Authors' Note: Mike Stephenson, M.A. (Texas A&M, 1993) is a doctoral candidate in the Department of Communication at the University of Kentucky. Kim Witte, Ph.D. (University of California, Irvine, 1991) is an associate professor in the Department of Communication at Michigan State University. This manuscript is part of the first author's thesis, which was directed by the second author while both were at Texas A&M University. This paper was selected as a Top Three Paper in the Health Communication Division for the 1997 International Communication Association, Montreal, Canada.

Fear appeal messages were designed and tested to ascertain their effectiveness in promoting skin protective behaviors. Theoretically guided by the Extended Parallel Process Model, the fear appeal messages in this study conveyed the seriousness of exposed and unprotected skin for college students in the Southwestern United States, and recommended a series of behaviors that would protect individuals from serious sunburns. The results demonstrated that fear appeals are effective in promoting skin protective behaviors. Specifically, highly threatening fear appeal messages were most effective when combined with a strong efficacy message emphasizing the effectiveness of the recommended behavior.

"Public awareness of the problem of melanoma is increasing. Despite this, a recent national survey..showed that only one third of Americans knew that melanoma was skin cancer and only 50% knew that it was a cancer at all" (Rigel, Friedman, and Kopf, 1996, p. 843).

The amount of time individuals spend in the sun continues to gain attention. Recent findings by the American Academy of Dermatology show that 1 out of every 87 Americans will get skin cancer in their lifetime (Rigel, Friedman, & Kopf, 1996). That means that an estimated 1,000,000 Americans will get some type of skin cancer in the next year. Further, the experts suggest that the skin cancer risk will increase to 1 in 75 by the turn of the century. That's a drastic turnaround from the lifetime risk sixty years ago, which was only 1 in 1500.

The news, however, is not as bleak for those who are willing to protect their skin from the ultraviolet rays. "Exposure to the sun constitutes the major preventable risk factor for skin cancer" (Crane, Marcus, & Pike, 1993, p. 232). Fortunately, most forms of skin cancer can be prevented with detection and treatment (Cody & Lee, 1990). Experts maintain that skin cancer disease "is estimated to be almost 80% preventable and almost 100% curable by the adoption of a correct and vigilant set of primary and secondary preventive behaviors" (Girgis, Sanson-Fisher, & Watson, 1994, p. 77). The task is to get people to adopt such precautions.

The increasing danger to sun exposure opens the door for new, effective, and innovative health promotion campaigns. Such campaigns should not only educate individuals about the damaging and deadly effects of the sun, but also motivate them to take the precautions necessary to save their skin, and in some cases, save their lives. Campaigns may mostly be faced with the prevailing attitude that a tan makes one look healthy (e.g., Boutwell, 1995; Newman, Agro, Woodruff, & Mayer, 1996), when in reality, the ultraviolet ray exposure required to get that tan may be fatal. This investigation reviews the some of the skin cancer awareness efforts thus far, reviews a theory utilizing fear as a method for persuading the public on skin protective behaviors, and tests the effectiveness of fear appeal skin cancer prevention messages.

Promotion campaigns have made an effort to begin teaching the younger population about the dangers of sun exposure. Research has linked melanoma skin cancer with severe, blistering childhood and adolescent sunburns (Weinstock, et al., 1989; Lombard, et al, 1993; Hughes, 1994). In response, Arizona, Texas, and Colorado, among others, have established skin cancer education in either early childhood education programs or public school systems in part because of their sun-intense locations (Crane, Marcus, & Pike, 1993). Arizona's "Sun and Skin Cancer" curriculum provided older-elementary, middle, and junior high school students with lessons about the sun, skin cancer, and skin protective measures. That curriculum featured a coyote named "Sun Cowboy" teaching a white duck called "Paleface" about skin cancer prevention and detection. Most recently, Arizona's older elementary school children are educated about the sun and skin protection with the "Sunny Days, Healthy Ways" curriculum, while preschool children learn to cover up, find shade, and ask for sun-safe things through puppet shows, songs, and skits in the "Be Sun Safe" curriculum (Loescher, Buller, Buller, Emerson, & Taylor, 1995). Evaluation efforts thus far have demonstrated their effectiveness. Programming, however, is not limited to Arizona. In Texas, schools have implemented a skin cancer awareness program for kindergarten through sixth grades. Evaluations of each program vary, but most are successful in increasing knowledge as well as changing attitudes and self-reported skin protective behaviors of children (Crane, et al., 1993).

Several field studies assess skin cancer campaigns aimed at the general population. For example, Lombard, Neubauer, Canfield, and Winett (1991) conducted the "SafeSun" skin cancer intervention at two community swimming pools in Virginia and attained modest results. "SafeSun" promoted the use of shade, hats, sunglasses, zinc oxide, sunscreen, and covering up with shirts through posters, fliers, and peer modeling by the pool's lifeguards. Additionally, sunscreen was given out free and commitment cards pledging to follow SafeSun recommendations were utilized. Results indicated much increased use of the shade and covering up with shirts with minimal change for children or adults for other recommended procedures. Girgis, Sanson-Fisher, and Watson's (1994) evaluation of a worksite intervention for outdoor electrical supply company workers was successful in implementing significantly greater knowledge and behaviors of those participating. Participants received a screening session by a dermatologist plus an educational session about skin cancer prevention. A theoretically based intervention by Rossi, Blais, and Weinstock's (1994) called the Rhode Island Sun Smart Project was implemented on Rhode Island beaches. Grounded in the Transtheoretical Model (Prochaska & DiClemente, 1986), this intervention delivered information to sunbathers at a teen, family, and community beach. Most unique to this study was the use of an ultraviolet light camera which provided sunbathers with pictures depicting skin pigmentation damage from the sun. This method "provided a convenient and highly personalized take-home reminder of the consequences of unprotected exposure to solar ultraviolet radiation" (Rossi, et al., p. 673). The drawback to their efforts is the expense for others wishing to conduct a similar campaign.

Quite clearly, some individuals are being informed about the dangers of exposing unprotected skin to the sun, and some are changing their behaviors. The bigger picture, however, reveals that these campaigns have not made a large contribution toward the decline of skin cancer (Rigel, et al., 1996). "There is no evidence that there are long-term incidence effects yet noted as a result of these programs" (p. 840). For example, a national survey reveals that only about one in two Americans knew that melanoma was cancer, and only one in three knew that melanoma was skin cancer. This lack of knowledge, combined with ozone depletion and lifestyle change means that skin cancer continues to rise. This clarifies the need to reconsider the manner in which Americans are being educated about being in the sun.

In considering campaign alternatives, a review of the campaign literature revealed that none have employed the use of fear to motivate changes in behavior. There are explanations for the neglect of fear in skin cancer prevention campaigns. First, there are a limited number of scientific evaluations of skin cancer interventions in the literature. Much of the scientific literature examines causes of skin cancer, while other articles explain how the skin can be protected. Still other preventive efforts in the form of fliers and pamphlets have not been empirically evaluated for their effectiveness. A second reason for the lack of fear in skin cancer campaigns is that early studies of fear and risk demonstrated that scaring people backfired and therefore was an ineffective means of persuasion (see Janis & Feshbach, 1953; Rogers & Mewborn, 1976). As a result, researchers and health educators have stigmatized fear and avoided using scare tactics for health interventions (see Covello, et al., 1986). Thus, many remain skeptical of the utility of fear for persuasive messages. Recent theoretical advances, however, have demonstrated that the correct use of fear in public health campaign messages can be an effective tool in attaining changes of knowledge and behavior (Witte, 1992a; Witte, 1992b), making fear appeals a viable alternative for health promotion campaigns, such as skin cancer.

Fear appeals are persuasive strategies that "attempt to change our attitudes by appealing to [the] unpleasant emotion of fear" (Rogers, 1983, p. 153). Fear appeals "scare people" through fear and describe "the terrible things that will happen to them" if they fail to comply with the message (Witte, 1992a, p. 3). The most recent fear appeal theory, the Extended Parallel Process Model (EPPM) has clarified the role of fear in persuasive messages by emphasizing the relationship between threat and efficacy (Witte, 1992a). Most promising, however, is the theory's demonstrated effectiveness in attaining behavior change across different populations and health issues. Thus, the EPPM is potentially useful for persuasive skin cancer prevention campaigns.

Fear appeals, according to Witte (1992a), typically contain two parts - a threat component and recommended response. Each component is appraised separately. Individuals first appraise the threat component, which is defined by the message's severity (e.g., "Skin cancer kills people") and its susceptibility (e.g., "You are at-risk for skin cancer because you spend a lot of time in the sun"). If the message initiates perceptions of threat that reach a certain threshold level, then individuals next appraise the efficacy of the recommended response. The level of efficacy perceived from a recommended response shapes how the individual can respond to the threat. Response efficacy (e.g., "Sunscreen will reduce the risk of getting skin cancer") and self efficacy (e.g., "I can use sunscreen to protect myself from the dangers of the sun") are the components of this variable.

What happens when one encounters a fear appeal? According to the EPPM, when individuals are exposed to a fear appeal, they do one of three things. Individuals will either engage in danger control, where they cognitively process the message and take action to avoid the threat. An alternative response is fear control, where individuals emotionally repress the message and ignore the threat. A third alternative is to ignore the message, which typically occurs because the threat is perceived as low.

Danger control, according to the EPPM, occurs when perceived threat is high (e.g., "I can die from skin cancer...") and perceived efficacy is high (e.g., "...but I can effectively take precautions to avoid getting skin cancer."). Danger control processes occur when individuals believe they are easily and effectively able to avoid the threat by following the recommendations in the message. Danger control motivates individuals to think about the ways the message recommends avoiding the threat (e.g., "I can use sunscreen when I go out to the pool today, and I can wear a hat for extra protection for my face"). Danger control processes lead to positive attitudes about the recommended responses (e.g. "It is good and beneficial to me to use sunscreen when I go outside today"), favorable intentions to follow the recommended responses (e.g., "I plan on using sunscreen and covering up when I'm out at the park today"), and a change in behaviors utilizing the message's recommendations (e.g., use of sunscreen, hat, something to cover the skin).

Alternatively, fear control occurs when perceived threat is high (e.g., "I can die from skin cancer...") and perceived efficacy is low (e.g., "...there's nothing I can do about this sunburn now"). Fear control processing occurs when individuals have low self efficacy (i.e., perceive the recommended responses in the message to be too time consuming, difficult, or troublesome) or low response efficacy (i.e., feel that the response is not effective in averting the threat). Essentially, individuals are at a critical point where perceived threat has exceeded perceived efficacy. Instead of focusing how to avoid the threat, individuals begin to assess how they can control their fear. To do this, individuals engage in denial (e.g., "I'm not going to get skin cancer, no one else I know has it"), defensive avoidance (e.g., "I'm just not going to think about these gross skin cancer pictures"), or message manipulation (e.g., "They are just trying to scare me, but it won't work on me").

In sum, when perceived efficacy is low, a high threat fear appeal fails to be effective and promotes fear control processing. The high threat, low efficacy fear appeal leads individuals only to control their fear through denial, message derogation, and defensive avoidance. Such a message fails to motivate individuals to think about ways to avoid the threat. In contrast, when perceived efficacy is high, a high threat message is effective in prompting individuals to avert the threat by considering the recommendations advanced by the fear appeal message. Typically, danger control outcomes include positive attitudes about the recommended responses, increased intentions to perform recommended behaviors, and actual change in behaviors using the recommended responses. Essentially, danger control responses indicate message acceptance; fear control processing evidences message rejection.

The tenets of the EPPM have been validated in some health contexts. The initial test of the model was confirmed, where a high threat, high efficacy message depicting the threat of AIDS and condom efficacy increased condom usage, while a high AIDS threat, low condom efficacy message led to fear control responses and failure to use condoms (Witte, 1992b; 1994). Since then, Witte, Berkowitz, Cameron, and McKeon (1995) successfully attained danger control outcomes by using high threat, high efficacy messages to prevent the spread of genital warts and promote condom usage. While not tests of the model, the EPPM has been used to analyze breast cancer messages (Kline, 1995), as a means of campaign design for tractor safety (Witte, et al., 1993) radon awareness (Witte, Berkowitz, Cameron, McKeon, Lapinski, & Liu, 1996) teenage pregnancy (Witte, 1997), in assessing and developing HIV/AIDS prevention programs in Africa (Witte, Cameron, & Nzyuko, 1995; Cameron, Witte, & Nzyuko, in press), and in message processing (Stephenson & Witte, 1996).

This study proposes testing the EPPM in the context of skin cancer. Note that there are relatively few studies that actually test and confirm the tenets of this relatively new theory. An additional experimental test of this model simply contributes toward theoretical validation and replication to measure the theory's robustness. Given the shaky past of fear appeal research as well as the early signs that this model is effective is promoting changes in cognitive and behavioral outcomes, such a study is desirable. While meta-analyses by Boster and Mongeau (1984) and Witte and Allen (1996) show, in general, that message acceptance increases with aroused fear and heightened perceptions of threat, not until Witte's (1992a) model emerged was there an attempt to reconcile the inconsistent fear appeal literature that provides a viable rationale why this was occurring. Since Witte's explanation and rationale for previous inconsistent outcomes from fear appeal experiments is initially promising, this study seeks to further test Witte's assertion that fear appeals can work if developed effectively. The context for this study, skin cancer messages, is one that is gaining prominence as a health threat, and this research could potentially be important in future skin cancer health campaigns.

Thus, the purpose of the present study is twofold. First, in light of past failures with fear appeal messages, a need exists to offer further validation of the EPPM as a viable means for health campaign message design. The tenets of the EPPM appear thus far to offer answers to the numerous past inconsistencies in the fear appeal literature, but the EPPM needs to be tested to determine its applicability to various health contexts. Second, this study seeks to assess the effectiveness of these theoretically-based fear appeal messages within the context of skin cancer. No skin cancer messages or existing campaign literature has been tested in the fear appeal domain. Given the continuing increase in the number of skin cancer cases being reported each year, this investigation focused on awareness and prevention among young adults living in the sun-intense region of South Central Texas. Young adults between 18 and 24 spend the most time in the sun for both leisure and tanning purposes (Newman, et al., 1996). "With more than 624 miles of coastline, a population attracted to outdoor activities, and a social norm that promotes tanning as a symbol of health and youth, Texas needs comprehensive public education and rigorous behavioral research programs supporting risk-reduction strategies - programs that are long overdue" (Boutwell, 1995, p. 657). This investigation offers a preliminary assessment of one possible preventive program that embraces fear as a means of reducing skin cancer risks in Texas among young adults.

Thus, fear appeal skin cancer messages were designed to increase perceptions of the threat of skin cancer and offered recommendations to individuals on how to effectively reduce the threat of skin cancer from occurring. The theoretical tenets of the EPPM suggest that danger control outcomes, which lead to message acceptance, are a product of high threat, high efficacy messages. Therefore,

H1: High threat, high efficacy messages will produce more positive attitudes, intentions toward skin cancer protection measures than high threat, low efficacy messages.

On the contrary, fear control responses lead to message rejection through denial, defensive avoidance, and message derogation. The EPPM posits that fear control outcomes are a result of individual's perceptions of low efficacy when combined with a high threat message. Therefore,

H2: High threat, low efficacy messages will produce greater defensive avoidance, perceived manipulation, and message derogation than high threat, high efficacy messages.

Additionally, separate studies have revealed the role of threat in a fear appeal. For instance, if a message is perceived to be of little threat, then individuals do not respond to the fear appeal. As an example, individuals with very dark skin who tan easily may perceive skin cancer to be an issue only for those who have light skin that burns easily. These individuals, then, may not feel threatened by the threat of skin cancer and are not motivated to take any action. Witte (1992a, 1992b) has confirmed that when perceived threat is low, individuals are not affected by the message. Additionally, meta-analyses by Boster and Mongeau (1984) and Witte and Allen (1996) have found that message processing increases with perceptions of threat. This suggests that the greater perception of threat, the more likely one is to respond to the message. Witte and Allen (1996) suggest that "while efficacy improves the effectiveness of fear appeals, it appears that threat drives the persuasive relationship" (p. 23). To give validity to the notion that threat is persuasive and motivates behavior change, it is predicted that,

H3: The more one perceives the threat of skin cancer, the more one will be motivated to accept the message's skin protection recommendations.

Consistent with the research goal of investigating theory-based fear appeal messages for skin cancer prevention is the desire to assess the logistics of skin cancer message design. Specifically, when designing a fear appeal message, what is the role of high threat pictures in conveying the threat and recommending preventive behaviors? No investigation has empirically determined the impact of pictures in fear-based messages, although Witte (1993) has suggested that "graphic and vivid" pictures are especially useful in conveying and making the threat substantial (p. 148). Some research on vividness supports this claim. For example, Nisbett and Ross (1980) note that "vivid information is more likely to be stored and remembered than pallid information is" (p. 45). Likewise, they define vividness as being emotionally interesting, concrete and imagery provoking, and proximate. Is it possible that pictures containing these characteristics enhances the persuasiveness of fear appeal messages? Intuition might suggest they would; previous research would not. Empirical investigations offer only mixed and inconclusive findings that do not provide any easy answers.

Still, the idea that carefully selected and validated graphic and vivid pictures could provide emotional stimulation that is relevant, concrete, and imagery provoking should not be discounted, although some have raised valuable questions about whether or not vividness enhances persuasiveness. Taylor and Thompson (1982) say that vividness is elusive and have concluded that the vividness effect is weak and inconclusive, citing problems with conceptualization, construct validity, faulty operationalizations, and laboratory effects. Taylor and Thompson (1982) say that vividness effects may be justified under certain conditions, such as differential attention, or message characteristics such as source credibility (Witte (1993) would consider source credibility a possible confound in fear appeal messages).

While agreeing that Taylor and Thompson's (1982) arguments should not be discounted, Sherer and Rogers (1984) suggest the critics would have reached different conclusions if vividness were operationalized according to the three criteria specified by Nisbett and Ross (1980). In fact, Sherer and Rogers (1984), in the only fear appeal study assessing vividness and persuasibility, determined that vividness played a significant role in one's intentions to moderate alcohol use. Additionally, the work of Reyes, Thompson, and Bower (1980) is acknowledged to show evidence in support of the vividness effect. Likewise, and especially relevant to this study, are portions of the empirical studies on vividness by Shedler and Manis (1986) noting that photographs contribute to vividness by enhancing recall and judgment. Thus, while the extant literature has a tendency to question vividness as a valid phenomenon for inquiry, there is sufficient evidence to probe this issue further in this fear appeal study "because of the conceptual importance of the vividness hypothesis and the fact that it is widely accepted (despite the Taylor & Thompson review)" (Shedler & Manis, 1986, p. 26).

Therefore, this study suggests investigating the role of pictures in skin cancer fear appeals for much the same reason Sherer and Rogers (1984) assessed vividness in their text-based fear appeals - because "no one has tracked...into the wild terrain of fear-based attitude change" (p. 323). There is virtually no insight available on the role of threatening pictures in fear appeals. The void of research in this specific domain raises enough questions to warrant an investigation of the picture-based vividness effect in fear appeal messages. Thus, consistent with Nisbett and Ross' (1980) assertion that vivid information should have more impact on attitudes than pallid information, as well as with Witte's (1993) guidelines for developing persuasive fear appeals, it is predicted that,

H4: Fear appeals containing pictures will lead to stronger perceptions of fear and threat, and thus greater message acceptance.

Design and Subjects

To accomplish the goals of this study, message type and efficacy were manipulated in a 2 (text only, text-picture combination) X 2 (high efficacy, low efficacy) factorial design. Only high threat messages were used because low threat messages have been shown to produce no effect (Witte, 1992a; note, however, that low threat messages are used and explained later during pretesting and message validation). The dependent variables were attitudes toward using skin-protective behaviors, intentions to use skin-protective measures, defensive avoidance, perceived manipulation, and message derogation. The dependent variables were assessed immediately after subjects read the message.

Ninety two undergraduates (41 males and 51 females) received extra credit for participating in the study. Of those participating, 13% were nonwhite and 2% were over 26 years old. The median age was 21. Additionally, 90% indicated growing up in Texas with 99% spending most of their time over the past three years in Texas. Of those participating, 98% feel the Texas sun in generally intense. Regarding leisure time, 92% indicated they participate or observe outdoor sporting activities. However, only 29% usually or always use sunscreen when they know they are going to be out in the sun, while 44% sometimes use sunscreen and 28% never or rarely use sunscreen outside. These trends are consistent with other studies examining amount of time spent in the sun (e.g., Newman, et al., 1996).

Procedure

Subjects were run in groups ranging up to six persons and were randomly assigned to one of the four experimental conditions, with approximately 23 in each condition. Participants were exposed to combinations of the independent variables by reading written messages. Participants were told that they were "evaluating messages for skin cancer advertisement campaigns" and that their input is needed for refinement of the messages.

After reading the skin cancer messages, subjects immediately completed a posttest questionnaire. Anonymity was ensured in order to obtain honest answers. When participants completed the posttest questionnaire, they were debriefed as to the purpose of the study, what conditions they experienced, and given skin cancer information provided by the American Cancer Society. Subjects indicating concern about potential skin cancer were referred to the campus' student health center.

Stimulus Materials

Subjects read a pre-validated message manipulation consisting of two parts - a threatening message and a message about the effectiveness of skin protective behaviors. The threat messages emphasized (a) the target population's susceptibility to skin cancer and (b) the severity of skin cancer with graphic language. Two versions of a threat message were used - one version containing only written text, while the other combined four pictures of individuals in the advanced stages of skin cancer with the written text from the other message. As the text discussed the threat of skin cancer, pictures of individuals in the advanced stages of skin cancer appeared on the page opposite of the text.

Tagged to the end of the high threat base message were one of two efficacy messages. The high efficacy message emphasized the prevention of skin cancer by emphasizing the effectiveness of sun block in preventing skin cancer, as well as the ease with which sun block can be used. The low efficacy message discussed detection, specifically stating that while sun block is effective in preventing any future skin damage, it is impossible to undo any past skin damage.

Message Validation

Before the main study could begin, all message components were pretested. Two separate stimulus components were selected for use in composing the skin cancer messages - written text and pictures. To ensure that each stimulus agent conveyed what it was originally designed to convey (i.e., high threat picture is actually highly threatening, high efficacy message conveys high response and self-efficacy, etc.), the materials were tested and validated in separate studies. Any necessary revisions were then made before combining the materials into a full message that included text and pictures.

Written Text. The written text of the message was factual with information taken from pamphlets published by the American Cancer Society ("Melanoma/Skin Cancer, You can recognize the signs," "Facts on skin cancer," and "Fry now. Pay later"). While only versions of high threat text were used in the actual study, both high and low threat messages were written to assure that the high threat induced stronger perceptions of threat than the low threat messages. Text in the high threat and low threat message varied in accordance with the definition of threat. The high threat message was more personalistic and used more vivid and threat-inducing words than the low threat message. The results confirmed that the high threat messages conveyed distinctly stronger perceptions of threat, severity, and susceptibility than did the low threat messages. Specifically, the high threat message (M = 5.63) induced significantly stronger perceptions of threat when compared to the low threat message (M = 4.21), t(1,165) = 8.89, p < .001.

Text in the high efficacy and the low efficacy message varied as well. The high efficacy message sought to increase individual's response and self-efficacy about sunscreen by emphasizing its effectiveness, its ease of use, its low cost, and its benefits for safer tanning. On the contrary, the low efficacy message discussed sunscreen's inconvenience, messiness, inability to stay on in the water, and extra time it takes to put on the sunscreen. Message pretesting validated that each efficacy message was significantly different from the other. The high efficacy message (M = 5.43) induced significantly stronger perceptions of efficacy when compared to the low efficacy message (M = 3.97), t(1,165) = 9.86, p < .001.

Pictures. The pictures used in the high threat message condition were of actual skin cancer patients. Ten skin cancer pictures, five conveying high threat and five conveying low threat, were selected from Atlas of Skin Cancer (Du Vivier, 1991) and Atlas of Tumors of the Skin (Kopf, 1978). While low threat pictures would not be used in the main study, it was necessary to determine if the high threat pictures induced significantly higher levels of threat than the low threat pictures.

High threat pictures operationalized threat by conveying a sense of severity. The items selected included depictions of individuals with red, open skin, yellow infected excretions on the forehead, an exposed nasal cavity and eye socket where the skin cancer has eaten away the skin, and an ear that is infected with dark, black scars and is decaying away. Also included in the high threat picture set was a "before-after" picture of Bridgette Bardot showing her in her younger years with unwrinkled, undamaged skin next to a picture of her many years later with very wrinkled, damaged, leathery skin. Low threat pictures showed moles of different shapes and colors in five different locations, including the cheek, head, chest, arm, and thigh.

Results indicated that individuals perceived the high threat pictures to convey a greater sense of fear, severity, and susceptibility than the low threat pictures. The overall means were computed for each picture to assess the threat rating. The high threat pictures and their respective means are included in Table 2. Duncan multiple-range tests (p < .05) revealed that the high threat pictures induced significantly stronger perceptions of threat than two of the five low threat pictures. Three of the low threat pictures were not significantly different from the other picture means. One high threat picture, the Bridgette Bardot image, fell within the same subset as two of the low threat pictures. Therefore, this picture was not included in the completed message. A total of four high threat pictures were selected for use from this validation.

Seven-point Likert-type response formats were used to assess participants' perceptions for each item, except where noted otherwise. Measures for internal consistency are given in Table 3. All measures are described briefly below.

Demographic Variables.

Demographic variables assessed included gender, ethnicity, age, and scholastic classification. Additionally, to account for possible confounds, the subject's home state, involvement with outdoor sporting events, outdoor leisure activities, outdoor employment, time spent in a tanning bed, skin complexion, and personal and family history of skin cancer were all assessed.

Perceptions

Consistent with previous tests of the EPPM, threat was measured separately through severity and susceptibility, and then combined to create the overall construct of threat. Similarly, efficacy is measured separately by self efficacy and response efficacy, but combined to assess the overall construct of efficacy (see Witte, 1992b). These conceptually distinct variables were measured separately and served as manipulation checks. Severity was assessed with 1) skin cancer is a dangerous disease, 2) skin cancer is a severe disease, and 3) skin cancer is a serious disease were used. Susceptibility was measured with 1) I am likely to get skin cancer sometime during my life, 2) I will get skin cancer sometime during my life, and 3) I am not susceptible to skin cancer (reverse scored). Self-efficacy was operationalized by the questions 1) I am able to use sunscreen and other sun-protective measures to prevent my getting skin cancer, 2) using sunscreen and other sun-protective measures is easy, 3) I think that using sun-protective measures and putting on sunscreen is inconvenient (reverse scored), and 4) I am capable of using sunscreen and other sun-protective measures to prevent my getting skin cancer. Response efficacy, the individual's belief that the recommended behavior is effective, was measured with 1) my using sunscreen and other sun-protective measures will keep me from getting skin cancer, 2) sunscreen and other suggested protective measures are effective in preventing skin cancer, and 3) I think that sunscreen and other suggested protective measures prevent skin cancer.

Fear.

To assess an arousal of fear, participants rated ("not at all" to "extremely") the following five mood adjectives: frightened, tense, anxious, comfortable, nervous.

Dependent Variables.

The dependent variables were attitudes toward skin protective behaviors, intentions to use skin protective behaviors, defensive avoidance, perceived manipulation, and message derogation. Each dependent variable was measured at least three times. See Table 3 for items and reliability coefficients.

Overview.

Hypotheses 1, 2, and 4 were tested with Analysis of Variance. To get an accurate assessment of the influence of threat on outcomes for Hypotheses 1, 2, and 4, we removed the influence of demographic or prior experience variables on the outcomes prior to assessing how the theoretical variables influenced outcomes. Thus, the influence of each demographic/prior experience variable on outcomes was assessed and when significant, its effects were removed through Analysis of Covariance in order to adequately assess the influence of our theoretical variables of interest. Finally, Hypotheses 3 was tested with Pearson correlations to examine the relationship between constructs.

Attitudes. A significant main effect of efficacy on attitudes was detected, F(1,44) = 10.35, p = .002, (r² = .19). Under high threat conditions, individuals reading a high efficacy message had more favorable attitudes toward skin protective behaviors (M = 6.45) than did individuals reading a low efficacy message (M = 5.56). This confirms the hypothesis that high threat, high efficacy messages lead to danger control responses.

Intentions. A significant main effect of efficacy on intentions was detected, F(2,39) = 4.39, p = .02 (r² = .18) (while controlling for the number of leisure hours spent in the sun). Under high threat conditions, individuals reading a high efficacy message had stronger intentions of following recommended skin protective measures (M = 4.91) than did individuals reading a low efficacy message (M = 4.20). This also confirms the hypothesis that high threat, high efficacy messages lead to danger control responses.

Perceived Manipulation. A significant main effect for efficacy on perceived manipulation was detected, F(1,44) = 6.73, p = .01 (r² = .13). Under high threat conditions, individuals reading a low efficacy message perceived more manipulation (M = 4.20) than did individuals reading a high efficacy message (M = 2.97). This confirms the hypothesis that high threat, low efficacy messages lead to fear control responses.

Message Derogation. A significant main effect for efficacy on message derogation was detected, F(1,44) = 4.37, p = .04 (r² = .09). Under high threat conditions, individuals reading a low efficacy message perceived more derogation of the skin cancer message (M = 4.16) than did individuals reading a high efficacy message (M = 2.99). This also confirms the hypothesis that high threat, low efficacy messages lead to fear control responses.

Defensive Avoidance. No significant main effect for efficacy was detected on defensive avoidance. Subjects reading a low efficacy message indicated slightly less defensive avoidance (M= 5.80) than did individuals reading a high efficacy message (M = 6.07). This fails to support the hypothesis that high threat, low efficacy leads to fear control responses.

Threat had a positive relation to attitudes, where the more perceived threat, the more favorable their attitudes toward skin protective behaviors (r = .25, p = .01, n = 92). Threat also had a positive relationship to intentions. Individuals perceiving more threat indicated greater intentions to complying with recommended skin protection behaviors (r = .19, p = .03, n = 91). As predicted, threat had an inverse relation to perceived manipulation, where the greater the perceived threat, the less individuals felt manipulated by the message (r = -.22, p = .02, n = 92). Additionally, threat had an inverse relation to message derogation. Individuals perceiving more threat indicated less subjects derogated the message (r = -.20, p = .03, n = 92). Unexpectedly, threat was found to have a positive relation to defensive avoidance. The more threat individuals perceived from the message, the more people defensively avoided skin protective measures (r = .42, p = .001, n = 91). The results mostly confirm hypothesis four indicating that the more threat individuals perceive, the more likely they are to accept recommended behaviors.

Fear. No significant main effect for message type was detected on fear. Individuals reading the message with text and pictures (M=4.12) perceived higher levels of fear than did individuals reading the text-only message (M=3.86). Results do not confirm the hypothesis.

Threat. No significant main effect for message type was detected on threat. Individuals reading the message with text and pictures (M=5.36) perceived relatively similar levels of threat than did individuals reading the text-only message (M=5.15). Results fail to confirm the hypothesis.

Attitudes. No significant main effect for message type was found on attitudes. Individuals reading the message with text and pictures (M=6.00) indicated slightly more favorable attitudes toward skin protective responses than did individuals reading the text-only message (M=5.81). Results fail to confirm the hypothesis.

Intentions. A significant main effect for message type was detected for intentions, F(2,87) = 4.55, p = .01 (r² = .09) (while controlling for those individuals presently spending time in a tanning salon). Individuals reading the message with text and pictures indicated significantly stronger intentions (M=4.65) of using recommended skin protective measures than did individuals reading the text-only message (M=4.19).

Perceived Manipulation. A significant main effect for message type was found on perceived manipulation, F(1,90) = 5.35, p = .02 (r² = .06). Individuals reading the message with text and pictures perceived significantly more manipulation (M=3.59) than did individuals reading the text-only message (M=2.78).

Message Derogation. A significant main effect for message type was found on message derogation, F(1,90) = 4.93, p = .03 (r² = .05). Individuals reading the message with text and pictures felt the message was more derogated (M=3.57) than did individuals reading the text-only message (M=2.74).

Defensive Avoidance. No significant main effect for message type was found for defensive avoidance. Individuals reading the message with text and pictures (M=5.47) indicated only slightly higher levels of defensive avoidance than did individuals reading the text-only message (M=5.43). Results fail to confirm the hypothesis.

Consistent with the initial goals of this research, this study has validated the major tenets of the EPPM as well as determined that fear appeals are a viable alternative for skin cancer prevention campaigns. This investigation suggests that fear appeal messages constructed according to the theoretical tenets of the EPPM can be effective in generating message acceptance, and that such messages can be effective in skin cancer campaigns. Scary messages describing and depicting the terrible things that can happen to one's unprotected skin were successful when combined with a high efficacy message offering easy and effective means of reducing that threat.

While most fear appeal studies suffer from conceptual confounding and inconsistent results, this investigation contributes to a trend that takes a more positive outlook on fear appeal research. Support for fear appeals here is manifest in two distinct ways. First, the overall results of this study contribute to our understanding of the usefulness of fear appeals. While fear appeals have a history of inconsistent results, working some of the time but failing others, our findings (specifically Hypothesis 3) generally confirm meta-analyses which have determined that the greater one perceives a threat to be to one's health, the more likely the individual is to respond to the message (see Boster & Mongeau, 1984; Witte and Allen, 1996). Second, the overall results of this study offer validation of the EPPM in a new context. Theoretically, this study determined that the major tenets guiding the EPPM are viable in a previously uninvestigated health context.

High threat messages that portray the consequences from a behavior can be effective as long as the message also includes a high efficacy component. As demonstrated in Hypothesis 1, the high threat-high efficacy skin cancer messages elicited danger control processes. While respondents indicated high levels of threat, they also cognitively appraised the message's recommendations. As a result, readers formed favorable attitudes about skin protective recommendations and indicated their intent to follow the recommended guidelines.

In contrast, Hypothesis 2 demonstrated that high threat-low efficacy messages elicited fear control responses. Because individuals felt that there was very little of anything significant that they could do to prevent skin cancer (low self efficacy), or that there is very little effectiveness to the protective measures that exist (low response efficacy), individuals responded in ways that demonstrated their attempts to control the fear elicited by the scary message. Characteristic of fear control responses, individuals rejected the message that skin cancer could affect them in ways described in the text and shown in the pictures. As a result, individuals felt the fear appeal was manipulative and blown out of proportion. These results confirm perceptions of efficacy must be higher than perceptions of threat for fear appeals to be accepted by the receiver of the message.

Hypotheses 1 and 2 clarify that for a fear appeal to be effective, the efficacy component is critical in determining which type of response, danger control or fear control, is elicited. As Witte (1992a) explains, a scary message will not be effective unless one's beliefs about the effectiveness of the recommended response, as well as the ease with which individuals can do the response must exceed one's perceptions of the harm they will incur by not complying with the threat. Because fear appeal campaigns often fail to address the perceived efficacy issue, individuals must rely on existing perceptions of efficacy with regard to the recommended response. Thus, it is critical that fear appeal message give special attention to the role of perceived efficacy.

Overall, the fear appeals were successful, although certain aspects of this study did not turn out as predicted. Hypothesis 4, which sought practical insight into the construction of fear appeals messages, was not confirmed. Interestingly, the pictures in these skin cancer fear appeal messages were generally no more persuasive than the messages containing only written text. Results indicate that individuals felt manipulated and that the pictures were blown out of proportion. There are possible explanations for these findings. First, individuals may have perceived the pictures as too unrealistic, as they were rather gruesome depictions of individuals in the advanced stages of skin cancer. Four pictures may have been too much for this message, where one or two pictures may have been sufficient. Second, even though the individuals in the pictures were very deformed from the cancer, it was clear that they were much older than the age of those reading the messages. Thus, it may have been more realistic to include pictures of individuals in the same age range that had been seriously affected by skin cancer. These pictures would have been more congruent with the text of the message, which described a fatal skin cancer incident of a college sophomore.

Thus, while pictures can be included in campaign literature, their role may contribute more toward the literature's aesthetic design than toward furthering the effectiveness of the fear appeal itself. Our results, however, indicate that the pictures in this study's messages did not contribute toward the overall effectiveness of the high threat fear appeal message. As a result, this study offers no clarification for the already puzzling and debatable literature on vividness.

Also of issue in this study is the measure of defensive avoidance and its erratic results. As a dependent variable, defensive avoidance in both Hypothesis 2 and 4 did not score as predicted, and its correlation with threat was substantial and unexpected. Re-examination of the items in the questionnaire for defensive avoidance may shed light into these responses. The messages were constructed so that the threat message was read first, and the efficacy message last. Since items in the questionnaire all began with "when I was first reading this message," the questions called upon the respondents to think about their feelings when they first began reading the highly threatening text, and for some, looking at threatening pictures. Initial impressions by individuals may simply have been more prone to defensively avoid the message, leaving them to want to ignore or avoid the scary message.

Additionally, the threat component may draw a more affective response, given the wording of the text, and for some, the gruesome pictures. The efficacy component may have elicited more cognitive responses, since the efficacy message presents ways that individuals can or cannot respond to the initial threat. Thus, the defensive avoidance questions may be prompting cognitive responses from a time when most individuals were responding affectively to the distressful, shocking threat pictures. As a result, the wording of the defensive avoidance questions was incorrect, leaving the measure to be more unstable as indicated by reliability measures and inaccurate in measuring the appropriate construct.

Nevertheless, defensive avoidance is a difficult concept to assess. Defensive avoidance indicates that individuals are either distorting or ignoring the incoming threatening information. Its nature as an internal perceptual defense makes it difficult to quantitatively measure. Witte (in press) sees the issue of measuring this construct equally as difficult, and suggests that a delayed measure of defensive avoidance may be the best way to determine whether information from the fear appeal's threat or efficacy components was blocked by an unconscious process. Additionally, cross-validation procedures, such as thought listing tasks, could be useful in assessing defensive avoidance. These conclusions imply that future studies must go beyond the immediate posttest assessment of message effects, carefully word questionnaires so as to not prompt participants to think about a specific point in the message, and include thought listing in order to improve the defensive avoidance measure.

Skin Cancer Prevention

It appears that fear appeal messages hold great potential for promoting skin protective recommendations and ultimately reducing the number of skin cancer incidences nationwide. High threat skin cancer messages should focus on making individuals believe that they are increasingly susceptible to skin cancer. While individuals with fair skin continue to be at greatest risk, darker skinned individuals tend to spend more time unprotected in the sun and are continually increasing their risk for getting skin cancer too. Additionally, high threat skin cancer messages should focus on making individuals believe that skin cancer is serious, and in some cases fatal. Increasing these perceptions will motivate individuals to act. Vivid message text accompanied with several pictures of individuals in the age range of the target audience may be most successful in increasing threat, and thus motivating individuals to protect themselves.

In order to channel this motivation in the proper direction, all fear appeal campaigns should be accompanied by high efficacy messages regarding the recommended response. For example, a campaign seeking to increase use of sunscreen should include strong efficacy messages that emphasize how easy, inexpensive, and effective sunscreen is for the user. Additionally, efficacy messages can emphasize new long lasting waterproof sunscreen, or that some zinc oxide sunscreens are colorful and fashionable for outside leisure activities. Finally, not only efficacy messages about sunscreen, but all recommended responses should seek to increase response efficacy (i.e., the effectiveness of the recommended response) and self-efficacy (i.e., the ease and ability of the individual to perform the recommended response) to maximize the persuasive impact of the efficacy message.

Limitations

This research does suggest that the EPPM is successful for skin cancer messages. But the design of this study does not allow us to suggest or offer support for any long-term effects from the skin cancer fear appeal messages. While the EPPM has been supported in longitudinal tests (e.g., Witte, 1994), this cross-sectional intervention only reflects how participants felt about protective skin cancer recommendations immediately after reading the message. Additionally, this study was conducted in a lab setting. The trade-off is that this study's clean experimental test of anxiety-provoking messages allowed for stronger internal validity, an initially important element in confirming a theory, especially one in the fear appeal domain. There is mounting evidence, including the results of this study, that fear messages designed with the guidelines of the EPPM work in the lab. However, future studies should aim to take fear appeal messages into an applied field research setting and build in a longitudinal measure. Thus, the results should be interpreted cautiously.

Second, the generalizability of this study is limited to the young adult population. Our findings can only be extended to similar demographic populations. While the time spent in the sun by this group leaves them at a high risk of getting skin cancer, research on other demographic populations is equally desirable.

Finally, this persuasion study is limited by the absence of a true control group. This study allows for comparisons only between high threat-high efficacy messages and high threat-low efficacy messages. It is not clear how a separate cohort of subjects who had not read a fear appeal message about skin cancer protection would have responded in comparison with the two other message groups used here. Nevertheless, we feel that this initial research is useful in suggesting that fear appeals can be effective in skin cancer interventions, but suggest that future research incorporate the deficiencies of this study.

In sum, this preliminary research suggests that fear appeals offer a promising alternative for developing future skin cancer prevention campaigns. Our findings add to the recent trail of research (e.g., Witte, 1992b; Witte, Berkowitz, Cameron, & McKeon, 1995) that shows fear appeals are effective and useful when designed following the tenets of the EPPM. As Hypothesis 1 demonstrated, high threat messages portraying the danger of skin cancer can be successful when incorporated with a strong efficacy component. These messages promoted positive attitudes toward the recommended behaviors, and the results have suggested that individuals are more likely to comply with the message's recommendations. To the contrary, the results from Hypothesis 2 have confirmed that high threat skin cancer messages with a low efficacy component do not work, and merely lead individuals to believe the message is manipulative in nature and blown out of proportion. By following the guidelines of the EPPM and incorporating strong efficacy components into scary messages, fear appeals appear to be an effective vehicle for use in future skin cancer prevention campaigns.

Boster, F. J., & Mongeau, P. (1984). Fear-arousing persuasive messages. In R.N. Bostrom & B.H. Westley (Eds.), Communication yearbook 8, (pp. 330-375). Newbury Park, CA: Sage.

Boutwell, W. B. (1995). The Under Cover skin cancer prevention project: a community-based program in four Texas cities. Cancer, 75, 657-660.

Cameron, K., Witte, K., & Nzyuko, S. (in press). Perceptions of condoms and barriers to condom use along the Trans-Africa Highway in Kenya. In W. Elwood (Ed.) Power in the Blood: AIDS, Politics, and Communication.

Cody, R., & Lee, C. (1990). Behaviors, beliefs, and intentions in skin cancer prevention. Journal of Behavioral Medicine, 13(4), 373-389.

Covello, V., von Winterfeldt, D., & Slovic, P. (1986). Risk communication: A review of the literature. Risk Abstracts, 3, 171-182.

Crane, L. A., Marcus, A. C., & Pike, D. K. (1993). Skin cancer prevention in preschools and day care centers. Journal of School Health, 63, 232-233.

Du Vivier, A. (1991). Atlas of skin cancer. London: Lippincott.

Girgis, A., Sanson-Fisher, R. W., & Watson, A. (1994). A workplace intervention for increasing outdoor workers' use of solar protection. American Journal of Public Health, 84, 77-81.

Hughes, A. S. (1994). Sun protection and younger children: Lessons from the "Living With Sunshine" program. Journal of School Health, 64, 201-204.

Janis, I. L., & Feshbach, S. (1953). Effects of fear arousal on attitude change: Recent developments in theory and experimental research. In L. Berkowitz (Ed.), Advances in experimental social psychology (Vol. 3, pp. 166-225). New York: Academic Press.

Kline, K. N. (1995, November). Applying Witte's extended parallel process model to pamphlets urging women to engage in BSE: Where are the efficacy messages? Paper presented at the annual meeting of the Speech Communication Association, San Antonio, Texas.

Kopf, A.W. (1978). Atlas of tumors of the skin. Philadelphia: Saunders.

Loescher, L. J., Buller, M. K., Buller, D. B., Emerson, J., & Taylor, A. M. (1995). The evolution of skin cancer prevention education for children at a comprehensive cancer center. Cancer, 75, 651-656.

Lombard, D., Neubauer, T. E., Canfield, D., & Winett, R. A. (1991). Behavioral community intervention to reduce the risk of skin cancer. Journal of Applied Behavioral Analysis, 24, 677-686.

Newman, W. G., Agro, A. D., Woodruff, S. I., & Mayer, J. A. (1996). A survey of recreational sun exposure of residents of San Diego, California. American Journal of Preventive Medicine, 12, 186-194.

Nisbett, R., & Ross, L. (1980). Human Inference: Strategies and Shortcomings of Social Judgment. Englewood Cliffs, NJ: Prentice-Hall.

Prochaska, J. O., & DiClemente, C. C. (1986). Toward a comprehensive model of change. In W. R. Miller & N. Heather (Eds.), Treating addictive behaviors: Processes of change (pp. 3-27). New York: Plenum Press.

Reyes, R. M., Thompson, W. L., & Bower, G. H. (1980). Judgmental biases resulting from differing availabilities of arguments. Journal of Personality and Social Psychology, 39, 2-12.

Rigel, D. S., Friedman, R. J., & Kopf, A. W. (1996). The incidence of malignant melanoma in the United States: Issues as we approach the 21st century. Journal of the American Academy of Dermatology, 34, 839-847.

Rogers, R. W. (1983). Cognitive and physiological processes in fear appeals and attitude change: A revised theory of protection motivation. In J. Cacioppo & R. Petty (Eds.), Social psychophysiology (pp. 153-176). New York: Guilford.

Rogers, R. W., & Mewborn, C. R. (1976). Fear appeals and attitude change: Effects of a threat's noxiousness, probability of occurrence, and the efficacy of the coping responses. Journal of Personality and Social Psychology, 34, 54-61.

Rossi, J. S., Blais, L. M., & Weinstock, M. A. (1994). The Rhode Island sun smart project: Skin cancer prevention reaches the beaches. American Journal of Public Health, 84(4), 672-674.

Shedler, J., & Manis, M. (1986). Can the availability heurist explain vividness effects. Journal of Personality and Social Psychology, 51, 26-36.

Sherer, M., & Rogers, R. W. (1984). The role of vivid information in fear appeals and attitude change. Journal of Research in Personality, 18, 321-334.

Stephenson, M. T., & Witte, K. (1996). Processing fear appeals outside of conscious awareness. Paper presented at the annual meeting of the Speech Communication Association, San Diego, CA.

Taylor, S. E., & Thompson, S. C. (1982). Stalking the elusive "vividness" effect. Psychological Review, 89, 155-181.

Weinstock, M. A., Stampfer, M. J., Lew, R. A., Willet, W. C., & Sober, A. J. (1992). Case-control study of melanoma and dietary vitamin D: implications for advocacy of sun protection and sunscreen use. Journal of Investigative Dermatology, 98, 809-811.

Witte, K. (1992a). Putting the fear back into fear appeals: The extended parallel process model. Communication Monographs, 59, 329-348.

Witte, K. (1992b). The role of threat and efficacy in AIDS prevention. International Quarterly of Community Health Education, 12, 225-249.

Witte, K. (1993). Message and conceptual confounds in fear appeals: The role of threat, fear, and efficacy. Southern Communication Journal, 58, 147-155.

Witte, K. (1994). Fear control and danger control: A test of the extended parallel process model. Communication Monographs, 61, 113-125.

Witte, K. (1997). Preventing teen pregnancy through persuasive communications: Realities, myths, and the hard-fact truths. Journal of Community Health, 22, 137-154.

Witte, K. (in press). Fear as motivator, fear as inhibitor: Using the extended parallel process model to explain fear appeal successes and failures. In P. A. Andersen and L. K. Guerrero (Eds.), The Handbook of Communication and Emotion. New York: Academic.

Witte, K. & Allen, M. (1996). When do scare tactics work? A meta-analysis of fear appeals. Paper presented at the annual meeting of the Speech Communication Association, San Diego, CA.

Witte, K., Cameron, K., Nzyuko, S. (1995). HIV/AIDS along the Trans-Africa Highway in Kenya: Examining risk perceptions, recommended responses, and campaign materials. Report submitted to the All-University Initiative Grant program, Michigan State University.

Witte, K., Berkowitz, J., Cameron, K., & McKeon, J. (1995). Preventing the spread of genital warts: Using fear appeals to promote self-protective behaviors. Paper presented at the annual meeting of the Speech Communication Association, San Antonio, TX.

Witte, K., Berkowitz, J., Cameron, K. A., McKeon, J., Lapinski, M. K., & Liu, W. Y. (1996). How African-Americans view radon awareness and reduction campaigns: A theoretically-based formative and summative evaluation. Paper presented at the annual meeting of the Speech Communication Association, San Diego, CA.

Witte, K., Peterson, T., Vallaban, S., Stephenson, M.T., Givens, V., Todd, J., Becktold, M., Hyde, M., & Jarrett, R. (1993). Preventing tractor-related injuries and deaths in rural populations: Using a persuasive health message framework in formative evaluation research. International Quarterly of Community Health Education, 13, 219-251.

Stimulus____________High Description ________Low Description______________________

Note. Single subscripts note significance at the p < .05 level from other pictures in the subset; two or more subscripts indicates that the picture is not significant from other pictures in the subset at the p < .05 level. The purpose was to validate the high threat pictures from the low threat pictures. Of the five previously selected high threat pictures, only one ("Bridgette") did not fall in Subset 4 with the others. While "Bridgette" did attain a significantly higher score than the other two low threat pictures in Subset 3, the other four high threat picture mean scores are higher and in a different subset. Thus, "Bridgette" was not selected for use. Only the four high threat pictures in Subset 4 were used in the main message. Low threat pictures were used for validation only; no low threat messages were included in this study.

THIS TABLE IS UNAVAILABLE HERE. PLEASE E-MAIL THE AUTHOR FOR A COPY.