Self-Efficacy Manipulations in Protection Motivation Research: A Meta-Analysis

By Melissa L. Lewis, Doctoral Student Department of Telecommunication, Information Studies, and Media Michigan State University Robert LaRose, Ph.D. Department of Telecommunication, Information Studies, and Media Michigan State University Nora J. Rifon, Ph.D. Department of Advertising, Public Relations, and Retailing Michigan State University Christina Wirth, Doctoral Student Department of Telecommunication, Information Studies, and Media Michigan State University

Self-Efficacy Manipulations in Protection Motivation Research: A Meta-Analysis

Self-efficacy is one of the most important predictors in the Protection Motivation Theory model. However, it is very rarely used as an independent variable and is more often measured rather than manipulated. This meta-analysis attempts to determine whether successful manipulation of self-efficacy as an independent variable is possible. As well, if manipulation is possible, what are the effect sizes that we can expect and what are the salient characteristics of successful manipulations. Effect sizes (using Cohen's d) ranged from .19 (relatively small) to 9.5 (extremely large) over 17 separate studies. Two separate sample populations emerged, and homogeneity of these is explained by the type of subjects used in the samples: undergraduate samples tend to have greater effect sizes than samples taken from the world at large. Other types of intervention categorization were not significant in explaining population differences. Reasons for this difference as well as the application of self-efficacy treatments in future research are discussed.

Keywords: self-efficacy, protection motivation theory, extended parallel process model, metaanalysis, effect size

2 Introduction

Self-efficacy is used to explain a wide variety of human behavior, and is often included in theories of health behavior to explain the actions taken (or not taken) by individuals. In spite of self-efficacys importance both in general and especially in the overall Protection Motivation Theory model (Bandura, 1992; Floyd, Prentice-Dunn, & Rogers, 2000; Schwarzer, 1992), there is a paucity of studies that include experimental interventions of self-efficacy. Rather, selfefficacy most often appears as a measured variable rather than a manipulated one. This is an adequate and valid way to study self-efficacy, but we should move beyond just measuring selfefficacy to determining whether or not effective self-efficacy interventions can be created and what type of effect size might we expect. Self-Efficacy Self-efficacy is defined specifically by Bandura as the belief in ones capability to engage in a particular course of action to achieve important attainments (Bandura, 1997). Bandura (1977) describes four sources of efficacy expectations: performance accomplishments, vicarious experience, verbal persuasion, and emotional arousal. Performance accomplishment, or progressive mastery of a task as it is referred to in this paper, is the best source of efficacy information, as it is individually tailored and tends to reduce future negative effects. Repeated successes emphasize an individuals mastery of tasks or experiences. If failure is experienced later, the negative effects are likely to be reduced because they are offset by the already-present self-efficacy felt due to the previous successes. Additionally, the positive feelings and increased self-efficacy associated with one particular task can transfer to increased self-efficacy with other experiences.

3 Vicarious experience explains how we learn by watching others. If an individual observes someone else being successful at an endeavor, and observes that individual getting through the experience with no adverse effects, then that individual may feel as if s/he will also experience success with continued effort and improvement. Modeling behavior of others, however, tends to lead to weaker efficacy expectations that are more vulnerable to change and reduction. Individuals may see others succeed, but individual differences and the capacity for social comparison may impede the sense of self-efficacy. In other words, just because someone else was able to succeed at a task doesnt mean I will be; Im not nearly as skilled at things as the other person might be. Verbal persuasion is the third source of efficacy expectation defined by Bandura. Persuasion is one of the oldest tricks in the book when it comes to getting people to do what is desired of them. Persuasion is easy to do and readily available; a few well-chosen words over a matter of even a few seconds is all that is needed, as opposed to the previous sources involving either personal experience (time and effort are needed to actually work through the task) or vicarious experience (more time and effort are needed to watch someone else work through the task). However, persuasion is also a less effective efficacy source than the others listed so far because there is no authentic experience, either by oneself or by a third party. Persuasion is still effective, though. Giving someone the tools to accomplish a task may not lead to mastery of the situation to the degree that giving them the tools with a persuasive message that yes, they too will be able to succeed, would. Unfortunately, if the individual ultimately fails at the task, not only will self-efficacy take a very serious hit, but the credibility of the persuaders will be diminished as well.

4 Finally, emotional arousal can be a source of efficacy expectation. It is much the same as the fight or flight response, however, and can have negative repercussions as well (discussed later in this paper with regard to fear appeals). An emotionally arousing situation might necessitate action and provoke individuals to act. On the flip side, individuals are often afraid of situations in which they know they lack skill and ability. For example, I am afraid of contracting a disease, but I do not feel I have the ability to properly defend myself against it; therefore, I just avoid thinking about the situation entirely. Thus, emotional arousal can instigate a fight response, where the individual processes the stressful situation and then responds accordingly, or it can cause a flight response instead, leading to avoidance and even more stress as the individual is not only upset at the initial arousal but increasingly upset at thoughts of his or her personal ineptitude (Bandura, 1977). Self-efficacy can therefore be categorized as well into either active or passive categories. Active involves situations in which the individual participates to increase self-efficacy, as in personal experience and progressive mastery of a task. Situations in which the individual is merely an observer or a recipient of information, as in the vicarious situations, or situations where the individual receives persuasive messages or emotionally arousing messages, are passive situations. Protection Motivation Theory & Self-Efficacy Protection Motivation Theory (PMT) might be characterized as a fear appeal theory. It was originally introduced by Rogers (1975, 1983) in an attempt to explain how individuals change their health attitudes and behaviors in response to health risk messages. Fear appeals arouse fear in order to promote adherence to the messages recommendation, and are perhaps the

5 most common type of persuasive message used in health communication campaigns, although not necessarily the most effective ones. PMT explains how individuals cognitively process perceived threats and respond with coping mechanisms. Specifically, when people perceive a message as posing a serious and relevant threat, they perceive the threat as fearful enough to motivate them to avert the threat. As a result, people often engage in protective action to reduce the threat. If people do not perceive the message as a significant and relevant one then they will not respond to the risk message. PMT suggests that fear appeals instigate two cognitive processes that mediate actual protection behavior: threat-appraisal process and coping-appraisal process. The threat appraisal assesses threat seriousness and personal susceptibility. The coping appraisal process assesses the effectiveness of potential responses and ones capabilities to undertake these successfully. Together, the two cognitive processes arouse protection motivation, which is an intervening variable that has the typical characteristics of a motive: it arouses, sustains, and directs activity (Rogers, 1975, p. 98). As a result, the appraisal of the threat and the appraisal of the coping responses result in performing adaptive response behavior, which is operationalized as an intention to adopt the communicators recommendation. Two additional constructs were incorporated in a later revision of PMT: rewards associated with maladaptive responses (i.e., smoking and relaxation) and costs associated with adaptive responses (i.e., non-smoking and gaining weight). Rogers (1983) added the two variables and redefined PMT as an attitude-based model (in which attitudes are the product of outcome expectations and evaluations of those outcomes), rather than as a derivative the health belief model (Becker, 1974; Janz & Becker, 1984). Consequently, the respecification of the cognitive appraisal processes parallels the Theory of Planned Behavior (Ajzen, 1991) and social

6 cognitive theory (Bandura, 1986). The revised PMT encompasses the cognitive constituents of outcome expectations (specified beliefs about seriousness, susceptibility, rewards, costs, and response efficacy) and self-efficacy expectations as the main determinants of intention to take self-protective action. Although originally conceived to explain health related behaviors, PMT has been applied to wide range of protective behaviors, including earthquake protection (Asgary & Willis, 1997; Palm, 1995) and traffic safety (Sonmez & Graefe, 1998). A previous exploratory study (Lee & LaRose, 2004) demonstrated its applicability to online safety and found the significant predictors of virus protection adoption behavior. When deciding whether or not to act on a perceived threat, users evaluate their ability to respond to it, by performing a coping appraisal. Coping self-efficacy (Rogers, 1983) is ones belief in his or her capability to carry out the recommended coping behavior. Perceived behavioral control, or the perceived ease or difficulty of performing the behavior of interest (Ajzen, 1991), is a related concept that builds on the notions of controllability. The other component of the coping appraisal is belief that the recommended preventive behavior will be effective, or response efficacy in the PMT framework. Self-efficacy and response efficacy have the most consistent impact on safe behavior across many safety issues (Abraham, Sheeran, & Johnson, 1998; Floyd et al., 2000). When people dont have the information necessary to make an adaptive coping choice, they will tend to make maladaptive coping choices (Eppright, Hunt, Tanner & Franke, 2002; Witte, 1994). In other words, fear appeals dont work unless you give people the skills and/or tools to deal with that fear. According to the Extended Parallel Processing Model (EPPM) when messages arouse fears but dont offer a rational means for dealing with the fear, people are likely to deny that the danger exists or that it is unlikely to affect them (Witte, 1992). Simply assuming

7 that people know how to handle that fear (when in reality, they may not) will most likely lead to maladaptive coping behaviors (Eppright et al., 2002). Reassurance has the greatest impact when it comes to taking precautions; if individuals feel reassured, they will engage in proper and careful precautionary behaviors (Ruiter, Abraham, & Kok, 2001). This reassurance along with information about how to cope with the treat raises self-efficacy which in turn leads to enacting of good behaviors (rather than avoidance behaviors). As well, threat appraisal is a weak element in the PMT model (Floyd et al., 2000; Milne, Sheeran, & Orbell, 2000), having no predictive power on individuals intention to engage in protective behaviors (LaRose, Rifon, Liu, & Lee, 2005). Self-efficacy in the PMT model can be thought of as a sort of self-confidence. If people believe they can do something, they are much more likely to do it than if they think their efforts will be for naught. Social Cognitive Theory & Self-Efficacy Banduras Social Cognitive Theory (SCT) has also been used in numerous health intervention efforts (Bandura, 1986; Bandura, 2004; Johnson, Witte, Boulay, Figueroa, Storey, & Tweedie, 2000). SCT explains how people acquire and maintain certain behavioral patterns, while also providing the basis for intervention strategies (Bandura, 2004). It provides a comprehensive theoretical framework for understanding human behavior, environment, and social interaction. It also offers an agentic perspective on human behavior: human agents intentionally make things happen through their actions by exercising forethought, reflecting on their behavior, and applying self-reactive motivating influences (Bandura, 2001). Outcome expectations regulate behavior. People can have both positive and negative expectations of the perceived consequences of a behavior (Bandura, 1986). The courses of action that are likely to produce positive outcomes are generally adopted and used. However, those

8 actions that bring unrewarding or punishing outcomes are generally discarded. Expectations of behavioral outcomes may be formed either by direct personal experience, in which case it is called enactive learning, or through vicarious experience of the consequences encountered by others, that is observational learning. Self-efficacy is another important determinant of behavior. Peoples judgments of personal efficacy are the most central and pervasive self-reference thoughts that influence human motivation, affect, and action (Bandura, 1997). Expected outcomes and self-efficacy are related. We are more likely to expect favorable outcomes for actions that we are confident we can perform. On the other hand, positive outcome expectations help to build self-esteem, selfsatisfaction, and pride, while negative outcomes produce self-devaluation (Bandura, 1986, 1997). Self-efficacy determines whether sufficient levels of effort will be invested to achieve successful outcomes when one is already motivated to do so. Given the importance of self-efficacy, it is strange that it is rarely used as an independent variable and instead appears more often as a moderator or a dependent variable. Therefore, a meta-analysis was conducted to determine first, how many studies in existence actually manipulated self-efficacy, second, how effective these efficacy manipulations were, and finally, what (if any) variables present in the self-efficacy manipulations made a difference on effect size.

Method

This meta-analysis was based on work originally done by Floyd, Prentice-Dunn, and Rogers (2000), and desired to both expand upon the years contained in the search as well as to

9 narrow down the resulting studies to those that only manipulated self-efficacy. Therefore, the original methods were repeated from Floyd et al. (2000). Literature Search The first step was to review all studies included in Floyd et al.s reference section to determine whether or not these studies contained manipulations of self-efficacy. These articles formed the first tier of the literature search. Next, the original search criteria used by Floyd et al. (2000) were altered slightly to focus solely on self-efficacy. A keyword search of PsycINFO including PsycARTICLES was conducted using search terms of self-efficacy, protection motivation, experiment, manipulation, and all possible combinations of these terms. An initial pool of articles contained 497 results, with the earliest article dated 1982 and the most recent dated 2007. As would be expected, the search returned some duplicates from Floyd et al.s (2000) original search in addition to all new articles available on PsycINFO up to the current year. Inclusion Requirements The main requirement for inclusion in this meta-analysis was that authors had to have manipulated levels of self-efficacy as an independent variable to determine the effectiveness of such manipulations. Each of the results of the searches performed as well as the articles gleaned from Floyd et al.s reference section was analyzed to determine whether or not self-efficacy had been manipulated as an independent variable. Studies that manipulated self-efficacy that were not done under the theoretical umbrella of protection motivation theory were also included, as the resultant sample size was extraordinarily small, and regardless of the theory utilized, it is still important to note whether or not self-efficacy could be sufficiently and successfully manipulated. (Only one study with two separate manipulations was included that did not specifically utilize

10 protection motivation theory, but instead, used self-efficacy in general as the theoretical umbrella.) The next step was to acquire each of the articles in the resultant list of self-efficacy manipulations to determine whether or not self-efficacy was indeed manipulated as an independent variable, and whether or not the manipulation was successful. Studies with nonsignificant (unsuccessful) manipulations of self-efficacy were not included in the subsequent analyses. Classification and Coding Studies were coded on a number of different variables. Due to the extremely small list of studies with significant self-efficacy manipulations, only one coder categorized the studies. (The limitations of this are noted in the discussion section.) Studies were coded for number of subjects used, the type of subjects used, the type of behavior desired, active versus passive interventions, the type of disease, and the type of selfefficacy appeal. Number of subjects was recorded as being the number of subjects that got the self-efficacy intervention. Type of subjects was coded as either being from an undergraduate sample or an at-large (real world) sample. Type of behavior desired refers to the behavior that the self-efficacy intervention was attempting to influence. These were coded into three types: self-examination, cessation of bad behavior (such as smoking), and beginning new healthy behavior (such as exercising). Active interventions were defined as those in which the subject needed to actively participate, either by discussion, providing information, or other forms of engaging participation. Passive interventions were defined as those in which the subject merely received information (for example, reading a paragraph or a pamphlet given to them). The type of disease was classified into serious (including cancer), not serious, and fabricated (as one study

11 used an imaginary disease). Self-efficacy appeals were coded using four categories: vicarious, persuasive, anxiety-reducing, and progressive mastery. Statistical Procedures Calculation of effect sizes was done in stages. Overall, the final effect size utilized was the d statistic, the standardized mean difference between groups (Cohen, 1988). Several different methods of calculating this statistic had to be used due to inconsistency in the type of reported data. For studies that provided means and standard deviations of self-efficacy measures, Cohens d was calculated as the mean difference between groups divided by the pooled standard deviation, which has become a common and acceptable practice (Rosnow & Rosenthal, 1996). Effect sizes were computed solely for divisions of high self-efficacy and low self-efficacy. Effect sizes were not computed for differences between self-efficacy levels as compared with various control measures, as these data would not be comparable with studies that did not include controls. Effect sizes for studies that used several types of self-efficacy (self-efficacy about two different but related subjects for example) were computed as a mean effect size, in essence an average of the two effect sizes However, six of the studies did not provide adequate information for calculating Cohens d (means and standard deviations). The first authors of these studies were contacted and the appropriate figures were requested. Unfortunately, many of the studies contained in this metaanalysis are over several decades old, and the authors contacted reported that the data were no longer available. F-test statistics were provided in the articles for these cases, however. Therefore, the effect size r was calculated for these studies using the F-statistic and the proper degrees of freedom for treatment and control groups. From this, d was computed from r, using

12 the accepted formula of d=2r / (1 - r). Thus, all studies were eventually analyzed using Cohens d as the effect size statistic. To calculate the weight for each study, the inverse variance weight was used. This way, larger studies with bigger sample sizes carry more weight than studies with smaller sample sizes. The inverse variance weight (w) was calculated as 1/SE2, where SE is the standard error. To calculate the Q-statistic to test for homogeneity of the sample, the formula of (w*ES2) [(w*ES)]2/w was used.

Results

Sample Characteristics 17 studies were included in the original coded dataset. However, three of the studies had multiple experiments or multiple manipulations of self-efficacy, bringing the total number of effect sizes calculated to 20. Self-efficacy was conceptualized in the studies used in the meta-analysis in an extremely consistent manner, as Banduras work provided the foundation, defining self-efficacy as the belief or perception that one could successfully perform the coping behavior. Operationally, self-efficacy was measured with various Likert-type scales asking participants to respond to statements such as This message leads me to believe that I can properly perform the TSE on a monthly basis (Morman, 2000, p. 100), and I am able to get a vaccination to prevent against meningitis (Gore & Bracken, 2005, p. 33). Effect Sizes

13 The mean effect size was 1.17, with a standard deviation of 2.01. The mean weighted effect size of the studies was calculated as 30.54, with a standard deviation of 29.52. The largest effect size calculated was 9.48 (an extraordinarily large effect size) and the smallest effect size was .12. Removing the largest of the effect sizes (9.48) as an outlier resulted in a mean effect size of .73, with a standard deviation of .49. Removing the next largest effect size at 2.00 (also a potential outlier), the mean effect size became .66 with a standard deviation of .39. ------------------------------------------------------------------------------------------------Insert Table 1 about here ------------------------------------------------------------------------------------------------A histogram of the effect sizes revealed two very prominent peaks in the distribution (suggesting two different populations, which will be discussed shortly) and one extreme outlier with an effect size of 9.48. When this outlier is removed from the dataset, the mean effect size is reduced dramatically to .73, with a standard deviation of .49. However, the histogram still displays two very prominent peaks, still suggesting samples of two separate populations. ------------------------------------------------------------------------------------------------Insert Figure 1 about here ------------------------------------------------------------------------------------------------Examples of Self-Efficacy Manipulations The three studies with the highest effect sizes (1.32, 2.00, and 9.48) all offered passive manipulations (subjects simply read material presented to them), but each of them used different appeals to self-efficacy: vicarious, persuasive, and anxiety reduction. Mormans study (2000) had the largest effect size by far of 9.48 in their study that attempted o get participants to perform the testicular self-examination (TSE). The study was a

14 2x2 design, giving participants either high or low efficacy message with either a narrative or a factual message. The difference between the high and low efficacy messages involved whether or not specific instruction was given as to how to perform the TSE. The step-by-step, simple instructions were present in the high efficacy condition, but were absent in the low efficacy condition. Vicariousness was achieved by using the fictional testicular cancer victims narrative, and anxiety reduction was present, as the experimenters notably used only high fear appeals; thus, anxiety reduction was always involved. The second largest effect size calculated (2.00) was achieved in a study using meningitis as the disease in question (Gore & Bracken, 2005). The manipulation was still a passive manipulation, although this time rather than having participants read a prepared text, a confederate gave participants a 12-minute, pre-prepared, memorized lecture about the topic, although the exact contents of this lecture were not made available. The intended outcome of the manipulations was to encourage the sample of undergraduates to get a preventative meningitis vaccination. Finally, an effect size of 1.32 was achieved in an experiment elaborating on the connection between exercise and colon cancer (Courneya & Hellsten, 2001). Participants in this study read a 415-word printed communication as their manipulation. Self-efficacy was manipulated within by informing readers that the amount of exercise needed to reduce the risk of colon cancer was one hour a day for five to six days per week at high intensity (low selfefficacy) or twenty minutes a day for two to three days per week at moderate intensity (high selfefficacy). Clearly, this was a passive manipulation (subjects simply read material) and intended to appeal to anxiety reduction. Sources of Heterogeneity

15 In the first analysis, the studies as a whole were heterogeneous in nature. The Q-statistic was calculated as 497.65, clearly exceeding the 31.41 cut-off in the Chi-square distribution table (n=20 and p-value = .05).1 In removing studies as well as referring to the histogram of the distribution of effect sizes, two distinct homogenous samples emerged, with two studies not falling into either a sample or forming their own sample. Studies with effect sizes between .12 and .71 fell into one distribution, while studies with effect sizes between .94 and 1.32 comprised the other distribution. Two outliers emerged here: first was the effect size of 9.5 discussed earlier. Additionally, an effect size of 2.0 was discarded to create the second distribution. The two did not form their own sub-population and are therefore regarded as outliers, analyzed in more detail in the discussion section. For the first distribution, the Q-statistic was calculated as 12.62, with the cut-off for homogeneity at 19.68 (n=11, p-value = .05). As the Q-statistic is below the cut-off, this indicates homogeneity for this sample. For the second distribution, the Q-statistic was 5.44, with the cutoff at 14.08 (n=7, p-value = .05). Again, the Q-statistic falls below the cut-off, indicating homogeneity here as well. The effect size of 2.0 was at first included here in this distribution, but the Q-statistic calculated was 21.69, which exceeds the cut-off of 15.51 (n=8, p-value = .05) and so it was not included. These two studies were analyzed, but together, they did not form a separate population, though it is doubtful that meaningful information could be gained from analyzing only two studies. The mean effect size for the first population (individuals taken from real-world samples) was .39 (SD=.18), while the mean effect size for the second population (undergraduates) was 1.09 (SD=.12).

For studies where ns were not reported for each of the categories of treatment, equal ns were assumed and calculated by dividing the overall n into the appropriate number of groups.

16 Discussion

The intent of this analysis was to determine whether or not self-efficacy can be significantly manipulated, and what manipulations of self-efficacy are most effective. Clearly, manipulations have an impact, from small (with effect sizes around .2) to very large (with effect sizes of .8 and above). Interestingly, there were no significant differences found between effect sizes and number of subjects used, the type of behavior desired, active versus passive interventions, the type of disease, and the type of self-efficacy appeal (vicarious, persuasive, anxiety reduction, and progressive mastery). However, this is also heartening, because it means that whatever the intention of the experiment, whatever the desired behavior, the options that investigators choose that are best suited to the experiment are not hindered by smaller (or inflated by larger) effect sizes simply because of these choices. The one variable that appeared to make a difference in effect size observed was the subject type: undergraduates versus individuals from a real world population. Unfortunately, the larger effect sizes were noted in the undergraduate samples, with the smaller effect sizes coming from the at-large samples. This may be due to the nature of using undergraduate samples. In most cases, subjects were offered compensation in the form of class credit, and thus were aware that they were participating in an academic study. This awareness may have led some participants to exaggerate their responses in an attempt to give the researchers their desired results, and this may have been conscious or unconscious on the part of the subjects. It may also be attributed to education level. Undergraduates are in the process of getting a college education, suggesting a certain level of education and a certain level of intelligence. Samples

17 external to undergraduate populations did not have data on education levels, but may have included individuals whose education was below that of an undergraduate. Lack of higher education might lead to decreased self-efficacy because individuals might not be as informed or as able to understand the material. However, this may also be a developmental effect more than anything else. Young people may be more open to self-efficacy interventions because they are still developing their competencies and self image across a variety of domains. They do not yet have a set system of beliefs about what they can and cannot do, and appeals to self-efficacy may be more effective. College students are also making a passage from a life in which others look out for their health into one where they are personally responsible. Hence, their perceived ability to enact protective behaviors is salient for them. Of course, further analysis is needed so for the present time, this explanation is speculative in nature. Overall effect sizes in PMT studies vary, but appear to be of moderate size. A previous meta-analysis of PMT literature conducted by Milne, Sheeran, and Orbell (2000) found that selfefficacy was the variable that was most often associated with significant intention to enact the specific behavior, with a medium effect size (r+ = .33). They also analyzed cognition change after experimental manipulation of a variety of PMT variables. Self-efficacy had, again, a medium effect (r+ = .32), and authors report that experimental manipulations appear to be more successful than health-education interventions in changing threat- and coping-appraisal cognitions (p. 133). The average of the real-world studies in this meta-analysis was an effect size of .38, also moderate. However, in studies that do not fall under the PMT umbrella and that operate outside the health communication arena, effect sizes tend to increase. A meta-analysis of work-related self-efficacy interventions (which we may interpret as using real world samples) yielded an overall estimated corrected population mean effect of .80, more than twice that of the

18 PMT studies self-efficacy interventions (McNatt, Campbell, & Hirschfeld, 2005). Clearly, bigger effect sizes are out there in terms of self-efficacy interventions, and we should attempt to capitalize on the differences between these studies and PMT-based studies to try and reach these larger effect sizes. One of this studys limitations is the use of only a single searchable index, PsycINFO. Obviously, not all studies that utilize protection motivation theory and include manipulations of self-efficacy will be indexed here. However, this studys intention was first and foremost to follow to method of Floyd et al. (2000), to then include studies that had been published in the seven years since the original article, and use the resultant database of studies as the pool from which to choose articles that contained successful self-efficacy manipulations in an experimental setting. As well, self-efficacy is a psychological variable, and PsycINFO, a database of psychological journals and articles, would be most likely to have a wider array of related articles than other more wide-spanning indexes. Finally, this study is intended not as the be-all end-all self-efficacy meta-analysis but as a starting point for further examination. Time constraints and limited resources necessitated keeping this first step of analysis to a smaller, more manageable study, but further consideration of other indexes as well as other specific journals, conference papers, and unpublished works is needed for a comprehensive review. Related to this, only one coder was used to code the variables of this analysis. Again, time and resource limitations required using only one coder. However, we dont believe this to be too problematic. As the final pool of studies only included seventeen (with twenty total effect sizes calculated), and the variables coded were straightforward, it is quite likely that including more than one coder would not have resulted in any differences in coding. Never the less, again we note that this is merely a first pass at what is likely an extensive body of research, so future

19 study should endeavor to both broaden the search as well as to include a minimum of three coders for accuracy and clarity. Since self-efficacy is such an important part of protection motivation theory (and since it is as well related to so very many outcomes), then we should be endeavoring to manipulate it and analyze the outcomes. There should be no limitations on manipulation, since this analysis has shown that regardless of the complicated nature of the manipulation (from elaborate participatory focus groups to simple typed paragraphs), effective change in self-efficacy can be initiated. As well, self-efficacy manipulation should extend beyond the health behavior boundaries to include any self-protection behaviors. If we extend this meta-analysis to include experiments with all self-efficacy interventions rather than ones that are specifically situated in Protection Motivation Theory and results are consistentthat is, all types of self-efficacy treatments can be equally effectivethen we should consider for future studies which treatments are the most efficient to administer.

20 References Abraham, C., Sheeran, P. & Johnson, M. (1998). From health beliefs to self-regulation: Theoretical advances in the psychology of action control. Psychology and Health, 13, 569-591. Ajzen, I. (1991). The theory of planned behavior. Organizational Behavior and Human Decision Processes, 50, 179-211. Asgary, A., & Willis, K.G. (1997). Household behavior in response to earthquake risk: An assessment of alternative theories. Disasters, 21(4), 354-364. Bandura, A. (1992). Exercise of personal agency through the self-efficacy mechanism. In R. Schwarzer (Ed.), Self-efficacy: Thought control of action (pp. 3-38). Washington, DC: Hemisphere. Bandura, A. (1977). Self-efficacy: Toward a unifying theory of behavioral change. Psychological Review, 84(2), 191-215. Bandura, A. (1986). Social foundations of thought and action: A social cognitive theory. Englewood Cliffs, NJ: Prentice Hall. Bandura, A. (1997). Self-efficacy: the exercise of control. New York: W.H. Freeman. Bandura, A. (1999). Social cognitive theory of personality. In L.A. Pervin & O.P. John (Eds.), Handbook of Personality theory and research (2nd ed.), (pp. 154-196). New York: the Guilford Press. Bandura, A. (2001). Social cognitive theory of mass communication. Media Psychology, 3, 265299. Bandura, A. (2004). Health Promotion by Social Cognitive Means. Health Education & Behavior, 31, 143-164.

21 Basen-Engquist, K. (1994). Evaluation of a theory-based HIV prevention intervention for college students. AIDS Education and Prevention, 6(5), 412-424. Becker, M. H. (Ed.). (1974). The health belief model and personal health behavior. Thorofare, NJ: Charles B. Slack. Inc. Brouwers, M. C., & Sorrentino, R. M. (1993). Uncertainty orientation and protection motivation theory: The role of individual differences in health compliance. Journal of Personality and Social Psychology, 65(1), 102-112. Cohen, J. (1988). Statistical power analysis for the behavioral sciences (2nd ed.). Hillsdale, NJ: Lawrence Earlbaum Associates. Courneya, K. S., & Hellsten, L. A. M. (2001). Cancer prevention as a source of exercise motivation: An experimental test using protection motivation theory. Psychology, Health & Medicine, 6(1), 59-64. Eppright, D. R., Hunt, J.B., Tanner, J.F., & Franke, G.R. (2002). Fear, coping, and information: A pilot study on motivating a healthy response. Health Marketing Quarterly, 20(1), 5173. Floyd, D. L., Prentice-Dunn, S. & Rogers, R.W. (2000). A meta-analysis of research on protection motivation theory. Journal of Applied Social Psychology, 30(2), 407-29. Fruin, D. J., Pratt, C., & Owen, N. (1992). Protection motivation theory and adolescents' perceptions of exercise. Journal of Applied Social Psychology, 22(1), 55-69. Gore, T. D., & Campanella Bracken, C. (2005). Testing the theoretical design of a health risk message: Reexamining the major tenets of the extended parallel process model. Health Education & Behavior, 32(1), 27-41. Graham, S. P., Prapavessis, H., & Cameron, L. D. (2006). Colon cancer information as a source

22 of exercise motivation. Psychology & Health, 21(6), 739-755. Hall, S., Bishop, A. J., & Marteau, T. M. (2006). Does changing the order of threat and efficacy information influence the persuasiveness of threat messages? British Journal of Health Psychology, 11(2), 333-343. Janz, N., & Becker, M. H. (1984). The health belief model: a decade later. Health Education Quarterly, 11, 1-47. Johnson, M., Witte, L., Boulay, M., Figueroa, M. E., Storey, D., Tweedie, I. (2000). Using health education theories to explain behavior change: A cross-country analysis. International Quarterly of Community Health Education, 20(4), 323-345. LaRose, R., Rifon, N.J., Liu, X., & Lee, D. (2005). Understanding online safety behavior: A multivariate model. Paper presented at the 2005 International Communication Association conference, New York, NY. Lee, D., & Larose, R. (2004). Keeping our network safe: A model of online safety behavior. Paper presented at the annual meeting of the Association for Education in Journalism and Mass Communication, Toronto, Canada. Litt, M. D., Nye, C., & Shafer, D. (1995). Preparation for oral surgery: Evaluating elements of coping. Journal of Behavioral Medicine, 18(5), 435-459. Maddux, J. E., & Rogers, R. W. (1983). Protection motivation and self-efficacy: A revised theory of fear appeals and attitude change. Journal of Experimental Social Psychology, 19(5), 469-479. McKay, D. L., Berkowitz, J. M., Blumberg, J. B., & Goldberg, J. P. (2004). Communicating cardiovascular disease risk due to elevated homocysteine levels: Using the EPPM to develop print materials. Health Education & Behavior, 31(3), 355-371.

23 McNatt, D.B., Campbell, S., & Hirschfeld, R.R. (2005). Building self-confidence: A metaanalysis of the effectiveness of self-efficacy interventions in work-related contexts. Poster session presented at the 20th annual conference of the Society for Industrial and Organizational Psychology, Los Angeles, CA. Milne, G., Sheeran, P. & Orbell, S. (2000). Prediction and intervention in health-related behavior: A meta-analytic review of protection motivation theory. Journal of Applied Social Psychology, 30(1), 106-43. Morman, M. T. (2000). The influence of fear appeals, message design, and masculinity on men's motivation to perform the testicular self-exam. Journal of Applied Communication Research, 28(2), 91-116. Palm, R. I. (1995). Earthquake Insurance: A Longitudinal Study of California Homeowners. Westview Press, Boulder. Rimal, R. N., & Real, K. (2003). Perceived risk and efficacy beliefs as motivators of change: Use of the risk perception attitude (RPA) framework to understand health behaviors. Human Communication Research, 29(3), 370-399. Rippetoe, P. A., & Rogers, R. W. (1987). Effects of components of protection-motivation theory on adaptive and maladaptive coping with a health threat. Journal of Personality and Social Psychology, 52(3), 596-604. Rogers, R. W. (1975). A Protection motivation theory of fear appeals and attitude change. The Journal of Psychology, 91, 93-114. 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 Press.

24 Rosnow, R. L., & Rosenthal, R. (1996). Computing contrasts, effect sizes, and counternulls on other people's published data: General procedures for research consumers. Psychological Methods, 1, 331-340. Ruiter, R. A. C., Abraham, C., & Kok, G. (2001). Scary warnings and rational precautions: A review of the psychology of fear appeals. Psychology and Health, 16(6), 613-30. Sanna, L. J. (1992). Self-efficacy theory: Implications for social facilitation and social loafing. Journal of Personality and Social Psychology, 62(5), 774-786. Schwarzer, R. (1992). Self-efficacy in the adoption and maintenance of health behaviors: Theoretical approaches and a new model. In R. Schwarzer (Ed.), Self-efficacy: Thought control of action (pp. 217-243). Washington, DC: Hemisphere Sonmez, S. F., & Graefe, A. R. (1998). Determining future travel behavior from past travel experience and perceptions of risk and safety. Journal of Travel Research, 37(2), 171178. Stanley, M. A., & Maddux, J. E. (1986). Cognitive processes in health enhancement: Investigation of a combined protection motivation and self-efficacy model. Basic and Applied Social Psychology, 7(2), 101-113. Turner, M. M., Rimal, R. N., Morrison, D., & Kim, H. (2006). The role of anxiety in seeking and retaining risk information: Testing the risk perception attitude framework in two studies. Human Communication Research, 32(2), 130-156. Witte, K. (1994). Fear control and danger controlA test of the extended parallel process model (eppm). Communication Monographs, 61(2), 113-134. Witte, K. (1992). Putting the fear back into fear appeals: The extended parallel process model. Communication Monographs, 59(4), 329-349.

25 Wurtele, S. K., & Maddux, J. E. (1987). Relative contributions of protection motivation theory components in predicting exercise intentions and behavior. Health Psychology, 6(5), 453466.

26 Table 1: Study, Effect Size, and Inverse Variance Weight Sorted By Effect Size
Study Effect Size .1231 Inverse Variance Weight 38.9131

Hall, S., Bishop, A. J., & Marteau, T. M. (2006). Does changing the order of threat and efficacy information influence the persuasiveness of threat messages? British Journal of Health Psychology, 11(2), 333-343. Fruin, D. J., Pratt, C., & Owen, N. (1992). Protection motivation theory and adolescents' perceptions of exercise. Journal of Applied Social Psychology, 22(1), 55-69. Litt, M. D., Nye, C., & Shafer, D. (1995). Preparation for oral surgery: Evaluating elements of coping. Journal of Behavioral Medicine, 18(5), 435-459. McKay, D. L., Berkowitz, J. M., Blumberg, J. B., & Goldberg, J. P. (2004). Communicating Cardiovascular Disease Risk Due to Elevated Homocysteine Levels: Using the EPPM to Develop Print Materials. Health Education & Behavior, 31(3), 355-371. Rippetoe, P. A., & Rogers, R. W. (1987). Effects of components of protection-motivation theory on adaptive and maladaptive coping with a health threat. Journal of Personality and Social Psychology, 52(3), 596-604. Wurtele, S. K., & Maddux, J. E. (1987). Relative contributions of protection motivation theory components in predicting exercise intentions and behavior. Health Psychology, 6(5), 453-466. Rimal, R. N., & Real, K. (2003). Perceived Risk and Efficacy Beliefs as Motivators of Change: Use of the Risk Perception Attitude (RPA) Framework to Understand Health Behaviors. Human Communication Research, 29(3), 370-399. Graham, S. P., Prapavessis, H., & Cameron, L. D. (2006). Colon cancer information as a source of exercise motivation. Psychology & Health, 21(6), 739-755. Basen-Engquist, K. (1994). Evaluation of a theory-based HIV prevention intervention for college students. AIDS Education and Prevention, 6(5), 412-424. Graham, S. P., Prapavessis, H., & Cameron, L. D. (2006). Colon cancer information as a source of exercise motivation. Psychology & Health, 21(6), 739-755. Brouwers, M. C., & Sorrentino, R. M. (1993). Uncertainty orientation and protection motivation theory: The role

.2172

149.4426

.2300

22.3573

.2878

18.6700

.3510

32.5709

.3511

38.3183

.4069

34.7333

.4441

12.0398

.4991

13.2047

.6107

9.7347

.7161

48.7850

27 of individual differences in health compliance. Journal of Personality and Social Psychology, 65(1), 102-112. Turner, M. M., Rimal, R. N., Morrison, D., & Kim, H. (2006). The Role of Anxiety in Seeking and Retaining Risk Information: Testing the Risk Perception Attitude Framework in Two Studies. Human Communication Research, 32(2), 130-156. Maddux, J. E., & Rogers, R. W. (1983). Protection motivation and self-efficacy: A revised theory of fear appeals and attitude change. Journal of Experimental Social Psychology, 19(5), 469-479. Sanna, L. J. (1992). Self-efficacy theory: Implications for social facilitation and social loafing. Journal of Personality and Social Psychology, 62(5), 774-786. Stanley, M. A., & Maddux, J. E. (1986). Cognitive processes in health enhancement: Investigation of a combined protection motivation and self-efficacy model. Basic and Applied Social Psychology, 7(2), 101-113. Sanna, L. J. (1992). Self-efficacy theory: Implications for social facilitation and social loafing. Journal of Personality and Social Psychology, 62(5), 774-786. Turner, M. M., Rimal, R. N., Morrison, D., & Kim, H. (2006). The Role of Anxiety in Seeking and Retaining Risk Information: Testing the Risk Perception Attitude Framework in Two Studies. Human Communication Research, 32(2), 130-156. Courneya, K. S., & Hellsten, L. A. M. (2001). Cancer prevention as a source of exercise motivation: An experimental test using protection motivation theory. Psychology, Health & Medicine, 6(1), 59-64. Gore, T. D., & Campanella Bracken, C. (2005). Testing the Theoretical Design of a Health Risk Message: Reexamining the Major Tenets of the Extended Parallel Process Model. Health Education & Behavior, 32(1), 27-41. Morman, M. T. (2000). The influence of fear appeals, message design, and masculinity on men's motivation to perform the testicular self-exam. Journal of Applied Communication Research, 28(2), 91-116.

.9444

39.3497

1.0159

33.7182

1.0592

31.7909

1.0606

42.8228

1.0908

31.6804

1.1739

27.6738

1.3165

91.7431

2.0041

23.0211

9.4764

9.1552

28 Figure 1: Histogram of Effect Sizes