Get 20M+ Full-Text Papers For Less Than $1.50/day. Start a 7-Day Trial for You or Your Team.

Learn More →

The Relative Persuasiveness of Gain-Framed and Loss-Framed Messages for Encouraging Disease Detection Behaviors: A Meta-Analytic Review

The Relative Persuasiveness of Gain-Framed and Loss-Framed Messages for Encouraging Disease... Abstract A meta-analytic review of 53 studies (N = 9,145) finds that in messages aimed at encouraging disease detection behaviors, loss-framed appeals (which emphasize the disadvantages of noncompliance with the communicator's recommendation) are only slightly, but statistically significantly, more persuasive than gain-framed appeals (which emphasize the advantages of compliance); the difference corresponds to a correlation of −.04. Loss-framed appeals enjoy a small statistically significant advantage for messages advocating breast cancer detection behaviors, but not for any other kind of detection behavior (detection of skin cancer, other cancers, dental problems, or miscellaneous other diseases) nor for all other kinds of detection behaviors combined. Thus, in advocacy of disease detection behaviors, using loss-framed rather than gain-framed appeals is unlikely to substantially improve persuasiveness. Early detection of disease conditions can potentially have substantial benefits, such as reduced risk of death and availability of a larger number of (and less aggressive) treatment options (Smith, Cokkinides, & Eyre, 2007). The chances for early detection are enhanced by performance of appropriate screening behaviors—and hence one persuasive task facing health professionals is that of encouraging such behaviors. Even when mass screenings may not be justified by the benefits (e.g., Helfand, Mahon, Eden, Frame, & Orleans, 2001; U.S. Preventive Services Task Force, 2001), screenings targeted to specific populations may still be beneficial (e.g., Swetter, Waddell, Vazquez, & Khosravi, 2003). Among the kinds of persuasive messages that might be used to encourage disease detection behaviors, loss-framed appeals have often been suggested to be especially effective. A loss-framed persuasive appeal emphasizes the disadvantages of failing to comply with the communicator's recommendation (e.g., “If you don't have regular mammograms, you reduce your chances of detecting breast cancer at an early, more treatable stage”); the contrast is a gain-framed appeal, which emphasizes the advantages of compliance (e.g., “If you have regular mammograms, you increase your chances of detecting breast cancer at an early, more treatable stage”). Previous primary research (e.g., Meyerowitz & Chaiken, 1987) and reviews and theoretical analyses (e.g., Edwards, Elwyn, Covey, Matthews, & Pill, 2001; Kuhberger, 1998; Rothman & Salovey, 1997; Salovey, Schneider, & Apanovitch, 2002; Wilson, Purdon, & Wallston, 1988) have suggested that loss-framed appeals will enjoy a significant persuasive advantage over gain-framed appeals in the domain of disease detection behaviors. The most common explanatory framework invokes prospect-theoretic reasoning to suggest that potential losses are more motivating than potential gains when risky actions are contemplated, whereas gains are more motivating than losses for low-risk behaviors (Kahneman & Tversky, 1979). The apparent implication is that for relatively high-risk actions such as disease detection behaviors, loss-framed messages should be more persuasive than gain-framed messages (see, e.g., Salovey et al., 2002; cf. O'Keefe & Jensen, 2006, p. 23). However, a recent broad meta-analytic review concluded that loss-framed appeals are not significantly more persuasive than gain-framed appeals for encouraging disease detection behaviors (O'Keefe & Jensen, 2006). The observed direction of effect favored loss-framed appeals, but the mean effect size (across 34 studies) was r = −.03, a value not significantly different from zero. Unfortunately, that review treated disease detection behaviors as a global category and did not distinguish effects for different specific detection behaviors; indeed, to date, no systematic review appears to have done so. Thus, the primary purpose of this review was to examine whether the relative persuasiveness of gain- and loss-framed appeals varies depending on the particular detection behavior being advocated. These behavior-specific analyses were motivated by two considerations. First, such analyses have considerable practical importance; the best evidence of whether loss-framed appeals are especially persuasive for encouraging (say) skin cancer detection behaviors comes from studies specifically concerning those behaviors, not from general research on disease detection behaviors. Second, a global nonsignificant effect might mask a substantial effect occurring for some particular detection behavior; without behavior-specific analyses, such effects could go unnoticed. The secondary purpose was to explore potential moderating factors. A number of variables have been suggested as possible moderators of gain–loss message framing effects, although many candidates are unsuitable for meta-analytic examination. For example, it is not feasible to code studies for characteristics such as audience mood (e.g., Martin & Lawson, 1998; Mitchell, 2001), attitudinal ambivalence (e.g., Broemer, 2002; Matwin, 2007), self-efficacy (e.g., van ‘t Riet, Ruiter, Werrij, & de Vries, in press), and the like. However, one particular potential moderator merited attention, namely, a variation in the way in which the consequences (of adopting or not adopting the recommended behavior) were expressed. As noticed by several commentators (e.g., Dillard & Marshall, 2003; Nan, 2007b; Rothman & Salovey, 1997; Wilson et al., 1988), gain-framed and loss-framed appeals can each take two forms, with the resulting four possibilities represented in a 2 × 2 array in which the contrasts are whether (a) the outcome described is a desirable or an undesirable one and (b) the outcome is described as one that is attained (acquired, made more likely) or avoided (averted, made less likely). That is, a gain-framed appeal might take the form “If you perform the advocated behavior, desirable outcome X will be obtained” or the form “If you perform the advocated behavior, undesirable outcome Y will be avoided.” A loss-framed appeal might take the form “If you do not perform the advocated behavior, desirable outcome X will be avoided” or the form “If you do not perform the advocated behavior, undesirable outcome Y will be obtained.” This variation has previously been described as a matter of the appeals differing in the linguistic representation of the “kernel state” of the consequence under discussion (O'Keefe & Jensen, 2006). The kernel state is the basic, root state mentioned in the message's description of the consequence. For example, in an appeal such as “If you have regular skin exams, you will reduce your risk of skin cancer,” the kernel state is “skin cancer,” which is plainly an undesirable state; that is, this appeal emphasizes the desirable consequences of compliance by discussing an undesirable kernel state (“skin cancer”) that will be avoided. By comparison, “If you have regular skin exams, you will increase your chances of having healthy skin” is an appeal describing a desirable kernel state (“healthy skin”) that will be attained by compliance. A gain-framed appeal thus could be phrased entirely in terms of avoided undesirable outcomes (“heart disease,”“skin cancer,”“premature death,” etc.) and a loss-framed appeal could be phrased entirely in terms of foregone desirable outcomes (“healthy heart,”“attractive skin,”“long life,” etc.)—and hence to sort out these possible evaluative confoundings, examination of this moderator seemed warranted. Method Identification of relevant investigations Literature search Relevant research reports were located through personal knowledge of the literature, examination of previous reviews, and inspection of reference lists in previously located reports. In addition, articles were identified through computerized database searches through at least August 2008 of ABI-INFORM, Cumulative Index of Nursing and Allied Health Literature (CINAHL), Current Contents, Dissertation Abstracts, EBSCO, Educational Resources Information Center (ERIC), Linguistics and Language Behavior Abstracts, MEDLINE, and PsycINFO, using various appropriate combinations of terms such as framing, framed, frame, appeal, message, persuasion, persuasive, gain, positive, positively, benefit, loss, negative, negatively, threat, and valence. Inclusion criteria We included a study if it met three criteria. First, the study had to compare gain- and loss-framed persuasive messages. A gain-framed message emphasizes the desirable consequences of compliance with the advocated view; a loss-framed message emphasizes the undesirable consequences of noncompliance. Excluded by this criterion were studies with imperfect realizations of this message contrast, such as studies that compared one framing form with unframed information (Abood, Coster, Mullis, & Black, 2002; Kalichman & Coley, 1995). Second, the messages had to advocate disease detection behaviors (breast cancer detection behaviors such as mammography and breast self-examination [BSE], skin cancer detection behaviors such as skin examinations, etc.). We excluded studies in which the messages advocated other behaviors, such as disease prevention behaviors (e.g., Block, 1993, sun exposure condition) or behaviors unrelated to health (e.g., Thorsteinson & Highhouse, 2003). As noted by Salovey and Wegener (2003, p. 61), some health-related behaviors, such as Pap tests and colonoscopies, might plausibly be described as either (or both) a disease-detection behavior or a disease-prevention behavior; in the interest of focusing on clear-cut studies of disease detection behaviors, studies of such “dual-function” behaviors were excluded (Miles, Brotherstone, Robb, Atkin, & Wardle, 2005; Mullins, 2005; Rivers, Salovey, Pizarro, Pizzaro, & Schneider, 2005; Wilkin, 2004). Following similar reasoning, we excluded studies in which a message advocated both detection behaviors and prevention behaviors (Block & Keller, 1995, Study 1) and studies in which the disease condition to be detected was not the audience's (Nan, 2007a, other target condition; Nan, 2007c; Roberto, Goodall, West, & Mahan, in press). Third, appropriate quantitative data relevant to persuasive effects (e.g., attitude change, intention, or behavior) had to be available; where it was not provided in the report, we made efforts to obtain information from authors. Excluded by this criterion were studies of effects on other outcome variables and studies for which appropriate quantitative information could not be obtained (e.g., Brunton, 2007; Miller et al., 1999; Salmon, Loken, & Finnegan, 1985; Steffen, Sternberg, Teegarden, & Shepherd, 1994; Trotto, 2001; Umphrey, 2001). Outcome variable and effect size measure Outcome variable The outcome variable was persuasion, as assessed through attitude change, postcommunication agreement, behavioral intention, behavior, and the like. When multiple indices of persuasion (e.g., assessments of attitude and of intention) were available, we averaged the effects to yield a single summary.1 Most studies reported only short-term effects; where both long- and short-term effect size information was available (e.g., Banks et al., 1995; Finney, 2001; Lalor, 1990), only short-term effects were included to maximize comparability across studies. Effect size measure Every comparison between a gain-framed message and its loss-framed counterpart was summarized using r as the effect size measure; differences indicating greater persuasion with gain-framed messages were given a positive sign. Results not reported as correlations were converted to r using formulas provided by Johnson (1993) and Rosenthal (1991). When correlations were averaged (e.g., across several indices of persuasive effect), we computed the average using the r-to-z-to-r transformation procedure, weighted by n. Wherever possible, multiple-factor designs were analyzed by reconstituting the analysis such that individual-difference factors (but not, e.g., other experimental manipulations) were put back into the error term (following the suggestion of Johnson, 1989, p. 16). Moderator variables Specific detection behavior Cases were classified by the kind of detection behavior advocated, with five broad categories distinguished: Breast cancer detection (e.g., mammography), skin cancer detection (e.g., skin self-examination), detection of other cancers (e.g., colon cancer screenings), detection of dental problems (e.g., plaque detection), and miscellaneous detection behaviors (screenings for cholesterol, hypertension, etc.). Kernel-state phrasing The kernel states in each appeal were identified; as discussed above, a kernel state is the basic, root state mentioned in the message's description of the consequence under discussion. We coded each appeal as containing exclusively desirable kernel states (e.g., “healthy heart,”“attractive skin”), exclusively undesirable kernel states (e.g., “heart disease,”“skin cancer”), a combination of desirable and undesirable kernel states, or as indeterminate with respect to kernel-state phrasing (as when insufficient detail was available about the messages). Coding reliabilities Codings for these moderator variables were completed independently by the authors for a sample of 30 cases. Intercoder reliabilities (percent agreement and Cohen's kappa, respectively) were.97 and.95 for advocated behavior,.93 and.89 for kernel state phrasing in gain appeals, and.97 and.95 for kernel state phrasing in loss appeals. Discrepancies were resolved by discussion. D. J. O'K. coded the remaining cases. Unit of analysis The unit of analysis was the message pair, that is, the pair composed of a gain-framed message and its loss-framed counterpart. We recorded an effect size for each distinguishable message pair found in the body of studies. Most message pairs were used only in a single investigation, so only one effect size estimate was associated with the pair. However in three instances, message pairs were used in more than one study, with the result that several effect size estimates could be associated with a given message pair. These multiple estimates were averaged to yield a single summary estimate before inclusion in the analysis. Specifically, data from Experiment 1 and Experiment 2 in Keller, Lipkus, and Rimer (2003) were combined and reported as Keller et al. (2003); data from Meyerowitz and Chaiken (1987) and Lalor (1990) were combined and reported as Meyerowitz and Chaiken (1987) combined; and data from Rothman, Martino, Bedell, Detweiler, and Salovey (1999). Experiment 1 detection condition and Sanchez (2006) were combined and reported as Rothman letrolisus detection combined. Whenever a study included more than one message pair and reported data separately for each pair, each pair was treated as providing a separate effect size estimate (e.g., Ying, 2001). Some studies included more than one message pair but did not report results in ways that permitted computing separate effect sizes for each pair (Apanovitch, McCarthy, & Salovey, 2003; Schmitt, 2004; Williams, Clarke, & Borland, 2001); we computed a single effect size in such cases, with the consequence that this analysis underrepresents the amount of message-to-message effect variability in these data. In some cases, the same primary data served as the basis for multiple reports (e.g., both a dissertation and a subsequent publication). When a given investigation was reported in more than one outlet, it was treated as a single study and analyzed accordingly. The same research was reported (in whole or in part) in: Block (1993) and Block and Keller (1995, Study 2), recorded under the former; Brenes (1998) and Brenes (1999), recorded under the latter; Cothran, Schneider, and Salovey (1998) and Schneider et al. (2001), recorded under the latter; Finney (2001) and Finney and Iannotti (2002), recorded under the former; Lalor (1990) and Lalor and Hailey (1990), with, as noted above, results recorded under Meyerowitz and Chaiken (1987) combined; Nan (2006, Experiment 2) and Nan (2007b, Experiment 2), recorded under the former; Shen (2005, Study 2, glaucoma exam), and Shen and Dillard (2007, Study 2, glaucoma exam), recorded under the former. Meta-analytic procedures The individual correlations (effect sizes) were initially transformed to Fisher's zs; the zs were analyzed using random-effects procedures (specifically, those of Borenstein & Rothstein, 2005; see also Hedges & Vevea, 1998; Shadish & Haddock, 1994), with results then transformed back to r. A random-effects analysis was used in preference to a fixed-effects analysis because of an interest in generalizing across messages (for some discussion, see Hedges & Vevea, 1998; Raudenbush, 1994; Shadish & Haddock, 1994). Results Overall effect Effect sizes were available for 53 cases, with a total of 9,145 participants. Details for each included case are contained in Table 1. Across all 53 cases, loss-framed appeals were significantly more persuasive than gain-framed appeals; mean r = −.039 (p =.02).2 Table 1 Cases Analyzed Study . r . N . Codingsa . Apanovitch et al. (2003) .064 425 5/3/3/100 Arora (2000) −.157 210 4/1/2/60 Banks et al. (1995) −.011 133 1/3/1/100 Block (1993) self-exam −.222 57 2/2/1/na Brenes (1999) −.037 142 1/4/4/100 Broemer (2004) Study 2 .167 60 1/3/3/100 Chang (2007) Study 1 disclosing gum −.250 66 4/4/4/na Chang (2007) Study 1 disclosing strips .443 66 4/4/4/na Cherubini et al. (2005) fear-evoking −.164 60 3/3/3/0 Cherubini et al. (2005) reassuring −.166 60 3/3/3/0 Consedine, Horton, Magai, and Kukafka (2007) −.024 89 1/3/3/100 Cox, Cox, and Zimet (2006) Study 1 detection −.078 142 2/3/3/41 Cox and Cox (2001) anecdotal −.306 55 1/3/3/100 Cox and Cox (2001) statistical .046 55 1/2/1/100 Finney (2001) −.044 628 1/2/1/100 Fischer and Nabi (2001) skin exam .144 87 2/3/1/54 Gintner, Rectanus, Achord, and Parker (1987) .051 177 5/4/4/59 Hsiao (2002) testing-prevention −.300 46 5/3/3/56 Hsiao (2002) testing-detection .308 46 5/3/3/56 Keller et al. (2003) −.024 162 1/3/3/100 Kostygina (2007) −.066 119 3/3/3/na Kwan (1996) −.476 33 1/3/3/100 Kyriakaki (2007) diabetes health −.182 63 5/3/3/na Kyriakaki (2007) diabetes psychosocial −.193 65 5/3/3/na Lauver and Rubin (1990) −.060 116 3/1/2/100 Lee, Brown, & Blood (2000) self-exam −.106 137 2/1/2/54 Lerman et al. (1992) −.011 203 1/4/4/100 Lotto, Tasso, Carnaghi, and Rumati (2006) −.018 80 1/4/4/100 Maheswaran and Meyers-Levy (1990) .023 98 5/3/3/na Matwin (2007) strong −.039 162 5/3/3/na Matwin (2007) weak −.018 160 5/3/1/na Meyerowitz and Chaiken (1987) combined −.219 91 1/3/3/100 Myers et al. (1991) −.035 2,201 3/4/4/47 Nan (2006) Experiment 2 gain-positive loss-positive .124 68 5/1/2/na Nan (2006) Experiment 2 gain-positive loss-negative .076 67 5/1/1/na Nan (2006) Experiment 2 gain-negative loss-positive .070 68 5/2/2/na Nan (2006) Experiment 2 gain-negative loss-negative .161 67 5/2/1/na Nan (2007a) self-target .105 64 5/3/3/100 Phelan (2003) .000 60 1/4/4/100 Rothman letrolisus detection combined −.218 321 5/3/3/87 Rothman et al. (1999) Experiment 2 detection −.305 60 4/2/1/74 Ruiter, Kok, Verplanken, and van Eersel (2003) −.099 110 1/4/4/100 Schmitt (2004) −.055 150 5/4/4/100 Schneider et al. (2001) multicultural −.138 264 1/4/4/100 Schneider et al. (2001) targeted .047 264 1/4/4/100 Shen (2005) Study 2 glaucoma exam .021 252 5/4/4/71 Spaderna (2003) −.019 162 2/3/3/60 Umphrey (2003) .085 128 3/3/3/0 Urban, Stout, Zimet, and Blake (2006) .083 35 5/3/3/100 van ‘t Riet, Ruiter, Werrij, and de Vries (in press) −.141 124 2/3/1/86 Williams, Clarke, and Borland (2001) −.089 307 1/4/4/100 Ying (2001) concrete −.021 140 2/3/3/na Ying (2001) abstract .069 140 2/3/3/na Study . r . N . Codingsa . Apanovitch et al. (2003) .064 425 5/3/3/100 Arora (2000) −.157 210 4/1/2/60 Banks et al. (1995) −.011 133 1/3/1/100 Block (1993) self-exam −.222 57 2/2/1/na Brenes (1999) −.037 142 1/4/4/100 Broemer (2004) Study 2 .167 60 1/3/3/100 Chang (2007) Study 1 disclosing gum −.250 66 4/4/4/na Chang (2007) Study 1 disclosing strips .443 66 4/4/4/na Cherubini et al. (2005) fear-evoking −.164 60 3/3/3/0 Cherubini et al. (2005) reassuring −.166 60 3/3/3/0 Consedine, Horton, Magai, and Kukafka (2007) −.024 89 1/3/3/100 Cox, Cox, and Zimet (2006) Study 1 detection −.078 142 2/3/3/41 Cox and Cox (2001) anecdotal −.306 55 1/3/3/100 Cox and Cox (2001) statistical .046 55 1/2/1/100 Finney (2001) −.044 628 1/2/1/100 Fischer and Nabi (2001) skin exam .144 87 2/3/1/54 Gintner, Rectanus, Achord, and Parker (1987) .051 177 5/4/4/59 Hsiao (2002) testing-prevention −.300 46 5/3/3/56 Hsiao (2002) testing-detection .308 46 5/3/3/56 Keller et al. (2003) −.024 162 1/3/3/100 Kostygina (2007) −.066 119 3/3/3/na Kwan (1996) −.476 33 1/3/3/100 Kyriakaki (2007) diabetes health −.182 63 5/3/3/na Kyriakaki (2007) diabetes psychosocial −.193 65 5/3/3/na Lauver and Rubin (1990) −.060 116 3/1/2/100 Lee, Brown, & Blood (2000) self-exam −.106 137 2/1/2/54 Lerman et al. (1992) −.011 203 1/4/4/100 Lotto, Tasso, Carnaghi, and Rumati (2006) −.018 80 1/4/4/100 Maheswaran and Meyers-Levy (1990) .023 98 5/3/3/na Matwin (2007) strong −.039 162 5/3/3/na Matwin (2007) weak −.018 160 5/3/1/na Meyerowitz and Chaiken (1987) combined −.219 91 1/3/3/100 Myers et al. (1991) −.035 2,201 3/4/4/47 Nan (2006) Experiment 2 gain-positive loss-positive .124 68 5/1/2/na Nan (2006) Experiment 2 gain-positive loss-negative .076 67 5/1/1/na Nan (2006) Experiment 2 gain-negative loss-positive .070 68 5/2/2/na Nan (2006) Experiment 2 gain-negative loss-negative .161 67 5/2/1/na Nan (2007a) self-target .105 64 5/3/3/100 Phelan (2003) .000 60 1/4/4/100 Rothman letrolisus detection combined −.218 321 5/3/3/87 Rothman et al. (1999) Experiment 2 detection −.305 60 4/2/1/74 Ruiter, Kok, Verplanken, and van Eersel (2003) −.099 110 1/4/4/100 Schmitt (2004) −.055 150 5/4/4/100 Schneider et al. (2001) multicultural −.138 264 1/4/4/100 Schneider et al. (2001) targeted .047 264 1/4/4/100 Shen (2005) Study 2 glaucoma exam .021 252 5/4/4/71 Spaderna (2003) −.019 162 2/3/3/60 Umphrey (2003) .085 128 3/3/3/0 Urban, Stout, Zimet, and Blake (2006) .083 35 5/3/3/100 van ‘t Riet, Ruiter, Werrij, and de Vries (in press) −.141 124 2/3/1/86 Williams, Clarke, and Borland (2001) −.089 307 1/4/4/100 Ying (2001) concrete −.021 140 2/3/3/na Ying (2001) abstract .069 140 2/3/3/na Note: a The codings are the advocacy topic (1 = breast cancer, 2 = skin cancer, 3 = other cancer, 4 = dental problems, 5 = miscellaneous); gain kernel-state language (1 = desirable states, 2 = undesirable states, 3 = both desirable and undesirable states, 4 = indeterminate); loss kernel-state language (1 = undesirable states, 2 = desirable states, 3 = both desirable and undesirable states, 4 = indeterminate); and the percentage of female participants (na = not available). Open in new tab Table 1 Cases Analyzed Study . r . N . Codingsa . Apanovitch et al. (2003) .064 425 5/3/3/100 Arora (2000) −.157 210 4/1/2/60 Banks et al. (1995) −.011 133 1/3/1/100 Block (1993) self-exam −.222 57 2/2/1/na Brenes (1999) −.037 142 1/4/4/100 Broemer (2004) Study 2 .167 60 1/3/3/100 Chang (2007) Study 1 disclosing gum −.250 66 4/4/4/na Chang (2007) Study 1 disclosing strips .443 66 4/4/4/na Cherubini et al. (2005) fear-evoking −.164 60 3/3/3/0 Cherubini et al. (2005) reassuring −.166 60 3/3/3/0 Consedine, Horton, Magai, and Kukafka (2007) −.024 89 1/3/3/100 Cox, Cox, and Zimet (2006) Study 1 detection −.078 142 2/3/3/41 Cox and Cox (2001) anecdotal −.306 55 1/3/3/100 Cox and Cox (2001) statistical .046 55 1/2/1/100 Finney (2001) −.044 628 1/2/1/100 Fischer and Nabi (2001) skin exam .144 87 2/3/1/54 Gintner, Rectanus, Achord, and Parker (1987) .051 177 5/4/4/59 Hsiao (2002) testing-prevention −.300 46 5/3/3/56 Hsiao (2002) testing-detection .308 46 5/3/3/56 Keller et al. (2003) −.024 162 1/3/3/100 Kostygina (2007) −.066 119 3/3/3/na Kwan (1996) −.476 33 1/3/3/100 Kyriakaki (2007) diabetes health −.182 63 5/3/3/na Kyriakaki (2007) diabetes psychosocial −.193 65 5/3/3/na Lauver and Rubin (1990) −.060 116 3/1/2/100 Lee, Brown, & Blood (2000) self-exam −.106 137 2/1/2/54 Lerman et al. (1992) −.011 203 1/4/4/100 Lotto, Tasso, Carnaghi, and Rumati (2006) −.018 80 1/4/4/100 Maheswaran and Meyers-Levy (1990) .023 98 5/3/3/na Matwin (2007) strong −.039 162 5/3/3/na Matwin (2007) weak −.018 160 5/3/1/na Meyerowitz and Chaiken (1987) combined −.219 91 1/3/3/100 Myers et al. (1991) −.035 2,201 3/4/4/47 Nan (2006) Experiment 2 gain-positive loss-positive .124 68 5/1/2/na Nan (2006) Experiment 2 gain-positive loss-negative .076 67 5/1/1/na Nan (2006) Experiment 2 gain-negative loss-positive .070 68 5/2/2/na Nan (2006) Experiment 2 gain-negative loss-negative .161 67 5/2/1/na Nan (2007a) self-target .105 64 5/3/3/100 Phelan (2003) .000 60 1/4/4/100 Rothman letrolisus detection combined −.218 321 5/3/3/87 Rothman et al. (1999) Experiment 2 detection −.305 60 4/2/1/74 Ruiter, Kok, Verplanken, and van Eersel (2003) −.099 110 1/4/4/100 Schmitt (2004) −.055 150 5/4/4/100 Schneider et al. (2001) multicultural −.138 264 1/4/4/100 Schneider et al. (2001) targeted .047 264 1/4/4/100 Shen (2005) Study 2 glaucoma exam .021 252 5/4/4/71 Spaderna (2003) −.019 162 2/3/3/60 Umphrey (2003) .085 128 3/3/3/0 Urban, Stout, Zimet, and Blake (2006) .083 35 5/3/3/100 van ‘t Riet, Ruiter, Werrij, and de Vries (in press) −.141 124 2/3/1/86 Williams, Clarke, and Borland (2001) −.089 307 1/4/4/100 Ying (2001) concrete −.021 140 2/3/3/na Ying (2001) abstract .069 140 2/3/3/na Study . r . N . Codingsa . Apanovitch et al. (2003) .064 425 5/3/3/100 Arora (2000) −.157 210 4/1/2/60 Banks et al. (1995) −.011 133 1/3/1/100 Block (1993) self-exam −.222 57 2/2/1/na Brenes (1999) −.037 142 1/4/4/100 Broemer (2004) Study 2 .167 60 1/3/3/100 Chang (2007) Study 1 disclosing gum −.250 66 4/4/4/na Chang (2007) Study 1 disclosing strips .443 66 4/4/4/na Cherubini et al. (2005) fear-evoking −.164 60 3/3/3/0 Cherubini et al. (2005) reassuring −.166 60 3/3/3/0 Consedine, Horton, Magai, and Kukafka (2007) −.024 89 1/3/3/100 Cox, Cox, and Zimet (2006) Study 1 detection −.078 142 2/3/3/41 Cox and Cox (2001) anecdotal −.306 55 1/3/3/100 Cox and Cox (2001) statistical .046 55 1/2/1/100 Finney (2001) −.044 628 1/2/1/100 Fischer and Nabi (2001) skin exam .144 87 2/3/1/54 Gintner, Rectanus, Achord, and Parker (1987) .051 177 5/4/4/59 Hsiao (2002) testing-prevention −.300 46 5/3/3/56 Hsiao (2002) testing-detection .308 46 5/3/3/56 Keller et al. (2003) −.024 162 1/3/3/100 Kostygina (2007) −.066 119 3/3/3/na Kwan (1996) −.476 33 1/3/3/100 Kyriakaki (2007) diabetes health −.182 63 5/3/3/na Kyriakaki (2007) diabetes psychosocial −.193 65 5/3/3/na Lauver and Rubin (1990) −.060 116 3/1/2/100 Lee, Brown, & Blood (2000) self-exam −.106 137 2/1/2/54 Lerman et al. (1992) −.011 203 1/4/4/100 Lotto, Tasso, Carnaghi, and Rumati (2006) −.018 80 1/4/4/100 Maheswaran and Meyers-Levy (1990) .023 98 5/3/3/na Matwin (2007) strong −.039 162 5/3/3/na Matwin (2007) weak −.018 160 5/3/1/na Meyerowitz and Chaiken (1987) combined −.219 91 1/3/3/100 Myers et al. (1991) −.035 2,201 3/4/4/47 Nan (2006) Experiment 2 gain-positive loss-positive .124 68 5/1/2/na Nan (2006) Experiment 2 gain-positive loss-negative .076 67 5/1/1/na Nan (2006) Experiment 2 gain-negative loss-positive .070 68 5/2/2/na Nan (2006) Experiment 2 gain-negative loss-negative .161 67 5/2/1/na Nan (2007a) self-target .105 64 5/3/3/100 Phelan (2003) .000 60 1/4/4/100 Rothman letrolisus detection combined −.218 321 5/3/3/87 Rothman et al. (1999) Experiment 2 detection −.305 60 4/2/1/74 Ruiter, Kok, Verplanken, and van Eersel (2003) −.099 110 1/4/4/100 Schmitt (2004) −.055 150 5/4/4/100 Schneider et al. (2001) multicultural −.138 264 1/4/4/100 Schneider et al. (2001) targeted .047 264 1/4/4/100 Shen (2005) Study 2 glaucoma exam .021 252 5/4/4/71 Spaderna (2003) −.019 162 2/3/3/60 Umphrey (2003) .085 128 3/3/3/0 Urban, Stout, Zimet, and Blake (2006) .083 35 5/3/3/100 van ‘t Riet, Ruiter, Werrij, and de Vries (in press) −.141 124 2/3/1/86 Williams, Clarke, and Borland (2001) −.089 307 1/4/4/100 Ying (2001) concrete −.021 140 2/3/3/na Ying (2001) abstract .069 140 2/3/3/na Note: a The codings are the advocacy topic (1 = breast cancer, 2 = skin cancer, 3 = other cancer, 4 = dental problems, 5 = miscellaneous); gain kernel-state language (1 = desirable states, 2 = undesirable states, 3 = both desirable and undesirable states, 4 = indeterminate); loss kernel-state language (1 = undesirable states, 2 = desirable states, 3 = both desirable and undesirable states, 4 = indeterminate); and the percentage of female participants (na = not available). Open in new tab Specific detection behaviors As Table 2 indicates, gain- and loss-framed appeals did not significantly differ in persuasiveness for most kinds of detection behavior. A dependable advantage for loss-framed appeals was apparent for breast cancer detection behaviors (mean r = −.056), but not for any other category of detection behaviors.3 For the 36 cases not focused on breast cancer detection behaviors, gain- and loss-framed appeals were not significantly different: Mean r = −.029, 95% confidence interval (CI) limits of −.072 and.013 (p =.18), power =.99; Q(35) = 76.2, p <.001.4 Table 2 Summary of Results . k . N . Mean r . 95% CI . Powera . Q (df) . All cases Specific detection behaviors 53 9,145 −.039 −.072, −.007 — 99.2(52)*  Breast cancer 17 2,736 −.056 −.104, −.009 — 22.2(16)  Skin cancer 8 989 −.041 −.110,.029 .59 8.5(7)  Other cancer 6 2,684 −.037 −.075,.001 .95 3.9(5)  Dental 4 402 −.065 −.363,.246 .29 25.7(3)*  Miscellaneous 18 2,334 −.005 −.069,.059 .92 35.5(17)** Gain message kernel language  Desirable only 5 598 −.057 −.155,.042 .40 5.6(4)  Undesirable only 6 935 −.046 −.166,.074 .57 10.0(5)  Both 30 3,402 −.042 −.099,.016 .98 74.7(29)*  Indeterminate 12 4,210 −.034 −.064, −.004 — 8.2(11) Loss message kernel language  Desirable only 5 599 −.058 −.155,.040 .40 5.5(4)  Undesirable only 10 1,438 −.031 −.106,.044 .76 14.3(9)  Both 26 2,898 −.048 −.113,.018 .96 70.4(25)*  Indeterminate 12 4,210 −.034 −.064, −.004 — 8.2(11) . k . N . Mean r . 95% CI . Powera . Q (df) . All cases Specific detection behaviors 53 9,145 −.039 −.072, −.007 — 99.2(52)*  Breast cancer 17 2,736 −.056 −.104, −.009 — 22.2(16)  Skin cancer 8 989 −.041 −.110,.029 .59 8.5(7)  Other cancer 6 2,684 −.037 −.075,.001 .95 3.9(5)  Dental 4 402 −.065 −.363,.246 .29 25.7(3)*  Miscellaneous 18 2,334 −.005 −.069,.059 .92 35.5(17)** Gain message kernel language  Desirable only 5 598 −.057 −.155,.042 .40 5.6(4)  Undesirable only 6 935 −.046 −.166,.074 .57 10.0(5)  Both 30 3,402 −.042 −.099,.016 .98 74.7(29)*  Indeterminate 12 4,210 −.034 −.064, −.004 — 8.2(11) Loss message kernel language  Desirable only 5 599 −.058 −.155,.040 .40 5.5(4)  Undesirable only 10 1,438 −.031 −.106,.044 .76 14.3(9)  Both 26 2,898 −.048 −.113,.018 .96 70.4(25)*  Indeterminate 12 4,210 −.034 −.064, −.004 — 8.2(11) a These are power figures for detecting a population effect size of r =.10, assuming large heterogeneity, with a random-effects analysis,.05 alpha, and a two-tailed test (Hedges & Pigott, 2001). * p <.001. ** p <.01. Open in new tab Table 2 Summary of Results . k . N . Mean r . 95% CI . Powera . Q (df) . All cases Specific detection behaviors 53 9,145 −.039 −.072, −.007 — 99.2(52)*  Breast cancer 17 2,736 −.056 −.104, −.009 — 22.2(16)  Skin cancer 8 989 −.041 −.110,.029 .59 8.5(7)  Other cancer 6 2,684 −.037 −.075,.001 .95 3.9(5)  Dental 4 402 −.065 −.363,.246 .29 25.7(3)*  Miscellaneous 18 2,334 −.005 −.069,.059 .92 35.5(17)** Gain message kernel language  Desirable only 5 598 −.057 −.155,.042 .40 5.6(4)  Undesirable only 6 935 −.046 −.166,.074 .57 10.0(5)  Both 30 3,402 −.042 −.099,.016 .98 74.7(29)*  Indeterminate 12 4,210 −.034 −.064, −.004 — 8.2(11) Loss message kernel language  Desirable only 5 599 −.058 −.155,.040 .40 5.5(4)  Undesirable only 10 1,438 −.031 −.106,.044 .76 14.3(9)  Both 26 2,898 −.048 −.113,.018 .96 70.4(25)*  Indeterminate 12 4,210 −.034 −.064, −.004 — 8.2(11) . k . N . Mean r . 95% CI . Powera . Q (df) . All cases Specific detection behaviors 53 9,145 −.039 −.072, −.007 — 99.2(52)*  Breast cancer 17 2,736 −.056 −.104, −.009 — 22.2(16)  Skin cancer 8 989 −.041 −.110,.029 .59 8.5(7)  Other cancer 6 2,684 −.037 −.075,.001 .95 3.9(5)  Dental 4 402 −.065 −.363,.246 .29 25.7(3)*  Miscellaneous 18 2,334 −.005 −.069,.059 .92 35.5(17)** Gain message kernel language  Desirable only 5 598 −.057 −.155,.042 .40 5.6(4)  Undesirable only 6 935 −.046 −.166,.074 .57 10.0(5)  Both 30 3,402 −.042 −.099,.016 .98 74.7(29)*  Indeterminate 12 4,210 −.034 −.064, −.004 — 8.2(11) Loss message kernel language  Desirable only 5 599 −.058 −.155,.040 .40 5.5(4)  Undesirable only 10 1,438 −.031 −.106,.044 .76 14.3(9)  Both 26 2,898 −.048 −.113,.018 .96 70.4(25)*  Indeterminate 12 4,210 −.034 −.064, −.004 — 8.2(11) a These are power figures for detecting a population effect size of r =.10, assuming large heterogeneity, with a random-effects analysis,.05 alpha, and a two-tailed test (Hedges & Pigott, 2001). * p <.001. ** p <.01. Open in new tab Kernel state phrasing Relatively few studies used messages containing exclusively desirable or undesirable kernel states, as indicated in Table 2. In those studies, gain- and loss-framed appeals did not significantly differ in persuasiveness either when the gain-framed appeal had exclusively desirable kernel states (mean r = −.057) or when it had exclusively undesirable kernel states (mean r = −.046); these two mean effects were not significantly different (p =.89). Gain- and loss-framed appeals also did not differ significantly in persuasiveness either when the loss-framed appeal had exclusively desirable kernel states (mean r = −.058) or when it had exclusively undesirable kernel states (mean r = −.031); these two mean effects were not significantly different (p =.67). Discussion Overall effect O'Keefe and Jensen's (2006) review concluded that gain- and loss-framed appeals do not significantly differ in persuasiveness concerning disease detection behaviors. Because our search included studies appearing after that review, the current review had a larger number of cases (53) than did that earlier review (34). The addition of these new cases has revealed a statistically significant advantage for loss-framed appeals—but that advantage is quite small (a mean r of −.039). So on one hand, there is good evidence for some nonzero population effect, but on the other hand these results indicate that loss-framed appeals do not have some substantial advantage for encouraging disease detection behaviors. Indeed, the 95% CI around the mean effect indicates that a population value of −.08 is implausible. This must be counted a rather disappointing conclusion. Given the benefits of early recognition of disease conditions, identifying effective means of encouraging disease detection behaviors is an important task, and loss-framed appeals have appeared to offer a promising tool in this regard. But sufficient research evidence has accumulated to make it plain that the use of loss-framed rather than gain-framed appeals will generally not make much difference to persuasive success in this enterprise. Those who have the responsibility of designing messages aimed at encouraging disease detection behaviors should not believe that choosing loss-framed appeals will substantially improve persuasiveness. Kernel-state phrasing These data give no reason to suppose that the relative persuasiveness of gain- and loss-framed appeals in this domain is systematically related to whether the appeals mention desirable or undesirable outcomes (as the outcomes that are obtained or foregone). No matter whether the appeals were expressed in terms of desirable or undesirable outcomes, no significant difference emerged between gain- and loss-framed appeals. There is not as much evidence as one might like on this matter, but the research in hand gives no hint that variations in kernel-state phrasing will produce systematic differences in the relative persuasiveness of gain-framed and loss-framed appeals for disease detection behaviors. Specific detection behaviors Loss-framed appeals were dependably more persuasive than their gain-framed counterparts concerning breast cancer detection behaviors.5 This straightforwardly implies that those designing breast cancer detection messages should favor loss-framed appeals. However, even though statistically significant, the advantage for loss-framed appeals was relatively small (r = −.056); the 95% CI around the effect suggests that a population value as large as r = −.11 is implausible. So although these results recommend the use of loss-framed appeals for breast cancer detection messages, such appeals provide only a small persuasive advantage.6 With respect to the advocacy of other specific disease detection behaviors, however, the evidence in hand is insufficient to recommend the use of loss-framed appeals. For no other kind of detection behavior (other than breast cancer detection) was a statistically significant difference observed. And not even across studies of all other detection behaviors combined (i.e., excluding studies of breast cancer detection) was there a significant difference, despite considerable statistical power. The challenge of generalization The present results may offer a useful object lesson concerning generalizations in communication research (or, indeed, social-scientific research generally). Given any individual study, it is always an open question just what generalizations it underwrites. Two aspects of generalization can be usefully distinguished: The scope of the effect (the cases to which it applies) and the size of the effect. The scope of the effect One of the earliest investigations of gain–loss message framing effects was Meyerowitz and Chaiken's (1987) study of messages urging BSE, which reported a significant persuasive advantage for loss-framed appeals. This specific result might be used in the service of any number of different generalizations varying in scope. Most broadly, Meyerowitz and Chaiken's (1987) research has sometimes been interpreted as suggesting that loss-framed appeals are more effective than gain-framed appeals in general: “Typically, loss frames are more persuasive than gain-frame messages (Meyerowitz & Chaiken, 1987)” (Johnson, Maio, & Smith-McLallen, 2005, p. 640). But several reviews of gain-loss framing effects have concluded that there is no such general effect (e.g., Rothman & Salovey, 1997). A narrower scope is represented by the suggestion that loss-framed appeals are more effective than gain-framed appeals for disease detection behaviors (e.g., Salovey et al., 2002). But the present results do not unambiguously underwrite such a conclusion. There is indeed a statistically significant overall effect favoring loss-framed appeals—but that effect is not statistically significant for any specific kind of detection behavior other than breast cancer detection. Nor is the effect statistically significant across all other detection behaviors combined (i.e., excluding breast cancer detection behaviors), despite excellent statistical power. Thus, the overall significant effect might properly be described as carried by the effects obtained in breast-cancer detection studies specifically. So, more narrowly still, Meyerowitz and Chaiken's (1987) findings might be interpreted as indicating that loss-framed appeals are more effective than gain-framed appeals for breast cancer detection behaviors (only). Restricting the scope of the generalization in this way would naturally encourage attempts to account for the difference (between the effect for breast cancer detection studies and the effects for other particular detection behaviors) by identifying what is distinctive about breast cancer detection behaviors, or about the gain–loss messages used in these breast cancer detection studies, or about some other potentially differentiating factor. But these impulses would be mistaken. The advantage of loss-framed appeals is statistically significant for breast cancer detection but not for other behaviors—but this does not mean that the effect for breast cancer detection is significantly different from the effects for other behaviors. In fact, the breast cancer detection behavior effect is not significantly different from the effects for skin cancer detection behaviors (p =.72), behaviors to detect other cancers (p =.54), dental problem detection behaviors (p =.96), or miscellaneous detection behaviors (p =.21). Indeed, the effect for breast cancer detection behavior is not significantly different from the effect for all other detection behaviors combined (p =.41). So perhaps the most appropriate conclusion—with respect to the scope of the effect—might be this: There is some very small nonzero overall population effect (as indicated by the dependably nonzero overall effect observed here), with potentially some similarly small population effect for each specific detection behavior. Examination of the confidence intervals in Table 2 reveals that these results are consistent with a belief that, for example, the population effect is −.02 both overall and for each specific kind of detection behavior; that is, a value of −.02 falls within the 95% CI both for all detection cases and for each of the specific behaviors broken out here. The sample effect emerges as dependable (statistically significant) only in the case of breast cancer detection behaviors, for which there is a relatively large number of studies, but the effect for breast cancer detection behaviors is not significantly different from that for other specific detection behaviors. Thus, both the overall population effect and each behavior-specific population effect might well be some very small (but nonzero) value. There is one other potential scope condition worth considering, however. Given that Meyerowitz and Chaiken's (1987) participants were female (as were the participants in other breast-cancer detection studies), one might suspect that on disease detection topics, loss-framed appeals are more persuasive than gain-framed appeals for females but not for males. To explore this hypothesis, we examined the relationship between a study's effect size and the proportion of female participants in that study (see Table 1). The generalization under discussion implies that as the proportion of female participants increases, so should the observed advantage of loss-framed appeals. Thus, the correlation between proportion-of-female-participants and effect size is expected to be significantly negative; that is, larger proportions of female participants should be dependably associated with increasingly negative effect sizes (which indicate greater persuasive advantage to loss-framed appeals). Among the 36 cases concerning detection behaviors other than breast cancer detection, 15 did not provide information about the proportion of female participants. Across the remaining 21 studies (where the proportion-of-female participants ranged from.00 to 1.00), proportion-of-female-participants and effect size were not significantly related, r(19) =.039, p =.87.7 Thus, there is little evidence that males and females are differentially susceptible to gain- and loss-framed appeals concerning disease detection behaviors. In sum, concerning the scope of the effect, the evidence in hand is consistent with a belief that loss-framed appeals may have some very small persuasive advantage over gain-framed appeals for encouraging disease detection behaviors (both in general and for any specific sort of detection behavior). But because the various observed effects are so tightly clustered (and hence not significantly different from each other) and so small (and hence often not statistically significant), the scope of the effect cannot be identified confidently. The size of the effect Ioannidis (2005) has observed that in clinical medical research, it is not uncommon for initial studies of an intervention to report relatively strong effects but for subsequent research to yield smaller effects (for a similar conclusion about marketing research, see Hubbard & Armstrong, 1994). This pattern seems apparent here: Meyerowitz and Chaiken's (1987) early research produced an effect noticeably more powerful than many of those in later studies.8 There is an obvious moral to be drawn about the prudent interpretation of isolated research results. The need for replication of social-scientific research findings has been emphasized for quite some time (e.g., Hunter, 2001; Smith, 1970). But the research community should be interested in replications not only for whether they reproduce the direction of a previously observed effect, but also for the extent to which they reproduce the size of that effect. In considering generalizations about effect size from isolated results, it will surely be useful to attend closely to sample size, as this affects the width of the confidence interval (i.e., the range of plausible population values). Where a small-sample study yields a large effect size—one sufficiently large to be statistically significant—the inevitably wide confidence interval will indicate some caution about the possible size of the population effect. Notably, the four studies reporting a relatively large advantage for loss-framed appeals (r≤−.30) averaged fewer than 50 participants: D. Cox and Cox (2001 anecdotal; n = 55), Hsiao (2002 testing-prevention; n = 46), Kwan (1996; n = 33), and Rothman et al. (1999 Experiment 2 detection; n = 60). By comparison, the mean sample size in the other 49 studies reviewed here was 183; the median was 119. That is to say, where large effects have been reported, they have commonly been obtained with relatively small samples. Larger-sample studies tended to produce smaller effects. Plainly, large effects in small-sample studies, even if statistically significant, are invitations to misunderstandings. Attending to confidence intervals might minimize misinterpretations of small-sample effects because the wide CIs attendant to such results will indicate the considerable uncertainty about the location of the population effect. Conclusion When a research study yields statistically significant findings that promise some consequential practical application, there is an understandable eagerness to generalize the obtained results. Against the backdrop of Meyerowitz and Chaiken's (1987) early research finding, there has been no dearth of assertions that loss-framed messages are more persuasive than are gain-framed messages for encouraging disease detection behaviors. And although there is indeed a statistically significant difference in persuasiveness between gain- and loss-framed messages advocating disease detection behaviors, the difference is remarkably small—and apparently largely attributable to studies of messages aimed at encouraging breast cancer detection behaviors. It now seems clear that in advocacy of disease detection behaviors, the use of loss-framed rather than gain-framed appeals will not dramatically improve persuasiveness. Acknowledgments Thanks to Gretchen Brenes, Debbie Chang, Paolo Cherubini, Dee Lisa Cothran, Tony Cox, Laurinda Kwan, Amy Latimer, Lorella Lotto, Sonia Matwin, Robin Nabi, Xiaoli Nan, Kathleen Phelan, Rob Ruiter, Peter Salovey, Lijiang Shen, Andrea Urban, and Jonathan van ‘t Riet for supplying primary-research information. A version of this paper was presented at the 2008 conference of the International Communication Association. Notes 1 " Although these various persuasion outcomes are distinct, they are nevertheless appropriately combined as indicators of relative persuasiveness. Even if, for example, attitudes are in general more easily changed than are behaviors, this does not mean that the difference in persuasiveness between one message form and another will vary between these outcomes. Notably, Witte and Allen's (2000) meta-analytic review of fear appeal research reported that the mean effects of variations in depicted threat severity (high vs. low depicted severity) were statistically indistinguishable for attitudinal (mean r =.15), intentional (mean r =.14), and behavioral (mean r =.13) outcomes. That is, conclusions about the relative persuasiveness of high- versus low-depicted-severity messages were identical regardless of whether attitude, intention, or behavior was the outcome variable. Witte and Allen reported a similar pattern of effects for variations in depicted threat susceptibility (mean rs of.12,.17, and.14, respectively), variations in depicted response efficacy (mean rs of.14,.17, and.13), and variations in depicted self-efficacy (mean rs of.12,.17, and.13). These results, albeit limited, suggest that when one's research question concerns the relative persuasiveness of two message forms (as in the present enterprise), one's conclusions may not be much affected by whether the persuasive outcome assessed is attitude, intention, or behavior—and hence effect sizes are appropriately combined across these indicators. 2 " Because the unit of analysis was the message pair, multiple effect size estimates were obtained from a given study when the research design contained more than one message pair (e.g., Chang, 2007). In every case save one, the obtained effect sizes were based on different human samples and thus were statistically independent. The exception arises from Nan (2006), whose design contained two gain-framed appeals and two loss-framed appeals; this provided a total of four message pairs and hence four effect sizes, but some of these effect sizes had some participants in common. If one replaces those four estimates with the appropriate composite effect size that ensures complete statistical independence (r =.143, N = 135), the results (with k = 50, N = 9,010) are substantively unchanged: Mean r = −.044, 95% CI limits of −.077 and −.010 (p =.01); Q(49) = 97.7, p <.001. 3 " The 18 cases coded as “miscellaneous” detection topics were a varied lot, precluding any useful further breakdown of these cases. Five concerned a hepatitis C test, but four of these were derived from Nan's (2006) design with overlapping participants, which recommended against separate examination of results for this topic; three concerned fictional (hypothetical) diseases; two each concerned diabetes testing, cholesterol screening, and cardiovascular disease testing; one each concerned screening for Chlamydia, glaucoma, HIV, and hypertension. 4 " Throughout this report, power figures are those for detecting a population effect size of r =.10, assuming large heterogeneity, with a random-effects analysis,.05 alpha, and a two-tailed test (Hedges & Pigott, 2001). 5 " These studies varied in the specific sort of breast cancer detection behavior being advocated, but there was no dependable difference (in observed gain–loss framing persuasiveness differences) between messages that advocated mammography and messages that advocated breast self-examination (BSE). Of the 17 breast cancer detection studies, 10 concerned mammography (N = 1,986, mean r = −.041, 95% CI limits of −.086 and.005 [p =.08], power =.88; Q(9) = 9.2, p =.42); five concerned BSE (N = 354, mean r = −.116, 95% CI limits of −.292 and.068 [p =.22], power =.26; Q(4) = 11.2, p =.02); and two advocated a variety of breast cancer detection behaviors (mammography, BSE, and clinical breast examination; N = 396, mean r = −.075, 95% CI limits of −.172 and.024 [p =.14], power =.29; Q(1) =.29, p =.59). The CIs for mammography and BSE overlap substantially (the former is entirely contained in the latter), giving no reason to suspect that it makes much difference (to these effects) which breast cancer detection method was being advocated. 6 " In Brunton's (2007) study of mammography appeals, the outcome assessment asked for participants' judgments about the relative likely persuasiveness of gain- and loss-framed appeals; our review excluded such cases on the grounds that assessments of observed persuasive effects are to be preferred over assessments of perceived or expected persuasiveness. If Brunton's (2007) results (r =.438, N = 596) had been added to the set of breast cancer detection cases, the advantage for loss-framed appeals on that topic would have become nonsignificant: k = 18, N = 3,332, mean r = −.033, 95% CI limits of −.143 and.077 (p =.55), power =.98; Q(17) = 155.4, p <.001. The point is not that the apparent advantage of loss-framed appeals for breast-cancer detection is not genuine, but rather that it can be dangerous to fail to distinguish assessments of observed persuasive effects from assessments of expected and perceived persuasive effects. Assessments of perceived and observed effects do covary (Dillard, Weber, & Vail, 2007), and there are circumstances in which assessment of perceived effects might appropriately and justifiably be used as a proxy for assessment of observed effects (as in formative-research circumstances where assessment of observed effects is impractical or impossible; Dillard et al., 2007), but where assessments of observed effects are available they should (ceteris paribus) be preferred. 7 " These 21 cases included four studies of sex-specific topics, namely, prostate cancer screening (Cherubini, Rumiati, Rossi, Nigro, & Calabro, 2005, fear-evoking and reassuring), colposcopy (Lauver & Rubin, 1990), and testicle self-examination (Umphrey, 2003). Excluding these cases, across the remaining 17 studies (where the proportion of female participants ranged from.41 to 1.00), proportion-of-female-participants and effect size were not significantly related, r(15) = −.038, p =.88. 8 " In Meyerowitz and Chaiken's (1987) study, the effect size was r = −.280 with N = 44. In Lalor's (1990) subsequent research using the same messages, the effect was −.160 with N = 47. The composite effect, as reported in Table 1, is r = −.219. References References marked with an asterisk indicate studies included in the meta-analysis. Abood , D. A. , Coster , D. C., Mullis , A. K., & Black , D. R. ( 2002 ). Evaluation of a “loss-framed” minimal intervention to increase mammography utilization among medically un- and under-insured women . Cancer Detection and Prevention , 26 , 394 – 400 . Google Scholar Crossref Search ADS PubMed WorldCat * Apanovitch , A. M. , McCarthy , D., & Salovey , P. ( 2003 ). Using message framing to motivate HIV testing among low-income, ethnic minority women . Health Psychology , 22 , 60 – 67 . Google Scholar Crossref Search ADS PubMed WorldCat * Arora , R . ( 2000 ). Message framing and credibility: Application in dental services . Health Marketing Quarterly , 18 , 29 – 44 . Google Scholar Crossref Search ADS PubMed WorldCat * Banks , S. M. , Salovey , P., Greener , S., Rothman , A. J., Moyer , A., Beauvais , J., et al. ( 1995 ). The effects of message framing on mammography utilization . Health Psychology , 14 , 178 – 184 . Google Scholar Crossref Search ADS PubMed WorldCat * Block , L. G . ( 1993 ). The effects of perceived efficacy, message framing and vividness on the persuasiveness of a fear appeal (Doctoral dissertation, Columbia University, 1993) . Dissertation Abstracts International, 54 (1994), 2652A . (UMI No. AAC-9333731). * Block , L. G. , & Keller , P. A. ( 1995 ). When to accentuate the negative: The effects of perceived efficacy and message framing on intentions to perform a health-related behavior . Journal of Marketing Research , 32 , 192 – 203 . Google Scholar Crossref Search ADS WorldCat Borenstein , M. , & Rothstein , H. ( 2005 ). Comprehensive meta-analysis (Version 2.2.023) [Computer software]. Englewood, NJ : Biostat . * Brenes , G. A . ( 1998 ). The effects of message frame, perceived risk, and response efficacy on mammography compliance (Doctoral dissertation, Washington University, 1998) . Dissertation Abstracts International, 59 (1999), 5071B . (UMI No. AAT-9905193). * Brenes , G. A . (1999, August). The effects of message framing on mammography compliance . Paper presented at the annual meeting of the American Psychological Association, Boston, MA. Broemer , P . ( 2002 ). Relative effectiveness of differently framed health messages: The influence of ambivalence . European Journal of Social Psychology , 32 , 685 – 703 . Google Scholar Crossref Search ADS WorldCat * Broemer , P . ( 2004 ). Ease of imagination moderates reactions to differently framed health messages . European Journal of Social Psychology , 34 , 103 – 119 . Google Scholar Crossref Search ADS WorldCat Brunton , M. A . ( 2007 ). One message for all? Framing public health messages to recognize diversity . International Journal of Intercultural Relations , 31 , 127 – 132 . Google Scholar Crossref Search ADS WorldCat * Chang , C.-T . ( 2007 ). Health-care product advertising: The influences of message framing and perceived product characteristics . Psychology and Marketing , 24 , 143 – 169 . Google Scholar Crossref Search ADS WorldCat * Cherubini , P. , Rumiati , R., Rossi , D., Nigro , F., & Calabro , A. ( 2005 ). Improving attitudes toward prostate examinations by loss-framed appeals . Journal of Applied Social Psychology , 35 , 732 – 744 . Google Scholar Crossref Search ADS WorldCat * Consedine , N. S. , Horton , D., Magai , C., & Kukafka , R. ( 2007 ). Breast screening in response to gain, loss, and empowerment framed messages among diverse, low-income women . Journal of Health Care for the Poor and Underserved , 18 , 550 – 566 . Google Scholar Crossref Search ADS PubMed WorldCat * Cothran , D. L. , Schneider , T., & Salovey , P. (1998, July). The effects of message framing and involvement on mammography use in low income women . Paper presented at the Yale Summer Fellows' Conference, New Haven, CT. * Cox , A. D. , Cox , D., & Zimet , G. ( 2006 ). Understanding consumer responses to product risk information . Journal of Marketing , 70 , 79 – 91 . Google Scholar Crossref Search ADS WorldCat * Cox , D. S. , & Cox , A. D. ( 2001 ). Communicating the consequences of early detection: The role of evidence and framing . Journal of Marketing , 65 , 91 – 103 . Google Scholar Crossref Search ADS WorldCat Dillard , J. P. , & Marshall , L. J. ( 2003 ). Persuasion as a social skill. In J. O. Greene & B. R. Burleson (Eds.), Handbook of communication and social interaction skills (pp. 479 – 513 ). Mahwah, NJ : Erlbaum . Google Scholar Google Preview OpenURL Placeholder Text WorldCat COPAC Dillard , J. P. , Weber , K. M., & Vail , R. G. ( 2007 ). The relationship between the perceived and actual effectiveness of persuasive messages: A meta-analysis with implications for formative campaign research . Journal of Communication , 57 , 613 – 631 . Google Scholar Crossref Search ADS WorldCat Edwards , A. , Elwyn , G., Covey , J., Matthews , E., & Pill , R. ( 2001 ). Presenting risk information: A review of the effects of “framing” and other manipulations on patient outcomes . Journal of Health Communication , 6 , 61 – 82 . Google Scholar Crossref Search ADS PubMed WorldCat * Finney , L. J . ( 2001 ). Health beliefs, message framing and mammography screening compliance: Measurement development and theory testing (Doctoral dissertation, Miami University, 2001) . Dissertation Abstracts International, 62 (2001), 1641B . (UMI No. AAT-3009844). * Finney , L. J. , & Iannotti , R. J. ( 2002 ). Message framing and mammography screening: A theory-driven intervention . Behavioral Medicine , 28 , 5 – 14 . Google Scholar Crossref Search ADS PubMed WorldCat * Fischer , J. C. , & Nabi , R. (2001, May). Priming frames: Can framing effects extend beyond message topic? Paper presented at the annual meeting of the International Communication Association, Washington, DC. * Gintner , G. G. , Rectanus , E. F., Achord , K., & Parker , B. ( 1987 ). Parental history of hypertension and screening attendance: Effects of wellness appeal versus threat appeal . Health Psychology , 6 , 431 – 444 . Google Scholar Crossref Search ADS PubMed WorldCat Hedges , L. V. , & Pigott , T. D. ( 2001 ). The power of statistical tests in meta-analysis . Psychological Methods , 6 , 203 – 217 . Google Scholar Crossref Search ADS PubMed WorldCat Hedges , L. V. , & Vevea , J. L. ( 1998 ). Fixed- and random-effects models in meta-analysis . Psychological Methods , 3 , 486 – 504 . Google Scholar Crossref Search ADS WorldCat Helfand , M. , Mahon , S. M., Eden , K. B., Frame , P. S., & Orleans , C. T. ( 2001 ). Screening for skin cancer . American Journal of Preventive Medicine , 20 , 47 – 58 . Google Scholar Crossref Search ADS PubMed WorldCat * Hsiao , E. T. Y . ( 2002 ). Using message framing to promote regular physical activity in college-age women and men (Doctoral dissertation, Ohio State University, 2002) . Dissertation Abstracts International, 63 (2003), 3461B . (UMI No. AAT-3059265). Hubbard , R. , & Armstrong , J. S. ( 1994 ). Replications and extensions in marketing: Rarely published but quite contrary . International Journal of Research in Marketing , 11 , 233 – 248 . Google Scholar Crossref Search ADS WorldCat Hunter , J. E . ( 2001 ). The desperate need for replications . Journal of Consumer Research , 28 , 149 – 158 . Google Scholar Crossref Search ADS WorldCat Ioannidis , J. P . ( 2005 ). Contradicted and initially stronger effects in highly cited clinical research . JAMA , 294 , 218 – 228 . Google Scholar Crossref Search ADS PubMed WorldCat Johnson , B. T . ( 1989 ). DSTAT: Software for the meta-analytic review of research literatures . Hillsdale, NJ : Erlbaum . Google Scholar Google Preview OpenURL Placeholder Text WorldCat COPAC Johnson , B. T . ( 1993 ). DSTAT (Version 1.10) [Computer software] . Hillsdale, NJ: Erlbaum. Johnson , B. T. , Maio , G. R., & Smith-McLallen , A. ( 2005 ). Communication and attitude change: Causes, processes, and effects. In D. Albarracin, B. T. Johnson, & M. P. Zanna (Eds.), The handbook of attitudes (pp. 617 – 669 ). Mahwah, NJ : Erlbaum . Google Scholar Google Preview OpenURL Placeholder Text WorldCat COPAC Kahneman , D. , & Tversky , A. ( 1979 ). Prospect theory: An analysis of decision under risk . Econometrica , 47 , 263 – 291 . Google Scholar Crossref Search ADS WorldCat Kalichman , S. C. , & Coley , B. ( 1995 ). Context framing to enhance HIV-antibody-testing messages targeted to African American women . Health Psychology , 14 , 247 – 254 . Google Scholar Crossref Search ADS PubMed WorldCat * Keller , P. A. , Lipkus , I. M., & Rimer , B. K. ( 2003 ). Affect, framing, and persuasion . Journal of Marketing Research , 40 , 54 – 64 . Google Scholar Crossref Search ADS WorldCat * Kostygina , G. Y . ( 2007 ). Mitigating Chernobyl's lingering threat: What messages might motivate Ukraine's radiation-exposed youth to seek cancer screening tests? (Doctoral dissertation, University of Southern California, 2007). Dissertation Abstracts International, 69 (2008). (UMI No. AAT-3291863). Kuhberger , A . ( 1998 ). The influence of framing on risky decisions: A meta-analysis . Organizational Behavior and Human Decision Processes , 75 , 23 – 55 . Google Scholar Crossref Search ADS PubMed WorldCat * Kwan , L. L . ( 1996 ). The effects of message framing on breast self-examination: Intentions and behaviours of university women . Unpublished undergraduate honors thesis, Department of Psychology, York University, Toronto, Ontario, Canada. * Kyriakaki , M . ( 2007 ). Promotion of physical health behaviours: “Framing” the persuasive message (Doctoral dissertation, University of Essex, 2007) . Dissertation Abstracts International, 68 (2007), 1155C . (UMI No. AAT-C829783). * Lalor , K. M . ( 1990 ). The effects of message framing and feelings of susceptibility to breast cancer on reported frequency of breast self-examination . Unpublished master's thesis, University of Southern Mississippi, Hattiesburg, Mississippi. * Lalor , K. M. , & Hailey , B. J. ( 1990 ). The effects of message framing and feelings of susceptibility to breast cancer on reported frequency of breast self-examination . International Quarterly of Community Health Education , 10 , 183 – 192 . Google Scholar Crossref Search ADS WorldCat * Lauver , D. , & Rubin , M. ( 1990 ). Message framing, dispositional optimism, and follow-up for abnormal Papanicolaou tests . Research in Nursing and Health , 13 , 199 – 207 . Google Scholar Crossref Search ADS PubMed WorldCat * Lee , C. K.-C. , Brown , R., & Blood , D. ( 2000 ). The effects of efficacy, cognitive processing and message framing on persuasion . Australasian Marketing Journal , 8 , 5 – 17 . Google Scholar Crossref Search ADS WorldCat * Lerman , C. , Ross , E., Boyce , A., Gorchov , P. M., McLaughlin , R., Rimer , B., et al. ( 1992 ). The impact of mailing psychoeducational materials to women with abnormal mammograms . American Journal of Public Health , 82 , 729 – 730 . Google Scholar Crossref Search ADS PubMed WorldCat * Lotto , L. , Tasso , A., Carnaghi , A., & Rumiati , R. ( 2006 ). Promuovere l’adesione alla mammografia: Effetti di frame e salienza delle informazioni [Promoting mammography adherence: Effects of frame and information salience] . Psicologia Sociale , 2 , 355 – 372 . OpenURL Placeholder Text WorldCat * Maheswaran , D. , & Meyers-Levy , J. ( 1990 ). The influence of message framing and issue involvement . Journal of Marketing Research , 27 , 361 – 371 . Google Scholar Crossref Search ADS WorldCat Martin , B. , & Lawson , R. ( 1998 ). Mood and framing effects in advertising . Australasian Marketing Journal , 6 , 35 – 50 . Google Scholar Crossref Search ADS WorldCat * Matwin , S . ( 2007 ). Ambivalence as a mediator of persuasive health appeals and the relative effectiveness of differently framed messages in regards to “risky” health decisions (Doctoral dissertation, University of Utah, 2007) . Dissertation Abstracts International, 68 (2008), 7712B . (UMI No. AAT-3288646). * Meyerowitz , B. E. , & Chaiken , S. ( 1987 ). The effect of message framing on breast self-examination attitudes, intentions, and behavior . Journal of Personality and Social Psychology , 52 , 500 – 510 . Google Scholar Crossref Search ADS PubMed WorldCat Miles , A. , Brotherstone , H., Robb , K., Atkin , W., & Wardle , J. ( 2005 ). Using loss-framed messages to promote adherence to colorectal cancer screening . Annals of Behavioral Medicine , 29 , S57 . OpenURL Placeholder Text WorldCat Miller , S. M. , Buzaglo , J. S., Simms , S. L., Green , V., Bales , C., Mangan , C. E., et al. ( 1999 ). Monitoring styles in women at risk for cervical cancer: Implications for the framing of health-relevant messages . Annals of Behavioral Medicine , 21 , 27 – 34 . Google Scholar Crossref Search ADS PubMed WorldCat Mitchell , M. M . ( 2001 ). Risk, threat, and information seeking about genital herpes: The effects of mood and message framing . Communication Studies , 52 , 141 – 152 . Google Scholar Crossref Search ADS WorldCat Mullins , R . ( 2005 ). Evaluation of a strategy to increase cervical screening through the use of personalised letters . CBRC Research Paper Series No. 15. Melbourne, Australia: Centre for Behavioural Research in Cancer, The Cancer Council Victoria. * Myers , R. E. , Ross , E. A., Wolf , T. A., Balshem , A., Jepson , C., & Millner , L. ( 1991 ). Behavioral interventions to increase adherence in colorectal cancer screening . Medical Care , 29 , 1039 – 1050 . Google Scholar Crossref Search ADS PubMed WorldCat * Nan , X . (2006, August). The relative persuasive effect of gain versus loss-framed messages: Exploring the moderating role of the desirability of end-states . Paper presented at the annual conference of the Association for Education in Journalism and Mass Communication, San Francisco, CA. * Nan , X . ( 2007a ). Does psychological reactance to loss-framed messages dissipate the negativity bias? An investigation of the message framing effect . Paper presented at the annual conference of the International Communication Association, San Francisco, CA. * Nan , X . ( 2007b ). The relative persuasive effect of gain- versus loss-framed messages: Exploring the moderating role of the desirability of end-states . Journalism and Mass Communication Quarterly , 84 , 509 – 524 . Google Scholar Crossref Search ADS WorldCat Nan , X . ( 2007c ). Social distance, framing, and judgment: A construal level perspective . Human Communication Research , 33 , 489 – 514 . Google Scholar Crossref Search ADS WorldCat O'Keefe , D. J. , & Jensen , J. D. ( 2006 ). The advantages of compliance or the disadvantages of noncompliance? A meta-analytic review of the relative persuasive effectiveness of gain-framed and loss-framed messages . Communication Yearbook , 30 , 1 – 43 . Google Scholar Crossref Search ADS WorldCat * Phelan , K . (2003, May). Effects of gain-loss message framing and knowledge of breast cancer on breast self-examinations amongst college women . Paper presented at the Psi Chi Midwestern Regional Convention, Chicago, IL. Raudenbush , S. W . ( 1994 ). Random effects models. In H. Cooper & L. V. Hedges (Eds.), The handbook of research synthesis (pp. 301 – 321 ). New York : Russell Sage Foundation . Google Scholar Google Preview OpenURL Placeholder Text WorldCat COPAC Rivers , S. E. , Salovey , P., Pizarro , D. A., Pizarro , J., & Schneider , T. R. ( 2005 ). Message framing and pap test utilization among women attending a community health clinic . Journal of Health Psychology , 10 , 65 – 77 . Google Scholar Crossref Search ADS PubMed WorldCat Roberto , A. J. , Goodall , C. E, West , P. M., & Mahan , J. D. (in press). Persuading physicians to test their patients' level of kidney functioning: The effects of message frame and point of view . Health Communication . OpenURL Placeholder Text WorldCat Rosenthal , R . ( 1991 ). Meta-analytic procedures for social research (2nd ed.). Beverly Hills, CA : Sage . Google Scholar Crossref Search ADS Google Scholar Google Preview WorldCat COPAC * Rothman , A. J. , Martino , S. C., Bedell , B. T., Detweiler , J. B., & Salovey , P. ( 1999 ). The systematic influence of gain- and loss-framed messages on interest in and use of different types of health behavior . Personality and Social Psychology Bulletin , 25 , 1355 – 1369 . Google Scholar Crossref Search ADS WorldCat Rothman , A. J. , & Salovey , P. ( 1997 ). Shaping perceptions to motivate healthy behavior: The role of message framing . Psychological Bulletin , 121 , 3 – 19 . Google Scholar Crossref Search ADS PubMed WorldCat * Ruiter , R. A. C. , Kok , G., Verplanken , B., & Van Eersel , G. ( 2003 ). Strengthening the persuasive impact of fear appeals: The role of action framing . Journal of Social Psychology , 143 , 397 – 400 . Google Scholar Crossref Search ADS PubMed WorldCat Salmon , C. T. , Loken , B., & Finnegan , J. Jr. ( 1985 ). Direct mail in a cardiovascular health campaign: Use and effectiveness . Evaluation and the Health Professions , 8 , 438 – 452 . Google Scholar Crossref Search ADS WorldCat Salovey , P. , Schneider , T. R., & Apanovitch , A. M. ( 2002 ). Message framing in the prevention and early detection of illness. In J. P. Dillard & M. Pfau (Eds.), The persuasion handbook: Developments in theory and practice (pp. 391 – 406 ). Thousand Oaks, CA : Sage . Google Scholar Crossref Search ADS Google Scholar Google Preview WorldCat COPAC Salovey , P. , & Wegener , D. T. ( 2003 ). Communicating about health: Message framing, persuasion, and health behavior. In J. Suls & K. Wallston (Eds.), Social psychology foundations of health and illness (pp. 54 – 81 ). Oxford, UK : Blackwell . Google Scholar Crossref Search ADS Google Scholar Google Preview WorldCat COPAC * Sanchez , J. C . ( 2006 ). Efectos de la presentacion del mensaje para realizar conductas saludables: el papel de la autoeficacia y de la motivacion cognitiva [Effects of message framing to produce healthy behaviour: The role of self-efficacy and cognitive motivation] . International Journal of Clinical and Health Psychology , 6 , 613 – 630 . OpenURL Placeholder Text WorldCat * Schmitt , S. K . ( 2004 ). The effects of message framing on women's goal and implementation intentions to obtain a blood cholesterol screening test (Doctoral dissertation, Walden University, 2004) . Dissertation Abstracts International, 65 (2004), 482B . (UMI No. AAT-3119877). * Schneider , T. R. , Salovey , P., Apanovitch , A. M., Pizarro , J., McCarthy , D., Zullo , J., et al. ( 2001 ). The effects of message framing and ethnic targeting on mammography use among low-income women . Health Psychology , 20 , 256 – 266 . Google Scholar Crossref Search ADS PubMed WorldCat Shadish , W. R. , & Haddock , C. K. ( 1994 ). Combining estimates of effect size. In H. Cooper & L. V. Hedges (Eds.), Handbook of research synthesis (pp. 261 – 281 ). New York : Russell Sage Foundation . Google Scholar Google Preview OpenURL Placeholder Text WorldCat COPAC * Shen , L . ( 2005 ). The interplay of message framing, cognition and affect in persuasive health communication (Doctoral dissertation, University of Wisconsin-Madison, 2005) . Dissertation Abstracts International, 66 (2005), 1561A . (UMI No. AAT-3175536). * Shen , L. , & Dillard , J. P. ( 2007 ). The influence of behavioral inhibition/approach systems and message framing on the processing of persuasive health messages . Communication Research , 34 , 433 – 467 . Google Scholar Crossref Search ADS WorldCat Smith , N. C. Jr. ( 1970 ). Replication studies: A neglected aspect of psychological research . American Psychologist , 25 , 970 – 975 . Google Scholar Crossref Search ADS WorldCat Smith , R. A. , Cokkinides , V., & Eyre , H. J. ( 2007 ). Cancer screening in the United States, 2007: A review of current guidelines, practices, and prospects . CA: A Cancer Journal for Clinicians , 57 , 90 – 104 . Google Scholar Crossref Search ADS PubMed WorldCat * Spaderna , H . ( 2003 ). Broschuren zur Fruherkennung von Hautkrebs: Personlichkeitsun-terschiede bei der Wirkung von Gewinn- und Verlustrahmung auf die Durchfuhrung von Hautselbstuntersuchung [Brochures for the early recognition of skin cancer: Personality differences and the effect of gain and loss framing on execution of skin self-examination] . Doctoral dissertation, Johannes Gutenberg University, Mainz, Germany. Steffen , V. J. , Sternberg , L., Teegarden , L. A., & Shepherd , K. ( 1994 ). Practice and persuasive frame: Effects on beliefs, intention, and performance of a cancer self-examination . Journal of Applied Social Psychology , 24 , 897 – 925 . Google Scholar Crossref Search ADS WorldCat Swetter , S. M. , Waddell , B. L., Vazquez , M. D., & Khosravi , V. S. ( 2003 ). Increased effectiveness of targeted skin cancer screening in the Veterans Affairs population of Northern California . Preventive Medicine , 36 , 164 – 171 . Google Scholar Crossref Search ADS PubMed WorldCat Thorsteinson , T. J. , & Highhouse , S. ( 2003 ). Effects of goal framing in job advertisements on organizational attractiveness . Journal of Applied Social Psychology , 33 , 2393 – 2412 . Google Scholar Crossref Search ADS WorldCat Trotto , P. A . ( 2001 ). The effect of message framing and imagination on breast self-examination compliance (Doctoral dissertation, DePaul University, 2001) . Dissertation Abstracts International, 62 (2002), 4250B . (UMI No. AAT-3025460). Umphrey , L. R . ( 2001 ). The effects of message framing and message processing on cognitive and behavioral outcomes: An examination of breast self-examination messages (Doctoral dissertation, University of Arizona, 2001) . Dissertation Abstracts International, 62 (2001), 2681B . (UMI No. AAT-3016500). * Umphrey , L. R . ( 2003 ). The effects of message framing and message processing on testicular self-examination attitudes and perceived susceptibility . Communication Research Reports , 20 , 97 – 105 . Google Scholar Crossref Search ADS WorldCat * Urban , A. B. , Stout , P. A., Zimet , G. D., & Blake , D. R. ( 2006 ). The influence of message framing on adolescent females' intention to obtain Chlamydia screening . Journal of Adolescent Health , 38 , 126 – 127 . Google Scholar Crossref Search ADS WorldCat U. S. Preventive Services Task Force . ( 2001 ). Screening for skin cancer: Recommendations and rationale . American Journal of Preventive Medicine , 20 , 44 – 46 . Crossref Search ADS PubMed WorldCat * Van‘t Riet , J. , Ruiter , R. A. C., Werrij , M. Q., & De Vries , H. (in press). Self-efficacy moderates message-framing effects: The case of skin-cancer detection . Psychology and Health . OpenURL Placeholder Text WorldCat Wilkin , H. A . (2004, November). Mixed messages: Implications of combining detection and prevention messages for HPV and cervical cancer . Paper presented at the annual meeting of the National Communication Association, Chicago, IL. * Williams , T. , Clarke , V., & Borland , R. ( 2001 ). Effects of message framing on breast-cancer-related beliefs and behaviors: The role of mediating factors . Journal of Applied Social Psychology , 31 , 925 – 950 . Google Scholar Crossref Search ADS WorldCat Wilson , D. K. , Purdon , S. E., & Wallston , K. A. ( 1988 ). Compliance to health recommendations: A theoretical overview of message framing . Health Education Research: Theory and Practice , 3 , 161 – 171 . Google Scholar Crossref Search ADS WorldCat Witte , K. , & Allen , M. ( 2000 ). A meta-analysis of fear appeals: Implications for effective public health programs . Health Education and Behavior , 27 , 591 – 615 . Google Scholar Crossref Search ADS PubMed WorldCat * Ying , J . ( 2001 ). The effects of persuasive message strategies on melanoma detection information . Unpublished master's thesis, University of Auckland, Auckland, New Zealand. © 2009 International Communication Association http://www.deepdyve.com/assets/images/DeepDyve-Logo-lg.png Journal of Communication Oxford University Press

The Relative Persuasiveness of Gain-Framed and Loss-Framed Messages for Encouraging Disease Detection Behaviors: A Meta-Analytic Review

Download PDF
21 pages

Loading next page...
 
/lp/oxford-university-press/the-relative-persuasiveness-of-gain-framed-and-loss-framed-messages-KeAVrghokh

References (71)

Publisher
Oxford University Press
Copyright
© 2009 International Communication Association
ISSN
0021-9916
eISSN
1460-2466
DOI
10.1111/j.1460-2466.2009.01417.x
Publisher site
See Article on Publisher Site

Abstract

Abstract A meta-analytic review of 53 studies (N = 9,145) finds that in messages aimed at encouraging disease detection behaviors, loss-framed appeals (which emphasize the disadvantages of noncompliance with the communicator's recommendation) are only slightly, but statistically significantly, more persuasive than gain-framed appeals (which emphasize the advantages of compliance); the difference corresponds to a correlation of −.04. Loss-framed appeals enjoy a small statistically significant advantage for messages advocating breast cancer detection behaviors, but not for any other kind of detection behavior (detection of skin cancer, other cancers, dental problems, or miscellaneous other diseases) nor for all other kinds of detection behaviors combined. Thus, in advocacy of disease detection behaviors, using loss-framed rather than gain-framed appeals is unlikely to substantially improve persuasiveness. Early detection of disease conditions can potentially have substantial benefits, such as reduced risk of death and availability of a larger number of (and less aggressive) treatment options (Smith, Cokkinides, & Eyre, 2007). The chances for early detection are enhanced by performance of appropriate screening behaviors—and hence one persuasive task facing health professionals is that of encouraging such behaviors. Even when mass screenings may not be justified by the benefits (e.g., Helfand, Mahon, Eden, Frame, & Orleans, 2001; U.S. Preventive Services Task Force, 2001), screenings targeted to specific populations may still be beneficial (e.g., Swetter, Waddell, Vazquez, & Khosravi, 2003). Among the kinds of persuasive messages that might be used to encourage disease detection behaviors, loss-framed appeals have often been suggested to be especially effective. A loss-framed persuasive appeal emphasizes the disadvantages of failing to comply with the communicator's recommendation (e.g., “If you don't have regular mammograms, you reduce your chances of detecting breast cancer at an early, more treatable stage”); the contrast is a gain-framed appeal, which emphasizes the advantages of compliance (e.g., “If you have regular mammograms, you increase your chances of detecting breast cancer at an early, more treatable stage”). Previous primary research (e.g., Meyerowitz & Chaiken, 1987) and reviews and theoretical analyses (e.g., Edwards, Elwyn, Covey, Matthews, & Pill, 2001; Kuhberger, 1998; Rothman & Salovey, 1997; Salovey, Schneider, & Apanovitch, 2002; Wilson, Purdon, & Wallston, 1988) have suggested that loss-framed appeals will enjoy a significant persuasive advantage over gain-framed appeals in the domain of disease detection behaviors. The most common explanatory framework invokes prospect-theoretic reasoning to suggest that potential losses are more motivating than potential gains when risky actions are contemplated, whereas gains are more motivating than losses for low-risk behaviors (Kahneman & Tversky, 1979). The apparent implication is that for relatively high-risk actions such as disease detection behaviors, loss-framed messages should be more persuasive than gain-framed messages (see, e.g., Salovey et al., 2002; cf. O'Keefe & Jensen, 2006, p. 23). However, a recent broad meta-analytic review concluded that loss-framed appeals are not significantly more persuasive than gain-framed appeals for encouraging disease detection behaviors (O'Keefe & Jensen, 2006). The observed direction of effect favored loss-framed appeals, but the mean effect size (across 34 studies) was r = −.03, a value not significantly different from zero. Unfortunately, that review treated disease detection behaviors as a global category and did not distinguish effects for different specific detection behaviors; indeed, to date, no systematic review appears to have done so. Thus, the primary purpose of this review was to examine whether the relative persuasiveness of gain- and loss-framed appeals varies depending on the particular detection behavior being advocated. These behavior-specific analyses were motivated by two considerations. First, such analyses have considerable practical importance; the best evidence of whether loss-framed appeals are especially persuasive for encouraging (say) skin cancer detection behaviors comes from studies specifically concerning those behaviors, not from general research on disease detection behaviors. Second, a global nonsignificant effect might mask a substantial effect occurring for some particular detection behavior; without behavior-specific analyses, such effects could go unnoticed. The secondary purpose was to explore potential moderating factors. A number of variables have been suggested as possible moderators of gain–loss message framing effects, although many candidates are unsuitable for meta-analytic examination. For example, it is not feasible to code studies for characteristics such as audience mood (e.g., Martin & Lawson, 1998; Mitchell, 2001), attitudinal ambivalence (e.g., Broemer, 2002; Matwin, 2007), self-efficacy (e.g., van ‘t Riet, Ruiter, Werrij, & de Vries, in press), and the like. However, one particular potential moderator merited attention, namely, a variation in the way in which the consequences (of adopting or not adopting the recommended behavior) were expressed. As noticed by several commentators (e.g., Dillard & Marshall, 2003; Nan, 2007b; Rothman & Salovey, 1997; Wilson et al., 1988), gain-framed and loss-framed appeals can each take two forms, with the resulting four possibilities represented in a 2 × 2 array in which the contrasts are whether (a) the outcome described is a desirable or an undesirable one and (b) the outcome is described as one that is attained (acquired, made more likely) or avoided (averted, made less likely). That is, a gain-framed appeal might take the form “If you perform the advocated behavior, desirable outcome X will be obtained” or the form “If you perform the advocated behavior, undesirable outcome Y will be avoided.” A loss-framed appeal might take the form “If you do not perform the advocated behavior, desirable outcome X will be avoided” or the form “If you do not perform the advocated behavior, undesirable outcome Y will be obtained.” This variation has previously been described as a matter of the appeals differing in the linguistic representation of the “kernel state” of the consequence under discussion (O'Keefe & Jensen, 2006). The kernel state is the basic, root state mentioned in the message's description of the consequence. For example, in an appeal such as “If you have regular skin exams, you will reduce your risk of skin cancer,” the kernel state is “skin cancer,” which is plainly an undesirable state; that is, this appeal emphasizes the desirable consequences of compliance by discussing an undesirable kernel state (“skin cancer”) that will be avoided. By comparison, “If you have regular skin exams, you will increase your chances of having healthy skin” is an appeal describing a desirable kernel state (“healthy skin”) that will be attained by compliance. A gain-framed appeal thus could be phrased entirely in terms of avoided undesirable outcomes (“heart disease,”“skin cancer,”“premature death,” etc.) and a loss-framed appeal could be phrased entirely in terms of foregone desirable outcomes (“healthy heart,”“attractive skin,”“long life,” etc.)—and hence to sort out these possible evaluative confoundings, examination of this moderator seemed warranted. Method Identification of relevant investigations Literature search Relevant research reports were located through personal knowledge of the literature, examination of previous reviews, and inspection of reference lists in previously located reports. In addition, articles were identified through computerized database searches through at least August 2008 of ABI-INFORM, Cumulative Index of Nursing and Allied Health Literature (CINAHL), Current Contents, Dissertation Abstracts, EBSCO, Educational Resources Information Center (ERIC), Linguistics and Language Behavior Abstracts, MEDLINE, and PsycINFO, using various appropriate combinations of terms such as framing, framed, frame, appeal, message, persuasion, persuasive, gain, positive, positively, benefit, loss, negative, negatively, threat, and valence. Inclusion criteria We included a study if it met three criteria. First, the study had to compare gain- and loss-framed persuasive messages. A gain-framed message emphasizes the desirable consequences of compliance with the advocated view; a loss-framed message emphasizes the undesirable consequences of noncompliance. Excluded by this criterion were studies with imperfect realizations of this message contrast, such as studies that compared one framing form with unframed information (Abood, Coster, Mullis, & Black, 2002; Kalichman & Coley, 1995). Second, the messages had to advocate disease detection behaviors (breast cancer detection behaviors such as mammography and breast self-examination [BSE], skin cancer detection behaviors such as skin examinations, etc.). We excluded studies in which the messages advocated other behaviors, such as disease prevention behaviors (e.g., Block, 1993, sun exposure condition) or behaviors unrelated to health (e.g., Thorsteinson & Highhouse, 2003). As noted by Salovey and Wegener (2003, p. 61), some health-related behaviors, such as Pap tests and colonoscopies, might plausibly be described as either (or both) a disease-detection behavior or a disease-prevention behavior; in the interest of focusing on clear-cut studies of disease detection behaviors, studies of such “dual-function” behaviors were excluded (Miles, Brotherstone, Robb, Atkin, & Wardle, 2005; Mullins, 2005; Rivers, Salovey, Pizarro, Pizzaro, & Schneider, 2005; Wilkin, 2004). Following similar reasoning, we excluded studies in which a message advocated both detection behaviors and prevention behaviors (Block & Keller, 1995, Study 1) and studies in which the disease condition to be detected was not the audience's (Nan, 2007a, other target condition; Nan, 2007c; Roberto, Goodall, West, & Mahan, in press). Third, appropriate quantitative data relevant to persuasive effects (e.g., attitude change, intention, or behavior) had to be available; where it was not provided in the report, we made efforts to obtain information from authors. Excluded by this criterion were studies of effects on other outcome variables and studies for which appropriate quantitative information could not be obtained (e.g., Brunton, 2007; Miller et al., 1999; Salmon, Loken, & Finnegan, 1985; Steffen, Sternberg, Teegarden, & Shepherd, 1994; Trotto, 2001; Umphrey, 2001). Outcome variable and effect size measure Outcome variable The outcome variable was persuasion, as assessed through attitude change, postcommunication agreement, behavioral intention, behavior, and the like. When multiple indices of persuasion (e.g., assessments of attitude and of intention) were available, we averaged the effects to yield a single summary.1 Most studies reported only short-term effects; where both long- and short-term effect size information was available (e.g., Banks et al., 1995; Finney, 2001; Lalor, 1990), only short-term effects were included to maximize comparability across studies. Effect size measure Every comparison between a gain-framed message and its loss-framed counterpart was summarized using r as the effect size measure; differences indicating greater persuasion with gain-framed messages were given a positive sign. Results not reported as correlations were converted to r using formulas provided by Johnson (1993) and Rosenthal (1991). When correlations were averaged (e.g., across several indices of persuasive effect), we computed the average using the r-to-z-to-r transformation procedure, weighted by n. Wherever possible, multiple-factor designs were analyzed by reconstituting the analysis such that individual-difference factors (but not, e.g., other experimental manipulations) were put back into the error term (following the suggestion of Johnson, 1989, p. 16). Moderator variables Specific detection behavior Cases were classified by the kind of detection behavior advocated, with five broad categories distinguished: Breast cancer detection (e.g., mammography), skin cancer detection (e.g., skin self-examination), detection of other cancers (e.g., colon cancer screenings), detection of dental problems (e.g., plaque detection), and miscellaneous detection behaviors (screenings for cholesterol, hypertension, etc.). Kernel-state phrasing The kernel states in each appeal were identified; as discussed above, a kernel state is the basic, root state mentioned in the message's description of the consequence under discussion. We coded each appeal as containing exclusively desirable kernel states (e.g., “healthy heart,”“attractive skin”), exclusively undesirable kernel states (e.g., “heart disease,”“skin cancer”), a combination of desirable and undesirable kernel states, or as indeterminate with respect to kernel-state phrasing (as when insufficient detail was available about the messages). Coding reliabilities Codings for these moderator variables were completed independently by the authors for a sample of 30 cases. Intercoder reliabilities (percent agreement and Cohen's kappa, respectively) were.97 and.95 for advocated behavior,.93 and.89 for kernel state phrasing in gain appeals, and.97 and.95 for kernel state phrasing in loss appeals. Discrepancies were resolved by discussion. D. J. O'K. coded the remaining cases. Unit of analysis The unit of analysis was the message pair, that is, the pair composed of a gain-framed message and its loss-framed counterpart. We recorded an effect size for each distinguishable message pair found in the body of studies. Most message pairs were used only in a single investigation, so only one effect size estimate was associated with the pair. However in three instances, message pairs were used in more than one study, with the result that several effect size estimates could be associated with a given message pair. These multiple estimates were averaged to yield a single summary estimate before inclusion in the analysis. Specifically, data from Experiment 1 and Experiment 2 in Keller, Lipkus, and Rimer (2003) were combined and reported as Keller et al. (2003); data from Meyerowitz and Chaiken (1987) and Lalor (1990) were combined and reported as Meyerowitz and Chaiken (1987) combined; and data from Rothman, Martino, Bedell, Detweiler, and Salovey (1999). Experiment 1 detection condition and Sanchez (2006) were combined and reported as Rothman letrolisus detection combined. Whenever a study included more than one message pair and reported data separately for each pair, each pair was treated as providing a separate effect size estimate (e.g., Ying, 2001). Some studies included more than one message pair but did not report results in ways that permitted computing separate effect sizes for each pair (Apanovitch, McCarthy, & Salovey, 2003; Schmitt, 2004; Williams, Clarke, & Borland, 2001); we computed a single effect size in such cases, with the consequence that this analysis underrepresents the amount of message-to-message effect variability in these data. In some cases, the same primary data served as the basis for multiple reports (e.g., both a dissertation and a subsequent publication). When a given investigation was reported in more than one outlet, it was treated as a single study and analyzed accordingly. The same research was reported (in whole or in part) in: Block (1993) and Block and Keller (1995, Study 2), recorded under the former; Brenes (1998) and Brenes (1999), recorded under the latter; Cothran, Schneider, and Salovey (1998) and Schneider et al. (2001), recorded under the latter; Finney (2001) and Finney and Iannotti (2002), recorded under the former; Lalor (1990) and Lalor and Hailey (1990), with, as noted above, results recorded under Meyerowitz and Chaiken (1987) combined; Nan (2006, Experiment 2) and Nan (2007b, Experiment 2), recorded under the former; Shen (2005, Study 2, glaucoma exam), and Shen and Dillard (2007, Study 2, glaucoma exam), recorded under the former. Meta-analytic procedures The individual correlations (effect sizes) were initially transformed to Fisher's zs; the zs were analyzed using random-effects procedures (specifically, those of Borenstein & Rothstein, 2005; see also Hedges & Vevea, 1998; Shadish & Haddock, 1994), with results then transformed back to r. A random-effects analysis was used in preference to a fixed-effects analysis because of an interest in generalizing across messages (for some discussion, see Hedges & Vevea, 1998; Raudenbush, 1994; Shadish & Haddock, 1994). Results Overall effect Effect sizes were available for 53 cases, with a total of 9,145 participants. Details for each included case are contained in Table 1. Across all 53 cases, loss-framed appeals were significantly more persuasive than gain-framed appeals; mean r = −.039 (p =.02).2 Table 1 Cases Analyzed Study . r . N . Codingsa . Apanovitch et al. (2003) .064 425 5/3/3/100 Arora (2000) −.157 210 4/1/2/60 Banks et al. (1995) −.011 133 1/3/1/100 Block (1993) self-exam −.222 57 2/2/1/na Brenes (1999) −.037 142 1/4/4/100 Broemer (2004) Study 2 .167 60 1/3/3/100 Chang (2007) Study 1 disclosing gum −.250 66 4/4/4/na Chang (2007) Study 1 disclosing strips .443 66 4/4/4/na Cherubini et al. (2005) fear-evoking −.164 60 3/3/3/0 Cherubini et al. (2005) reassuring −.166 60 3/3/3/0 Consedine, Horton, Magai, and Kukafka (2007) −.024 89 1/3/3/100 Cox, Cox, and Zimet (2006) Study 1 detection −.078 142 2/3/3/41 Cox and Cox (2001) anecdotal −.306 55 1/3/3/100 Cox and Cox (2001) statistical .046 55 1/2/1/100 Finney (2001) −.044 628 1/2/1/100 Fischer and Nabi (2001) skin exam .144 87 2/3/1/54 Gintner, Rectanus, Achord, and Parker (1987) .051 177 5/4/4/59 Hsiao (2002) testing-prevention −.300 46 5/3/3/56 Hsiao (2002) testing-detection .308 46 5/3/3/56 Keller et al. (2003) −.024 162 1/3/3/100 Kostygina (2007) −.066 119 3/3/3/na Kwan (1996) −.476 33 1/3/3/100 Kyriakaki (2007) diabetes health −.182 63 5/3/3/na Kyriakaki (2007) diabetes psychosocial −.193 65 5/3/3/na Lauver and Rubin (1990) −.060 116 3/1/2/100 Lee, Brown, & Blood (2000) self-exam −.106 137 2/1/2/54 Lerman et al. (1992) −.011 203 1/4/4/100 Lotto, Tasso, Carnaghi, and Rumati (2006) −.018 80 1/4/4/100 Maheswaran and Meyers-Levy (1990) .023 98 5/3/3/na Matwin (2007) strong −.039 162 5/3/3/na Matwin (2007) weak −.018 160 5/3/1/na Meyerowitz and Chaiken (1987) combined −.219 91 1/3/3/100 Myers et al. (1991) −.035 2,201 3/4/4/47 Nan (2006) Experiment 2 gain-positive loss-positive .124 68 5/1/2/na Nan (2006) Experiment 2 gain-positive loss-negative .076 67 5/1/1/na Nan (2006) Experiment 2 gain-negative loss-positive .070 68 5/2/2/na Nan (2006) Experiment 2 gain-negative loss-negative .161 67 5/2/1/na Nan (2007a) self-target .105 64 5/3/3/100 Phelan (2003) .000 60 1/4/4/100 Rothman letrolisus detection combined −.218 321 5/3/3/87 Rothman et al. (1999) Experiment 2 detection −.305 60 4/2/1/74 Ruiter, Kok, Verplanken, and van Eersel (2003) −.099 110 1/4/4/100 Schmitt (2004) −.055 150 5/4/4/100 Schneider et al. (2001) multicultural −.138 264 1/4/4/100 Schneider et al. (2001) targeted .047 264 1/4/4/100 Shen (2005) Study 2 glaucoma exam .021 252 5/4/4/71 Spaderna (2003) −.019 162 2/3/3/60 Umphrey (2003) .085 128 3/3/3/0 Urban, Stout, Zimet, and Blake (2006) .083 35 5/3/3/100 van ‘t Riet, Ruiter, Werrij, and de Vries (in press) −.141 124 2/3/1/86 Williams, Clarke, and Borland (2001) −.089 307 1/4/4/100 Ying (2001) concrete −.021 140 2/3/3/na Ying (2001) abstract .069 140 2/3/3/na Study . r . N . Codingsa . Apanovitch et al. (2003) .064 425 5/3/3/100 Arora (2000) −.157 210 4/1/2/60 Banks et al. (1995) −.011 133 1/3/1/100 Block (1993) self-exam −.222 57 2/2/1/na Brenes (1999) −.037 142 1/4/4/100 Broemer (2004) Study 2 .167 60 1/3/3/100 Chang (2007) Study 1 disclosing gum −.250 66 4/4/4/na Chang (2007) Study 1 disclosing strips .443 66 4/4/4/na Cherubini et al. (2005) fear-evoking −.164 60 3/3/3/0 Cherubini et al. (2005) reassuring −.166 60 3/3/3/0 Consedine, Horton, Magai, and Kukafka (2007) −.024 89 1/3/3/100 Cox, Cox, and Zimet (2006) Study 1 detection −.078 142 2/3/3/41 Cox and Cox (2001) anecdotal −.306 55 1/3/3/100 Cox and Cox (2001) statistical .046 55 1/2/1/100 Finney (2001) −.044 628 1/2/1/100 Fischer and Nabi (2001) skin exam .144 87 2/3/1/54 Gintner, Rectanus, Achord, and Parker (1987) .051 177 5/4/4/59 Hsiao (2002) testing-prevention −.300 46 5/3/3/56 Hsiao (2002) testing-detection .308 46 5/3/3/56 Keller et al. (2003) −.024 162 1/3/3/100 Kostygina (2007) −.066 119 3/3/3/na Kwan (1996) −.476 33 1/3/3/100 Kyriakaki (2007) diabetes health −.182 63 5/3/3/na Kyriakaki (2007) diabetes psychosocial −.193 65 5/3/3/na Lauver and Rubin (1990) −.060 116 3/1/2/100 Lee, Brown, & Blood (2000) self-exam −.106 137 2/1/2/54 Lerman et al. (1992) −.011 203 1/4/4/100 Lotto, Tasso, Carnaghi, and Rumati (2006) −.018 80 1/4/4/100 Maheswaran and Meyers-Levy (1990) .023 98 5/3/3/na Matwin (2007) strong −.039 162 5/3/3/na Matwin (2007) weak −.018 160 5/3/1/na Meyerowitz and Chaiken (1987) combined −.219 91 1/3/3/100 Myers et al. (1991) −.035 2,201 3/4/4/47 Nan (2006) Experiment 2 gain-positive loss-positive .124 68 5/1/2/na Nan (2006) Experiment 2 gain-positive loss-negative .076 67 5/1/1/na Nan (2006) Experiment 2 gain-negative loss-positive .070 68 5/2/2/na Nan (2006) Experiment 2 gain-negative loss-negative .161 67 5/2/1/na Nan (2007a) self-target .105 64 5/3/3/100 Phelan (2003) .000 60 1/4/4/100 Rothman letrolisus detection combined −.218 321 5/3/3/87 Rothman et al. (1999) Experiment 2 detection −.305 60 4/2/1/74 Ruiter, Kok, Verplanken, and van Eersel (2003) −.099 110 1/4/4/100 Schmitt (2004) −.055 150 5/4/4/100 Schneider et al. (2001) multicultural −.138 264 1/4/4/100 Schneider et al. (2001) targeted .047 264 1/4/4/100 Shen (2005) Study 2 glaucoma exam .021 252 5/4/4/71 Spaderna (2003) −.019 162 2/3/3/60 Umphrey (2003) .085 128 3/3/3/0 Urban, Stout, Zimet, and Blake (2006) .083 35 5/3/3/100 van ‘t Riet, Ruiter, Werrij, and de Vries (in press) −.141 124 2/3/1/86 Williams, Clarke, and Borland (2001) −.089 307 1/4/4/100 Ying (2001) concrete −.021 140 2/3/3/na Ying (2001) abstract .069 140 2/3/3/na Note: a The codings are the advocacy topic (1 = breast cancer, 2 = skin cancer, 3 = other cancer, 4 = dental problems, 5 = miscellaneous); gain kernel-state language (1 = desirable states, 2 = undesirable states, 3 = both desirable and undesirable states, 4 = indeterminate); loss kernel-state language (1 = undesirable states, 2 = desirable states, 3 = both desirable and undesirable states, 4 = indeterminate); and the percentage of female participants (na = not available). Open in new tab Table 1 Cases Analyzed Study . r . N . Codingsa . Apanovitch et al. (2003) .064 425 5/3/3/100 Arora (2000) −.157 210 4/1/2/60 Banks et al. (1995) −.011 133 1/3/1/100 Block (1993) self-exam −.222 57 2/2/1/na Brenes (1999) −.037 142 1/4/4/100 Broemer (2004) Study 2 .167 60 1/3/3/100 Chang (2007) Study 1 disclosing gum −.250 66 4/4/4/na Chang (2007) Study 1 disclosing strips .443 66 4/4/4/na Cherubini et al. (2005) fear-evoking −.164 60 3/3/3/0 Cherubini et al. (2005) reassuring −.166 60 3/3/3/0 Consedine, Horton, Magai, and Kukafka (2007) −.024 89 1/3/3/100 Cox, Cox, and Zimet (2006) Study 1 detection −.078 142 2/3/3/41 Cox and Cox (2001) anecdotal −.306 55 1/3/3/100 Cox and Cox (2001) statistical .046 55 1/2/1/100 Finney (2001) −.044 628 1/2/1/100 Fischer and Nabi (2001) skin exam .144 87 2/3/1/54 Gintner, Rectanus, Achord, and Parker (1987) .051 177 5/4/4/59 Hsiao (2002) testing-prevention −.300 46 5/3/3/56 Hsiao (2002) testing-detection .308 46 5/3/3/56 Keller et al. (2003) −.024 162 1/3/3/100 Kostygina (2007) −.066 119 3/3/3/na Kwan (1996) −.476 33 1/3/3/100 Kyriakaki (2007) diabetes health −.182 63 5/3/3/na Kyriakaki (2007) diabetes psychosocial −.193 65 5/3/3/na Lauver and Rubin (1990) −.060 116 3/1/2/100 Lee, Brown, & Blood (2000) self-exam −.106 137 2/1/2/54 Lerman et al. (1992) −.011 203 1/4/4/100 Lotto, Tasso, Carnaghi, and Rumati (2006) −.018 80 1/4/4/100 Maheswaran and Meyers-Levy (1990) .023 98 5/3/3/na Matwin (2007) strong −.039 162 5/3/3/na Matwin (2007) weak −.018 160 5/3/1/na Meyerowitz and Chaiken (1987) combined −.219 91 1/3/3/100 Myers et al. (1991) −.035 2,201 3/4/4/47 Nan (2006) Experiment 2 gain-positive loss-positive .124 68 5/1/2/na Nan (2006) Experiment 2 gain-positive loss-negative .076 67 5/1/1/na Nan (2006) Experiment 2 gain-negative loss-positive .070 68 5/2/2/na Nan (2006) Experiment 2 gain-negative loss-negative .161 67 5/2/1/na Nan (2007a) self-target .105 64 5/3/3/100 Phelan (2003) .000 60 1/4/4/100 Rothman letrolisus detection combined −.218 321 5/3/3/87 Rothman et al. (1999) Experiment 2 detection −.305 60 4/2/1/74 Ruiter, Kok, Verplanken, and van Eersel (2003) −.099 110 1/4/4/100 Schmitt (2004) −.055 150 5/4/4/100 Schneider et al. (2001) multicultural −.138 264 1/4/4/100 Schneider et al. (2001) targeted .047 264 1/4/4/100 Shen (2005) Study 2 glaucoma exam .021 252 5/4/4/71 Spaderna (2003) −.019 162 2/3/3/60 Umphrey (2003) .085 128 3/3/3/0 Urban, Stout, Zimet, and Blake (2006) .083 35 5/3/3/100 van ‘t Riet, Ruiter, Werrij, and de Vries (in press) −.141 124 2/3/1/86 Williams, Clarke, and Borland (2001) −.089 307 1/4/4/100 Ying (2001) concrete −.021 140 2/3/3/na Ying (2001) abstract .069 140 2/3/3/na Study . r . N . Codingsa . Apanovitch et al. (2003) .064 425 5/3/3/100 Arora (2000) −.157 210 4/1/2/60 Banks et al. (1995) −.011 133 1/3/1/100 Block (1993) self-exam −.222 57 2/2/1/na Brenes (1999) −.037 142 1/4/4/100 Broemer (2004) Study 2 .167 60 1/3/3/100 Chang (2007) Study 1 disclosing gum −.250 66 4/4/4/na Chang (2007) Study 1 disclosing strips .443 66 4/4/4/na Cherubini et al. (2005) fear-evoking −.164 60 3/3/3/0 Cherubini et al. (2005) reassuring −.166 60 3/3/3/0 Consedine, Horton, Magai, and Kukafka (2007) −.024 89 1/3/3/100 Cox, Cox, and Zimet (2006) Study 1 detection −.078 142 2/3/3/41 Cox and Cox (2001) anecdotal −.306 55 1/3/3/100 Cox and Cox (2001) statistical .046 55 1/2/1/100 Finney (2001) −.044 628 1/2/1/100 Fischer and Nabi (2001) skin exam .144 87 2/3/1/54 Gintner, Rectanus, Achord, and Parker (1987) .051 177 5/4/4/59 Hsiao (2002) testing-prevention −.300 46 5/3/3/56 Hsiao (2002) testing-detection .308 46 5/3/3/56 Keller et al. (2003) −.024 162 1/3/3/100 Kostygina (2007) −.066 119 3/3/3/na Kwan (1996) −.476 33 1/3/3/100 Kyriakaki (2007) diabetes health −.182 63 5/3/3/na Kyriakaki (2007) diabetes psychosocial −.193 65 5/3/3/na Lauver and Rubin (1990) −.060 116 3/1/2/100 Lee, Brown, & Blood (2000) self-exam −.106 137 2/1/2/54 Lerman et al. (1992) −.011 203 1/4/4/100 Lotto, Tasso, Carnaghi, and Rumati (2006) −.018 80 1/4/4/100 Maheswaran and Meyers-Levy (1990) .023 98 5/3/3/na Matwin (2007) strong −.039 162 5/3/3/na Matwin (2007) weak −.018 160 5/3/1/na Meyerowitz and Chaiken (1987) combined −.219 91 1/3/3/100 Myers et al. (1991) −.035 2,201 3/4/4/47 Nan (2006) Experiment 2 gain-positive loss-positive .124 68 5/1/2/na Nan (2006) Experiment 2 gain-positive loss-negative .076 67 5/1/1/na Nan (2006) Experiment 2 gain-negative loss-positive .070 68 5/2/2/na Nan (2006) Experiment 2 gain-negative loss-negative .161 67 5/2/1/na Nan (2007a) self-target .105 64 5/3/3/100 Phelan (2003) .000 60 1/4/4/100 Rothman letrolisus detection combined −.218 321 5/3/3/87 Rothman et al. (1999) Experiment 2 detection −.305 60 4/2/1/74 Ruiter, Kok, Verplanken, and van Eersel (2003) −.099 110 1/4/4/100 Schmitt (2004) −.055 150 5/4/4/100 Schneider et al. (2001) multicultural −.138 264 1/4/4/100 Schneider et al. (2001) targeted .047 264 1/4/4/100 Shen (2005) Study 2 glaucoma exam .021 252 5/4/4/71 Spaderna (2003) −.019 162 2/3/3/60 Umphrey (2003) .085 128 3/3/3/0 Urban, Stout, Zimet, and Blake (2006) .083 35 5/3/3/100 van ‘t Riet, Ruiter, Werrij, and de Vries (in press) −.141 124 2/3/1/86 Williams, Clarke, and Borland (2001) −.089 307 1/4/4/100 Ying (2001) concrete −.021 140 2/3/3/na Ying (2001) abstract .069 140 2/3/3/na Note: a The codings are the advocacy topic (1 = breast cancer, 2 = skin cancer, 3 = other cancer, 4 = dental problems, 5 = miscellaneous); gain kernel-state language (1 = desirable states, 2 = undesirable states, 3 = both desirable and undesirable states, 4 = indeterminate); loss kernel-state language (1 = undesirable states, 2 = desirable states, 3 = both desirable and undesirable states, 4 = indeterminate); and the percentage of female participants (na = not available). Open in new tab Specific detection behaviors As Table 2 indicates, gain- and loss-framed appeals did not significantly differ in persuasiveness for most kinds of detection behavior. A dependable advantage for loss-framed appeals was apparent for breast cancer detection behaviors (mean r = −.056), but not for any other category of detection behaviors.3 For the 36 cases not focused on breast cancer detection behaviors, gain- and loss-framed appeals were not significantly different: Mean r = −.029, 95% confidence interval (CI) limits of −.072 and.013 (p =.18), power =.99; Q(35) = 76.2, p <.001.4 Table 2 Summary of Results . k . N . Mean r . 95% CI . Powera . Q (df) . All cases Specific detection behaviors 53 9,145 −.039 −.072, −.007 — 99.2(52)*  Breast cancer 17 2,736 −.056 −.104, −.009 — 22.2(16)  Skin cancer 8 989 −.041 −.110,.029 .59 8.5(7)  Other cancer 6 2,684 −.037 −.075,.001 .95 3.9(5)  Dental 4 402 −.065 −.363,.246 .29 25.7(3)*  Miscellaneous 18 2,334 −.005 −.069,.059 .92 35.5(17)** Gain message kernel language  Desirable only 5 598 −.057 −.155,.042 .40 5.6(4)  Undesirable only 6 935 −.046 −.166,.074 .57 10.0(5)  Both 30 3,402 −.042 −.099,.016 .98 74.7(29)*  Indeterminate 12 4,210 −.034 −.064, −.004 — 8.2(11) Loss message kernel language  Desirable only 5 599 −.058 −.155,.040 .40 5.5(4)  Undesirable only 10 1,438 −.031 −.106,.044 .76 14.3(9)  Both 26 2,898 −.048 −.113,.018 .96 70.4(25)*  Indeterminate 12 4,210 −.034 −.064, −.004 — 8.2(11) . k . N . Mean r . 95% CI . Powera . Q (df) . All cases Specific detection behaviors 53 9,145 −.039 −.072, −.007 — 99.2(52)*  Breast cancer 17 2,736 −.056 −.104, −.009 — 22.2(16)  Skin cancer 8 989 −.041 −.110,.029 .59 8.5(7)  Other cancer 6 2,684 −.037 −.075,.001 .95 3.9(5)  Dental 4 402 −.065 −.363,.246 .29 25.7(3)*  Miscellaneous 18 2,334 −.005 −.069,.059 .92 35.5(17)** Gain message kernel language  Desirable only 5 598 −.057 −.155,.042 .40 5.6(4)  Undesirable only 6 935 −.046 −.166,.074 .57 10.0(5)  Both 30 3,402 −.042 −.099,.016 .98 74.7(29)*  Indeterminate 12 4,210 −.034 −.064, −.004 — 8.2(11) Loss message kernel language  Desirable only 5 599 −.058 −.155,.040 .40 5.5(4)  Undesirable only 10 1,438 −.031 −.106,.044 .76 14.3(9)  Both 26 2,898 −.048 −.113,.018 .96 70.4(25)*  Indeterminate 12 4,210 −.034 −.064, −.004 — 8.2(11) a These are power figures for detecting a population effect size of r =.10, assuming large heterogeneity, with a random-effects analysis,.05 alpha, and a two-tailed test (Hedges & Pigott, 2001). * p <.001. ** p <.01. Open in new tab Table 2 Summary of Results . k . N . Mean r . 95% CI . Powera . Q (df) . All cases Specific detection behaviors 53 9,145 −.039 −.072, −.007 — 99.2(52)*  Breast cancer 17 2,736 −.056 −.104, −.009 — 22.2(16)  Skin cancer 8 989 −.041 −.110,.029 .59 8.5(7)  Other cancer 6 2,684 −.037 −.075,.001 .95 3.9(5)  Dental 4 402 −.065 −.363,.246 .29 25.7(3)*  Miscellaneous 18 2,334 −.005 −.069,.059 .92 35.5(17)** Gain message kernel language  Desirable only 5 598 −.057 −.155,.042 .40 5.6(4)  Undesirable only 6 935 −.046 −.166,.074 .57 10.0(5)  Both 30 3,402 −.042 −.099,.016 .98 74.7(29)*  Indeterminate 12 4,210 −.034 −.064, −.004 — 8.2(11) Loss message kernel language  Desirable only 5 599 −.058 −.155,.040 .40 5.5(4)  Undesirable only 10 1,438 −.031 −.106,.044 .76 14.3(9)  Both 26 2,898 −.048 −.113,.018 .96 70.4(25)*  Indeterminate 12 4,210 −.034 −.064, −.004 — 8.2(11) . k . N . Mean r . 95% CI . Powera . Q (df) . All cases Specific detection behaviors 53 9,145 −.039 −.072, −.007 — 99.2(52)*  Breast cancer 17 2,736 −.056 −.104, −.009 — 22.2(16)  Skin cancer 8 989 −.041 −.110,.029 .59 8.5(7)  Other cancer 6 2,684 −.037 −.075,.001 .95 3.9(5)  Dental 4 402 −.065 −.363,.246 .29 25.7(3)*  Miscellaneous 18 2,334 −.005 −.069,.059 .92 35.5(17)** Gain message kernel language  Desirable only 5 598 −.057 −.155,.042 .40 5.6(4)  Undesirable only 6 935 −.046 −.166,.074 .57 10.0(5)  Both 30 3,402 −.042 −.099,.016 .98 74.7(29)*  Indeterminate 12 4,210 −.034 −.064, −.004 — 8.2(11) Loss message kernel language  Desirable only 5 599 −.058 −.155,.040 .40 5.5(4)  Undesirable only 10 1,438 −.031 −.106,.044 .76 14.3(9)  Both 26 2,898 −.048 −.113,.018 .96 70.4(25)*  Indeterminate 12 4,210 −.034 −.064, −.004 — 8.2(11) a These are power figures for detecting a population effect size of r =.10, assuming large heterogeneity, with a random-effects analysis,.05 alpha, and a two-tailed test (Hedges & Pigott, 2001). * p <.001. ** p <.01. Open in new tab Kernel state phrasing Relatively few studies used messages containing exclusively desirable or undesirable kernel states, as indicated in Table 2. In those studies, gain- and loss-framed appeals did not significantly differ in persuasiveness either when the gain-framed appeal had exclusively desirable kernel states (mean r = −.057) or when it had exclusively undesirable kernel states (mean r = −.046); these two mean effects were not significantly different (p =.89). Gain- and loss-framed appeals also did not differ significantly in persuasiveness either when the loss-framed appeal had exclusively desirable kernel states (mean r = −.058) or when it had exclusively undesirable kernel states (mean r = −.031); these two mean effects were not significantly different (p =.67). Discussion Overall effect O'Keefe and Jensen's (2006) review concluded that gain- and loss-framed appeals do not significantly differ in persuasiveness concerning disease detection behaviors. Because our search included studies appearing after that review, the current review had a larger number of cases (53) than did that earlier review (34). The addition of these new cases has revealed a statistically significant advantage for loss-framed appeals—but that advantage is quite small (a mean r of −.039). So on one hand, there is good evidence for some nonzero population effect, but on the other hand these results indicate that loss-framed appeals do not have some substantial advantage for encouraging disease detection behaviors. Indeed, the 95% CI around the mean effect indicates that a population value of −.08 is implausible. This must be counted a rather disappointing conclusion. Given the benefits of early recognition of disease conditions, identifying effective means of encouraging disease detection behaviors is an important task, and loss-framed appeals have appeared to offer a promising tool in this regard. But sufficient research evidence has accumulated to make it plain that the use of loss-framed rather than gain-framed appeals will generally not make much difference to persuasive success in this enterprise. Those who have the responsibility of designing messages aimed at encouraging disease detection behaviors should not believe that choosing loss-framed appeals will substantially improve persuasiveness. Kernel-state phrasing These data give no reason to suppose that the relative persuasiveness of gain- and loss-framed appeals in this domain is systematically related to whether the appeals mention desirable or undesirable outcomes (as the outcomes that are obtained or foregone). No matter whether the appeals were expressed in terms of desirable or undesirable outcomes, no significant difference emerged between gain- and loss-framed appeals. There is not as much evidence as one might like on this matter, but the research in hand gives no hint that variations in kernel-state phrasing will produce systematic differences in the relative persuasiveness of gain-framed and loss-framed appeals for disease detection behaviors. Specific detection behaviors Loss-framed appeals were dependably more persuasive than their gain-framed counterparts concerning breast cancer detection behaviors.5 This straightforwardly implies that those designing breast cancer detection messages should favor loss-framed appeals. However, even though statistically significant, the advantage for loss-framed appeals was relatively small (r = −.056); the 95% CI around the effect suggests that a population value as large as r = −.11 is implausible. So although these results recommend the use of loss-framed appeals for breast cancer detection messages, such appeals provide only a small persuasive advantage.6 With respect to the advocacy of other specific disease detection behaviors, however, the evidence in hand is insufficient to recommend the use of loss-framed appeals. For no other kind of detection behavior (other than breast cancer detection) was a statistically significant difference observed. And not even across studies of all other detection behaviors combined (i.e., excluding studies of breast cancer detection) was there a significant difference, despite considerable statistical power. The challenge of generalization The present results may offer a useful object lesson concerning generalizations in communication research (or, indeed, social-scientific research generally). Given any individual study, it is always an open question just what generalizations it underwrites. Two aspects of generalization can be usefully distinguished: The scope of the effect (the cases to which it applies) and the size of the effect. The scope of the effect One of the earliest investigations of gain–loss message framing effects was Meyerowitz and Chaiken's (1987) study of messages urging BSE, which reported a significant persuasive advantage for loss-framed appeals. This specific result might be used in the service of any number of different generalizations varying in scope. Most broadly, Meyerowitz and Chaiken's (1987) research has sometimes been interpreted as suggesting that loss-framed appeals are more effective than gain-framed appeals in general: “Typically, loss frames are more persuasive than gain-frame messages (Meyerowitz & Chaiken, 1987)” (Johnson, Maio, & Smith-McLallen, 2005, p. 640). But several reviews of gain-loss framing effects have concluded that there is no such general effect (e.g., Rothman & Salovey, 1997). A narrower scope is represented by the suggestion that loss-framed appeals are more effective than gain-framed appeals for disease detection behaviors (e.g., Salovey et al., 2002). But the present results do not unambiguously underwrite such a conclusion. There is indeed a statistically significant overall effect favoring loss-framed appeals—but that effect is not statistically significant for any specific kind of detection behavior other than breast cancer detection. Nor is the effect statistically significant across all other detection behaviors combined (i.e., excluding breast cancer detection behaviors), despite excellent statistical power. Thus, the overall significant effect might properly be described as carried by the effects obtained in breast-cancer detection studies specifically. So, more narrowly still, Meyerowitz and Chaiken's (1987) findings might be interpreted as indicating that loss-framed appeals are more effective than gain-framed appeals for breast cancer detection behaviors (only). Restricting the scope of the generalization in this way would naturally encourage attempts to account for the difference (between the effect for breast cancer detection studies and the effects for other particular detection behaviors) by identifying what is distinctive about breast cancer detection behaviors, or about the gain–loss messages used in these breast cancer detection studies, or about some other potentially differentiating factor. But these impulses would be mistaken. The advantage of loss-framed appeals is statistically significant for breast cancer detection but not for other behaviors—but this does not mean that the effect for breast cancer detection is significantly different from the effects for other behaviors. In fact, the breast cancer detection behavior effect is not significantly different from the effects for skin cancer detection behaviors (p =.72), behaviors to detect other cancers (p =.54), dental problem detection behaviors (p =.96), or miscellaneous detection behaviors (p =.21). Indeed, the effect for breast cancer detection behavior is not significantly different from the effect for all other detection behaviors combined (p =.41). So perhaps the most appropriate conclusion—with respect to the scope of the effect—might be this: There is some very small nonzero overall population effect (as indicated by the dependably nonzero overall effect observed here), with potentially some similarly small population effect for each specific detection behavior. Examination of the confidence intervals in Table 2 reveals that these results are consistent with a belief that, for example, the population effect is −.02 both overall and for each specific kind of detection behavior; that is, a value of −.02 falls within the 95% CI both for all detection cases and for each of the specific behaviors broken out here. The sample effect emerges as dependable (statistically significant) only in the case of breast cancer detection behaviors, for which there is a relatively large number of studies, but the effect for breast cancer detection behaviors is not significantly different from that for other specific detection behaviors. Thus, both the overall population effect and each behavior-specific population effect might well be some very small (but nonzero) value. There is one other potential scope condition worth considering, however. Given that Meyerowitz and Chaiken's (1987) participants were female (as were the participants in other breast-cancer detection studies), one might suspect that on disease detection topics, loss-framed appeals are more persuasive than gain-framed appeals for females but not for males. To explore this hypothesis, we examined the relationship between a study's effect size and the proportion of female participants in that study (see Table 1). The generalization under discussion implies that as the proportion of female participants increases, so should the observed advantage of loss-framed appeals. Thus, the correlation between proportion-of-female-participants and effect size is expected to be significantly negative; that is, larger proportions of female participants should be dependably associated with increasingly negative effect sizes (which indicate greater persuasive advantage to loss-framed appeals). Among the 36 cases concerning detection behaviors other than breast cancer detection, 15 did not provide information about the proportion of female participants. Across the remaining 21 studies (where the proportion-of-female participants ranged from.00 to 1.00), proportion-of-female-participants and effect size were not significantly related, r(19) =.039, p =.87.7 Thus, there is little evidence that males and females are differentially susceptible to gain- and loss-framed appeals concerning disease detection behaviors. In sum, concerning the scope of the effect, the evidence in hand is consistent with a belief that loss-framed appeals may have some very small persuasive advantage over gain-framed appeals for encouraging disease detection behaviors (both in general and for any specific sort of detection behavior). But because the various observed effects are so tightly clustered (and hence not significantly different from each other) and so small (and hence often not statistically significant), the scope of the effect cannot be identified confidently. The size of the effect Ioannidis (2005) has observed that in clinical medical research, it is not uncommon for initial studies of an intervention to report relatively strong effects but for subsequent research to yield smaller effects (for a similar conclusion about marketing research, see Hubbard & Armstrong, 1994). This pattern seems apparent here: Meyerowitz and Chaiken's (1987) early research produced an effect noticeably more powerful than many of those in later studies.8 There is an obvious moral to be drawn about the prudent interpretation of isolated research results. The need for replication of social-scientific research findings has been emphasized for quite some time (e.g., Hunter, 2001; Smith, 1970). But the research community should be interested in replications not only for whether they reproduce the direction of a previously observed effect, but also for the extent to which they reproduce the size of that effect. In considering generalizations about effect size from isolated results, it will surely be useful to attend closely to sample size, as this affects the width of the confidence interval (i.e., the range of plausible population values). Where a small-sample study yields a large effect size—one sufficiently large to be statistically significant—the inevitably wide confidence interval will indicate some caution about the possible size of the population effect. Notably, the four studies reporting a relatively large advantage for loss-framed appeals (r≤−.30) averaged fewer than 50 participants: D. Cox and Cox (2001 anecdotal; n = 55), Hsiao (2002 testing-prevention; n = 46), Kwan (1996; n = 33), and Rothman et al. (1999 Experiment 2 detection; n = 60). By comparison, the mean sample size in the other 49 studies reviewed here was 183; the median was 119. That is to say, where large effects have been reported, they have commonly been obtained with relatively small samples. Larger-sample studies tended to produce smaller effects. Plainly, large effects in small-sample studies, even if statistically significant, are invitations to misunderstandings. Attending to confidence intervals might minimize misinterpretations of small-sample effects because the wide CIs attendant to such results will indicate the considerable uncertainty about the location of the population effect. Conclusion When a research study yields statistically significant findings that promise some consequential practical application, there is an understandable eagerness to generalize the obtained results. Against the backdrop of Meyerowitz and Chaiken's (1987) early research finding, there has been no dearth of assertions that loss-framed messages are more persuasive than are gain-framed messages for encouraging disease detection behaviors. And although there is indeed a statistically significant difference in persuasiveness between gain- and loss-framed messages advocating disease detection behaviors, the difference is remarkably small—and apparently largely attributable to studies of messages aimed at encouraging breast cancer detection behaviors. It now seems clear that in advocacy of disease detection behaviors, the use of loss-framed rather than gain-framed appeals will not dramatically improve persuasiveness. Acknowledgments Thanks to Gretchen Brenes, Debbie Chang, Paolo Cherubini, Dee Lisa Cothran, Tony Cox, Laurinda Kwan, Amy Latimer, Lorella Lotto, Sonia Matwin, Robin Nabi, Xiaoli Nan, Kathleen Phelan, Rob Ruiter, Peter Salovey, Lijiang Shen, Andrea Urban, and Jonathan van ‘t Riet for supplying primary-research information. A version of this paper was presented at the 2008 conference of the International Communication Association. Notes 1 " Although these various persuasion outcomes are distinct, they are nevertheless appropriately combined as indicators of relative persuasiveness. Even if, for example, attitudes are in general more easily changed than are behaviors, this does not mean that the difference in persuasiveness between one message form and another will vary between these outcomes. Notably, Witte and Allen's (2000) meta-analytic review of fear appeal research reported that the mean effects of variations in depicted threat severity (high vs. low depicted severity) were statistically indistinguishable for attitudinal (mean r =.15), intentional (mean r =.14), and behavioral (mean r =.13) outcomes. That is, conclusions about the relative persuasiveness of high- versus low-depicted-severity messages were identical regardless of whether attitude, intention, or behavior was the outcome variable. Witte and Allen reported a similar pattern of effects for variations in depicted threat susceptibility (mean rs of.12,.17, and.14, respectively), variations in depicted response efficacy (mean rs of.14,.17, and.13), and variations in depicted self-efficacy (mean rs of.12,.17, and.13). These results, albeit limited, suggest that when one's research question concerns the relative persuasiveness of two message forms (as in the present enterprise), one's conclusions may not be much affected by whether the persuasive outcome assessed is attitude, intention, or behavior—and hence effect sizes are appropriately combined across these indicators. 2 " Because the unit of analysis was the message pair, multiple effect size estimates were obtained from a given study when the research design contained more than one message pair (e.g., Chang, 2007). In every case save one, the obtained effect sizes were based on different human samples and thus were statistically independent. The exception arises from Nan (2006), whose design contained two gain-framed appeals and two loss-framed appeals; this provided a total of four message pairs and hence four effect sizes, but some of these effect sizes had some participants in common. If one replaces those four estimates with the appropriate composite effect size that ensures complete statistical independence (r =.143, N = 135), the results (with k = 50, N = 9,010) are substantively unchanged: Mean r = −.044, 95% CI limits of −.077 and −.010 (p =.01); Q(49) = 97.7, p <.001. 3 " The 18 cases coded as “miscellaneous” detection topics were a varied lot, precluding any useful further breakdown of these cases. Five concerned a hepatitis C test, but four of these were derived from Nan's (2006) design with overlapping participants, which recommended against separate examination of results for this topic; three concerned fictional (hypothetical) diseases; two each concerned diabetes testing, cholesterol screening, and cardiovascular disease testing; one each concerned screening for Chlamydia, glaucoma, HIV, and hypertension. 4 " Throughout this report, power figures are those for detecting a population effect size of r =.10, assuming large heterogeneity, with a random-effects analysis,.05 alpha, and a two-tailed test (Hedges & Pigott, 2001). 5 " These studies varied in the specific sort of breast cancer detection behavior being advocated, but there was no dependable difference (in observed gain–loss framing persuasiveness differences) between messages that advocated mammography and messages that advocated breast self-examination (BSE). Of the 17 breast cancer detection studies, 10 concerned mammography (N = 1,986, mean r = −.041, 95% CI limits of −.086 and.005 [p =.08], power =.88; Q(9) = 9.2, p =.42); five concerned BSE (N = 354, mean r = −.116, 95% CI limits of −.292 and.068 [p =.22], power =.26; Q(4) = 11.2, p =.02); and two advocated a variety of breast cancer detection behaviors (mammography, BSE, and clinical breast examination; N = 396, mean r = −.075, 95% CI limits of −.172 and.024 [p =.14], power =.29; Q(1) =.29, p =.59). The CIs for mammography and BSE overlap substantially (the former is entirely contained in the latter), giving no reason to suspect that it makes much difference (to these effects) which breast cancer detection method was being advocated. 6 " In Brunton's (2007) study of mammography appeals, the outcome assessment asked for participants' judgments about the relative likely persuasiveness of gain- and loss-framed appeals; our review excluded such cases on the grounds that assessments of observed persuasive effects are to be preferred over assessments of perceived or expected persuasiveness. If Brunton's (2007) results (r =.438, N = 596) had been added to the set of breast cancer detection cases, the advantage for loss-framed appeals on that topic would have become nonsignificant: k = 18, N = 3,332, mean r = −.033, 95% CI limits of −.143 and.077 (p =.55), power =.98; Q(17) = 155.4, p <.001. The point is not that the apparent advantage of loss-framed appeals for breast-cancer detection is not genuine, but rather that it can be dangerous to fail to distinguish assessments of observed persuasive effects from assessments of expected and perceived persuasive effects. Assessments of perceived and observed effects do covary (Dillard, Weber, & Vail, 2007), and there are circumstances in which assessment of perceived effects might appropriately and justifiably be used as a proxy for assessment of observed effects (as in formative-research circumstances where assessment of observed effects is impractical or impossible; Dillard et al., 2007), but where assessments of observed effects are available they should (ceteris paribus) be preferred. 7 " These 21 cases included four studies of sex-specific topics, namely, prostate cancer screening (Cherubini, Rumiati, Rossi, Nigro, & Calabro, 2005, fear-evoking and reassuring), colposcopy (Lauver & Rubin, 1990), and testicle self-examination (Umphrey, 2003). Excluding these cases, across the remaining 17 studies (where the proportion of female participants ranged from.41 to 1.00), proportion-of-female-participants and effect size were not significantly related, r(15) = −.038, p =.88. 8 " In Meyerowitz and Chaiken's (1987) study, the effect size was r = −.280 with N = 44. In Lalor's (1990) subsequent research using the same messages, the effect was −.160 with N = 47. The composite effect, as reported in Table 1, is r = −.219. References References marked with an asterisk indicate studies included in the meta-analysis. Abood , D. A. , Coster , D. C., Mullis , A. K., & Black , D. R. ( 2002 ). Evaluation of a “loss-framed” minimal intervention to increase mammography utilization among medically un- and under-insured women . Cancer Detection and Prevention , 26 , 394 – 400 . Google Scholar Crossref Search ADS PubMed WorldCat * Apanovitch , A. M. , McCarthy , D., & Salovey , P. ( 2003 ). Using message framing to motivate HIV testing among low-income, ethnic minority women . Health Psychology , 22 , 60 – 67 . Google Scholar Crossref Search ADS PubMed WorldCat * Arora , R . ( 2000 ). Message framing and credibility: Application in dental services . Health Marketing Quarterly , 18 , 29 – 44 . Google Scholar Crossref Search ADS PubMed WorldCat * Banks , S. M. , Salovey , P., Greener , S., Rothman , A. J., Moyer , A., Beauvais , J., et al. ( 1995 ). The effects of message framing on mammography utilization . Health Psychology , 14 , 178 – 184 . Google Scholar Crossref Search ADS PubMed WorldCat * Block , L. G . ( 1993 ). The effects of perceived efficacy, message framing and vividness on the persuasiveness of a fear appeal (Doctoral dissertation, Columbia University, 1993) . Dissertation Abstracts International, 54 (1994), 2652A . (UMI No. AAC-9333731). * Block , L. G. , & Keller , P. A. ( 1995 ). When to accentuate the negative: The effects of perceived efficacy and message framing on intentions to perform a health-related behavior . Journal of Marketing Research , 32 , 192 – 203 . Google Scholar Crossref Search ADS WorldCat Borenstein , M. , & Rothstein , H. ( 2005 ). Comprehensive meta-analysis (Version 2.2.023) [Computer software]. Englewood, NJ : Biostat . * Brenes , G. A . ( 1998 ). The effects of message frame, perceived risk, and response efficacy on mammography compliance (Doctoral dissertation, Washington University, 1998) . Dissertation Abstracts International, 59 (1999), 5071B . (UMI No. AAT-9905193). * Brenes , G. A . (1999, August). The effects of message framing on mammography compliance . Paper presented at the annual meeting of the American Psychological Association, Boston, MA. Broemer , P . ( 2002 ). Relative effectiveness of differently framed health messages: The influence of ambivalence . European Journal of Social Psychology , 32 , 685 – 703 . Google Scholar Crossref Search ADS WorldCat * Broemer , P . ( 2004 ). Ease of imagination moderates reactions to differently framed health messages . European Journal of Social Psychology , 34 , 103 – 119 . Google Scholar Crossref Search ADS WorldCat Brunton , M. A . ( 2007 ). One message for all? Framing public health messages to recognize diversity . International Journal of Intercultural Relations , 31 , 127 – 132 . Google Scholar Crossref Search ADS WorldCat * Chang , C.-T . ( 2007 ). Health-care product advertising: The influences of message framing and perceived product characteristics . Psychology and Marketing , 24 , 143 – 169 . Google Scholar Crossref Search ADS WorldCat * Cherubini , P. , Rumiati , R., Rossi , D., Nigro , F., & Calabro , A. ( 2005 ). Improving attitudes toward prostate examinations by loss-framed appeals . Journal of Applied Social Psychology , 35 , 732 – 744 . Google Scholar Crossref Search ADS WorldCat * Consedine , N. S. , Horton , D., Magai , C., & Kukafka , R. ( 2007 ). Breast screening in response to gain, loss, and empowerment framed messages among diverse, low-income women . Journal of Health Care for the Poor and Underserved , 18 , 550 – 566 . Google Scholar Crossref Search ADS PubMed WorldCat * Cothran , D. L. , Schneider , T., & Salovey , P. (1998, July). The effects of message framing and involvement on mammography use in low income women . Paper presented at the Yale Summer Fellows' Conference, New Haven, CT. * Cox , A. D. , Cox , D., & Zimet , G. ( 2006 ). Understanding consumer responses to product risk information . Journal of Marketing , 70 , 79 – 91 . Google Scholar Crossref Search ADS WorldCat * Cox , D. S. , & Cox , A. D. ( 2001 ). Communicating the consequences of early detection: The role of evidence and framing . Journal of Marketing , 65 , 91 – 103 . Google Scholar Crossref Search ADS WorldCat Dillard , J. P. , & Marshall , L. J. ( 2003 ). Persuasion as a social skill. In J. O. Greene & B. R. Burleson (Eds.), Handbook of communication and social interaction skills (pp. 479 – 513 ). Mahwah, NJ : Erlbaum . Google Scholar Google Preview OpenURL Placeholder Text WorldCat COPAC Dillard , J. P. , Weber , K. M., & Vail , R. G. ( 2007 ). The relationship between the perceived and actual effectiveness of persuasive messages: A meta-analysis with implications for formative campaign research . Journal of Communication , 57 , 613 – 631 . Google Scholar Crossref Search ADS WorldCat Edwards , A. , Elwyn , G., Covey , J., Matthews , E., & Pill , R. ( 2001 ). Presenting risk information: A review of the effects of “framing” and other manipulations on patient outcomes . Journal of Health Communication , 6 , 61 – 82 . Google Scholar Crossref Search ADS PubMed WorldCat * Finney , L. J . ( 2001 ). Health beliefs, message framing and mammography screening compliance: Measurement development and theory testing (Doctoral dissertation, Miami University, 2001) . Dissertation Abstracts International, 62 (2001), 1641B . (UMI No. AAT-3009844). * Finney , L. J. , & Iannotti , R. J. ( 2002 ). Message framing and mammography screening: A theory-driven intervention . Behavioral Medicine , 28 , 5 – 14 . Google Scholar Crossref Search ADS PubMed WorldCat * Fischer , J. C. , & Nabi , R. (2001, May). Priming frames: Can framing effects extend beyond message topic? Paper presented at the annual meeting of the International Communication Association, Washington, DC. * Gintner , G. G. , Rectanus , E. F., Achord , K., & Parker , B. ( 1987 ). Parental history of hypertension and screening attendance: Effects of wellness appeal versus threat appeal . Health Psychology , 6 , 431 – 444 . Google Scholar Crossref Search ADS PubMed WorldCat Hedges , L. V. , & Pigott , T. D. ( 2001 ). The power of statistical tests in meta-analysis . Psychological Methods , 6 , 203 – 217 . Google Scholar Crossref Search ADS PubMed WorldCat Hedges , L. V. , & Vevea , J. L. ( 1998 ). Fixed- and random-effects models in meta-analysis . Psychological Methods , 3 , 486 – 504 . Google Scholar Crossref Search ADS WorldCat Helfand , M. , Mahon , S. M., Eden , K. B., Frame , P. S., & Orleans , C. T. ( 2001 ). Screening for skin cancer . American Journal of Preventive Medicine , 20 , 47 – 58 . Google Scholar Crossref Search ADS PubMed WorldCat * Hsiao , E. T. Y . ( 2002 ). Using message framing to promote regular physical activity in college-age women and men (Doctoral dissertation, Ohio State University, 2002) . Dissertation Abstracts International, 63 (2003), 3461B . (UMI No. AAT-3059265). Hubbard , R. , & Armstrong , J. S. ( 1994 ). Replications and extensions in marketing: Rarely published but quite contrary . International Journal of Research in Marketing , 11 , 233 – 248 . Google Scholar Crossref Search ADS WorldCat Hunter , J. E . ( 2001 ). The desperate need for replications . Journal of Consumer Research , 28 , 149 – 158 . Google Scholar Crossref Search ADS WorldCat Ioannidis , J. P . ( 2005 ). Contradicted and initially stronger effects in highly cited clinical research . JAMA , 294 , 218 – 228 . Google Scholar Crossref Search ADS PubMed WorldCat Johnson , B. T . ( 1989 ). DSTAT: Software for the meta-analytic review of research literatures . Hillsdale, NJ : Erlbaum . Google Scholar Google Preview OpenURL Placeholder Text WorldCat COPAC Johnson , B. T . ( 1993 ). DSTAT (Version 1.10) [Computer software] . Hillsdale, NJ: Erlbaum. Johnson , B. T. , Maio , G. R., & Smith-McLallen , A. ( 2005 ). Communication and attitude change: Causes, processes, and effects. In D. Albarracin, B. T. Johnson, & M. P. Zanna (Eds.), The handbook of attitudes (pp. 617 – 669 ). Mahwah, NJ : Erlbaum . Google Scholar Google Preview OpenURL Placeholder Text WorldCat COPAC Kahneman , D. , & Tversky , A. ( 1979 ). Prospect theory: An analysis of decision under risk . Econometrica , 47 , 263 – 291 . Google Scholar Crossref Search ADS WorldCat Kalichman , S. C. , & Coley , B. ( 1995 ). Context framing to enhance HIV-antibody-testing messages targeted to African American women . Health Psychology , 14 , 247 – 254 . Google Scholar Crossref Search ADS PubMed WorldCat * Keller , P. A. , Lipkus , I. M., & Rimer , B. K. ( 2003 ). Affect, framing, and persuasion . Journal of Marketing Research , 40 , 54 – 64 . Google Scholar Crossref Search ADS WorldCat * Kostygina , G. Y . ( 2007 ). Mitigating Chernobyl's lingering threat: What messages might motivate Ukraine's radiation-exposed youth to seek cancer screening tests? (Doctoral dissertation, University of Southern California, 2007). Dissertation Abstracts International, 69 (2008). (UMI No. AAT-3291863). Kuhberger , A . ( 1998 ). The influence of framing on risky decisions: A meta-analysis . Organizational Behavior and Human Decision Processes , 75 , 23 – 55 . Google Scholar Crossref Search ADS PubMed WorldCat * Kwan , L. L . ( 1996 ). The effects of message framing on breast self-examination: Intentions and behaviours of university women . Unpublished undergraduate honors thesis, Department of Psychology, York University, Toronto, Ontario, Canada. * Kyriakaki , M . ( 2007 ). Promotion of physical health behaviours: “Framing” the persuasive message (Doctoral dissertation, University of Essex, 2007) . Dissertation Abstracts International, 68 (2007), 1155C . (UMI No. AAT-C829783). * Lalor , K. M . ( 1990 ). The effects of message framing and feelings of susceptibility to breast cancer on reported frequency of breast self-examination . Unpublished master's thesis, University of Southern Mississippi, Hattiesburg, Mississippi. * Lalor , K. M. , & Hailey , B. J. ( 1990 ). The effects of message framing and feelings of susceptibility to breast cancer on reported frequency of breast self-examination . International Quarterly of Community Health Education , 10 , 183 – 192 . Google Scholar Crossref Search ADS WorldCat * Lauver , D. , & Rubin , M. ( 1990 ). Message framing, dispositional optimism, and follow-up for abnormal Papanicolaou tests . Research in Nursing and Health , 13 , 199 – 207 . Google Scholar Crossref Search ADS PubMed WorldCat * Lee , C. K.-C. , Brown , R., & Blood , D. ( 2000 ). The effects of efficacy, cognitive processing and message framing on persuasion . Australasian Marketing Journal , 8 , 5 – 17 . Google Scholar Crossref Search ADS WorldCat * Lerman , C. , Ross , E., Boyce , A., Gorchov , P. M., McLaughlin , R., Rimer , B., et al. ( 1992 ). The impact of mailing psychoeducational materials to women with abnormal mammograms . American Journal of Public Health , 82 , 729 – 730 . Google Scholar Crossref Search ADS PubMed WorldCat * Lotto , L. , Tasso , A., Carnaghi , A., & Rumiati , R. ( 2006 ). Promuovere l’adesione alla mammografia: Effetti di frame e salienza delle informazioni [Promoting mammography adherence: Effects of frame and information salience] . Psicologia Sociale , 2 , 355 – 372 . OpenURL Placeholder Text WorldCat * Maheswaran , D. , & Meyers-Levy , J. ( 1990 ). The influence of message framing and issue involvement . Journal of Marketing Research , 27 , 361 – 371 . Google Scholar Crossref Search ADS WorldCat Martin , B. , & Lawson , R. ( 1998 ). Mood and framing effects in advertising . Australasian Marketing Journal , 6 , 35 – 50 . Google Scholar Crossref Search ADS WorldCat * Matwin , S . ( 2007 ). Ambivalence as a mediator of persuasive health appeals and the relative effectiveness of differently framed messages in regards to “risky” health decisions (Doctoral dissertation, University of Utah, 2007) . Dissertation Abstracts International, 68 (2008), 7712B . (UMI No. AAT-3288646). * Meyerowitz , B. E. , & Chaiken , S. ( 1987 ). The effect of message framing on breast self-examination attitudes, intentions, and behavior . Journal of Personality and Social Psychology , 52 , 500 – 510 . Google Scholar Crossref Search ADS PubMed WorldCat Miles , A. , Brotherstone , H., Robb , K., Atkin , W., & Wardle , J. ( 2005 ). Using loss-framed messages to promote adherence to colorectal cancer screening . Annals of Behavioral Medicine , 29 , S57 . OpenURL Placeholder Text WorldCat Miller , S. M. , Buzaglo , J. S., Simms , S. L., Green , V., Bales , C., Mangan , C. E., et al. ( 1999 ). Monitoring styles in women at risk for cervical cancer: Implications for the framing of health-relevant messages . Annals of Behavioral Medicine , 21 , 27 – 34 . Google Scholar Crossref Search ADS PubMed WorldCat Mitchell , M. M . ( 2001 ). Risk, threat, and information seeking about genital herpes: The effects of mood and message framing . Communication Studies , 52 , 141 – 152 . Google Scholar Crossref Search ADS WorldCat Mullins , R . ( 2005 ). Evaluation of a strategy to increase cervical screening through the use of personalised letters . CBRC Research Paper Series No. 15. Melbourne, Australia: Centre for Behavioural Research in Cancer, The Cancer Council Victoria. * Myers , R. E. , Ross , E. A., Wolf , T. A., Balshem , A., Jepson , C., & Millner , L. ( 1991 ). Behavioral interventions to increase adherence in colorectal cancer screening . Medical Care , 29 , 1039 – 1050 . Google Scholar Crossref Search ADS PubMed WorldCat * Nan , X . (2006, August). The relative persuasive effect of gain versus loss-framed messages: Exploring the moderating role of the desirability of end-states . Paper presented at the annual conference of the Association for Education in Journalism and Mass Communication, San Francisco, CA. * Nan , X . ( 2007a ). Does psychological reactance to loss-framed messages dissipate the negativity bias? An investigation of the message framing effect . Paper presented at the annual conference of the International Communication Association, San Francisco, CA. * Nan , X . ( 2007b ). The relative persuasive effect of gain- versus loss-framed messages: Exploring the moderating role of the desirability of end-states . Journalism and Mass Communication Quarterly , 84 , 509 – 524 . Google Scholar Crossref Search ADS WorldCat Nan , X . ( 2007c ). Social distance, framing, and judgment: A construal level perspective . Human Communication Research , 33 , 489 – 514 . Google Scholar Crossref Search ADS WorldCat O'Keefe , D. J. , & Jensen , J. D. ( 2006 ). The advantages of compliance or the disadvantages of noncompliance? A meta-analytic review of the relative persuasive effectiveness of gain-framed and loss-framed messages . Communication Yearbook , 30 , 1 – 43 . Google Scholar Crossref Search ADS WorldCat * Phelan , K . (2003, May). Effects of gain-loss message framing and knowledge of breast cancer on breast self-examinations amongst college women . Paper presented at the Psi Chi Midwestern Regional Convention, Chicago, IL. Raudenbush , S. W . ( 1994 ). Random effects models. In H. Cooper & L. V. Hedges (Eds.), The handbook of research synthesis (pp. 301 – 321 ). New York : Russell Sage Foundation . Google Scholar Google Preview OpenURL Placeholder Text WorldCat COPAC Rivers , S. E. , Salovey , P., Pizarro , D. A., Pizarro , J., & Schneider , T. R. ( 2005 ). Message framing and pap test utilization among women attending a community health clinic . Journal of Health Psychology , 10 , 65 – 77 . Google Scholar Crossref Search ADS PubMed WorldCat Roberto , A. J. , Goodall , C. E, West , P. M., & Mahan , J. D. (in press). Persuading physicians to test their patients' level of kidney functioning: The effects of message frame and point of view . Health Communication . OpenURL Placeholder Text WorldCat Rosenthal , R . ( 1991 ). Meta-analytic procedures for social research (2nd ed.). Beverly Hills, CA : Sage . Google Scholar Crossref Search ADS Google Scholar Google Preview WorldCat COPAC * Rothman , A. J. , Martino , S. C., Bedell , B. T., Detweiler , J. B., & Salovey , P. ( 1999 ). The systematic influence of gain- and loss-framed messages on interest in and use of different types of health behavior . Personality and Social Psychology Bulletin , 25 , 1355 – 1369 . Google Scholar Crossref Search ADS WorldCat Rothman , A. J. , & Salovey , P. ( 1997 ). Shaping perceptions to motivate healthy behavior: The role of message framing . Psychological Bulletin , 121 , 3 – 19 . Google Scholar Crossref Search ADS PubMed WorldCat * Ruiter , R. A. C. , Kok , G., Verplanken , B., & Van Eersel , G. ( 2003 ). Strengthening the persuasive impact of fear appeals: The role of action framing . Journal of Social Psychology , 143 , 397 – 400 . Google Scholar Crossref Search ADS PubMed WorldCat Salmon , C. T. , Loken , B., & Finnegan , J. Jr. ( 1985 ). Direct mail in a cardiovascular health campaign: Use and effectiveness . Evaluation and the Health Professions , 8 , 438 – 452 . Google Scholar Crossref Search ADS WorldCat Salovey , P. , Schneider , T. R., & Apanovitch , A. M. ( 2002 ). Message framing in the prevention and early detection of illness. In J. P. Dillard & M. Pfau (Eds.), The persuasion handbook: Developments in theory and practice (pp. 391 – 406 ). Thousand Oaks, CA : Sage . Google Scholar Crossref Search ADS Google Scholar Google Preview WorldCat COPAC Salovey , P. , & Wegener , D. T. ( 2003 ). Communicating about health: Message framing, persuasion, and health behavior. In J. Suls & K. Wallston (Eds.), Social psychology foundations of health and illness (pp. 54 – 81 ). Oxford, UK : Blackwell . Google Scholar Crossref Search ADS Google Scholar Google Preview WorldCat COPAC * Sanchez , J. C . ( 2006 ). Efectos de la presentacion del mensaje para realizar conductas saludables: el papel de la autoeficacia y de la motivacion cognitiva [Effects of message framing to produce healthy behaviour: The role of self-efficacy and cognitive motivation] . International Journal of Clinical and Health Psychology , 6 , 613 – 630 . OpenURL Placeholder Text WorldCat * Schmitt , S. K . ( 2004 ). The effects of message framing on women's goal and implementation intentions to obtain a blood cholesterol screening test (Doctoral dissertation, Walden University, 2004) . Dissertation Abstracts International, 65 (2004), 482B . (UMI No. AAT-3119877). * Schneider , T. R. , Salovey , P., Apanovitch , A. M., Pizarro , J., McCarthy , D., Zullo , J., et al. ( 2001 ). The effects of message framing and ethnic targeting on mammography use among low-income women . Health Psychology , 20 , 256 – 266 . Google Scholar Crossref Search ADS PubMed WorldCat Shadish , W. R. , & Haddock , C. K. ( 1994 ). Combining estimates of effect size. In H. Cooper & L. V. Hedges (Eds.), Handbook of research synthesis (pp. 261 – 281 ). New York : Russell Sage Foundation . Google Scholar Google Preview OpenURL Placeholder Text WorldCat COPAC * Shen , L . ( 2005 ). The interplay of message framing, cognition and affect in persuasive health communication (Doctoral dissertation, University of Wisconsin-Madison, 2005) . Dissertation Abstracts International, 66 (2005), 1561A . (UMI No. AAT-3175536). * Shen , L. , & Dillard , J. P. ( 2007 ). The influence of behavioral inhibition/approach systems and message framing on the processing of persuasive health messages . Communication Research , 34 , 433 – 467 . Google Scholar Crossref Search ADS WorldCat Smith , N. C. Jr. ( 1970 ). Replication studies: A neglected aspect of psychological research . American Psychologist , 25 , 970 – 975 . Google Scholar Crossref Search ADS WorldCat Smith , R. A. , Cokkinides , V., & Eyre , H. J. ( 2007 ). Cancer screening in the United States, 2007: A review of current guidelines, practices, and prospects . CA: A Cancer Journal for Clinicians , 57 , 90 – 104 . Google Scholar Crossref Search ADS PubMed WorldCat * Spaderna , H . ( 2003 ). Broschuren zur Fruherkennung von Hautkrebs: Personlichkeitsun-terschiede bei der Wirkung von Gewinn- und Verlustrahmung auf die Durchfuhrung von Hautselbstuntersuchung [Brochures for the early recognition of skin cancer: Personality differences and the effect of gain and loss framing on execution of skin self-examination] . Doctoral dissertation, Johannes Gutenberg University, Mainz, Germany. Steffen , V. J. , Sternberg , L., Teegarden , L. A., & Shepherd , K. ( 1994 ). Practice and persuasive frame: Effects on beliefs, intention, and performance of a cancer self-examination . Journal of Applied Social Psychology , 24 , 897 – 925 . Google Scholar Crossref Search ADS WorldCat Swetter , S. M. , Waddell , B. L., Vazquez , M. D., & Khosravi , V. S. ( 2003 ). Increased effectiveness of targeted skin cancer screening in the Veterans Affairs population of Northern California . Preventive Medicine , 36 , 164 – 171 . Google Scholar Crossref Search ADS PubMed WorldCat Thorsteinson , T. J. , & Highhouse , S. ( 2003 ). Effects of goal framing in job advertisements on organizational attractiveness . Journal of Applied Social Psychology , 33 , 2393 – 2412 . Google Scholar Crossref Search ADS WorldCat Trotto , P. A . ( 2001 ). The effect of message framing and imagination on breast self-examination compliance (Doctoral dissertation, DePaul University, 2001) . Dissertation Abstracts International, 62 (2002), 4250B . (UMI No. AAT-3025460). Umphrey , L. R . ( 2001 ). The effects of message framing and message processing on cognitive and behavioral outcomes: An examination of breast self-examination messages (Doctoral dissertation, University of Arizona, 2001) . Dissertation Abstracts International, 62 (2001), 2681B . (UMI No. AAT-3016500). * Umphrey , L. R . ( 2003 ). The effects of message framing and message processing on testicular self-examination attitudes and perceived susceptibility . Communication Research Reports , 20 , 97 – 105 . Google Scholar Crossref Search ADS WorldCat * Urban , A. B. , Stout , P. A., Zimet , G. D., & Blake , D. R. ( 2006 ). The influence of message framing on adolescent females' intention to obtain Chlamydia screening . Journal of Adolescent Health , 38 , 126 – 127 . Google Scholar Crossref Search ADS WorldCat U. S. Preventive Services Task Force . ( 2001 ). Screening for skin cancer: Recommendations and rationale . American Journal of Preventive Medicine , 20 , 44 – 46 . Crossref Search ADS PubMed WorldCat * Van‘t Riet , J. , Ruiter , R. A. C., Werrij , M. Q., & De Vries , H. (in press). Self-efficacy moderates message-framing effects: The case of skin-cancer detection . Psychology and Health . OpenURL Placeholder Text WorldCat Wilkin , H. A . (2004, November). Mixed messages: Implications of combining detection and prevention messages for HPV and cervical cancer . Paper presented at the annual meeting of the National Communication Association, Chicago, IL. * Williams , T. , Clarke , V., & Borland , R. ( 2001 ). Effects of message framing on breast-cancer-related beliefs and behaviors: The role of mediating factors . Journal of Applied Social Psychology , 31 , 925 – 950 . Google Scholar Crossref Search ADS WorldCat Wilson , D. K. , Purdon , S. E., & Wallston , K. A. ( 1988 ). Compliance to health recommendations: A theoretical overview of message framing . Health Education Research: Theory and Practice , 3 , 161 – 171 . Google Scholar Crossref Search ADS WorldCat Witte , K. , & Allen , M. ( 2000 ). A meta-analysis of fear appeals: Implications for effective public health programs . Health Education and Behavior , 27 , 591 – 615 . Google Scholar Crossref Search ADS PubMed WorldCat * Ying , J . ( 2001 ). The effects of persuasive message strategies on melanoma detection information . Unpublished master's thesis, University of Auckland, Auckland, New Zealand. © 2009 International Communication Association

Journal

Journal of Communication – Oxford University Press

Published: Jun 1, 2009

There are no references for this article.