How to persuade people to engage in preventive health behavior has beena matter of concern to professional health workers. This theoreticalanalysis is devoted to the role of fear though it is only a partof the picture.

FEAR APPEALS AND PERSUASION: THE DIFFERENTIATION

OF A MOTIVATIONAL CONSTRUCT

Howard Leventhal, Ph.D.

THE goal of the present paper is toinvite new thoughts about the effects

of fear-arousing communications on per-suasion. As a consequence, it is un-ashamedly theoretical and may seem torun counter to the insistent cries for thepractical and relevant. But it is mythought that the theories and assumptionswe make about other people are an im-portant determinant of what we say whenwe try to influence or educate them. Todo a better job in this role, we must besure that we are using the most adequatetheory in developing our persuasive andeducational appeals. The best way to in-crease the quality of our theories is todevelop and compare different ways ofthinking about the same problem. Thus,to convince you that nothing is so prac-tical as a good theory, I shall review twosets of ideas that can be used to under-stand the persuasive effects of fear-arous-ing health messages, discuss how wellthese ideas fit empirical data, and thenleave you to decide which provides themost creative suggestions for practicalaction.The health professional's unique

knowledge gives him the special respon-sibility to inform others of the imminenceof health dangers. Because he wants hisaudience to believe and to act upon hisrecommendations, he is very likely to

wonder about the most effective way ofpresenting his information. At this pointthe practitioner has to consider hisoptions. Should he "hit them hard" andarouse strong fear with graphic descrip-tions of death and danger, or should heuse a mild, unfrightening approach andavoid any mention of the serious hazardsthat may occur if one fails to executea protective action? Which technique ismost effective? If he has read the earlystudy by Janis and Feshbach (1953) hewould find support for a mild approach.This study suggests that increasing thevividness of the threat information ondental disease and thereby raising thelevel of fear decreases the proportion ofsubjects who follow a program of dentalcare.

If he reads further, he is likely tonotice that nearly all subsequent studiesshow that communications which stimu-late higher levels of fear [on issues suchas tetanus (Leventhal, Singer and Jones,1965; Dabbs and Leventhal, 1966),smoking and lung cancer (Janis andMann, 1965; Leventhal, Watts andPagano, 1967)] lead to more acceptanceof the communicator's recommendations.2Indeed, he may notice that the number ofstudies with positive effects, high-fearmessages producing more attitude andbehavior change than low-fear messages,

VOL. 61, NO. 6. A.J.P.H.1 208

FEAR APPEALS AND PERSUASION

is far greater than the number with theopposite effect. He may also note thathigh-fear messages are much more effec-tive in getting people to say they will actthan in producing action. Even thepowerful emotional experience of play-ing the role of a lung cancer patient(Janis and Mann, 1965; Mann and Janis,1968) produces only a 2.5 cigarette perday greater reduction for role-playersthan for control subjects. The data arenot consistent. Perhaps high fear is abetter persuader than low fear, but maybeit is superior only for some issues, somepopulations and some, but not all, recom-mended actions. It has also been sug-gested that if we compared many mes-sages which differed in fear level, wewould note that at the low end of thescale, where the messages were low ormild, increases in threat level would beassociated with increases in persuasion(compliance with recommendations tostop smoking, take tetanus shots, etc.).At the high end of the scale, going frommild to very strong messages, increasesin threat level would be associated withdecreases in persuasion. The problemseems to be that of selecting the rightfear level (see Janis, 1967; Janis andLeventhal, 1968).

Contrasting Theories

The drive pcradigmMy discussion of the effects of fear

upon persuasion assumed that fear is adrive or a motivator of behavior. The

fear-drive theory is a member of the classof Darwinian models that emphasize that"man is an animal." These models see thehuman organism as govemed by a set ofinternal needs which activate motivatingmechanisms. Once a need is present, be-havior must be forthcoming; starve theanimal and it must eat. Dri'ves like fearare aroused by external stimulation. Thus,when the normal human animal is con-fronted by danger, he becomes fearful,and when this happens, he will be mo-tivated to protect himself. When a self-protective act is executed successfully, itwill reduce fear. When fear is completelyremoved there is no longer any push tobehave. Fear is a necessary interveningstep between the stimulus and the action.Without it, action will not occur (seeFigure 1). In the simplistic world ofthe fear model, our main job is to pro-duce and evaluate fluctuations in fear.There are many reasons for rejecting

the serial or fear-drive position. Two ofthese reasons are: (1) the available datashow that the variables which are sup-posed to change fear, to turn it on andoff, do not have any important effectsupon persuasions; (2) concentration onthe concept of fear keeps us from askingrelevant questions about variables thatdo effect preparation for danger. Thus,neither the practitioner nor the experi-mentalist can use the fear model as areliable guide to influence others.

Before reviewing the available data onfacts which influence fear, I would liketo examine an alternative model. This

Figure 1-The Drive paradigm: Fear Behavior provides the important cues for startingand stopping of motivated behavior. The action that reduces fear stops motivation.This action may be an appropriate health act or a defensive act.

PEAR STI TMEPRESENT l ~RTIONEXTERUIAL El!MOTIONAL >R ERCEIVE I TAKEDANGER PR3SPONSE TENSION ACTION

FEAR ABSENTISTOP

JUNE. 1971 1209

Figure 2-Basic features of Parallel Model. Figure a shows that the three basic typesof response-action, attitude and emotion-are dependent upon the situation. Fearis not a necessary link between Danger and action. Figure b illustrates the com-plexity of the Danger situation and that different aspects of it may effect action,attitude and emotion. There may, of course, be some stimulus elements which effect2 of these factors.

DangerSituationi4-

protectiveaction

attitudestoward dangerand action

emotionalbehavior

(a.)

components of danger situation

IS _4 Protective2 actions3

S4 _ AttitudesS5 >, toward dangers6 -_ and action

57 ~ -4 DEmotionals8 behavior

(b.)

will give the reader two perspectives fromwhich to view the data.

The porallel response paradigm

The basic assumption of this alterna-tive paradigm is that emotional responses(such as fear) and adaptive responses(such as belief changes and protectivehealth acts) are arranged in a parallelrather than serial relationship. Theparallel model is related to a general classof models which are called informationprocessing models. Situational stimuli aresources of information and the interpre-tation, given these stimuli, leads to avariety of response processes. For ex-ample, if a situation is interpreted asdangerous, the interpretation can giverise to coping efforts to minimize theimpact of the danger, and it can simul-taneously give rise to feelings or emotionsof fear. Thus, in a parallel arrangementeach of the functional units, fear andcoping activity, are activated by the in-put of situational stimuli (see Figure 2a).

Perhaps we can clarify the differencebetween serial and parallel arrangementsby an analogy to a set of electric lights

that are arranged in serial or parallelorder. In series arrangements, commonin old Christmas tree lights, each func-tional unit or light is dependent upon theothers. If one burns out, the entire seriesgoes out. If one light has high resistanceto the flow of current, all the lights inthe series get little current. With aparallel arrangement each light completesits own circuit so that each one respondsto the current according to its own char-acteristics. Thus, if one light goes out theremainder stay active, and if one passesmuch current and burns brightly, it willnot change the light output of the others.They are relatively independent of oneanother, and all are responding to thesame current input. The key point in theanalogy is that fear and adaptive activityare like response units, each of whichare turned on by the danger situation.Both responses are to the warning mes-sage and there is no need to assume thatfear arousal is a necessary prior link inthe causal sequence that leads to adaptiveacts. Part b of Figure 2 shows that thestimulus situation actually consists of avariety of elements. This way of depict-ing the parallel model not only shows

VOL. 61, NO. 6. A.J.P.H.1210

FEAR APPEALS AND PERSUASION

that the stimulus is responsible for emo-tional, attitudinal and adaptive actions,but suggests that different aspects of thesituation may be more important ineliciting one or another type of response.

In Figure 3 is added another part tothe parallel model, the cognitive encoder.It is placed between the stimulus and theresponse (emotional and coping) outputs.This illustrates that it is not the externalstimulus per se, but an encoded or elabo-rated version of it that leads to behavior.The diagram is also extended from leftto right to illustrate that the process ofadaptation to danger is extended in time.The elaborated version also suggests thatthe person perceives and responds to hisown behavior. If he acts to control thedanger, he can see the effects of his ac-tions upon the environment and he canthen respond to the changed situation. Ifhe becomes frightened, he will experiencehis bodily reactions as well as the ex-ternal danger. These inner reactions mayget stronger or weaker; they too can pro-vide information to guide behavior. Thus,the perceptual-cognitive operator obtainsinformation from three sources: (1) theexternal danger object, (2) the person'semotional behavior and (3) the person'scoping behavior.Under certain conditions, an individ-

ual will respond to danger; he will thinkabout and act to control the outer world,and ignore or be inattentive to his fear.This type of control process can be calledDanger Control (DC). It is problem-solving activity. On the other hand, it isalso possible to respond to internal cuesgenerated by one's own emotions andselect behaviors whose aim is Fear Con-trol (FC).

Neither Danger Control nor Fear Con-trol are likely to appear in pure form,though one or the other process maydominate under different conditions. Forexample, decreased clarity of externaldanger cues should shift control to in-ternal or emotional cues. Secondly, it islikely that different types of coping re-sponses will be used to control fear thanto control danger. Responses to controldanger are aimed at manipulating theexternal environment, by changing thedanger agent and its ability to strike us.Responses to control fear are concernedwith avoiding contact with stimuli whichare fear-producing and to engage in re-actions, such as eating, sleeping, intenselaughter, etc., that will interfere with ordisrupt unpleasant emotional responses.Of course, there will be individual differ-ences in the way people respond to theirfear, or to danger. Some people may

Figure 3-Danger and fear control processes over time: the danger encoder is addedto the parallel model. This refers to processes of identifying danger and evaluatingthe value of actions at a later time (Time 3). Danger and fear control are the twotypes of motivational control that may follow the recognition of danger. Aspectsof these could be simultaneous, though they will frequently appear in some alter-nating order.

Time 1.

Dangercontrol

External CognitiveDanger i encoder

(ilessage or actualdanger)

Fearresponse'

Time 2.

Attitudesand

Action

ExternalDanger -

Feelings ofFear.

,s Desire tocqntrol fear

JUNE. 1971

Time 3.

Continue

etc.

Continue

121

respond to their own fear by searchingfor external cues in order to engage indanger-control reactions.

The Empirical Adequacy of the Models

Because past research was developedwithin the drive framework, it is easierto demonstrate the inadequacies of thedrive model than to demonstrate the pre-dictive power of the parallel-responsemodel. Despite this difficulty, we can notethat the parallel model is reasonably post-dictive.The simple version of the emotional-

drive model cannot predict whether in-creased fear would increase or decreasethe persuasive impact of a health com-munication; other factors have to bespecified. These factors can be identifiedby their effects upon fear level. Anyfactor which reduces fear should rewardor strengthen the persuasive impact ofthe communication. Any recommendationwhich increases fear and leaves doubtshould be ignored. For example: (1) Ifa person cannot take protective actionimmediately after a warning message, hemust find some other way to reduce hisfear-perhaps he will defend himselfagainst fear by denying his vulnerabilityto danger (Janis and Feshbach, 1953 );(2) If a recommended action is not com-pletely effective in eliminating danger,there will be residual fear and onceagain, this fear may be reduced by denialof vulnerability (Janis, 1967); (3)People who think well of themselves tendto be optimistic and are more likely todeny vulnerability to danger. Thus, theyare less likely to be accepting of recom-mendations with high-fear messages asthey will be motivated to maintain afavorable world view and to deny therelevance of the threat (Cohen, 1957;Leventhal and Perloe, 1962); (4) If wemake it difficult for a person to think ofhimself as invulnerable, by having himplay the part of someone with cancer(Janis and Mann, 1965), he will find it

more difficult to reduce fear by denialand will more readily accept the recom-mendation (to stop smoking) made bythe communicator. If the above hypothe-ses are true, persuasion will be increasedby high-threat messages when: (1) ac-tions are immediately available, (2) ac-tions are relatively effective, (3) subjectsare low in esteem, and (4) we have sup-pressed the ability to deny personal vul-nerability. Each of the hypotheses listedabove have important, common features.They all attempt to account for the in-consistencies in the relationship betweenfear and persuasion. Second, each of thehypotheses is based upon key assumptionsof the fear-drive model. All assume thatfear motivates both the acceptance andthe rejection of the communicator's mes-sages. The distinctive feature of eachhypothesis is that it suggests a particularfactor that helps channel fear so that itleads to the acceptance of the recom-mendation.

In the next section I shall review datapertinent to each of those hypotheses.The data suggest that most of the hy-potheses are false. The variables specifiedeither have no effect upon attitude andbehavior, or, when they do have effects,they are different from those predicted bythe fear-drive model. Moreover it is im-portant to note that many of the factorsdo reduce fear as the drive model saysthey should. But if a variable changesfear and then has no effect upon attitudeand behavior, we can only conclude thatfear is to a great degree separate fromor in parallel to the evaluative and actionprocesses involved in coping.

Recommendation effectiveness

If tetanus shots are seen as a perfectpreventive, then raising fear level abouttetanus should facilitate favorable atti-tudes toward shots (Leventhal, Singerand Jones, 1965; Leventhal, Jones andTrembly, 1966). If dental hygiene pro.grams are seen as an inadequate preven-

VOL. 61. NO. 6. A.J.P.H.1212

FEAR APPEALS AND PERSUASION

tive, then raising fear level on toothdecay should decrease reported compli-ance to dental hygienic recommendations(Janis and Feshbach, 1953). Accordingto the drive model, differences in recom-mendation effectiveness should accountfor the differences in results. To test thisassumption, Dabbs and Leventhal (1966)conducted an experiment where tetanusshots were described as virtually a per-fect preventive or as the only, though lessthan perfect, preventive. The subjectsreceiving each description of the shotwere further subdivided; some received anon-threatening talk, others a mild oneand others a strong one.The results showed that both fear and

effectiveness of the communication in-fluenced attitudes and intentions. Thehigh-fear message created stronger beliefsin the importance of shots and the needfor getting shots. High-fear messageswere superior to low-fear messages re-gardless of the ascribed effectiveness ofshots. The description of effectiveness wasimportant, however; more people wantedshots if they thought them perfect. Itwas also true that more people took shotsafter exposure to the high-fear messagethan after exposure to the messages inthe less fearful communication condi-tions.To reduce the perceived effectiveness

of tetanus shots, Dabbs and Leventhaldescribed them as "less than perfect."While this did reduce confidence in theadequacy of shots, perhaps a much biggerreduction in perceived effectiveness isnecessary to produce the reversal pre-dicted by drive theory, low-fear messagesmore persuasive than high. Fortunately,Godwin Chu (1966) conducted a studywhich did include a very extreme changein the described adequacy of the protec-tive measure. The topic and recommenda-tion concerned a drug to combat parasiticworms. His audience was a population ofTaiwanese schoolchildren for whom thisissue is relevant. For a third of his sub-jects the drug was described as extremely

effective (cures 90% of cases), for an-other third as moderately effective (cures60% of cases) and for the last third asrather ineffective (cures less than 30%of cases). Fear was also varied over threelevels (low, moderate, and high) makingnine messages in all. Chu's subjects werequite rational: (1) effective recommen-dations were accepted more strongly thanineffective recommendations; (2) high-fear messages led to more acceptancethan low-fear messages; and (3) therewas absolutely no tendency for low-fearmessages to be more persuasive thanhigh-fear messages at the lowest level ofdrug effectiveness.The most reasonable conclusion from

the effectiveness studies is the following:when people are faced with danger, theyprefer to do something rather thannothing, and they always prefer the mosteffective means of control regardless ofthe intensity of their emotions. This con-clusion is consistent with the parallel-response model. When a threat is well-defined, people want to control it-Dan-ger Control. The more vivid the threat,the more important it is to do something.But even when fear is high and recom-mendations poor, they show no signs ofdenial. The results are consistent withthe parallel model.

Immediacy of response

When protective acts are immediatelyavailable and there is no time for denialof vulnerability (Janis and Feshbach,1953), the drive model predicts moreacceptance for high- than for low-fearmessages. The parallel-response modelsuggests an opposite outcome. Interfer-ence with realistic action should begreatest immediately after exposure tothreat messages at the time when fearis strongest and the processes of FearControl are most salient. Under theseconditions, Danger Control reactions willbe inhibited by fear-controlling avoid-ance behaviors. The Janis and Feshbach

JUNE, 1971 1213

(1953) study reports that a low-fear mes-sage is more persuasive than a high-fearmessage after a delay. It is the only studythat reports such a finding and its resultis consistent with the fear-drive modeland inconsistent with the parallel model.In all other cases where a low-fear mes-sage seems to produce more persuasionthan a high-fear message, the effectoccurs immediately after the communica-tion. When a delay is introduced betweencommunication and the measurement ofaction the trend is for high-fear messagesto be more persuasive than low. For ex-ample, Leventhal and Watts (1966) usedlow, mild and strong threat messages(on lung cancer) to influence attitudestoward chest x-rays and stoppingsmoking. Their high-fear movie includedclose-ups from a lung surgery opera-tion. The three groups were comparedfor the number of people who took x-raysimmediately after the communications;an x-ray booth was available right out-side the theater. The results showed thatthe smokers in the audience did not takex-rays after exposure to the high-threatfilm; they did take x-rays after the lowand moderate threat films. Reports ofsuccess in reducing smoking were col-lected three months later and a greaterproportion of subjects exposed to thestrong high-fear film claimed success incutting back.The low rate of x-ray taking after

high-fear can be attributed to the needto control a strong state of fear. But itcould also reflect the fact that taking anx-ray leads to danger detection (findingcancer) rather than to protection. Thus,the comparison of x-rays and stoppingsmoking involves the factor of movingtoward danger, as well as the factor ofimmediacy. Both of these conditionswould favor the control of fear emotionby the avoidance of the detection of thedanger.

In a second study, Kornzweig (1967)used the same response, taking atetanus shot, for the immediate and de-

layed condition. He told half his subjectsthat shots were immediately available(downstairs in the same building) andthe other half that shots were availableon the following day. Information onthe availability of shots was presented atthe end of a booklet. Different bookletsgave strong, mild or no threat informa-tion on tetanus. Kornzweig found a tend-ency for more shots to be taken afterhigh than after low-threat messages, andmore shots were taken when they wereimmediately available. But, there was alsoa complex interaction of the threat levelof the communication, with the self-esteem of the subject and delay. Immedi-ately after the messages, individuals highin self-esteem were much more likely totake tetanus shots if they were exposedto the high-threat message. Low self-esteem subjects, on the other hand, weremuch more likely to take shots if theywere exposed to the no threat or mildthreat messages. Although fewer subjectstook shots on the following day, high andlow self-esteem subjects no longer differedin their response to the fear messages;both groups took more shots if they hadbeen exposed to the high-threat message.

Kornzweig's data also show that high-fear messages do less well than low-fearmessages when action is immediate. Thegeneralization is qualified by the fact thatthe failure to take the shot-an avoidanceor Fear Control response-is visible onlyin subjects with inadequate copingmechanisms, the low self-esteem peoplewho feel overwhelmed by the danger. Re-sults reported by Dabbs and Leventhal(1966) and Leventhal and Trembly(1968) also support this coping hypoth-esis. In all cases, low self-esteem subjectsaccept the reality of the threat but do notconform in beliefs or action to the rec-ommendations made to deal with it.When their sense of invulnerability isshattered, low-esteem subjects becomepassive in the face of danger. Perhapsthey are reluctant to act to controldanger because acting may bring them

VOL. 61, NO. 6. A.J.P.H.1214

FEAR APPEALS AND PERSUASION

closer to danger cues and increase theirfear. The need to avoid or to controlfear will be strong when emotional dis-turbance is strong, immediately after thehigh-threat message. But, as time passesand emotional disturbance abates, manylow-esteem subjects will be able to facethe danger and take protective action.High-esteem subjects, on the other hand,seem to focus upon the danger at bothearly and late points in time. Thus, in-stead of showing fear motivated avoid-ance behaviors, they attempt to increasetheir invulnerability to danger by takingthe recommended action.

Invulnerability defenses

Janis and his associates (Janis, 1967;Janis and Mann, 1965) state that it isimportant to break through the "facadeof defensive denial" and suggest that onecan do this by making the subject be-lieve that he is personally vulnerable todanger. They attempt to arouse fear andbreak through defenses by having thesubject play the role of a person withlung cancer. In two reports of an initialstudy (Janis and Mann, 1965; Mannand Janis, 1968), the claim is madethat emotional role-playing has a startlingimpact on smoking behavior. But a care-ful examination of their data shows thatthe final differences between their role-play and their control groups is ex-tremely small (2.5 cigarettes per day)and more appropriate probability testswould probably not yield statistical sig-nificance. More to the point, however,is the fact that later studies (Mann,1967) have failed to replicate this find-ing. An extremely interesting result isreported by Mausner and his associates(Mausner and Platt, 1970). Four groupsof subjects were used in their study:(1) the subject role-played the part of apatient; (2) he played the doctor in-forming the patient of the dangers ofsmoking; (3) the smoker observed therole-playing sessions; and (4) control

subjects were unexposed to any informa-tion on smoking and cancer. The resultsshowed essentially no reduction in smok-ing for the control subjects or for thesubjects who played the part of the pa-tient. But subjects acting as observers orplaying the part of the doctor showedsignificant reductions in smoking. Thus,the subjects who were frightened whentheir invulnerability defenses were dis-rupted (the patients) did not respond tothe anti-smoking communication.

There is much experimental data sug-gesting that increasing vulnerabilitywhile stimulating fear reduces acceptanceof protective recommendations. For ex-ample, subjects who do feel vulnerableto cancer are less likely to stop smokingwhen exposed to high-fear messages(Niles, 1964). They seem convinced thatthey will get cancer no matter what theydo. Leventhal and Niles (1964) foundthat fewer smokers (vulnerable subjects)took x-rays after high-fear messages.Brock and Balloun (1967) also find thatsmokers are resistant to information onvulnerability and Berkowitz and Cotting-ham (1960) found that occasional andnot regular drivers were most persuadedby high-fear messages. Further, nationalsurvey data reported by Daniel Hornshow that high personalization of thecancer threat is associated with a failureto reduce smoking over the long term.

In an experimental study on lung can-cer, Jean Watts (1966) exposed threegroups of smokers to motion pictures andused a fourth unexposed control group.One group saw a fear message (surgeryfor cancer), another a vulnerability mes-sage, and the third group received boththe vulnerability and fear messages. Allsubjects were given instructions on howto stop smoking. Subjects seeing thevulnerability movie only, and subjectsseeing the fear movie only, significantlyreduced their smoking. Control subjectsand subjects seeing both the vulnerabilityand the fear movie did not. As in theMausner role-playing study, combining

JUNE. 1971 1215

strong images of a danger with explicitindications of vulnerability, arouses fearand undermines the person's ability toconceive of himself as coping with andcontrolling the danger. The conditionfavors avoidance and Fear Control ratherthan confronting and manipulating theexternal danger. All of these studies sug-gest that the subject thinks the dangerterrible, but feels hopeless about con-trolling it.

Reducing drive

In one of our studies, Leventhal andSinger (1966), we attempted to test thehypothesis that persuasion would be in-creased when the communication reducedthe subject's fear. The hypothesis is adirect prediction from the emotional-drive theory. The fear produces motiva-tion and a desire to take protective ac-tions. By making sure that the fear iseliminated by the presentation of ourrecommendations, we should prevent theappearance of defensive denial. The ex-periment used high- and low-fear mes-sages describing the dangers of dentalneglect. One high- and one low-feargroup did not hear protective recom-mendations. This made it unlikely thatthey would reduce their fear by thinkingabout hygenic practices. Three otherpairs of high- and low-fear groups hearddetailed recommendations on dental hy-giene. One pair received the recommen-dations before the fear material, anotherpair received the recommendations inter-spersed with the fear material and thethird pair received the recommendationsafter the fear material. Recommendationsgiven before fear is aroused seem lesslikely to reduce fear than recommenda-tions that come along with the threaten-ing information. Listening to the recom-mendations and thinking about themafter fear is aroused should be most fear-reducing. The results were clear: First,all of the groups receiving high-fear mes-sages were more persuaded of the impor-tance of the hygienic practices than were

the groups receiving the low-fear mes-sages. Second, the positioning of the hy-gienic practices affected fear reduction;when the recommendations were last,subjects were least fearful. All of thegroups were in the appropriate order.But subjects were not more persuaded iftheir fear was reduced! The drive-modelprediction is wrong. Instead, the recom-mendations are somewhat more stronglyaccepted when they are presented in anorganized group, either before or afterthe fear message. This happens regard-less of the fear-level of the message (highor low). Thus the clarity of the informa-tion, presenting the practices in an unin-terrupted unit, is what increases per-suasion.We should also note that the topic in the

Leventhal-Singer study (1966) was den-tal hygiene, and the fear messages werenearly identical to those used by Janisand Feshbach; the results, however, wereopposite. Two additional studies withhigh school students, one by RobertHaefner (1965) the other by RobertSinger (1965), also found high-fear mes-sages more effective than low-fear mes-sages and contradict the Janis and Fesh-bach results. Both used actual records ofbehavior, in addition to questionnaires.Their results also showed that increasingthe amount of fear information addedlittle to the message's persuasive powerfor students in college preparatory pro-grams. But students in non-college pro-grams showed more belief change as thecommunication became more threatening.The new dental studies point to an in-

formational interpretation of persuasion.The Leventhal-Singer study suggestedthat fear reduction is irrelevant but thatclarity of information is relevant: knowl-edge, not feeling, affects belief. The Haef-ner and Singer studies gave further sup-port to the informational interpretationof the persuasion effects. They found thatstudents in the academic program of ahigh school were equally responsive tohigh- and low-fear messages. These stu-dents usually score higher in tests of in-

VOL. 61. NO. 6. A.J.P.H.1216

FEAR APPEALS AND PERSUASION

telligence and achievement. They alsohave a socialization history where parentsuse explanations rather than punishmentsand threat (Aronfreed, 1961). Thesestudents can be expected, therefore, torespond to information. The non-aca-demically oriented child, on the otherhand, is more responsive to threat. Thus,individual difference factors modify thedegree to which the fear communicationpersuades through informational chan-nels (Danger Control) or through fear(Fear Control).

Specific Action Instructions

Early in our research program, I gavesome attention to creating conditionswhich could maximize a communicationsimpact upon behavior. On the basis of apersonal hunch, I formulated the idea of"specific action instructions." This is in-formation which encourages the audienceto rehearse various aspects of the recom-mended action. One purpose of the re-hearsal is to commit the person to thedecision to act. Another, and I think thekey aspect, is to help the subject attendto the series of cues and specific acts thatare needed to complete the action se-quence.We first studied specific instructions

in an investigation of tetanus inocula-tions. Four experimental and two controlconditions were used. The experimentalconditions consisted of two high-feargroups and two low-fear groups. Afterthe fear material was presented, all sub-jects were told to get a shot at the stu-dent health service. One high. and onelow-fear group was also given specificaction instructions. The instructions con-sisted of a map of the campus with thestudent health service circled. Exampleswere given of routes that could be takento pass the health service as one changedclasses. The student was asked to thinkover his daily schedule and to plan atleast one class change so that he wouldwalk by the health service. None of this

information about routes was novel asthe students were seniors. The idea wasto insert the health act into the ongoingschedule of activities.The data showed that after high-fear

messages, subjects were more frightened,thought shots more important and ex-pressed stronger intentions to be inocu-lated. But there were no differences be-tween low- and high-fear messages foractual shot taking. The drive model ledus to expect that action instructionswould interact with fear level. By makingaction seem easier, they would reducefear and encourage more subjects to getshots after high-fear messages. Action in-structions, however, had no effect uponfear and no effect upon attitudes, butthey did effect action. Roughly 30% ofthe subjects took shots if they had re-ceived either a low- or a high-fear mes-sage and the action instructions versus3% in the fear groups without actioninstructions. The level of fear did notmatter, though some level of motivationwas necessary. No subject took shotsfrom a control group that was givenonly action instructions, nor did anysubject take shots in a control groupthat was not exposed to tetanus in-formation.The above findings, and a replication

by Leventhal, Jones and Trembly (1966),suggest that the action component isseparate not only from the emotional re-sponse but from the attitudinal response.It seems to require special information tolink attitudes to action and the processof tying the two together is not affectedby fear intensity. The result is clearlyout of step with the emotional-drive posi-tion, but quite consistent with the em.phasis the parallel-response model placeson the independence of different be-haviors.A study on smoking behavior by

Leventhal, Watts and Pagano (1967)revealed the long-term importance of ac-tion instructions. The experimental de-sign was similar to that of the tetanus

JUNE. 1971 1217

study; two high-fear conditions, one withspecific instructions, one without, and asimilar pair of low-fear conditions. Acontrol group that received only the spe-cific instructions was also included. Thefear material was based on our standardsurgery film. The action instructions weredesigned to help the student identify spe-cific stimuli which elicited buying andsmoking of cigarettes and to substitutenew acts for each of these cues; e.g. buy-ing a magazine instead of cigarettes,having an excuse for rejecting a cigaretteoffered at a party, etc.The results showed that high-fear mes-

sages produced stronger desires to quitsmoking, and that specific instructionshad no effect upon attitudes or desire toquit. Smoking behavior was followed forthree months. One week after the com-munications, all groups, except the con-trol, showed significant reductions insmoking. Thus, all of the fear messages,high and low and with or without specificinstructions, inhibited smoking behaviorfor the first week of the follow-up. But astime passed many subjects returned totheir pre-communication smoking levels.The great majority of these were the sub-jects who had been exposed to eitherhigh- or low-fear communications withoutspecific action instructions. Those sub-jects who cut down smoking after a fearmessage with specific instructions main-tained their reduced smoking levels fora three month period. Thus, the effect ofthe action instructions is visible onlyafter a relatively prolonged period oftime.The idea of action instructions is not

new; Dorwin Cartwright (1949) talkedabout the role of information in develop-ing behavior structures in addition tomotivational structures. But much morework is needed to understand this processof linking attitudes to action.

Theory and PracticeA theory or model is a conceptual

device that is useful for understanding,

predicting and influencing events. The-ory, whether it is explicit or implicit, isthe basis of practice. It guides us in whatwe say and what we do when we try toinfluence people. In the experiments I'vereported, the independent variables offear, specific instructions, etc. representactions taken by the experimenter to in-fluence his subjects. These actions weredictated by his theory. But how usefulare the experimental results in guidingthe action of the practitioner?

It is not simple to move from experi-mental findings into action programs.One might assume that the experiment isrepresentative of reality; that its com-munications, subjects, atmosphere, re-sults, etc. are on a small scale, butbasically the same as the world of thepractitioner. If we accepted this idea, pro-grams could be developed by simply bor-rowing the procedures used in the ex-periment. But the atmosphere of the ex-perimental situation and the control itprovides make it different from moretypical communication settings (see Hov-land, 1954). The very fact that it is anexperiment makes it different. Thus, wecannot generalize directly from the ex-periment to the field setting. The experi-ment is designed to have high internalvalidity; it compares conditions thatdiffer on specifiable factors. This is neces-sary to test an hypothesis that is derivedfrom a theory or model. But the experi-ment is seldom representative of externalreality. The value of the experiment isthat it serves to increase or to decreaseour belief in the adequacy of the theo-retical model and to add new conceptsor components to the model. The model,however, is far more complete than theexperiment and takes into account agreater range of variables. Thus, it is themodel that is the bridge from the experi-ment to the field setting. It is used bythe practitioner to analyze the relevantfeatures of reality, to appraise the char-acteristics of his sample, and then togenerate his communications. Once he

VOL. 61. NO. 6. A.J.P.H.1218

FEAR APPEALS AND PERSUASION

develops his program he may conduct afield trial, and if it shows that his com-munications change behavior, he can feelconfidence in the model he selected andbe satisfied with his skill in applying it.

In summary, the practitioners test ofa model involves two steps: (1) usingthe model to generate a communicationprogram and (2) testing the effectivenessof the program in changing behavior. Inthe remainder of this paper, I will touchbriefly upon some of the issues involvedin using the drive and parallel models togenerate health programs. I will saynothing about program evaluation as thetopic is covered in many other sources(e.g., Campbell and Stanley, 1963).

The drive model

According to the drive model, thearousal of fear is the source of motiva-tion to cope with danger. Fear motiva-tion can lead one to adapt or to denyones vulnerability to threat. Thus, fearis a necessary antecedent to action andwe must be sure that our health programstimulates at least some level of this emo-tion (see Janis, 1967). How do we stimu-late fear? The model tells us little ornothing about that. Should we surpriseour audience with a frightening explosionand then recommend that they brushtheir teeth? Common sense may lead usto suspect that scaring someone for norelevant reason may make him angry andresistant to our appeals. But the drivemodel gives no clear prescriptions norany abstract rationale to guide us in thesechoices. This is illustrated by the con-tradictory statements that two differentinvestigators have advanced from thefear-drive model, Janis (1967) suggeststhat realistic or relevant fear, fear pro-duced by the danger object, facilitatespersuasion while Sigall and Helmreich(1969) suggest that irrelevant fear fa-cilitate persuasion and that relevant fearproduces defense.

The fear-drive model does point to a

number of variables that will influencewhen and how we stimulate fear. Forexample, we should not arouse fear underconditions where our recommendationcannot be fear reducing. If we make feartoo strong so that it is not reduced byour protective messages, if we permit adelay so that fear is reduced by denial,or if we present ineffective recommenda-tions, fear will lead to denial of vulner-ability and failure to accept recommenda-tions. But we have already reviewed thesehypotheses and I have tried to show thatthey are disconfirmed by the empiricalevidence. At best, the variables elabo-rated by the fear model are irrelevant tofear arousal and have effects that arecompletely independent of the threat levelof the communication. At worst, thevariables produce results that are op-posite to the model's expectations. Ithink it reasonable to conclude, therefore,that the fear-drive model is inadequate tobridge the world of the experiment withthe world of practice. The focus uponfear level provides an inaccurate andoverly limited conceptual base for theconstruction of complex communicationprograms.

Parallel-response model

The parallel-response model is con-cerned with a person's attention to differ-ent sources of information. When I warnsomeone of danger or directly expose himto it, he is confronted with informationfrom the outer world, and with informa-tion from his emotional behavior and hiscoping reactions. The interpretation ofthe external situation, and the appraisalof it as threatening (see Lazarus, 1966),is a precursor of fear arousal and of theinitiation of coping reactions. But thetaking of action seems to depend upona second process of evaluation that isconcerned with the selection and plan-ning of a response program; Lazarus(1966) calls this a process of secondaryappraisal. This selection and planning

JUNE, 1971 1219

process is influenced by informationalfactors which are largely independent ofthe informational variables that effectthe perception of threat and the arousalof fear. What does the model suggestabout the construction of a health pro-gram?

Danger appraisal

To initiate Danger Control activity aperson must be convinced that a threatexists, that it is serious and that it isrelevant to him. Our experiments haveshown, however, that we risk arousingvarious forms of resistance to influenceif we present communications that com-bine vivid information on a threat withclear infornation on one's vulnerabilityto it. This combination appears to stimu-late loss of hope and feelings of resigna-tion and inability to cope with danger.

But our parallel model suggests that wemay be able to differentiate between in-formation that gives rise to fear andinformation that is needed for the ap-praisal of danger. If so, we should beable to maximize Danger Control andeliminate the avoidance behavior andhopelessness that seems part of the proc-ess of Fear Control. For example, strongnegative emotions (including disgust,depression and fear), are likely to bestimulated by vivid scenes of the con-sequences of threats, i.e. pictures ofmutilated accident victims, or lungsurgery operations (Leventhal andTrembly, 1968). Moreover, these scenesof the consequences of danger providerelatively little information about thethreat agent, the cause of danger. Indeed,fear messages often omit information onhow the danger agent operates; informa-tion which convinces the observer thatthe agent is present, that it can attackindividuals like himself, and that whenit does he will find that only some actscan avert serious damage.Of course, if people think that a par-

ticular danger is harmless, if they believe

that they can easily resist or recoverfrom lung cancer, TB, diabetes, tetanus,etc., it may be necessary to illustrate theconsequences of these illnesses when theyare uninterested. This might be donewithout arousing strong fear if simpleverbal statements are adequate to con.vince people of the seriousness of thedanger. But it may be necessary to usepictorial and other more vivid materialto convincingly elaborate upon the con.sequences. In this event, the communica-tion will stimulate strong fear and wewill again risk various forms of fear-con-trolling avoidance behaviors. Fear con-trol seems especially likely if we alsomake the message recipient feel person-ally vulnerable. How can we eliminatethe negative effects of fear?A first suggestion would be to avoid

any direct mention of personal vulner-ability while convincing the individual ofthe severity of the danger. Once he isconvinced of its harmfulness and has anegative attitude toward the threat, wecan approach the problem of personalrelevance. But the best approach to per-sonal relevance would probably be toimply relevance without explicitly statingit. This could be done by concentratingon how the audience can avoid the threat.A second way of handling fear would beto prepare for it. For example, we mightwarn people that most, if not all, viewersare frightened by the message and thatit makes them want to avoid facing theissue. Such preparation might actuallyreduce or inhibit fear. Even if it doesn'treduce fear, it will provide an explana-tion and justification for being fearful."If a message frightens everyone whoviews it, then I needn't be frightened bymy own fear." Fear itself can threatenand alarm us, as it may suggest that weare inadequate to cope with danger.Many experiments show how providingan explanation of a warning for fearserves as a cushion or protective coverand actually reduces the tendency to fleeblindly from the danger setting (examples

VOL. 61, NO. 6, A.J.P.H.1220

FEAR APPEALS AND PERSUASION

include studies by Davison and Valins,1969; Nomikos, et al., 1968; Nisbettand Schachter, 1966).

In fact, most any approach that pro-vides clear expectations about ones emo-tionality, the danger itself, and how tocope with danger, is likely to decreasethe occurrence of fear and facilitate atask-oriented approach in which the sub-ject focuses upon the danger and its con-trol. When we attend to a danger andobtain adequate information about it, itis easier to maintain a realistic perspec-tive and to avoid endowing the threatwith supernatural properties. Moreover,clear expectations about a danger willencourage searching and experimentingwith techniques for controlling it.

Attitudes to action

Convincing our audience of the seri-ousness of a danger, and giving theminformation respecting their personalrelevance provides the first part of aset of cognitive mediators needed foraction. But if people are to act, they musthave a clear idea of the action alterna-tives that are available to them, and de-tailed action instructions to help bridgethe gap between their attitudes and theactual performance of the behavior. Ac-tion instructions direct the individual tospecific environment cues and preparehim to respond to these cues. I am muchmore likely to take a tetanus shot if Iarrange my lunch appointment so that Ipass my doctor's office than if I have tomake a special trip for my inoculation.

Designing action instructions can bea formidable task. Differences in individ-ual life situations and the complexity ofsome health practices (e.g. dieting) mayprohibit presenting such instructions ina mass communication. One should notassume, however, that it is necessary totailor programs to each individual untilone has explored the life situation of asample of individuals drawn at randomfrom the target population. These inter-

views could reveal common features ofpeoples' life space and suggest typicalways as to how, when and where ahealth action could be completed. But,even if it proves impossible to fit actioninstructions to any group of people, itmight be possible to provide people withclues to help them generate their ownaction instructions. For example, wecould suggest that smokers keep a recordof their daily smoking behavior and ofthe details of the situations in whichsmoking occurs. This could help thesmoker classify the various situationswhich seem to control his smoking be-havior. He could then develop plans fordealing with each of these types of situa-tions. The communications could alsoprovide examples as to how people wentabout controlling their behavior in eachof these situations and the smoker couldthen improvise and elaborate a programthat fits his own special needs.

There are times when we persuadepeople of the importance of a danger,only to find them taking actions otherthan those which are recommended. In-dividuals build up relatively complexhabits of self-diagnosis and self-treat.ment. These behavioral patterns are well-learned and well-practiced and they maybe barriers to more meaningful and effec-tive actions. A mother may decide thatthe best protection against polio is stay-ing away from beaches in hot weather,not the taking of polio shots. Effectiveself-generated acts provide useful alterna-tives to the communicator's recommenda-tions. When these acts are inadequatethey can interfere with prevention. Thepractitioner will have to determine, there-fore, the types of self-suggested actionsthat are available to his audience anddevelop communications to point out theinadequacy of those measures whicA areineffective.

Finally, a person's own actions rein-force or undercut his ability to carrythrough a health program. For example,people often expect the initial steps of

JUNE, 1971 1221

dieting or smoking withdrawal to besimpler than later ones. This could beunfortunate as the early success wouldthen be unlikely to increase confidencein one's ability to execute the total pro-gram. But, if a smoker believes that it isespecially difficult to take the first step,to decline a proferred cigarette, his re-fusal can increase his belief that he cancontrol his habit. The actual difficulty ofvarious steps may matter less than theirperceived difficulty. The important pointis that a person will attribute skill orwill power to himself if he believes hehas mastered a difficult act (Bem, 1967).Communicators could judiciously encour-age this process by a carefully designedaction program.

Summary

I have covered a wide range oftheoretical and empirical issues. First, Ihave tried to convince the reader thattheory is of critical importance forpractice; it is the tool for the analysisof the practice situation. Second, I havetried to show that experiments are crit-ical for evaluating theory; they showwhether the theory can survive a com-parison to data. It is clearly to one'sadvantage to approach the analysis ofthe practice setting with a conceptualtool that has met this test, rather thanwith one which has failed. In makingthe above arguments I presented twoparticular theoretical approaches to theproblem of fear in health communica-tions. The first was the traditional drivemodel; a theoretical model that assumesthat the emotion of fear and fear re-duction provides the motivation for theacceptance and the rejection of per-suasive messages. Experiments to test thetheory concentrated upon fear level andwhether the variables that served to re-duce or eliminate fear increased the per-suasive impact of health messages. Thepractice implications of the theory,whether in mass communications (Janis,1967; Leventhal, 1965) or in the re-

sponse of surgical patients to stress ofsurgery (Janis, 1967), is upon the ob-servation and manipulation of fear level.The goal is to provoke sufficient fearto motivate acceptance while preventingdefensive denial or other neurotic fearresponses. As pointed out in the reviewof the empirical literature, the findingsof a lengthy series of studies give littlesupport to the idea that the level of emo-tional motivation and its reduction isresponsible for the acceptance of preven-tive health practices. The model doesnot seem to fit the data.

Finally, I presented a second theoret-ical model which I have called the paral-lel response model. This particular modelshares many properties with so calledinformation processing models. Its pointsare fairly simple; when a person re-ceives a warning communication or con-fronts a danger situation, his appraisalof the threat gives rise to two relativelyindependent processes: Danger Controland Fear Control. Some features of thesetwo processes may appear simultane-ously; a person may be coping withdanger at the very same time that heis frightened. At other times these proc-esses may appear sequentially; a personmay not become frightened until afterhe successfully copes with a danger. Ihave emphasized that controlling dan-ger is largely a cognitive process andthat is seems to have two separablecomponents: 1) a clearly perceived dan-ger agent which one wishes to avoid;2) an awareness of alternative ways ofcoping with the agent and a programof action instructions to actually under-take danger control behavior.The second process I have labeled

Fear Control; it refers to the fact thatthe recognition of danger can give riseto strong emotional behaviors and to thesubjective awareness of unpleasant emo-tion. When emotional cues dominate ex-perience, the individual will be moti-vated to minimize or control fear.

It is important to emphasize that bothof these processes are a consequence of

VOL. 61, NO. 6, A.J.P.H.1222

FEAR APPEALS AND PERSUASION

the appraisal of danger; they are a func-tion of that perception rather thancausally linked to one another. Sinceboth forms of behavior may appear atthe same time or in different patternsover time, I have attempted to specifysome of the factors which will determinewhich of the two processes will be domi-nant. For example, fear control will beencouraged by: 1) ambiguity of thedanger agent, 2) information on con-sequences of the threat which stimulatesfear but gives little information aboutthe agent, 3) inability to devise anactioii program, so that one is helplessand unable to manipulate the danger,etc. Each of these conditions can bebrought about by features of the com-munication, long-standing characteristicsof the personality such as self-esteem,and the timing of the occasion foraction.The alternation between fear and dan-

ger control and the assembling of thevarious attitudes and action instructionsneeded for danger control, occur overtime. Deciding to stop smoking andstopping, deciding to get a tetanus inocu-lation and doing so, etc., do not usuallyoccur at one point in time. Therefore,at any given time, different people willbe at different places in respect to theircommitment to and execution of a pre-ventive health program. The healthpractitioner must be sensitive to the dif-ferent informational needs related tothese inter- and intra-individual differ-ences. There is information for con-vincing people of danger, informationon the relevance of danger, informationon controlling danger, information onconstructing action programs and in-formation to help prevent the arousalof strong avoidance motives. The prac-titioner must direct himself to all ofthese tasks, as each is relevant to thecontinuing process of protective action.

Despite the complexity of the prob-lem, I think it is clear that I have onlytouched upon one aspect of persuasion

for preventive health behavior. Social,economic, and a variety of other per-sonal goals can serve as a basis forhealth action. Fear and Danger are onlya part of the picture. Though my focushas been narrow, it is my hope that theparallel model will raise new questions,encourage new experiments and even-tually help us to communicate moreeffectively with others as to how theycan go about solving their health prob-lems.

ACKNOWLEDGMENT-The research discussedin this paper has been supported hv granttfrom the United States Public Health Service,CH-00371-01 and 0371-02.A detailed review of this literature will

soon be available (Leventhal, H., Findingsand Theory in the Study of Fear Communica-tions. In L. Berkowitz, (Ed.), Advances inExperimental Social Psychology. Vol. 5. NewYork: Academic Press, in press.

BIBLIOGRAPHY

Aronfreed, J. The nature, variety, and socialpatterning of moral responses to transgres-sion. J. abnorm. soc. Psychol., 1961, 63, 223-240.

Bem, D. J. Self perception: An alternative in-terpretation of cognitive dissonance phe-nomenon. Psychol. Rev., 1967, 74, 183-200.

Berkowitz, L. and Cottingham, D. R. The in-terest value and relevance of fear-arousingcommunications. J. abnorm. soc. Psychol.,1960, 60, 3743.

Brock, T. C. and Balloun, J. L. Behavioral re-ceptivity to dissonant information. J. pers.soc. Psychol., 1967, 6, 413-428.

Campbell, D. T. and Stanley, J. C. Experi-mental and quasi-experimental designs forresearch. Chicago: Rand, McNally and Co.,1963.

Cartwright, D. Some principles of mass per-suasion. Selected findings of research in thescale of United States war bonds. Hum.Relat., 1949, 2, 252-267.

Cohen, A. R. Need for cognition and order ofcommunication as determinants of opinionchange. In C. Hovland (Ed.), Order of pre-sentation. New Haven: Yale UniversityPress, 1957, pp. 79-97.

Chu, G. C. Fear arousal, efficacy, and im-minency. J. pers. soc. Psychol., 1966, 4, 517-524.

Dabbs, J. M. and Leventhal, H. Effects ofvarying the recommendations in a fear-

JUNE, 1971 1223

arousing communication. J. pers. soc. Psy-chol., 1966, 4, 525-531.

Davison, G. and Valins, S. Maintenance of self-attributed and drug-attributed behaviorchange. J. pers. soc. Psychol., 1969, 11, 25-33.

Epstein, S. Toward a unified theory of anxiety.In B. A. Mahrer (Ed.), Progress in experi-mental personality research. Vol. 4, NewYork: Academic Press, 1967, pp. 1-89.

Haefner, Don P. Arousing fear in DentalHealth Education. J. Publ. Hlth. Dentistry,1965, 25, 140-146.

Hovland, C. I. Effects of the mass media ofcommunication. In G. Lindzey (Ed.), Hand-book of social psychology. Vol. 2. Cam-bridge: Addison-Wesley Publishing Com-pany, Inc., 1954, pp. 1062-1103.

Janis, I. L. Effects of fear arousal on attitudechange: Recent developments in theory andexperimental reseach. In L. Berkowitz (Ed.),Advances in Experimental Social Psychology.Vol. 3. New York: Academic Press, 1967,pp. 167-222.

Janis, I. L. and Feshbach, S. Effects of fear.arousing communications. J. abnorm. soc.Psychol., 1953, 48, 78-92.

Janis, I. L. and Leventhal, H. Human reac-tions to stress. In E. Borgatta and W. Lam.bert (Eds.), Handbook of personality theoryand research. Boston: Rand McNally andCo., 1968.

Janis, I. L. and Mann, L. Effectiveness of emo-tional role-playing in modifying smokinghabits and attitudes. J. exp. res. Pers., 1965,1, 84-90.

Komzweig, N. D. Behavioral change as a func-tion of fear arousal and personality. Unpub.lished doctoral dissertation, Yale University,1967.

Lazarus, R. S. Psychological Stress and theCoDing Process. New York: McGraw-Hill,1966.

Leventhal, H. Fear communications in the ac-ceptance of preventive health practices.Bulletin of the New York Academy of Medi-cine, 1965, 41, 1144-1168.

Leventhal, H., Jones, Susan and Trembly,Grevilda. Sex differences in attitude and be-havior change under conditions of fear andspecific instruction. J. exp. soc. Psychol.,1966, 2, 387-399.

Leventhal, H. and Niles, Patricia. Persistenceof influence for varying durations of ex-posure to threat stimuli. Psychol. Rep., 1965,16, 223-233.

Leventhal, H. and Perloe, S. A relationship

between self-esteem and persuasibility. J.abnorm. soc. Psychol., 1962, 64, 385-388.

Leventhal, H. and Singer, R. P. Affect arousaland positioning of recommendations in per-suasive communications. J. pers. soc. Psy-chol., 1966, 4, 137-146.

Leventhal, J., Singer, R. and Jones, Susan.Effects of fear and specificity of recom-mendation upon attitudes and behavior. J.pers. soc. Psychol., 1965, 2, 20-29.

Leventhal, H. and Trembly, Grevilda. Nega-tive emotions and persuasion. J. Pers., 1968,36, 154-168.

Leventhal, H. and Watts, Jean. Sources of re-sistance to fear-arousing communications onsmoking and lung cancer. J. Pers., 1966, 34,155-175.

Leventhal, H., Watts, Jean C., and Pagano,Francia. Effects of fear and instructions onhow to cope with danger. J. pers. soc. Psy-chol., 1967, 6, 313-321.

Mann, L. The effects of emotional role-playingon desire to modify smoking habits. J. exp.soc. Psychol., 1967, 3, 334-348.

Mann, L. and Janis, I. L. A follow-up studyon the long term effects of emotional role-playing. J. pers. soc. Psychol., 1968, 8, 339-342.

Mausner, B. and Platt, Ellen S. Smoking: ABehavioral Analysis. In press.

Niles, Patricia. The relationships of suscepti-bility and anxiety to acceptance of fear-arousing communications. Unpublished doc-toral dissertation, Yale University, 1964.

Nisbett, R. E. and Schacter, S. S. Cognitivemanipulation of pain. J. exp. soc. Psychol.,1966, 2, 227-236.

Nomikos, M. S., Opton, E., Averill, J. R. andLazarus, R. S. Surprise versus suspense inthe production of stress reaction. J. pers. soc.Psychol., 1968, 8, 204-208.

Platt, Ellen S., Krassen, Ellen and Mausner, B.Individual variation in behavioral changefollowing role-playing. Mimeo paper: BeaverCollege, 1967.

Sigall, H. and Helmreich, R. Opinion changeas a function of stress and communicationcredibility. J. exp. soc. Psychol., 1969, 5,70-78.

Singer, R. P. The effects of fear-arousing com-munications on attitude change and be-havior. Unpublished doctoral dissertation,University of Connecticut, 1965.

Watts, Jean C. The role of vulnerablity inresistance to fear-arousing communications.Unpublished doctoral dissertation, BrynMawr College, 1966.

Dr. Leventhal is with the Department of Psychology, University of Wisconsin,Madison, Wisc. 53706

This paper was submitted for publication in September, 1969.

1224 VOL. 61, NO. 6. A.J.P.H.