University of Amsterdam
Faculty of Social and Behavioural Sciences Communication Science
Moving beyond explicit measures: Implicit measures and
the effectiveness of gain- and loss-framed messages in the
promotion of dental hygiene behavior
- Masterthesis -
Research Master Communication Science
Inka Kleinschmidt 5695643
Supervisor: dr. Gert-Jan de Bruijn
Abstract
Based on prospect theory, it has been suggested that messages highlighting potential benefits of preventive health behavior (a gain frame) are more effective than messages highlighting potential losses of not engaging in the behavior (a loss frame). However, past results
regarding message framing and preventive health behaviors have been highly inconsistent and a recent meta-analysis suggests that message framing might have a small effect on behavior, but not on attitude or behavioral intention. Two studies are presented, which incorporate implicit measures in order to reach a better understanding of message framing effects. Both studies focus on one specific kind of preventive behavior, namely dental flossing. In the first study, the effect of framed messages on implicit attitude was investigated. In the second study, risk priming was applied in order to test the assumption that risk perception moderates the effectiveness of gain- and loss-framed messages. In addition, the effect of message framing on explicit attitude, behavioral intention, perceived severity, susceptibility and affect was tested and ambivalence was assessed as a moderator. A loss-framed message was found to slightly increase behavioral intention whereas a gain-framed message was found to increase positive affect, compared to the oppositely framed message. Furthermore, an interaction effect emerged between message framing and priming, whereby those primed with
risk-avoidant words perceived higher severity of not flossing after reading the loss-framed message whereas no difference between messages emerged for those primed with
risk-seeking words. No other main or interaction effect could be established and the assumption of risk-perception could not be supported. It is therefore concluded that the usefulness of
Moving beyond explicit measures: Implicit measures and the effectiveness of gain- and loss-framed messages in the promotion of dental hygiene behavior
Introduction
Persuasive messages regarding health-related behavior can generally be distinguished into two categories: The message either promotes the potential benefits of engaging in the
proposed health behavior (a gain frame) or focuses on potential losses related to not engaging in the advocated behavior (a loss frame). Prospect theory suggests that considering potential losses increases one’s willingness to take risks, whereas considering potential benefits of a certain behavior increases risk-avoidance (Tversky, & Kahnemann, 1981). Concerning health-related messages, it has therefore been suggested that the degree to which a health behavior is perceived as risky moderates the effectiveness of messages emphasizing either potential losses or gains (Rothman, & Salovey, 1997). Hereby, prevention behaviors are thought to differ from detection behaviors: Whereas prevention behavior is a mean to avoid future aversive health outcomes, hereby providing a relatively certain and desirable outcome, detection behaviors carry the risk of actually detecting a disease, a relatively uncertain and undesirable outcome. Therefore, it is argued that emphasizing potential benefits (a gain frame) is more effective for prevention behaviors, whereas emphasizing potential losses (a loss frame) is more effective regarding detection behaviors (Rothman, Bartels, Wlaschin, & Salovey, 2006).
Dental hygiene behavior, like dental flossing, is one of the commonly used examples of prevention behavior. Many dental and oral diseases, like caries and periodontal problems, are preventable and self-care behaviors like regular brushing and flossing were found to play a very important role in the prevention of those diseases (Ciancio, 2003; Hujoel, Cunha-Cruz, Banting, & Loesche, 2006). In addition, it has been suggested that dental hygiene behavior may carry extraordinary low levels of risk because the expected positive outcome is perceived
as relatively certain (O’Keefe, & Jensen, 2007). Despite the promising nature of preventive dental hygiene behaviors, 60-90 per cent of all children and nearly 100 per cent of all adults worldwide have dental cavities (WHO, 2012). In the Netherlands, 79 per cent of the
population was found to brush their teeth twice a day, whereas only 31 percent engage in any form of daily interdental cleaning (Schuller, 2009). Effective persuasive messages regarding dental hygiene behavior, and especially regarding interdental cleaning, are therefore
warranted.
Whereas theory suggests that gain-framed messages should be superior to loss-framed messages in the promotion of preventive dental hygiene behavior, meta-analyses found only limited support for the effect of message framing on the persuasiveness of health-related messages. O’Keefe and Jensen (2007) concluded that differences in the effectiveness of gain- and loss-framed messages regarding preventive behavior are generally very small and largely attributable to studies regarding dental hygiene behaviors. In contrast, Gallagher and
Updegraff (2012) did not find any difference in the effectiveness of message framing between dental hygiene behaviors and other prevention behaviors. They did, however, find large differences between the included outcome variables: Whereas message framing does not appear to have an effect on attitude and intention, it does appear to influence behavior. Hereby, gain-framed messages were found to be more effective than loss-framed messages when promoting prevention behavior. This finding, however, contradicts most prominent social-cognitive theories (e.g. the theory of planned behavior), which postulate that behavior is a result of both attitude towards the respective behavior and behavioral intention.
Therefore, the mechanism by which message framing is thought to influence behavior remains unclear. Previous studies have solely relied on explicit measures and self-report. However, many processes are not open to introspection and might therefore not be reported accurately by participants. Moving beyond explicit self-report measures by incorporating
implicit measures in the study of the effectiveness of gain- and loss-framed messages is therefore necessary.
The aim of this study is to investigate the persuasiveness of gain- and loss-framed messages regarding dental flossing. Hereby, the usefulness of incorporating implicit
measurements will be assessed. In the first study, based on preexisting data, the effect of gain- and loss-framed messages on implicit attitude towards dental flossing will be investigated. In the second study, it will be tested whether risk priming enhances the effectiveness of framed messages regarding a variety of outcome variables.
Theoretical background
Past research
Theory suggests that a gain-framed message should outperform a loss-framed message when applied to prevention behavior. However, former research on preventive dental hygiene behavior led to mixed and inconsistent results. For example, in a recent intervention study among Iranian schoolchildren (Pakpour, Yekaninejad, Sniehotta, Updegraff, & Dombrowski, 2014), a loss frame was found to outperform a gain framing regarding attitude towards flossing, but no differences were found regarding intention. In addition, a loss frame was found to increase flossing behavior, compared to the control group. Regarding tooth brushing, a gain frame as well as a loss frame was found to increase attitude towards brushing and behavioral intention after 2 weeks, compared to the control group, but no difference between the two message frames emerged. After 24 weeks, however, only the loss frame was found to have a significant effect on both attitude and intention. In addition, only the loss-framed message was found to increase actual brushing behavior. In contrast, a study on gum diseases and plaque-fighting mouth rinse (Rothman, Martino, Bedell, Detweiler, & Salovey, 1999) found that a gain frame increases both intention and behavior, compared to a loss-framed message. However, a loss frame was found to lead to more favorable thoughts, higher
perceived risk and severity, and a more positive attitude towards preventive mouth rinse than a gain-framed message. In contrast, another study on preventive mouth rinse could not detect any main effect for message framing on perceived risk, severity or behavioral intention (Chang, 2003). Several other studies could not establish any main effect of message framing on either flossing behavior (Mann, Sherman, & Updegraff, 2004; Sherman, Mann, &
Updegraff, 2006; Updegraff, Sherman, Luyster, & Mann, 2007), visits at the dentist
(Altmann, & Traxler, 2014), perception of the message (Chang, 2003; Sherman et al., 2006; Updegraff et al., 2007), flossing-related self-efficacy (Sherman et al., 2006; Updegraff et al., 2007), intention regarding flossing (Sherman et al., 2006; Updegraff et al., 2007; Uskul, Sherman, & Fitzgibbon, 2009), perceived norm (Updegraff et al., 2007), or attitude towards flossing (Updegraff et al., 2007; Uskul et al., 2009).
Overall, most studies did not show any effect of message framing on preventive dental hygiene behavior and related constructs. Significant results are often inconsistent between studies, between different kinds of dental hygiene behavior and between different outcome variables. In order to reach a better understanding of the effectiveness of message framing regarding preventive dental hygiene behaviors, the following two studies will focus on one specific kind of behavior, namely dental flossing, and the message will held constant between studies. In addition, both explicit as implicit dependent and independent measures will be applied.
Explicit vs. implicit attitude
Most social-cognitive theories suggest that behavior is the product of a positive attitude towards that specific behavior and a positive behavioral intention (Ajzen, 1991). In addition, persuasion theory proposes that exposure to a persuasive message influences behavior through a change in attitude (McGuire, 1989). Following this line of thought, a framed message successful in altering health behavior should also have a positive effect on both
attitude and intention. Prospect theory hereby suggests that regarding preventive health behavior, a gain-framed message should outperform a loss-framed message. However, prior results regarding dental flossing were highly mixed and a meta-analysis on a variety of preventive health behaviors (Gallagher, & Updegraff, 2012) suggests that, although message framing affects behavior, no relationship between message frame and attitude or intention can be found. In order to test this finding with regard to dental flossing, the following hypothesis, in line with prospect theory, is proposed:
H1: Participants presented with a gain-framed message show more positive change in attitude towards dental flossing (H1a) and behavioral intention (H1b) than those presented with a loss-framed message.
Whereas message framing might not alter explicit attitudes, it remains possible that message framing does alter implicit attitudes, hereby affecting subsequent behavior. Information regarding positive or negative outcomes, as provided in framed messages, can affect behavior in a rapid and automatic fashion and without explicit awareness (Ferguson, 2007). As
proposed by persuasion theory, persuasive messages might therefore indeed influence behavior through attitude change (McGuire, 1989), but these changes might take place on an implicit rather than on an explicit level. Given that many attitudinal processes happen out of conscious awareness and are not accessible through introspection, sole dependence on explicit measures might be insufficient. In addition, people might ‘know’ that engaging in preventive health behaviors is considered ‘good’, ‘healthy’ or ‘socially desirable’ and might therefore show an overall positive explicit attitude towards these behaviors in self-report measures, regardless of the type of persuasive message presented.
None of the previous studies has investigated the effect of message framing on implicit attitudes. Former research on attitude formation, however, has shown that implicit
and explicit attitudes are not necessarily closely related and can be influenced independently of each other (e.g. Olson, & Fazio, 2006). In addition, it has been suggested that the
effectiveness of message framing might not only depend on the type of advocated behavior (prevention vs. detection) but also on ‘how the individual thinks and feels about the behavior’ (p. 646; Latimer, Salovey, & Rothman, 2007) and that a search for relevant moderators is therefore warranted. Regarding relevant moderators of implicit attitude, Gibson (2008) demonstrated that evaluative conditioning can be used to alter implicit but not explicit attitudes toward mature brands, but that this effect only occurred for participants with an initially neutral attitude towards those brands and not for those with a strong brand
preference. In addition, a recent systematic review on message framing effects showed that ambivalence moderates the relation between message framing and explicit outcome variables (Covey, 2014). It therefore appears feasible that the effect of message framing on implicit attitude might also depend on the presence or absence of a strong behavioral preference. Hereby, those with either a positive of negative initial attitude towards flossing are considered to show a form of preference, namely towards flossing or not flossing their teeth. The same reasoning might apply to those with an initial low or high intention to floss their teeth.
However, those with a neutral attitude or moderate intention are thought to show ambivalence regarding dental flossing and their implicit attitude might therefore be suggestible to framed messages. The following hypothesis is therefore proposed:
H2: Participants presented with a gain-framed message show more positive change in implicit attitude towards dental flossing than those presented with a loss-framed message (H2a). However, this effect is more pronounced for those with an initial ambivalent attitude (H2b) or initial ambivalent intention (H2c), than for those with an initial positive or negative attitude or for those with an initial high or low intention.
Risk priming
Implicit measures can be used as dependent variable, as employed in Study 1, but might also be useful as independent variable (Study 2), namely as risk priming. Priming refers to an implicit method in which the accessibility of mental representations is enhanced, in this case either risk-seeking or risk-avoidant representations, which are in turn thought to influence motivation and behavior without conscious awareness (Bargh, & Chartrand, 2000). Prospect theory argues that considering potential gains increases risk-avoidance (Tversky, &
Kahnemann, 1981) and it has therefore been suggested that gain-framed messages increase the motivation to engage in low-risk prevention behaviors (Rothman, & Salovey, 1997). However, whether a prevention behavior is indeed perceived as ‘low risk’, and a detection behavior is in contrast always perceived as ‘high risk’ remains questionable and may vary between persons as well as between specific prevention and detection behaviors. Given that the notion of risk is central to both prospect theory and message framing theory, individual differences in risk perception might (partially) explain the mixed results regarding the effectiveness of gain- and loss-framed messages concerning preventive dental hygiene behavior.
Previous studies on individual risk perceptions and message framing all relied on self-report measures and, again, led to mixed results: One study found that when risk implications are low, a gain-framed message outperforms a loss-framed message, but that when risk implications are high, a loss frame outperforms a gain frame, regardless whether detection or prevention behaviors are promoted (Bartels, Kelly, & Rothman, 2010). In contrast, a study on framed messages regarding donor appeals found slight advantages for a loss-framed message, compared to a gain-framed message, when perceived risks are low (Cohen, 2010).
Concerning breast cancer prevention, it was found that loss-framed messages outperform gain-framed messages among women with high and moderate perceived risk and no effect of
message framing was found among women with low perceived risk (Gallagher, Updegraff, Rothman, & Sims, 2011). Yet another study investigated framing effects for six preventive as well as detective health behaviors and could not find any interaction between message
framing and perceived risk (Van ‘t Riet et al., 2014). A recent systematic review also concluded that perceived risk is not a consistent moderator of message framing effects (Covey, 2014). The sole reliance on explicit measures might hereby form a serious problem, given that personal risk perceptions regarding health hazards were found to be largely biased and overly optimistic (Weinstein, 1987). Introducing a risk-avoidant or risk-seeking mindset through priming might overcome these problems with self-reported measures. In line with prospect theory, increasing the accessibility of risk-avoidant representations should hereby enhance the persuasiveness of gain-framed messages regarding preventive dental hygiene behavior. The following hypothesis is therefore proposed:
H3: Participants presented with a gain-framed message show a more positive attitude towards dental flossing (H3a) and a higher behavioral intention (H3b) than those presented with a loss-framed message. However, this effect on attitude (H3c) and intention (H3d) is more pronounced for those who were primed with risk-avoidant words than for those who were primed with risk-seeking words.
Severity, susceptibility and affect
Previous studies on dental hygiene behavior have mainly focused on attitude, intention and behavior as outcome variables. Only two studies included perceived severity and affect (Rothman, Martino, Bedell, Detweiler, & Salovey, 1999; Chang, 2003) and no study included perceived susceptibility. Hereby, one study did not find any effect of message framing on affect or perceived severity (Chang, 2003). The other study, however, found that a gain-framed message leads to more positive affect than a gain-framed message whereas a
loss-framed message leads to a higher perceived severity of gum diseases than a gain-loss-framed message (Rothman et al., 1999). The finding that a message highlighting potential benefits leads to more positive affect than a message highlighting potential losses does not appear surprising. However, protection motivation theory suggests that a high threat appraisal, consisting of both high perceived severity and high perceived susceptibility, should lead to high protection motivation and ultimately protective behavior (Rogers, 1975). Taken together with prospect theory, one would expect that, regarding prevention behavior, a gain-framed message should lead to a higher perceived severity and higher perceived susceptibility than a loss-framed message. Therefore, the following hypothesis is proposed:
H4: Participants presented with a gain-framed message show higher perceived severity (H4a), higher perceived susceptibility (H4b) and more positive affect (H4c) than those presented with a loss-framed message. However, this effect on severity (H4d), susceptibility (H4e) and affect (H4f) is more pronounced for those who were primed with risk-avoidant words than for those who were primed with risk-seeking words.
Study 1 Method
Participants
The experiment was conducted in Spring 2012. 249 Dutch students were recruited during class and via flyers and posters in the university building. 11 participants were excluded due to invalid data at baseline (T0), 2 participants did not provide their student ID and could therefore not be matched with the according follow-up (T1) data. 28 participants (12.2%) were lost to follow-up, leading to a final sample of 208 students (see also Table 1, Appendix A, and Figure 1, Appendix B). Of the remaining participants, 111 participants (53.4%) were randomly allocated to the gain-framed message and 97 participants (46.6 %) were allocated to
the loss-framed message. The mean age of the final sample was 21.05 years (SD = 2.55), with a range from 18 to 33 years, and 78.8% of the sample was female. 75.5% of the sample reported no use of dental floss at baseline. Drop-out at T1 did not depend on the assigned condition (χ2 = 3.08, p = .079) and the two conditions did not differ with respect to age (F (1, 206) = 0.12; p = .730), gender (F (1, 206) = 3.50; p = .063) or past use of dental floss (F (1, 206) = 0.02; p = .881). Randomization was therefore considered successful.
Research Design
A 2 (loss frame vs. gain frame) x 3 (negative/low vs. ambivalent vs. positive/high attitude/intention) factorial design was applied. There were two independent variables: Participants were randomly assigned to either the loss frame or the gain frame condition. In addition, 3 groups were created based on the respective mean score on attitude and intention at T0, leading to 3 attitude groups (negative vs. ambivalent vs. positive) and 3 intention groups (low vs. ambivalent vs. high). The baseline measurement (T0) took place in March and April 2012, the follow-up measurement (T1) was conducted approximately one week after completion of the baseline measurement. The dependent variables were implicit attitude towards flossing, assessed using the Implicit Associations Test (IAT), explicit attitude and behavioral intention.
Procedure
Students were recruited on campus and invited to the communication laboratory at the University of Amsterdam. They were informed about the basic content of the study and provided informed consent. At baseline (T0), participants first completed a computer-assisted, flossing-specific IAT, masked as a reaction time test. All participants were seated in
individual cubicles and the distance to the screen was held constant at 35-45cm. Afterwards, participants were asked to provide information regarding some basic demographics and past
use of dental floss. In addition, an online survey regarding explicit attitude and intention was completed. A research assistant was present outside the cubicle in order to answer any questions. One week after the baseline measurement, participants were contacted via e-mail and invited to the follow-up measurement (T1) at the laboratory. Participants were randomly assigned to either the loss frame or the gain frame condition and were asked to attentively read the respective message and to mark any important information. After reading the message, participants were again asked to complete the IAT and the online survey regarding explicit attitude and intention. Participants were debriefed about the true purpose of the IAT and received study points for their participation.
Measures
Independent variables
Gain-framed vs. loss-framed message
Participants were randomly assigned to either the or the loss-framed message. The gain-framed message emphasized the potential benefits of regular dental flossing, e.g. the positive effect on dental health and a fresh breath (see Appendix C). In contrast, the loss-framed message emphasized the potential losses associated with not using dental floss on a regular basis, e.g. the negative effect on dental health and a bad breath (see Appendix C). The two messages differed only in their emphasis on potential gains or losses, all other elements of the message (e.g. length and style) were held constant. Participants were asked to read the
message carefully and to highlight important information.
T0 Attitude
Attitude was measured using six items on a 7-point Likert scale (e.g. very healthy – very unhealthy; very pleasant – very unpleasant)1. The scale was recoded so that 0 represents an
ambivalent attitude (e.g. neither healthy nor unhealthy), -3 represents the most negative attitude (e.g. very unhealthy) and +3 represents the most positive attitude (e.g. very healthy). All items were averaged in order to obtain a single measure for attitude towards flossing at T0. The scale showed good reliability (Cronbach’s alpha = .83; M = 0.65; SD = 0.93). Allocation to groups was based on the respondent’s mean score at T0: Participants with a mean score equal to or below -0.5 were labeled as having a negative attitude (n = 26), participants with a score above -0.5 but below or equal to 1 were labeled as having an ambivalent attitude (n = 124) and participants with an attitude of above 1 were labeled as having a positive attitude (n = 56)2. 2 participants were excluded due to missing data on this specific variable. For distribution of groups across conditions, see Figure 1 (Appendix B).
T0 Intention
Intention was measured using three items on a 7-point Likert scale ranging from ‘strongly disagree’ to ‘strongly agree’ (e.g. ‘I plan to floss my teeth every day during the upcoming week’)1. The scale was recoded so that 0 represents an ambivalent intention (‘maybe yes, maybe no’), -3 represents the lowest intention (‘no, surely not’) and +3 represents the highest intention (‘yes, surely’). All items were averaged in order to obtain a single measure for intention at T0. The scale showed very good reliability (Cronbach’s alpha = .97; M = -1.03;
SD = 1.71). Allocation to groups was based on the respondent’s mean score at T0:
Participants with a mean score below -1 were labeled as having a low intention (n = 104), participants with a score equal to or above -1 but below or equal to 1 were labeled as having an ambivalent intention (n = 75) and participants with an attitude of above 1 were labeled as having a high intention (n = 29)2. For distribution of groups across conditions, see also Table 1 (Appendix A) and Figure 1 (Appendix B).
2 Different assignments to groups (e.g. based on tertiles) were tested. The different solutions all led to the same results.
Dependent variables
Implicit Associations Task
The Implicit Associations Task (IAT) followed standard procedures as outlined in Greenwald, Nosek and Banaji (2003). Photos were used as targets. 8 target photos depicted flossing-related scenes (e.g. a person flossing in front of a mirror) and 8 photos depicted objects unrelated to dental flossing (e.g. a bus, a telephone). The following 8 words were used as positive attributes: (1) superb, (2) wonderful, (3) pleasure, (4) beautiful, (5) joy, (6) peace, (7) delightful, and (8) fantastic. The 8 words used as negative attributes were (1) tragic, (2) horrible, (3) fear, (4) pain, (5) dreadful, (6) disgust, (7) humiliate, and (8) filthy. D-scores were calculated based on the reaction time in milliseconds and the error percentage. Scores ranged from -2 to +2 with higher scores indicating a more positive implicit attitude.
T1 Attitude
The same attitude measurement was applied as at T0 (see Independent Variables). Again, the scale was recoded so that 0 represents a neutral attitude (e.g. neither healthy nor unhealthy), -3 represents the most negative attitude (e.g. very unhealthy) and +3 represents the most positive attitude (e.g. very healthy). All items were averaged in order to obtain a single measure for attitude towards flossing at T1. The scale showed good reliability (Cronbach’s alpha = .82; M = 0.67; SD = 0.90).
T1 Intention
The same intention measurement was applied as at T0 (see Independent Variables). Again, the scale was recoded so that 0 represents a moderate intention (e.g. maybe yes, maybe no), -3 represents the lowest intention (‘no, surely not’) and +3 represents the highest intention (‘yes, surely’). All items were averaged in order to obtain a single measure for intention at T1. The
scale showed very good reliability (Cronbach’s alpha = .97; M = -0.76; SD = 1.65). For all original scales (in Dutch), see Appendix D.
Results
Manipulation Check
Participants were asked to indicate on a 7-point Likert scale to which extent the presented message highlighted positive or negative consequences of flossing (‘The information you just read gave information about the consequences of flossing. Do you think that the message highlighted positive or negative consequences?’). The scale was recoded so that 0 represents neither positive nor negative consequences, -3 represents very negative consequences and +3 represents very positive consequences. It was expected that participants in the gain frame condition would identify more positive consequences than participants in the loss frame condition. A univariate analysis of variance (ANOVA) was performed with message framing (gain frame vs. loss frame) as independent and the identified positive or negative
consequences as dependent variable. Levene’s F-test indicated that equal variances can be assumed (F (1, 206) = 3.06; p = .169). As expected, the analysis revealed that those in the gain frame condition rated the presented consequences more positively (M = 1.23; SD = 1.50) than those in the loss frame condition (M = -1.42; SD = 1.65; F (1, 206) = 146.00, p < .001). Manipulation was therefore considered successful.
Hypotheses
All hypotheses were tested using univariate analyses of variance (ANOVA) with the baseline (T0) measurement as covariate. Since groups regarding attitude and intention were not randomly assigned but created based on differences at baseline, an ANOVA with change rather than an analysis of covariance was applied. When groups are not randomly assigned but created based on preexisting differences at baseline, an ANOVA with change was found
to be less biased than an analysis of covariance (Van Breukelen, 2006). All reported means and corresponding standard errors refer to estimated marginal means, taking the initial measurement at T0 into account. See also Table 1 (Appendix A) for means and standard errors for the whole sample and per condition for all variables under study.
Hypothesis 1. Participants who were presented with the gain-framed message were expected
to show more positive change in explicit attitude towards dental flossing than participants who were presented with the loss-framed message (H1a). In order to test this hypothesis, an ANOVA was performed with message frame as independent and explicit attitude at T1 as dependent variable. In addition, explicit attitude at T0 was included as covariate. Levene’s F-test indicated that equal variances can be assumed across groups (F (1, 204) = 0.26; p = .600). No main effect of message framing on explicit attitude could be established (F (1, 203) = 0.93; p = .337; η2 = .01): The explicit attitude of participants who read the gain-framed message (M = 0.74; S.E. = 0.06) did not significantly differ from the explicit attitude of those who read the loss-framed message (M = 0.65; S.E. = 0.07). Hypothesis 1a was therefore not supported.
In addition, it has been expected that participants who were presented with the gain-framed message show more positive change in behavioral intention than participants who were presented with the loss-framed message (H1b). An ANOVA was performed with message frame as independent variable, behavioral intention at T1 as dependent variable and intention at T0 as covariate. Levene’s F-test indicated that equal variances can be assumed across groups (F (1, 206) = 0.00; p = .973). No main effect of message framing on intention could be established (F (1, 205) = 0.00; p = .998; η2 = .00): The behavioral intention of participants who read the gain-framed message (M = -0.77; S.E. = 0.12) did not significantly differ from the intention of those who read the loss-framed message (M = -0.77; S.E. = 0.11). Hypothesis 1b was therefore also not supported.
Hypothesis 2. Participants who were presented with the gain-framed message were expected
to show more positive change in implicit attitude towards dental flossing than participants who were presented with the loss-framed message (H2a). In order to test this hypothesis, a univariate ANOVA was performed with message frame as independent and implicit attitude at T1 as dependent variable. In addition, implicit attitude at T0 was included as covariate. Levene’s F-test indicated that equal variances can be assumed (F (1, 206) = 0.16; p = .693). No main effect of message framing on implicit attitude towards dental flossing could be established (F (1, 205) = 0.03; p = .864; η2 = .00): The implicit attitude of participants who
read the gain-framed message (M = 0.70; S.E. = 0.34) did not significantly differ from the implicit attitude of those who read the loss-framed message (M = 0.69; S.E. = 0.36). Hypothesis 2a was therefore not supported.
In addition, it has been expected that the difference in implicit attitude between those who were confronted with the gain-framed message and those confronted with the loss-framed message is more pronounced for participants with an ambivalent attitude (H2b). An ANOVA was performed with message frame and explicit attitude at T0 (positive vs. ambivalent vs. negative) as independent variables, implicit attitude at T1 as dependent
variable and implicit attitude at T0 as covariate. Levene’s F-test indicated that equal variances could still be assumed (F (5, 200) = 1.45; p = .208). Again, no main effect of message
framing on implicit attitude could be established (F (1, 199) = 0.17; p = .680; η2 = .00). In addition, no main effect of explicit attitude on implicit attitude was found (F (2, 199) = 0.59;
p = .556; η2 = .00). Furthermore, no interaction effect between message framing and explicit attitude on implicit attitude emerged (F (2, 199) = 0.42; p = .658; η2 = .00): Regardless of their prior explicit attitude, participants who read the gain-framed message did not report more change in implicit attitude (Negative Attitude: M = 0.79; S.E. = 0.10; Ambivalent Attitude: M = 0.71; S.E. = 0.04; Positive Attitude: M = 0.63; S.E. = 0.07) than those who read
the loss-framed message (Negative Attitude: M = 0.67; S.E. = 0.10; Ambivalent Attitude: M = 0.71; S.E. = 0.05; Positive Attitude: M = 0.67; S.E. = 0.07). Hypothesis 2b was therefore also not supported.
Additionally, it has been tested whether an ambivalent intention, rather than an ambivalent attitude, moderates the relation between message framing and implicit attitude change (H2c). An ANOVA was performed with message frame and behavioral intention at T0 (high vs. ambivalent vs. low) as independent variables, implicit attitude at T1 as dependent variable and implicit attitude at T0 as covariate. Levene’s F-test indicated that equal variances could still be assumed (F (5, 202) = 0.53; p = .751). Again, no main effect of message
framing on implicit attitude could be established (F (1, 201) = 0.06; p = .811; η2 = .00). In addition, no main effect of intention on implicit attitude was found (F (2, 201) = 0.04; p = .957; η2 = .00). Furthermore, no interaction effect between message framing and behavioral intention on implicit attitude could be established (F (2, 201) = 0.32; p = .729; η2 = .00):
Regardless of their prior behavioral intention, participants who read the gain-framed message did not report more change in implicit attitude (Low Intention: M = 0.72; S.E. = 0.05;
Ambivalent Intention: M = 0.69; S.E. = 0.05; High Intention: M = 0.65; S.E. = 0.10) than those who read the loss-framed message (Low Intention: M = 0.69; S.E. = 0.05; Ambivalent Intention: M = 0.68; S.E. = 0.06; High Intention: M = 0.74; S.E. = 0.09). Therefore,
Hypothesis 2c was not supported. See also Table 2 (Appendix A) for the results of all analyses.
Study 2 Method
Participants
513 participants were recruited from the extended social network of the researcher and via social media. 117 participants (22.8%) dropped out during completion of the online
experiment, mainly during the priming task. Another 178 participants (34.7%) were excluded from further analyses due to one or more mistake(s) on the priming task. Priming was
considered insufficient for these participants. In addition, 5 participants (1.0%) indicated awareness of the prime and its connection to the framed messages and were therefore also excluded. Furthermore, 5 participants (1.0%) were excluded because they had a full denture and 1 participant (0.2%) was unable to floss his or her teeth due to a dental brace. 2
participants (0.4%) experienced technical problems regarding the presentation of the framed message and were therefore also excluded.
Of the remaining 205 participants, 48 participants (23.4%) were presented with the risk-avoidant prime and the gain-framed message, 54 participants (26.3%) were presented with the risk-seeking prime and gain-framed message, 55 (26.8%) participants were presented with the risk-avoidant prime and the loss-framed message and 48 participants (23.4%) were presented with the risk-seeking prime and loss-framed message (see also Table 3, Appendix A). The mean age of the final sample was 36.19 years (SD = 13.50), with a range from 18 to 83 years, and 80.5% of the sample was female. Participants had largely completed academic education (46.8%), followed by higher vocational training (27.8%), higher secondary education (12.2%) and mediocre vocational training (10.7.%). 70.7% of the sample reported no use of dental floss during the past week and only 7.3% reported daily use of dental floss. The conditions did not differ with respect to age (F (1, 201) = 2.75; p = .099), gender (F (1, 201) = 0.95; p = .332), education (F (1, 201) = 1.25; p = .264) or past use of dental floss (F (1, 206) = 0.01; p = .909). In addition, drop-out during the online experiment did not depend on the assigned condition (F (1, 321) = 1.20; p = .274) and mistakes on the scrambled sentences task did not depend on the version of the task (risk-seeking vs. risk-avoidant; F (1, 331) = 2.21; p = .138). Randomization was therefore considered successful.
Research Design
A 2 (risk-avoidant vs. risk-seeking prime) x 2 (gain-framed vs. loss-framed message) factorial design was applied. There were two independent variables: Participants were randomly assigned to either the loss frame or the gain frame condition and to either the risk-avoidant or risk-seeking priming task. The dependent variables were attitude, intention, perceived
severity, perceived susceptibility and affect.
Procedure
Participants were invited to the online experiment via e-mail and social media and completed the experiment at home at their own leisure. They were informed about the general content of the study and provided informed consent. First, participants were asked information about some basic demographics and past behavior regarding dental hygiene. Afterwards,
participants were randomly assigned to either the risk-seeking or risk-avoidant condition and completed the respective scrambled sentences task, masked as a language test. Participants were then randomly assigned to either the gain- or the loss-framed message, which was presented as an ‘advertisement’ in form of a flyer. Participants were asked to read the
message carefully and afterwards completed a manipulation check and indicated how they felt while or directly after reading the message. In addition, an online survey regarding attitude towards flossing, behavioral intention, perceived severity and perceived susceptibility was completed. In the end, participants were asked whether they had any idea about the relation between the ‘language test’ and the ‘advertisement’. Participants were debriefed about the true nature of the priming task and its relation to the framed message. They were thanked for their participation and offered the possibility to join a lottery for a gift voucher.
Measures
Independent variables
Gain- and loss-framed message
The same gain- and loss-framed messages as in Study 1 were employed. However, instead of in a plain text, the information was somewhat shortened and incorporated into a flyer. In the gain frame condition, the text on the flyer focused on the advantages of daily flossing and was accompanied by a happy, healthy tooth and the slogan ‘Keep your smile!’ (see Appendix E). In contrast, the text on the flyer of the loss-framed message focused on the disadvantages of not flossing daily, accompanied by an unhappy, unhealthy tooth and the slogan ‘Don’t lose your smile!’ (see Appendix E). All other aspects of the message (e.g. length and layout) were held constant.
Risk priming
The Scrambled Sentences Task (SST) was applied in order to introduce a risk-avoidant or risk-seeking mindset in the participants. SST was used because it was found to produce strong priming effects (Bargh, & Chartrand, 2000) and is feasible for application in an online
experiment. Participants were randomly assigned to either the avoidant or the risk-seeking condition. Each condition consisted of 12 puzzles of which 6 contained the actual prime and 6 served as filler. The puzzles were presented in random order, one at a time. The participant’s task was to form grammatically correct sentences by rearranging the presented words and leaving one of the presented words out. No time limit was given but participants were asked to complete the task as quick as possible. The risk-avoidant primes were (1) alert, (2) worried, (3) conscientious, (4) thoughtful, (5) pensively and (6) thought-out. The risk-seeking primes were (1) daring, (2) adventurous, (3) risky, (4) courageous, (5) fearless and (6) reckless. For the complete task (in Dutch), see Appendix E. Responses on the priming tasks
were coded as either correct (grammatically correct sentence, leaving one word out) or incorrect and only participants with six correct answers were included in the analyses.
Dependent variables
Attitude
Attitude was measured using six items on a 7-point Likert scale (e.g. very healthy – very unhealthy; very pleasant – very unpleasant)1. The scale was recoded so that 0 represents a neutral attitude (e.g. neither healthy nor unhealthy), -3 represents the most negative attitude (e.g. very unhealthy) and +3 represents the most positive attitude (e.g. very healthy). All items were averaged in order to obtain a single measure for attitude towards flossing. The scale showed good reliability (Cronbach’s alpha = .83; M = -0.40; SD = 1.23).
Intention
Intention was measured using five items on a 7-point Likert scale ranging from ‘strongly disagree’ to ‘strongly agree’ (e.g. ‘When I think about the upcoming week, I intent to floss my teeth every day’)1. The scale was recoded so that 0 represents a moderate intention (‘maybe yes, maybe no’), -3 represents the lowest intention (‘no, surely not’) and +3 represents the highest intention (‘yes, surely’). All items were averaged in order to obtain a single measure for intention. The scale showed very good reliability (Cronbach’s alpha = .97;
M = -0.70; SD = 1.78).
Perceived severity
Perceived severity was measured using five items on a 7-point Likert scale ranging from ‘strongly disagree’ to ‘strongly agree’ (e.g. ‘If I don’t floss my teeth every day, the consequences negative consequences will be severe’)1. The scale was recoded so that 0 represents a moderate severity (e.g. ‘neither agree nor disagree’), -3 represents the lowest severity (e.g. ‘strongly disagree’) and +3 represents the highest severity (e.g. ‘strongly
agree’). All items were averaged in order to obtain a single measure for perceived severity. The scale showed excellent reliability (Cronbach’s alpha = .98; M = -0.58; SD = 1.71).
Perceived susceptibility
Susceptibility was measured using six items on a 7-point Likert scale ranging from ‘strongly disagree’ to ‘strongly agree’ (‘If I don’t floss my teeth every day, I will suffer from
cavities’)1. The scale was recoded so that 0 represents moderate susceptibility (‘neither agree nor disagree’), -3 represents the lowest susceptibility (‘strongly disagree’) and +3 represents the highest susceptibility (‘strongly agree’). All items were averaged in order to obtain a single measure for perceived susceptibility. The scale showed very good reliability (Cronbach’s alpha = .96; M = -0.12; SD = 1.71).
Affect
Affect was measured using nine items on a 7-point Likert scale ranging from ‘strongly disagree’ to ‘strongly agree’ (e.g. ‘During or directly after reading the message, I felt relieved.’)1. Five items assessed positive affect and four items assessed negative affect. Coding for the negative affect items was reversed in order to let higher scores reflect more positive affect. The scale was recoded so that 0 represents neutral affect (‘neither agree nor disagree’), -3 represents the most negative affect (‘strongly disagree’) and +3 represents the most positive affect (‘strongly agree’). All items were averaged in order to obtain a single measure for affect. The scale showed good reliability (Cronbach’s alpha = .75; M = -0.40; SD = 1.23). For all original scales (in Dutch), see Appendix F.
Results
Manipulation Check
Participants were asked to indicate on a 7-point Likert scale to which extent the presented message highlighted positive or negative consequences of daily flossing. The wording of the
question depended on the assigned condition: In the gain frame condition, participants were asked to indicate the degree to which the message highlighted positive or negative
consequences of daily dental flossing. In the loss frame condition, participants were asked to indicate the degree to which consequences of not flossing every day were highlighted (see also Appendix F). The scale was recoded so that 0 represents neither positive nor negative consequences, -3 represents very negative consequences and +3 represents very positive consequences. It was expected that participants in the gain frame condition would identify more positive consequences than participants in the loss frame condition. Priming, however, was expected to be unrelated to the identification of consequences within the message. A univariate analysis of variance (ANOVA) was performed with message framing (gain frame vs. loss frame) and priming (risk-seeking vs. risk-avoidant) as independent variables and the identified positive or negative consequences as dependent variable. Levene’s F-test indicated that equal variances can be assumed (F (2, 331) = 2.33; p = .075). As expected, the analysis revealed that those in the gain frame condition rated the presented consequences more
positively (M = 1.63; SD = 1.18) than those in the loss frame condition (M = -1.87; SD = 1.04;
F (1, 201) = 519.23, p < .001). In addition, priming was found to be unrelated to the
perception of presented consequences (F (1, 201) = 4.37, p = .060) and no interaction effect between priming and framing on the rating of consequences was observed (F (1, 201) = 0.02,
p = .888). Manipulation was therefore considered successful. Hypotheses
Hypothesis 3. Participants who were presented with the gain-framed message were expected
to show a more positive attitude towards dental flossing (H3a) and a higher behavioral intention (H3b) than participants who were presented with the loss-framed message. In addition, it has been hypothesized that these differences will be more pronounced for those who were presented with the risk-avoidant prime than for those who were presented with the
risk-seeking prime (H3c+d). In order to test these hypotheses, a univariate ANOVA was performed with message frame and priming as independent variables and attitude and intention, respectively, as dependent variable. Levene’s F-test indicated that equal variances can be assumed across groups in both cases (Attitude: F (3, 201) = 0.25; p = .861; Intention: F (3, 201) = 0.44; p = .436).
No main effect of message framing on attitude could be established (F (1, 201) = 2.55;
p = .112; η2 = .01): The attitude of participants who read the gain-framed message (M = -0.55;
S.E. = 0.12)3 did not significantly differ from the attitude of those who read the loss-framed message (M = -0.27; S.E. = 0.12). Hypothesis 3a was therefore not supported.
However, a main effect of message framing on intention was found (F (1, 201) = 4.99;
p = .027; η2 = .02): Contrary to the proposed hypothesis, participants who read the loss-framed message reported a higher intention to floss their teeth (M = -0.43; S.E. = 0.18) than participants who read the gain-framed message (M = -0.98; S.E. = 0.18). Hypothesis 3b was therefore not supported.
No main effect of priming on either attitude (F (1, 201) = 0.00; p = .950; η2 = .00) or intention was found (F (1, 201) = 0.07; p = .796; η2 = .00) and no interaction effect could be established between message framing and priming on either attitude (F (1, 201) = 0.72; p = .398; η2 = .00) or intention (F (1, 201) = 0.06; p = .806; η2 = .00): Regardless whether
participants were primed with risk-seeking or risk-avoidant words, those who read the gain-framed message did not report a more positive attitude (Risk-seeking: M = -0.47; S.E. = 0.17; Risk-avoidant: M = -0.63; S.E. = 0.18) than those who read the loss-framed message (Risk-seeking: M = -0.34; S.E. = 0.18; Risk-avoidant: M = -0.43; S.E. = 0.24). Regarding behavioral intention, those who read the loss-framed message showed a higher intention to floss their teeth than those who read the gain-framed message, regardless whether they were primed with
3 All reported means and according standard errors refer to estimated marginal means (see also Study 1).
risk-seeking (Gain Frame: M = -0.92; S.E. = 0.24; Loss Frame: M = -0.43; S.E. = 0.26) or risk-avoidant words (Gain Frame: M = -1.05; S.E. = 0.26; Loss Frame: M = -0.06; S.E. = 0.10). Therefore, hypotheses 3c and 3d were not supported. See also Table 3 (Appendix A) for means and standard errors and Table 4 (Appendix A) for all results of the analyses.
Hypothesis 4. It has been hypothesized that participants presented with the gain-framed
message report a higher perceived severity (H4a), higher perceived susceptibility (H4b) and more positive affect (H4c) than participants presented with the loss-framed message. In addition, it has been expected that these differences will be more pronounced for those who were presented with the avoidant prime than for those who were presented with the risk-seeking prime (H4d-f). In order to test these hypotheses, a univariate ANOVA was performed with message frame and priming as independent variables and severity, susceptibility and affect, respectively, as dependent variable. Levene’s F-test indicated that equal variances across groups can be assumed in all cases (Severity: F (3, 201) = 0.75; p = .524;
Susceptibility: F (3, 201) = 0.48; p = .696; Affect: F (3, 201) = 0.85; p = .467).
The expected main effect of message framing on perceived severity (F (1, 201) = 2.46;
p = .118; η2 = .01) and susceptibility (F (1, 201) = 1.33; p = .251; η2 = .01) could not be
established: Participants presented with the gain-framed message did not perceive the
consequences of not flossing their teeth as more severe (M = -0.78; S.E. = 0.17) or themselves as more susceptible (M = -0.27; S.E. = 0.17) than those presented with the loss-framed
message (Severity: M= -0.41; S.E. = 0.17; Susceptibility: M= 0.01; S.E. = 0.17). Hypothesis 4a and 4b were therefore not supported.
However, message framing was found to have a main effect on affect (F (1, 201) = 6.74; p = .010; η2 = .03): As expected, those presented with the gain-framed message experienced more positive affect (M = 0.21; S.E. = 0.08) than those presented with the loss-framed message (M = -0.07; S.E. = 0.08). Hypothesis 4c was therefore supported.
No main effect of priming on severity (F (1, 201) = 1.89; p = .171; η2 = .01),
susceptibility (F (1, 201) = 0.51; p = .475; η2 = .00) or affect (F (1, 201) = 0.00; p = .951; η2 = .00) was found. Regarding perceived severity, an interaction effect between message framing and priming emerged (F (1, 201) = 4.19; p = .042; η2 = .02): Participants who were primed with risk-avoidant words perceived a higher severity of not flossing their teeth after reading the loss-framed message (M = -0.01; S.E. = 0.23), compared to the gain-framed message (M = -0.86; S.E. = 0.24), whereas no difference between loss-framed (M = -0.82; S.E. = 0.24) and gain-framed message (M = -0.70; S.E. = 0.23) could be established for those who were primed with the risk-seeking words (see also Figure 2, Appendix B). These results run counter to the proposed hypothesis and hypothesis 4d was therefore not supported.
No interaction effect could be established between message framing and priming on either perceived susceptibility (F (1, 201) = 1.42; p = .235; η2 = .01) or affect (F (1, 201) = 0.05; p = .826; η2 = .00): Regardless whether participants were primed with risk-seeking or
risk-avoidant words, those who read the gain-framed message did not report a higher
perceived susceptibility (Risk-seeking: M = -0.21; S.E. = 0.23; Risk-avoidant: M = -0.33; S.E. = 0.25) those who read the loss-framed message (seeking: M = -0.22; S.E. = 0.25; Risk-avoidant: M = 0.24; S.E. = 0.23). Participants who read the gain-framed message reported more positive affect than those who read to loss-framed message, regardless whether they were primed with risk-seeking (Gain Frame: M = 0.22; S.E. = 0.10; Loss Frame: M = -0.08;
S.E. = 0.11) or risk-avoidant words (Gain Frame: M = 0.19; S.E. = 0.11; Loss Frame: M =
-0.06; S.E. = 0.10).
Discussion
Theory predicts that a gain-framed message should be more persuasive than a loss-framed message when applied to the promotion of preventive health behavior, such as dental flossing (e.g. the prevention of caries and gum diseases). This prediction, however, was not supported.
Overall, most of the predicted effects proved insignificant and the few significant results showed a slight advantage for loss-framed messages. A gain-framed message was only found superior with respect to the activation of positive affect, a result that is not really surprising and not directly related to the expected superior persuasiveness of gain-framed messages. Effect sizes were overall small. Implicit measures were incorporated in order to shed more light on the proposed relation between message framing and preventive health behavior. However, no effect of message framing on implicit attitude could be detected. Whereas these results contradict message framing theory, they are in line with many results of former research regarding preventive dental hygiene behaviors. Most of the previous studies on dental hygiene behavior did also not find any difference in the persuasiveness of gain- and loss-framed messages (e.g. Chang, 2003; Mann, Sherman, & Updegraff, 2004; Sherman, Mann, & Updegraff, 2006; Updegraff, Sherman, Luyster, & Mann, 2007) and studies with significant results also showed slight advantages for loss-framed messages (e.g. Pakpour, Yekaninejad, Sniehotta, Updegraff, & Dombrowski, 2014).
In addition, risk priming was applied in order to test the underlying assumption of message framing that the level of risk associated with the promoted behavior moderates the effectiveness of framed messages. This assumption was also not supported. Again, this result contradicts theory but does fall in line with previous research results. Prior studies on message framing effects and perceived risk, employing explicit measures, led to inconsistent and contradictory results (e.g. Bartels, Kelly, & Rothman, 2010; Cohen, 2010) or did not show any interaction effect between message framing and perceived risk (Covey, 2014; Van ‘t Riet et al., 2014). In conclusion, both implicit and explicit measures have largely failed to prove the superiority of gain-framed messages in the promotion of preventive dental hygiene behavior and the assumption of perceived risk as moderator of message framing effects could not be supported. The validity of message framing theory for the promotion of preventive
health behaviors, at least regarding dental hygiene behaviors, therefore seems questionable. Gallagher and Updegraff (2012) concluded in their meta-analysis that message
framing does influence behavior without affecting attitude or intention. Study 2, in contrast to the finding of Gallagher and Updegraff (2012) and the results of Study 1, did show a small effect of message framing on intention. This result, however, contradicts theory because slight advantages were found for the loss-framed message, rather than for the gain-framed message. In addition, the effect size was small and intention was relatively low in both conditions. Neither implicit nor explicit attitude was affected by message framing and mediators between message framing and behavior are therefore unlikely to be found with attitudinal constructs. Yet, perceived severity and susceptibility were also not affected by message framing. Regarding susceptibility, a framed message might not have enough persuasive power to overcome one’s unrealistic optimism regarding individual susceptibility to health hazards (Weinstein, 1987). Regarding severity, an unexpected interaction effect was found between risk priming and message framing, whereby those primed with risk-avoidant words were found to perceive higher severity after reading the loss-framed message, compared to the gain-framed message. This result runs counter to theory and the proposed hypothesis, and effect size was, again, small and overall perceived severity low. Over-interpretation of this result does therefore not appear appropriate.
Several limitations of the presented studies should be considered. In both studies, women were overrepresented and generalization to a male population should therefore be made with caution. However, there are no theoretical considerations why message framing regarding dental flossing should work differently for women and men. In addition, the superior persuasiveness of gain-framed messages could neither be supported in a student sample of highly educated young adults (Study 1), nor in a more diverse sample of somewhat older and less highly educated adults (Study 2). Specific sample characteristics do therefore
not appear to alter the results. Study 2 employed an online experiment with relatively little control and it might be argued that priming did not work properly in this setting. These problems were addressed by applying strict inclusion criteria and by only including participants who made no mistake during the priming task. Study 1, in contrast, was
conducted in a highly controlled laboratory setting and could still not provide any support for the effect of message framing. Furthermore, no measurement of subsequent flossing behavior was applied. The finding that message frame does influence behavior while not affecting attitude and intention (Gallagher, & Updegraff, 2012) could therefore only partially be replicated.
Regardless of these limitations, the results do have important implications for the promotion of preventive dental hygiene behavior. Whereas message framing theory offers an attractive framework for the study of health promotion messages, past and present research produced largely insignificant or highly inconsistent results. Especially the notion that gain-framed messages are more persuasive for prevention behaviors whereas loss-gain-framed messages are more effective for detection behaviors seems questionable. Future research should
therefore continue to search for individual moderators rather than focusing on the distinction between protection and detection behaviors. Personal tailoring of messages, for example in congruence with one’s regulatory focus, might lead to more promising results than a one-message-fits-all-approach. In addition, careful testing of the underlying assumptions of prospect theory and message framing theory is warranted.
References
Ajzen, I. (1991). The theory of planned behavior. Organizational behavior and human
decision processes, 50(2), 179-211.
Altmann, S., & Traxler, C. (2012). Nudges at the dentist. European Economic Review, 72, 19-38.
Bargh, J. A., & Chartrand, T. L. (2000). The mind in the middle. Handbook of research
methods in social and personality psychology, 253-285.
Bartels, R. D., Kelly, K. M., & Rothman, A. J. (2010). Moving beyond the function of the health behaviour: The effect of message frame on behavioural decision-making.
Psychology and Health, 25(7), 821-838.
Chang, C. T. (2007). Health-care product advertising: The influences of message framing and
perceived product characteristics. Psychology & Marketing, 24(2), 143-169. Ciancio, S. (2003). Improving oral health: current considerations. Journal of Clinical
Periodontology, 30(5), 4-6.
Cohen, E. L. (2010). The role of message frame, perceived risk, and ambivalence in individuals' decisions to become organ donors. Health communication, 25(8), 758-769.
Covey, J. (2014). The role of dispositional factors in moderating message framing effects.
Health Psychology, 33(1), 52.
Ferguson, M. J. (2007). The automaticity of evaluation. Social psychology and the
unconscious: The automaticity of higher mental processes, 219-64.
Gallagher, K. M., & Updegraff, J. A. (2012). Health message framing effects on attitudes, intentions, and behavior: a meta-analytic review. Annals of behavioral medicine,
43(1), 101-116.
Gallagher, K. M., Updegraff, J. A., Rothman, A. J., & Sims, L. (2011). Perceived
susceptibility to breast cancer moderates the effect of gain-and loss-framed messages on use of screening mammography. Health Psychology, 30(2), 145.
Gibson, B. (2008). Can evaluative conditioning change attitudes toward mature brands? New evidence from the Implicit Association Test. Journal of Consumer Research, 35(1), 178-188.
Hujoel, P. P., Cunha-Cruz, J., Banting, D. W., & Loesche, W. J. (2006). Dental flossing and interproximal caries: a systematic review. Journal of dental research, 85(4), 298-305. Latimer, A. E., Salovey, P., & Rothman, A. J. (2007). The effectiveness of gain-framed
messages for encouraging disease prevention behavior: Is all hope lost?. Journal of
Health Communication, 12(7), 645-649.
McGuire, W. J. (1989). Theoretical foundations of campaigns. In R. E. Rice & C. K. Atkin
(Eds.), Public communication campaigns (2nd ed., pp. 43-54). Newbury Park, CA:
Sage.
Mann, T., Sherman, D., & Updegraff, J. (2004). Dispositional motivations and message framing: a test of the congruency hypothesis in college students. Health Psychology,
23(3), 330.
O'Keefe, D. J., & Jensen, J. D. (2007). The relative persuasiveness of gain-framed loss-framed messages for encouraging disease prevention behaviors: A meta-analytic review. Journal of health communication, 12(7), 623-644.
Olson, M. A., & Fazio, R. H. (2006). Reducing automatically activated racial prejudice through implicit evaluative conditioning. Personality and Social Psychology Bulletin,
32(4), 421-433.
Pakpour, A. H., Yekaninejad, M. S., Sniehotta, F. F., Updegraff, J. A., & Dombrowski, S. U. (2014). The Effectiveness of Gain-Versus Loss-Framed Health Messages in
Improving Oral Health in Iranian Secondary Schools: A Cluster-Randomized Controlled Trial. Annals of Behavioral Medicine, 47(3), 376-387.
Rogers, R. W. (1975). A protection motivation theory of fear appeals and attitude change. The
Journal of Psychology, 91(1), 93-114.
Rothman, A. J., Bartels, R. D., Wlaschin, J., & Salovey, P. (2006). The Strategic Use of Gain- and Loss-Frames Messages to Promote Healthy Behavior: How Theory Can Inform
Practice. Journal of Communication, 56(1), 202-220.
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(11), 1355-1369.
Rothman, A. J., & Salovey, P. (1997). Shaping perceptions to motivate healthy behavior: the role of message framing. Psychological bulletin, 121(1), 3.
Schuller, A. A. (2009). Mondgezondheid volwassenen 2007. Leiden: TNO Kwaliteit van
Leven.
Sherman, D. K., Mann, T., & Updegraff, J. A. (2006). Approach/avoidance motivation, message framing, and health behavior: Understanding the congruency effect.
Motivation and Emotion, 30(2), 164-168.
Tversky, A., & Kahneman, D. (1981). The framing of decisions and the psychology of choice.
Science, 211(4481), 453-458.
Updegraff, J. A., Sherman, D. K., Luyster, F. S., & Mann, T. L. (2007). The effects of message quality and congruency on perceptions of tailored health communications.
Journal of Experimental Social Psychology, 43(2), 249-257.
Uskul, A. K., Sherman, D. K., & Fitzgibbon, J. (2009). The cultural congruency effect: Culture, regulatory focus, and the effectiveness of gain-vs. loss-framed health messages. Journal of Experimental Social Psychology, 45(3), 535-541.
Van’t Riet, J., Cox, A. D., Cox, D., Zimet, G. D., De Bruijn, G. J., Van den Putte, B., De Vries, H., Werrij, M. Q., & Ruiter, R. A. (2014). Does perceived risk influence the effects of message framing? A new investigation of a widely held notion. Psychology
& Health, (ahead-of-print), 1-17.
Conclusions from a community-wide sample. Journal of behavioral medicine, 10(5), 481-500.
World Health Organization (2012). Oral Health. Retrieved 12 December 2014 from http://www.who.int/mediacentre/factsheets/fs318/en.
Appendix A – Tables
Table 1
Study 1: Mean scores and corresponding standard errors
n IAT Attitude (T1) Intention (T1) Gain Frame Intention High 14 0.65 (0.10)
Ambivalent 43 0.69 (0.05) Low 54 0.72 (0.05) Attitude Positive 29 0.63 (0.07) Ambivalent 68 0.71 (0.04) Negative 13 0.79 (0.10) Total 111 0.70 (0.03) 0.74 (0.06) -0.77 (0.11)
Loss Frame Intention High 15 0.74 (0.09) Ambivalent 32 0.68 (0.06) Low 50 0.69 (0.05) Attitude Positive 27 0.67 (0.07) Ambivalent 56 0.71 (0.05) Negative 13 0.67 (0.10) Total 97 0.69 (0.04) 0.65 (0.07) -0.77 (0.12) Total 208 0.70 (0.03) 0.69 (0.05) -0.77 (0.08)
Note. IAT = Implicit Associations Test. IAT scores could range from -2 (most negative
implicit attitude) to +2 (most positive implicit attitude). Attitude and intention at T1 could range from -3 (lowest attitude/intention) to +3 (highest attitude/intention). No interaction effects between message frame and attitude/intention at T0 were tested (attitude/intention at T0 was included as covariate).
Table 2
Study 1: Results from the univariate analyses of variance
Sum of Squares df Mean Square F p η2
DV: Attitude (T1) Message Frame 0.41 1 0.41 0.93 0.337 0.01 Covariate: Attitude (T0) 64.74 1 64.74 147.31 < 0.001 0.42 Error 89.22 203 0.44 Total 254.14 206 DV: Intention (T1) Message Frame 0.00 1 0.00 0.00 0.998 0.00 Covariate: Intention (T0) 302.17 1 302.17 233.34 < 0.001 0.53 Error 265.48 205 1.30 Total 690.33 208 DV: IAT (T1) Message Frame 0.00 1 0.00 0.03 0.864 0.00 Frame*Attitude (T0) 0.15 2 0.08 0.59 0.556 0.01 Frame*Intention (T0) 0.01 2 0.01 0.04 0.957 0.00 Covariate: IAT (T0) 18.31 1 18.31 145.84 < 0.001 0.42 Error 25.73 205 0.13 Total 144.88 208
Note. DV = Dependent variable. IAT = Implicit Associations Test.
Table 3
Study 2: Mean scores and corresponding standard errors
Type of Message Type of Prime n Attitude Intention Severity Susceptibility Affect Gain Frame Risk-seeking 54 -0.47
(0.17) -0.92 (0.24) -0.70 (0.23) -0.21 (0.23) 0.22 (0.10) Risk avoidant 48 -0.63 (0.18) (0.26) -1.05 (0.24) -0.86 (0.25) -0.33 (0.11) 0.19 Total 102 -0.55 (0.12) -0.98 (0.18) -0.78 (0.17) -0.27 (0.17) 0.21 (0.08) Loss Frame Risk-seeking 48 -0.34
(0.18) (0.26) -0.43 (0.24) -0.82 (0.25) -0.22 (0.11) -0.08 Risk-avoidant 55 -0.21 (0.17) (0.24) -0.43 (0.23) -0.01 (0.23) 0.24 (0.10) -0.06 Total 103 -0.27 (0.12) -0.43 (0.18) -0.41 (0.17) 0.01 (0.17) -0.07 (0.08) Total 205 -0.41 (0.09) -0.71 (0.12) -0.60 (0.12) -0.13 (0.12) 0.07 (0.05)
Note. All scores could range from -3 (e.g. most negative attitude, lowest perceived severity) to
Table 4
Study 2: Results from the univariate analyses of variance
Sum of Squares df Mean Square F p η2
DV: Attitude Message Frame 3.83 1 3.83 2.55 0.112 0.01 Risk Priming 0.01 1 0.01 0.00 0.950 0.00 Frame*Priming 1.01 1 1.01 0.71 0.398 0.00 Error 301,47 201 1.50 Total 340.00 205 DV: Intention Message Frame 15.65 1 15.65 4.99 0.027 0.02 Risk Priming 0.21 1 0.21 0.07 0.796 0.00 Frame*Priming 0.19 1 0.19 0.06 0.806 0.00 Error 630.67 201 3.14 Total 747.72 205 DV: Severity Message Frame 7.03 1 7.03 2.46 0.118 0.01 Risk Priming 5.40 1 5.40 1.89 0.171 0.01 Frame*Priming 11.96 1 11.96 4.18 0.042 0.02 Error 574,46 201 2.86 145.84 < 0.001 0.42 Total 668.92 205 0.13 DV: Susceptibility Message Frame 3.88 1 3.88 1.33 0.251 0.01 Risk Priming 1.50 1 1.50 0.51 0.475 0.00 Frame*Priming 4.16 1 4.16 1.42 0.235 0.01 Error 588.91 201 2.93 Total 601.80 205 DV: Affect Message Frame 3.85 1 3.85 6.74 0.010 0.03 Risk Priming 0.00 1 0.00 0.00 0.951 0.00 Frame*Priming 0.03 1 0.03 0.05 0.826 0.00 Error 114.95 201 0.57 Total 119.80 205
