milholland (1954) compared theverbatim recall of high-fear versus low-fearinformation and, although they too found nodifferences in the recall of the major contenttheme or the recommended conclusion, theydid find a selective-recall tendency for cer-tain specific information. That is, althoughthere was apparently no marked loss in over-all learning efficiency as a result of strongthreat appeals, there may have been a selec-tive learning of certain types of information.However, this result may have been con-founded with the different lengths of theirmessages. It is not too surprising that sub-jects would tend to recall different materialfrom two messages, one of which containsalmost 50% more material than the other.

The best conclusion that can be drawnfrom the research relevant to learning of mes-sage content is that level of threat is notsignificantly related to learning of factualcontent. Thus, comprehension of the messageis probably not a variable mediating differing


persuasiveness of different levels of feararousal.

Interest

One situation in which strong fear maybe effective in persuading may arise whenone wishes to arouse interest before per-suading. This suggestion was made by Janisand Feshbach (19S3), even though theyfound low fear more effective in their study.Some subsequent studies have supported thissuggestion. Berkowitz and Cottingham (1960)presented high- and low-threat messages onthe use of safety belts and found that strongfear was indeed more persuasive than weakfear when the communication was initiallylow in interest value and that the dramaticvalue of the strong-fear communicationmade it more interesting than the weak-fearmessage.

Robbins (1962b) also found that fear levelappeared to be positively related to attention;subjects seemed to show more interest inlistening to a recording concerning smokingand cancer (developed by Robbins, 1962a)as the recording increased in anxiety-arousinglevel.

As with most variables related to fear-arousing communications, the research on therole of interest is not entirely consistent.Nunnally and Bobren (19S9) found thathigh-anxiety messages actually depressedpublic interest concerning communications onmental health; and Cannell and MacDonald(1956) suggested that threat-avoidance maybe an explanation for their findings thatsmokers read fewer articles about cigarettesand cancer than do nonsmokers.

There is thus no clear relationship betweenfear level and interest value. However, it maybe noted that the above studies suggestingthat high fear may arouse interest were in alaboratory setting, whereas those indicatingthat high fear may depress interest were in afield setting. It may be that with a captiveaudience high fear arouses the interest ofthose audience members who have no choicebut to receive the message, whereas in naturalsettings people have the choice of avoidingthe communication altogether. High fear maybe more effective in getting the attention ofboth groups of people but may depress interest

in exposing oneself to such a message if onehas the choice.

POSSIBLE SOURCES op INCONSISTENCYAMONG STUDIES

The studies reviewed thus far have beenconcerned with the nature of the recommen-dations, personality characteristics of the re-cipients, source credibility, learning of factualcontent, and interest value of threat. In eachof these areas, as well as in the general con-clusion regarding the persuasiveness of highversus low threat, there has been conflictingevidence. There may be some considerationsthat have not been explicitly studied whichmay have contributed to the inconsistency offindings across studies. One conceptual con-sideration and four methodological considera-tions are suggested here.

Fear oj What?

An important assumption underlying theresearch on fear-arousing appeals is that thestudies have been investigating the same phe-nomenon because they have all been con-cerned with "fear arousal" and "threat."However, it may be that the only thing allof these studies really have in common istheir concern with the arousal of differentkinds of negative affective responses todifferent topics.

Thus, the question may be asked of the"fear arousal" literature—fear of what?2

This question consists of two parts, which areconsidered separately. The first is, what isthis thing called "fear"? The second is, whatis it that the subjects fear?

Concerning the first question, the natureof the "fear" that is aroused by threat ap-peals, Leventhal (196S) and Janis and Leven-thal (1968) have suggested that the reasonfear motivates some people, but not others,to act is that the messages may arouse dif-ferent kinds of fear. Thus, the arousal of"neurotic anxiety" by fear-arousing stimulimay cause subjects to attempt to reduce fearby eliminating thoughts about danger (viarepression, denial, aggression, etc.), whereasthose subjects in whom "realistic fear" is

2 This consideration was suggested to the author byW. C. Redding (personal communication, September1968).


aroused may take realistic action (e.g., adoptthe communication's recommendations) toeliminate or avoid the danger.

Further support for the suggestion thatthe "fear" in all fear-arousal studies may notbe the same thing is given in a study byLeventhal and Trembly (1968). These au-thors felt that stress films on automotiveaccidents, which they showed to high schoolstudents, created at least two distinct statesof fear. "Anticipation fear" was a conse-quence of descriptions of threat agents, theirapproach, and their method of attack, andwas characterized by muscular tension andattention to the environment and to methodsof avoiding danger. "Inhibitory fear" seemedto follow descriptions of destruction and wasaccompanied by inner tension (nausea) andattention to actual or potential damage tothe self. Their study suggested that a focuson anticipation fear would facilitate copingactivity, whereas the focus on inhibitory fearwould lead to depression, anger, and a senseof loss.

Consistent with the above suggestions byLeventhal (1965), Janis and Leventhal(1968), and Leventhal and Trembly (1968)concerning two kinds of fear, an analysis ofthe published studies on fear appeals sug-gests that these studies may have involvedthe manipulation of at least two differentkinds of fear. The first type, corresponding toneurotic anxiety or inhibitory fear, may bea nauseated, sick feeling aroused by grue-some, vivid descriptions and pictures of suchscenes as automobile accidents (e.g., Berko-witz & Cottingham, 1960), infected gangrenemouths (e.g., Janis & Feshbach, 1953), andsurgery on cancerous lungs (e.g., Leventhal& Niles, 1964). The second type of fear,perhaps corresponding to realistic fear oranticipation fear, appears to be somewhatmore concerned with the likelihood of thesubjects' experiencing the threat than withthe gruesome seriousness of the threat (e.g.,Chu, 1966; Gollob & Dittes, 1965; Hewgill &Miller, 1965). That is, the threat to the sub-jects is that of a high probability of experi-encing the portrayed consequences if theydo not accept the communication's recom-mendations.

Not only may the experimental manipula-tions of fear have produced different kinds offear but the criteria used for measuring theeffectiveness of the fear-arousal manipulationsalso indicate that "fear" may mean differentthings in different studies. Thus, severalstudies have measured fear as the sum of aseries of reported reactions such as fear,anxiety, nausea, depression, panic, anger,tension, disgust, nervousness, and discomfort(see Leventhal et al., 1966; Leventhal &Singer, 1966; Leventhal et al., 1965; Leven-thal & Trembly, 1968; Leventhal & Watts,1966; Leventhal et al., 1967). Some studieshave considered fear primarily as the sub-jects' reports of anxiety (see Goldstein, 1959;Miller & Hewgill, 1966; Moltz & Thistle-thwaite, 1955; Powell & Miller, 1967). Otherstudies have defined fear arousal primarilyin terms of the subjects' reported worry orconcern (see Chu, 1966; Hewgill & Miller,1965; Janis & Feshbach, 1953, 1954; Powell,1965). Finally, in some studies fear has beenmeasured mainly in terms of reported fear(see Dabbs & Leventhal, 1966; Leventhal &Niles, 1965).

The second part of the question—fear ofwhat?—concerns just what it is that thesubjects fear. The consideration here is morewith the object of the fear, whereas thepreceding discussion was more concerned withthe feeling of fear itself. There has been con-siderable variation in the topics of the fear-arousing communications (see Table 2). Itseems unlikely to the present author thatsuch fears as losing one's teeth, contractingtuberculosis or lung cancer, being killed inan automobile accident, or being killed by anatomic holocaust are perceived as equivalentfears with regard to such factors as serious-ness, imminence, or realism. However, thesefears do have at least one thing in common;they all involve some kind of physical dangerto the subjects.

Some fear-arousal studies have used threatsinvolving undesirable consequences other thanphysical danger. Threats have included socialdisapproval (Powell & Miller, 1967), un-pleasant army life (Leventhal & Perloe,1962), undesirable grading system (Cohen,1957), undesirable effects of overpopulation(Frandsen, 1963; Millman, 1968), mental


illness (Millman, 1968; Nunnally & Bobren,1959), and harm to other people valued bythe subject (e.g., family—Hewgill & Miller,1965; Powell, 1965). It seems that thesetypes of threats may not be equivalent toeach other nor to the physical threatsenumerated above. If it is true that the"threat" in various fear-appeal studies is notequivalent, then it is not surprising that thestudies have yielded different findings on theeffectiveness of high threat versus low threat.

Further support for the suggestion that thetopics studied may not yield equivalent threatis indicated by some studies which haveshown that fear of physical dangers and fearof other types of danger may not be equiva-lent. For example, Miller and Hewgill (1966)found that, for a sample of college students,fears of losing loved ones by accident orseparation scored higher in "meaningfulness"than fears of death or of annihilation of thehuman race in a nuclear war. Similarly, afactor analysis of a fear survey schedule byRubin, Katkin, Weiss, and Efran (1968)indicated that fear of death and illness consti-tuted a factor distinct from fear of inter-personal events. Finally, Hodges (1968) foundthat threats to self-esteem (ego threats) andthreats of physical pain led to differentdegrees of anxiety arousal.

Thus, it may be that the concept of"threat" or "fear arousal" has been viewedin so many different ways that the studieson fear-arousing communications have notactually been studying the same phenomenon.Differences in the nature of the fear and inthe objects of the fear may have contributedto the inconsistency of the findings on thearousal of fear in persuasion.

Topics

The variability in topics used for threat-appeal studies can be seen in Table 2. Eventhough threat-appeal studies have used atleast 16 different topics, this diversity hasseldom been considered as an importantsource of inconsistency in findings. For ex-ample, Duke (1967) designed a study as atest of the Janis and Feshbach (1953) find-ings. However, Duke used a different topic(syphilis rather than dental hygiene), as wellas different subjects (college students ratherthan high school students). It is not sur-prising to this author that Duke's findingsconcerning learning of factual content weredifferent from those of Janis and Feshbach.

One reason that topic areas may contributeto differences in results among studies maybe the differences in the subjects' knowledgeof a topic. It is probable that such diversetopics as smoking, dental hygiene, tetanus,

TABLE 2

TOPICS USED IN RESEARCH ON THREAT APPEALS

Topics Representative Studies

Dental hygiene

Smoking

TetanusSafe drivingFallout sheltersPopulation growthMental healthCancerSafety beltsRoundwormsGradesTuberculosisSyphilisViewing sunArmy lifeDonating blood

Goldstein (1959), Janis & Feshbach (19S3), Janis & Feshback (19S4), Janis &Milholland (1954), Leventhal & Singer (1966), Moltz & Thistlethwaite (1955)

Insko et al. (1965), Janis & Terwilliger['(1962), Leventhal & Niles (1964),Leventhal & Watts (1966), Leventhal et al. (1967), Stern et al. (1965)

Dabbs & Leventhal (1966), Leventhal et al. (1966), Leventhal et al. (1965)Fischer et al. (1967), Leventhal & Niles (1965), Leventhal & Trembly (1968)Hewgill & Miller (1965), Powell (1965)Frandsen (1963), Millman (1968)Millman (1968), Nunnally & Bobren (1959)Gollob & Dittes (1965), Robbins (1962a, 1962b)Berkowitz & Cottingham (1960)Chu (1966)Cohen (1957)DeWolfe & Governdale (1964)Duke (1967)Kraus et al. (1966)Leventhal & Perloe (1962) ""Powell & Miller (1967)


safe driving, fallout shelters, populationgrowth, and mental health are not equallyfamiliar to a given group of people. Hovland,Janis, and Kelley (1953) state that the emo-tional reactions evoked by communicationswhich announce a threatening event will tendto be reduced by prior exposure to informa-tion about the event. Consistent with thissuggestion is the finding of Lewan and Stot-land (1961) that neutral information prior toan emotional appeal (although not specificallya threat appeal) lessened the effectiveness ofthe appeal. Lewan and Stotland suggestedthat the subjects' possession of prior informa-tion may have influenced the findings of Janisand Feshbach (19S3). Also, in their studyof the relationship between knowledge of lungcancer and smoking habits, Spelman and Ley(1966) indicated that heavy smokers dis-torted the information available to them.These authors suggested that this distortionmay be a reason for the lack of effectivenessof some fear-producing propaganda againstcigarette smoking.

As a check on the possibility that the topicsmay vary in familiarity, the author con-structed a brief familiarity survey, in whichsubjects were instructed, for each of five topicareas, "based on your exposure to informa-tion concerning each of these areas, rate yourfamiliarity with each of them by placing acheck in the appropriate space along thescale." Each topic was followed by a 7-pointscale ranging from "not at all familiar" to"extremely familiar." The topics included thethree most commonly used in threat-appealresearch (see Table 2): (1) dental neglectand improper tooth care, (2) cigarettesmoking, (3) tetanus, (4) atomic fallout, and(5) careless automobile driving.

Since there might be a difference betweensubjects (see below) as well as betweentopics, the scale was administered to highschool (ninth grade) males (n — 20, mean[M] age = 14.8 years), college males (n = 20,M age =19.2), college females («= 13,M age = 18.9), and adult females (n - 13, Mage = 40.6). Two two-way analyses of vari-ance were computed—Subjects (2 between)X Topics (5 within). In Analysis A, the highschool and college males were compared, andin Analysis B, the college and adult females

were compared. The results of both analyseswere strikingly parallel in all respects. Inneither analysis was there a significant maineffect for subjects, nor an interaction betweensubjects and topics. However, the main effectfor topics was significant in both analyses(Analysis A, F = 19.89, df = 4/1S2, p <.001; Analysis B, F = 21.23, df = 4/96,P < .001).

Topics thus differed in their perceivedfamiliarity for all subject groups. To deter-mine which pairs of topics were rated assignificantly different, the mean ratings weresubjected to a Newman-Keuls analysis. ForAnalysis A, the order of familiarity (in de-creasing order, the numbers referring to thefive topics numbered above) was S, 2, 1, 3, 4;for Analysis B the order was S, 2, 1, 4, 3.The Newman-Keuls method indicated, forboth analyses, that all differences betweenpairs of topics were significant (p < .01) ex-cept for 1-2, 3-4, and 2-5. In addition, Dif-ference 1-5 was significant in Analysis A, butnot in Analysis B. Thus, safe driving, dentalhygiene, and smoking appear to be topicswhich are perceived as more familiar thantetanus and atomic fallout.

In summary, both analyses indicated thatthere are differences in perceived familiaritywith the topics used in threat-appeal research.Greenwald and Sakumura (1967) found thatdifferences in judged familiarity of statementson the United States in Vietnam were notrelated to learning of the statements, butthat familiarity was positively related toacceptability of the statements. Perhaps ac-ceptability of a threatening persuasive appealvaries with the subjects' familiarity with thetopic of the appeal. Or high fear may havemore interest value (see above) for topicswith which the subjects are relatively un-familiar. Further research on topic familiaritymay be beneficial.

Besides differing in their familiarity, topicsmay differ in their importance. Colburn(1967) found a strong interaction betweenfear level and topics: The high-fear speechon cancer, the medium-fear speech on tuber-culosis, and the low-fear speech on toothdecay were the most successful speeches insecuring acceptance of their recommendations.Colburn attributed this to the differing im-


portance of the topics for his subjects (collegestudents). Thus, high-fear appeals may bemore effective than low-fear appeals for atopic which is important to the subjects (suchas cancer) but may be less effective for atopic which is not too important to thesubjects (such as tooth decay).

Two reasons have been suggested why topicdifferences may be an important source of in-consistencies in findings on fear appeals—familiarity and importance. In addition,topics may vary on such other dimensions asego-involvement and nearness to the subject(in time or space). Such considerations sug-gest that the differences in topics used inthreat appeals may have contributed to thedifferences in findings concerning the effec-tiveness of threat.

Subjects

Of course different studies must use dif-ferent subjects; and it is common for severalstudies on one topic to involve different kindsof subjects. However, the diversity of sub-jects in studies on threat appeals may havebeen great enough to contribute to differentfindings on the effectiveness of these appeals.As can be seen in Table 3, subjects haveranged from elementary school children tocollege students to adults.

Leventhal and Niles (1965) mentionedsubject sampling as a possible source of in-consistencies in findings on fear appeals, butmany investigators do not appear to consider

this consideration as being too important.Thus, for example, Moltz and Thistlethwaite(19SS) attempted to replicate the findings ofJanis and Feshbach (19S3) but failed toarouse differential fear levels in their sub-jects. This failure may have been due to thefact that army recruits (the subjects forMoltz and Thistlethwaite) may react differ-ently to threats of dental neglect than dohigh school freshmen (the subjects for Janisand Feshbach).

Similarly, Leventhal and Singer (1966) at-tempted to replicate the Janis and Feshbach(19S3) study using the same topic (dentalhygiene) and obtained results contradictoryto those of Janis and Feshbach. Perhaps theresults were affected by the fact that thesubjects for the Leventhal and Singer studywere visitors to a state fair, varying over awide range of ages and education levels.Again, it is not improbable that such a sampleof subjects may react differently from highschool freshmen to attempts at fear arousalon dental neglect. Singer (1965) found thateven within high school freshmen there weredifferences between general-course studentsand college-preparatory students in dentalconcern and in responses to fear arousal.

Differences in familiarity with the topicareas was suggested above as a possible reasonwhy differences in topics may be important.This same consideration may be a reason forthe importance of differences in subjects.Subjects of wide age and education-level

TABLE 3

SUBJECTS USED IN RESEARCH ON THREAT APPEALS

Subjects Representative studies

College students

Adults

Population cross-section(junior high to adults)

Junior high or high schoolstudents

Army recruitsStudent nursesTaiwan elementary school

students

Berkowitz & Cottingham (1960), Cohen (1957), Dabbs & Leventhal (1966), Duke(1967), Frandsen (1963), Gollob & Dittes (1965), Leventhal & Niles (1965),Leventhal & Perloe (1962), Leventhal et al. (1965), Leventhal et al. (1967),Millman (1968), Robbins (1962a, 1962b), Stern et al. (1965)

Hewgill & Miller (1965), Janis & Milholland (1954), Janis & Terwilliger (1962),Kraus et al. (1966), Leventhal et al. (1966), Powell (1965), Powell & Miller (1967)

Leventhal & Niles (1964), Leventhal & Singer (1966), Leventhal & Watts (1966),Nunnally & Bobren (1959)

Fischer et al. (1967), Goldstein (1959), Insko et al. (1965), Janis & Feshbach (1953),Janis & Feshbach (1954), Leventhal & Trembly (1968)

Moltz & Thistlethwaite (1955)DeWolfe & Governdale (1964)Chu (1966)

438 KENNETH L. H1GBEE

ranges may differ in their knowledge concern-ing any given topic (although it was foundabove they do not seem to differ in theirperceived familiarity with the topics). Forexample, Cannel and MacDonald (1956)found that education level interacts withsmoking habits in the perception of the rela-tionship between smoking and lung cancer.Among nonsmokers, the higher the educationthe more likely the person was to accept thisrelationship as proved; but among smokers,the higher the education the less likely theywere to accept it.

Besides differences in age and education,there is some evidence that sex differencesbetween subjects may account partially fordifferences in reactions to fear arousal(Fischer et al., 1967; Leventhal et al., 1966).Few studies have considered this aspect ofsubject differences, even though one studyused female subjects only (DeWolfe & Gov-erndale, 1964) and several studies have usedmale subjects only (Duke, 1967; Leventhal& Perloe, 1962; Leventhal & Trembly, 1968;Moltz & Thistlethwaite, 19SS; Powell, 1965).

Media

The various media for presentation of thecommunications in studies on threat appealsare shown in Table 4. The most commonmedium is the printed word, probably be-cause it is easiest to control extraneous vari-ables with a printed communication. Theleast common medium is live presentation,probably the most difficult situation in whichto control extraneous variables.

Frandsen (1963) made a direct attempt to

determine the differential effects of media oftransmitting threat appeals, using taped, tele-vised, and live presentations. He found nodifferences but suggested that further experi-ments are necessary before any generaliza-tions can be made. There is some evidencethat live presentations are more persuasivethan recorded messages, which, in turn, aremore efficient than print (see Klapper, 1960),but this evidence does not specifically pertainto fear-arousing messages.

In one study which did specifically con-cern fear-arousing messages, Leventhal andTrembly (1968) found that different aspects,such as size and color, of the same medium(film) may have unanticipated informationsignificance to the subjects. It seems even morelikely that different aspects of different kindsof media may have important effects on re-sponses to fear appeals. For example, the useof film, or of slides to accompany a verbaldescription, may be more conducive to anausea-type fear than a concern-type fear(see discussion on kinds of fear above),because gruesomeness may be more vividlydepicted via color pictures than via verbaldescriptions. Verbal presentation of threatmaterial may, in turn, be more conducive toconcern-type fear.

Criteria

A final possible source of inconsistency inthe threat-appeal findings lies in the criterionof persuasive effectiveness of a threat appeal.This may be an important practical considera-tion because, for example, in most appliedsettings a communication which changes only

TABLE 4

COMMUNICATION MEDIA USED IN RESEARCH ON THREAT APPEALS

Media Representative studies

Printed

Tape recorded

Tape recorded plus slides

Film

Live, oral

Dabbs & Leventhal (1966), Fischer et al. (1967), Gollob & Dittes (196S), Janis &Milholland (1954), Janis & Terwilliger (1962), Leventhal et al. (1966), Leven-thal & Perloe (1962), Leventhal et al. (1965), Nunnally & Bobren (1959)

Hewgill & Miller (1965), Millman (1968), Powell (1965), Powell & Miller (1967),Robbins (1962a, 1962b)

Berkowitz & Cottingham (1960), Duke (1967), Goldstein (1959), Insko et al.(1965), Janis & Feshbach (1953), Janis & Feshbach (1954), Leventhal & Singer(1966), Moltz & Thistlethwaite (1955)

Leventhal & Niles (1964), Leventhal & Niles (1965), Leventhal & Trembly (1968),Leventhal & Watts (1966), Leventhal et al. (1967)

Chu (1966), Cohen (1957), Stern et al. (1965)

FEAR AROUSAL AND TtfREAT APPEALS 439

TABLE 5

CRITERIA OF PERSUASION USED IN RESEARCH ON THREAT APPEALS

Criteria Representative studies

Attitude or opinion change

Conformity, or reported con-formity, to recommendedactions

Desire or intentions to act

Attitude change, intentions,and action

Berkowitz & Cottingham (I960), Cohen (1957), DeWolfe & Governdale (1964),Frandsen (1963), Gollob & Dittes (1965), Hewgill & Miller (1965), Janis &Terwilliger (1962), Leventhal & Perloe (1962), Millman (1968), Powell (1965),Powell & Miller (1967), Stern et al. (1965)

Goldstein (1959), Janis & Feshbach (1953), Janis & Feshbach (1954), Kraus et al.(1966), Moltz & Thistlethwaite (1955)

Chu (1966), Insko et al. (1965), Leventhal & Niles (1965), Leventhal & Singer(1966), Leventhal & Trembly (1968)

Dabbs & Leventhal (1966), Leventhal et al. (1966), Leventhal & Niles (1964),Leventhal et al. (1965), Leventhal & Watts (1966), Leventhal et al. (1967)

an attitude or an intention to act wouldprobably not be equatable with one whichalso causes the recipients to take action onthe recommendations presented.

Table 5 shows that in assessing the effectsof fear arousal, studies have used severalcriteria: attitude or opinion change; con-formity to the recommended actions; desireor intentions to act; and attitude change,intentions, and action. Whether high or lowfear is more effective may be a function ofwhich measure of effectiveness was used. Forexample, Leventhal et al. (1966) found thathigh fear produced more favorable intentionstoward taking action but was unrelated toaction taken. Similarly, Dabbs and Leventhal(1966) obtained an interaction effect betweenthreat and self-esteem, but the results heldonly for attitudes, not for action.

Evans, Rozelle, Lasater, Dembroski, andAllen (1968) have developed what they feelis an adequate measure of behavioral con-formity to certain recommendations on dentalhygiene. This may make possible the use ofactual conformity to the recommendations,rather than reported conformity to the recom-mendations, as a measure of effectiveness forthreat appeals concerning dental hygiene.

A CURVILINEAR RELATIONSHIP?

This review thus far has been concernedwith variables which may influence the ef-fectiveness of threat in persuasion and withfactors which may have contributed to someof the inconsistency in findings. The basicconsideration underlying this analysis (i.e.,the main area in which conflict occurs among

the findings) has been the fact that somestudies indicate that high fear is more per-suasive and others indicate that low fear ismore persuasive.

Recent attempts by some authors to recon-cile the conflicting findings on threat appealshave led to the suggestion of a curvilinear(quadratic), rather than a linear, relationshipbetween fear level and persuasion (see Janis,1967, 1968a, 1968b; Janis & Leventhal, 1968;McGuire, 1963, 1966, 1968; Millman, 1968).The suggestion is that attitude change, orbehavior change, increases with increases infear level up to a certain point then de-creases with further increases in fear level.The postulated relationship is illustrated inFigure 1.

The way in which such a hypothesizedrelationship may serve to reconcile the differ-ent findings on high versus low fear may bedescribed as follows. If the low and high fearlevels in one study were at, for example,Levels A and B, respectively, on the fear-level axis, then the results of that studywould indicate a positive relationship betweenfear level and persuasion. However, if thelow and high fear levels in another studywere at Levels B and C, or A and C, respec-tively, then the results of that study wouldshow a negative relationship between fearlevel and persuasion.

Unfortunately, there is no accurate wayof determining the comparability of fear levelsfrom one study to another. As indicated inTable 1, fear levels used in various studieshave been described in such terms as "low,""weak," "minimal," "mild," "medium,"


V)V)LUzLU

HIGHoUJu_u.UJ

LJ

00<

COorLUCL

LOWLOW HIGH

FEAR LEVEL OF COMMUNICATIONFIG. 1. Postulated curvilinear relationship between

level of fear arousal and persuasive effectiveness ofthe communication.

"moderate," "high," and "strong." What mayhave been labeled "high fear" in one studymay have been equivalent to "medium fear"in another study—if the two kinds of "fear"were even equivalent (see above).

Also, there has been little possibility fordiscovery of a curvilinear relationship in moststudies which have been published in the last15 years. As indicated in Table 1, most suchstudies have used only two fear levels(omitting the no-fear control groups as afear level). Of the 27 studies which experi-mentally manipulated the fear levels, only 5used three fear levels. The best way to deter-mine if a curvilinear relationship exists be-tween fear level and persuasive effectivenessis through future studies using more thantwo fear levels.

The hypothesis of a curvilinear relationshipbetween fear arousal and persuasion does notexplain anything in itself. Theorists have ap-parently resorted to the curvilinear hypothe-sis primarily because research has not beenvery successful in yielding variables whichinteract with fear consistently enough to ac-count for conflicting findings (see precedingdiscussion of such variables). Thus, for thecurvilinear hypothesis to be of use, one mustexplain the curve in terms of some underlyingvariables which cause it.

Several underlying variables which maycause a curvilinear relationship between feararousal and persuasion have been suggested.For McGuire (1966), the factors determiningthe point on the curve at which further fear

increases lead to decreases in persuasion arechronic level of anxiety, concern over theissue, and message complexity. Research hasindicated that chronic anxiety level is prob-ably not an important variable, althoughconcern with the topic may be important. Onestudy has looked at message complexity(Gollob & Dittes, 196S). Millman (1968)suggested that comprehension of the messagecontent is an important variable. Again, theresearch summarized does not support thisconclusion. Janis (1967, 1968a, 1968b) main-tained that defensive avoidance is the under-lying variable. That is, increases in fear leadto increases in persuasion until the fear be-comes so unpleasant that the subject rejectsthe message. Other than the Janis and Fesh-bach studies (19S3, 1954), this interpretationlacks empirical support.

Higbee and Heslin (1968) have suggesteda variable which the author feels may accountfor the curvilinear relationship, and whichmay be somewhat more parsimonious thansome of those previously suggested. Thisvariable is the probability that the undesiredconsequence will occur if no preventive actionis taken. This consideration may be importantbecause if the recipient feels that the conse-quence is unlikely to occur, he will not bemotivated to take action to avert it (i.e.,he will not adopt the recommendations of thecommunication).

The severity of the consequences (threatlevel) and the probability of their occurrencemay be negatively related. Thus, a highlythreatening consequence (e.g., paralysis orblindness) may not be seen as too likely toresult from not brushing one's teeth, whereasit may be seen as more likely that not brush-ing one's teeth could lead to cavities. If sucha negative relationship exists, then increases infear level could lead to decreases in perceivedprobability of occurrence of the threat andthus to decreased persuasiveness.

Thus, there may be two main factors in-fluencing the persuasiveness of a fear-arousingcommunication. The first is the threat level ofthe threatened consequence, which (accordingto most previous studies) facilitates opinionor behavior change as it increases. Thesecond factor is the likelihood of occurrenceof the threat, which may be negatively related


to the threat level of the undesirable conse-quences. The interaction of these two factorsmay result in a curvilinear relationship be-tween fear level and persuasion: As fearful-ness of the threatened consequence increases,motivation to avoid it increases. But per-ceived probability of occurrence may bedecreasing at the same time, until some pointis reached where the threat seems so unlikelyto the subjects that the improbability ofocurrence outweighs the effects of the fear-fulness of the consequences and reduces thepersuasiveness of the communication. Fromthis point, the curve dips downward.

This hypothesis of an interaction be-tween fear level and probability of occur-rence does not rest upon the model of fearas a drive, which model has been shownto be inconsistent with several empirical rela-tionships established. Rather, the Higbee andHeslin hypothesis may be viewed as theinteraction of two cognitive variables—per-ceived magnitude of danger and perceivedlikelihood of occurrence. Such an approachis not the first to attempt to explain fear-appeal findings as being mediated by vari-ables other than a fear emotion. Leventhal(1967), after expressing his dissatisfactionwith the fear-drive model, has developed andbeen working on a "parallel response para-digm," which depicts the fear appeal as con-sisting of components which may influenceacceptance responses or emotional responsesor both—however, these responses are viewedas parallel and do not necessarily influenceone another.

Although two unpublished studies on per-ceived probability of occurrence as a variableunderlying a curvilinear relationship betweenfear arousal and persuasion (Snider, 1962;Wuebben, 1966) have not been especially en-couraging, this hypothesis seems reasonableenough to warrant further study.

SUMMARY AND CONCLUSIONS

Research on fear-arousing appeals from19S3 through 1968 has yielded conflictingfindings concerning the relative effectivenessof high versus low threat in persuasion. Thisresearch has also yielded some inconsistenciesin attempting to determine what variables

may interact with fear level to cause low fearto be more effective in some situations andhigh fear to be more effective in other situa-tions. From the research reviewed in thepresent paper, the author feels that the fol-lowing conclusions concerning the presentstate of knowledge are justified:

1. The widely cited conclusion that highfear arousal creates a defensive-avoidancereaction which causes high threat to be lesspersuasive than low threat is not true in mostsituations. Most relevant research has indi-cated that high threat is superior to lowthreat in persuasion.

2. The nature of the recommendations hasbeen the subject of considerable study. Thespecificity and ease of implementation of therecommendations appear to increase the ef-fectiveness of a fear appeal, regardless of thelevel of fear. Efficacy may also be importantas a main effect but has received only weaksupport as interacting with fear level. Posi-tioning of the recommendations relative tothe threat within the communication does notappear to be important.

3. Responses to fear arousal differ amongindividuals. Characteristics of the recipientswhich may influence how they respond to fearappeals include self-esteem, coping style, andfeelings of vulnerability to danger. Chronicanxiety level has been suggested as an im-portant variable, but its interaction with fearlevel is not supported by research findings.

4. High fear arousal appears to be moreeffective than low fear arousal when the mes-sage source is perceived by the recipients ashighly credible.

5. There is apparently no differential learn-ing of the factual materials in high- and low-threat communications.

6. The available evidence on the interestvalue of high fear is not consistent. Furtherresearch may help in determining the situa-tions in which high fear may arouse interestand the situations in which high fear maydepress interest.

One conceptual consideration and fourmethodological considerations were suggestedas sources of the inconsistency in the findingson fear appeals: (a) the nature and the ob-jects of the "fear" in different studies, (b)the diversity in topics, (c) the diversity in


subjects, (d) the variability in media ofpresentation of the communications, and (e)the differences among studies in criteria ofpersuasive effectiveness.

Despite the considerable amount of re-search in the area of fear appeals, Miller(1968) has suggested that "the surface ofthe fear-arousal problem has only beenscratched [p. 270]." In making this observa-tion he may inadvertently be hinting at onereason we know so little for certain after somuch research. Many researchers in the areahave only been "scratching" at the problem,conducting one or two studies in which theymanipulate and measure what they interpretas "fear," and perhaps using whatever sub-jects, topic, and media may be most con-venient at the time. Few researchers havesettled down to dig a little more into theproblem, the only really concerted efforts ata program of study in the area being that ofLeventhal and his associates. The result ofso many researchers scratching at the problemis the diversity in topics, subjects, media, andcriteria which makes interpretation and com-parison across studies difficult.

Future research should do a little lessscratching on the surface of the problem anda little more digging in one place. Suchdigging would involve at least two considera-tions: First, if one is interested in testingthe findings of a particular study (i.e., repli-cation), he should vary only the variable ofinterest and not also use a different topic,subjects, medium, and/or criterion. Second,if one wishes to do research which may bemeaningfully compared with previous re-search, he should consider using a topic, sub-jects, medium, and criterion used in severalprevious studies (unless, of course, his maininterest is in one of these as the dependentvariable), so that his results can be com-pared with the results of other studies. It istrue that added generality may be given toa finding if it is replicated in a different con-text, but when different results are obtainedin different contexts the cause is difficult tointerpret.

What the author is suggesting is similarto the suggestion of Dabbs and Leventhal(1966) that fear appeals differ along a largenumber of different dimensions, and that

with such confounding it is impossible to attributeincreases in attitude change solely to increases infear. This is an inherent difficulty when fear ismanipulated by the use of differing descriptions ofdanger. Another approach should be developed byanyone wishing to manipulate fear independently ofother aspects of a persuasive communication [p. 531.

What Dabbs and Leventhal have suggestedconcerning the threat communication itself,the present author suggests should also beapplied to the other situational variablesinvolved in any study on fear appeals.

REFERENCES

BERELSON, B., & STEINER, G. A. Human behavior.New York: Harcourt, Brace & World, 1964.

BERKOWITZ, L., & COTTINGHAM, D. R. The interestvalue and relevance of fear-arousing communica-tions. Journal oj Abnormal and Social Psychology,1960, 60, 37-43.

CANNELL, C. F., & MACDONALD, J. C. The impact ofhealth news on attitudes and behavior. JournalismQuarterly, 1956, 33, 315-323.

CARMICHAL, C. W., & CRONKHITE, G. L. Frustrationand language intensity. Speech Monographs, 1965,32, 107-111.

CHIT, G. C. Fear arousal, efficacy, and imminency.Journal of Personality and Social Psychology, 1966,4, 517-524.

COHEN, A. R. Need for cognition and order of com-munication as determinants of opinion change. InC. I. Hovland (Ed.), The order oj presentation inpersuasion. New Haven: Yale University Press,1957.

COLBURN, C. W. II. An experimental study of therelationship between fear appeal and topic impor-tance in persuasion. Unpublished doctoral dis-sertation, Indiana University, 1967.

DABBS, J. M., JR., & LEVENTHAL, H. Effects of vary-ing the recommendations in a fear-arousing com-munication. Journal oj Personality and Social Psy-chology, 1966, 4, 525-531.

DEWOLFE, A. S., & GOVERNDALE, C. N. Fear andattitude change. Journal of Abnormal and SocialPsychology, 1964, 69, 119-123.

DUKE, J. D. Critique of the Janis and Feshbachstudy. Journal of Social Psychology, 1967, 72,71-80.

EVANS, R. I., ROZELLE, R. M., LASATER, T. M.,DEMBROSKI, T. M., & ALLEN, B. P. New measureof effects of persuasive communications: A chem-ical indicator of toothbrushing behavior. Psycho-logical Reports, 1968, 23, 731-736.

FISCHER, E. H., COHEN, S. L., SCHLESINGES, L. E.,& BLOOMER, R. H. Some effects of relevant storiesportraying danger on retention of informationassociated with the stories. Journal of Social Psy-chology, 1967, 73, 75-87.

FRANDSEN, K. Effects of threat appeals and mediaof transmission. Speech Monographs, 1963, 30,101-104.


GOLDSTEIN, M. J. The relationship between copingand avoiding behavior and response to fear-arousing propaganda. Journal oj Abnormal and,Social Psychology, 1959, 58, 247-252.

GOLDSTEIN, M. J., JONES, R. B., CLEMENS, T. L.,FLAGG, G. W., & ALEXANDER, F. G. Coping styleas a factor in psychophysiological response to atension-arousing film. Journal oj Personality andSocial Psychology, 1965, 1, 290-302.

GOLLOB, H. F., & DITTES, J. E. Effects of manipu-lated self-esteem on persuasibility depending onthreat and complexity of communication. Journalof Personality and Social Psychology, 196S, 2,195-201.

GREENWALD, A. G., & SAKUMUEA, J. S. Attitude andselective learning: Where are the phenomena ofyesteryear? Journal of Personality and SocialPsychology, 1967, 7, 387-397.

HAEFNER, D. P. Some effects of guilt-arousingand fear-arousing persuasive communications onopinion change. American Psychologist, 1956, 11,359. (Abstract)

HAEFNER, D. P. Arousing fear in dental health edu-cation. Journal of Public Health Dentistry, 1965,25, 140-146.

HEWOILL, M. A., & MILLER, G. R. Source credibilityand response to fear-arousing communications.Speech Monographs, 1965, 32, 95-101.

HIGBEE, K. L., & HESLIN, R. Fear-arousing com-munications and the probability of occurrence ofthe threatened consequences. Paper presented atthe meeting of the Indiana Psychological Associa-tion, French Lick, April 1968.

HILGARD, E. R. Introduction to psychology. (2nded.) New York: Harcourt, Brace & World, 1962.

HODGES, W. F. Effects of ego threat and threat ofpain on state anxiety. Journal of Personality andSocial Psychology, 1968, 8, 364-372.

HOVLAND, C. I., JANIS, I. L., & KELLEY, H. H.Communication and persuasion. New Haven: YaleUniversity Press, 1953.

INSKO, C. A., ARKOFF, A., & INSKO, V. M. Effectsof high and low fear-arousing communicationsupon opinions toward smoking. Journal of Experi-mental Social Psychology, 1965, 1, 256-266.

JANIS, I. L. Effects of fear arousal on attitude:Recent developments in theory and experimentalresearch. Advances in Experimental Social Psy-chology, 1967, 4, 166-224.

JANIS, I. L. The contours of fear. New York: Wiley,1968. (a)

JANIS, I. L. When fear is healthy., PsychologyToday, 1968, 1(11), 46-49, 60-61. (b)

JANIS, I. L., & FESHBACH, S. Effects of fear-arousingcommunications. Journal of Abnormal and SocialPsychology, 1953, 48, 78-92.

JANIS, I. L., & FESHBACH, S. Personality differencesassociated with responsiveness to fear-arousingcommunications. Journal of Personality, 1954, 23,154-166.

JANIS, I. L., & LEVENTHAL, H. Human reactions tostress. In E. F. Borgatta & W. Lambert (Eds.),Handbook of personality theory and research.Chicago: Rand McNally, 1968.

JANIS, I, L., & MILHOIXAND, H. C., JR. The influ-ence of threat-appeals on selective learning of thecontent of a persuasive communication. Journal ofPsychology, 1954, 37, 75-80.

JANIS, I. L., & TERWILLIGER, R. F. An experimentalstudy of psychological resistances to fear-arousingcommunications. Journal of Abnormal and SocialPsychology, 1962, 65, 403-410.

KATZ, D. The functional approach to the study ofattitudes. Public Opinion Quarterly, 1960, 24,163-204.

KLAPPER, J. T. The effects of mass communication.New York: Free Press, 1960.

KORNZWEIG, N. D. Behavior change as a functionof fear-arousal and personality. Unpublisheddoctoral dissertation, Yale University, 1968.

KRAUS, S., EL-ASSAL, E., & DEFLEUR, M. L. Fear-threat appeals in mass communications: An ap-parent contradiction. Speech Monographs, 1966,33, 23-29.

KRETCH, D., CRUTCHFIELD, R. S., & BALLACHEY, E. L.Individual in society. New York: McGraw-Hill,1962.

LEIBLER, S. N. An analysis of some characteristicsof respondents and nonrespondents to two mailedquestionnaires. Unpublished doctoral dissertation,University of Rochester, 1967.

LEVENTHAL, H. Fear communications in the accept-ance of preventive health practices. Bulletin of theNew York Academy of Medicine, 1965, 41, 1144-1168.

LEVENTHAL, H. Fear: For your health. PsychologyToday, 1967, 1(5), 54-58.

LEVENTHAL, H., JONES, S., & TREMBLY, G. Sex dif-ferences in attitude and behavior change underconditions of fear and specific instructions. Journaloj Experimental Social Psychology, 1966, 2, 387-399.

LEVENTHAL, H., & NILES, P. A field experiment onfear arousal with data on the validity of question-naire measures. Journal of Personality, 1964, 32,459-479.

LEVENTHAL, H., & NILES, P. Persistence of influencefor varying durations of exposure to threat stimuli.Psychological Reports, 1965, 16, 223-233.

LEVENTHAL, H., & PERLOE, S. I. A relationship be-tween self-esteem and persuasibility. Journal ofAbnormal and Social Psychology, 1962, 64, 385-388.

LEVENTHAL, H., & SINGER, R. P. Affect arousal andpositioning of recommendations in persuasive com-munications. Journal of Personality and SocialPsychology, 1966, 4, 137-146.

LEVENTHAL, H., SINGER, R. P., & JONES, S. Effects offear and specificity of recommendation upon atti-tudes and behavior. Journal of Personality andSocial Psychology, 1965, 2, 20-29.

LEVENTHAL, H., & TREMBLY, G. Negative emotionsand persuasion. Journal of Personality, 1968, 36,154-168.

LEVENTHAL, H., & WATTS, J. C. Sources of resistanceto fear-arousing communications on smoking andlung cancer. Journal of Personality, 1966, 34, 155-175.


LEVKNTHAL, H., WATTS, J. C., & PAOANO, F. Effectsof fear and instructions on how to cope withdanger. Journal of Personality and Social Psychol-ogy, 1967, 6, 313-321.

LEWAN, P. C., & STOTLAND, E, The effects of priorinformation on susceptibility to an emotional ap-peal. Journal of Abnormal and Social Psychology,1961, 62, 450-453.

LUNDY, R. M., SIMONSON, N. R., & LANDERS, A. D.Conformity, persuasibility, and irrelevant fear.Journal of Communication, 1967, 17, 39-54.

McGuiRE, W. J. Effectiveness of fear appeals inadvertising. New York: Advertising ResearchFoundation, 1963.

McGuiRE, W. J. Attitudes and opinions. AnnualReview of Psychology, 1966, 17, 475-514.

McGuiRE, W. J. Personality and susceptibility tosocial influence. In E. F. Borgatta & W. Lambert(Eds.), Handbook of personality theory and re-search. Chicago: Rand McNally, 1968.

McNuLTY, J. A., & WALTERS, R, H. Emotionalarousal, conflict, and susceptibility to social influ-ence. Canadian Journal of Psychology, 1962, 16,211-220.

MILLER, G. R. Studies on the use of fear appeals:A summary and analysis. Central States SpeechJournal, 1963, 14, 117-124.

MILLER, G. R. Communication and persuasion re-search: Current problems and prospects. QuarterlyJournal of Speech, 1968, 54, 268-276.

MILLER, G. R., & HEWOILL, M. A. Some recentresearch on fear-arousing message appeals. SpeechMonographs, 1966, 33, 377-391.

MILLMAN, S. Anxiety, comprehension, and suscepti-bility to social influence. Journal of Personalityand Social Psychology, 1968, 9, 251-256.

MOLTZ, H., & THISTLETHWAITE, D. L. Attitude modi-fication and anxiety reduction. Journal of Ab-normal and Social Psychology, 1955, SO, 231-237.

MOORE, H. M., JR. Effects of fear-arousing com-munications on driving safety attitudes and drivingbehavior. Unpublished doctoral dissertation. Uni-versity of Florida, 1965.

MORGAN, C. T., & KINO, R. A, Introduction to psy-chology. (3rd ed.) New York: McGraw-Hill, 1966.

NILES, P. The relationship of susceptibility and anx-iety to acceptance of fear-arousing communica-tions. Unpublished doctoral dissertation, YaleUniversity, 1964.

NUNNALLY, J. C., & BOBREN, H. M. Variables gov-erning the willingness to receive communicationson mental health. Journal of Personality, 1959,27, 38-46.

PAYNE, I. R. Persuasibility and its relationship toaffect, sex, and type of persuasive appeal. Unpub-lished doctoral dissertation, Pennsylvania StateUniversity, 1963.

PICCOLINO, E. B. Depicted threat, realism, and speci-ficity: Variables governing safety poster effective-ness. Unpublished doctoral dissertation, IllinoisInstitute of Technology, 1966.

POWELL, F. A. The effect of anxiety-arousing mes-sages when related to personal, familial, and im-personal referents. Speech Monographs, 1965, 32,102-106.

POWELL, F. A., & MILLER, G. R. Social approval anddisapproval cues in anxiety-arousing communica-tions. Speech Monographs, 1967, 34, 152-159.

RADELFQJOER, S. F. Some effects of fear-arousingcommunications on preventive health behavior.Unpublished doctoral dissertation, Stanford Uni-versity, 1963.

ROBBINS, P. R. An application of the method ofsuccessive intervals to the study of fear-arousinginformation. Psychological Reports, 1962, 11, 757-760. (a)

ROBBINS, P. R. Self-reports of reactions to fear-arousing information. Psychological Reports, 1962,11, 761-764. (b)

ROSENBLATT, P. C. Persuasive value of threat andamount of attitude change advocated. Unpublisheddoctoral dissertation, Northwestern University,1962.

RUBIN, B. M., KATKIN, E. S., WEBS, B. W., &EFRAN, J. S. Factor analysis of a fear surveyschedule. Behavior Research and Therapy, 1968,6, 65-75.

SASTROHAMIOJOJO, R. A. S. Institutional credibilityand emotionality of appeals as determinants ofattitude change. Unpublished doctoral dissertation,Indiana University, 1968.

SECORD, P. F., & BACKMAN, C. W. Social psychology.New York: McGraw-Hill, 1964.

SIMONSON, N. R., & LUNDY, R. M. The effectivenessof persuasive communication presented under con-ditions of irrelevant fear. Journal of Communica-tion, 1966, 16, 32-37.

SINGER, R. P. The effects of fear-arousing communi-cations on attitude change and behavior. Unpub-lished doctoral dissertation, University of Con-necticut, 1965.

SNIDER, M. The relationship between fear arousaland attitude change. Unpublished doctoral dis-sertation, Boston University Graduate School,1962.

SPELMAN, M. S., & LEY, P. Knowledge of lungcancer and smoking habits. British Journal ofSocial and Clinical Psychology, 1966, 5, 207-210.

STERN, G. S., LANA, R. E., & PAULINO, F. J. Feararousal and order of presentation of persuasivecommunications. Psychological Reports, 1965, 16,789-795.

WATTS, J. C. The role of vulnerability in resistanceto fear-arousing communications. Unpublisheddoctoral dissertation, Bryn Mawr College, 1967.

WUEBBEN, P. L. Experimental analysis of threat-based persuasive appeals. Unpublished doctoraldissertation, University of Wisconsin, 1966.

ZEMACH, M. The effects of guilt-arousing communi-cations on acceptance of recommendations. Un-published doctoral dissertation, Yale University,1966.

(Received February 12, 1969)

Documents

FIFTEEN YEARS OF FEAR AROUSAL: RESEARCH ON THREAT