The War on Prevention

Metaphors Shape Thought

Conceptual metaphor theory (Lakoff & Johnson, 1980) posits that metaphors structure thinking by providing conceptual mappings between concrete and abstract concepts. Concrete concepts highlight relevant aspects of metaphorically related abstract concepts, deemphasize irrelevant aspects, and ultimately guide knowledge of and reasoning about the abstract concept. Since the initial work in cognitive linguistics, extensive experimental research has illuminated how abstract concepts are understood in terms of metaphorically related concrete domains (for a review, see Landau, Meier, & Keefer, 2010). For instance, interpersonal warmth is often understood in terms of physical warmth (Williams & Bargh, 2008), importance is understood in terms of physical weight (Ackerman, Nocera, & Bargh, 2010; Chandler, Reinhard, & Schwarz, 2012; Jostmann, Lakens, & Schubert, 2009), valence and power are related to verticality (Meier & Robinson, 2004; Schubert, 2005), and so are God and Devil related concepts (Meier, Hauser, Robinson, Friesen, & Schjeldahl, 2007). As many conceptual metaphors are learned via linguistic experience, linguistic framing of an abstract concept via the use of metaphoric expressions can also activate a metaphoric representation of the abstract concept and influence reasoning (Lakoff & Johnson, 1980; Landau, Sullivan, & Greenberg, 2009; for reviews, see Gibbs, 2014; Ottati, Renstrom, & Price, 2014). For instance, metaphorically framing crime as either a virus plaguing a city or as a beast ravaging a city causes people to propose different solutions to a hypothetical crime wave (Thibodeau & Boroditsky, 2011). Similarly, metaphoric framing of America as a body harshens Americans’ attitudes toward immigration when they are threatened by physical contamination (Landau et al., 2009). In short, varying the accessibility of concrete metaphors for abstract concepts via metaphoric framing can guide reasoning in the abstract domain in metaphor-consistent ways.

Further illustrating the power of metaphors, a meta-analysis of persuasive messages found that the use of metaphors reliably increases persuasion when they metaphorically frame a familiar target early in the message in terms of a single novel source domain (Sopory & Dillard, 2002). The persuasive influence of metaphors can unfold through multiple pathways (for reviews, see Ottati & Renstrom, 2010; Ottati et al., 2014). For instance, metaphoric conceptualizations of abstract concepts can be activated in multiple ways (Ottati et al., 2014), such as by directly stating the root metaphor (“Cancer is an enemy”) or through more subtle means like evoking the root metaphor through surface metaphoric utterances (“We must win the war on cancer”). In addition, once a root metaphor is activated, metaphors can affect judgments through multiple processes. Metaphors can affect the amount of message elaboration when they link the target to a domain that is of interest to the recipient; for example, sports metaphors increase elaboration of persuasive messages for sports fans and decrease elaboration for non-sports fans (Ottati, Rhoads, & Graesser, 1999). Metaphors also increase persuasion when they match the recipient’s lay metaphoric representation of the topic (Sopory, 2005) and hence increase fluent access to related knowledge. Metaphors can also contribute to attitude change by directly mapping attributes from the source to the target domain, as is the case for the NATION IS A BODY metaphor, which maps attributes of physical contamination onto the abstract concept of national immigration policy (Landau et al., 2009).

Present Research

Drawing on these insights from metaphoric framing research, the present studies investigate whether conceptual metaphors that relate cancer to an enemy influence people’s reasoning about cancer and their willingness to engage in a variety of preventive behaviors. The studies also bear on whether key theoretical findings of metaphoric framing research extend across different manipulations and into socially relevant content domains. As noted in recent discussions (Stroebe & Strack, 2014), consistent effects of multiple operationalizations of a conceptual variable across diverse content domains are a crucial criterion for the robustness of a theoretical approach.

The prevailing wisdom in health communication is that fear raised by enemy framing will motivate people to engage in beneficial preventive behaviors (Kato et al., 2008; Khalil, 2012; Mukherjee, 2010; Stephenson & Witte, 1998). However, we hypothesized that, in line with the theory of metaphoric framing (Lakoff & Johnson, 1980), framing cancer as an enemy should cause people to bring attributes of enemies to bear on their representation of cancer. While this may enhance motivation to engage in prevention behaviors related to an enemy metaphor, it may hurt the motivation to engage in prevention behaviors that are difficult to reconcile with this metaphor. Unfortunately, the latter applies to the bulk of behaviors that support cancer prevention.

The enemy metaphor emphasizes power, strength, masculinity, and taking aggressive actions toward an enemy (Harrington, 2012; Reisfield & Wilson, 2004). As such, behaviors that entail active engagement are particularly suited to the enemy metaphor because they help bolster these attributes and promote attacking an enemy. However, behaviors which entail limitation and restraint are less applicable to the enemy metaphor because they often undermine attributes of power, strength, and masculinity, and they do not promote attacking an enemy. Cancer prevention recommendations promote either engagement or limitation and, accordingly, differ in their applicability to fighting enemies. Table 1 displays the 11 cancer prevention recommendations that an expert review identified as efficient in reducing the risk of developing cancer (World Cancer Research Fund & the American Institute for Cancer Research, 2007); the table also notes whether the recommendation promotes engagement or limitation. Other organizations have arrived at similar prevention recommendations (see Kushi et al., 2012).

OPEN IN VIEWER

Table 1.

How Common Cancer Prevention Recommendations Fit With Engagement and Limitation.

Prevention recommendation (American Institute for Cancer Research, 2012)	Engagement or limitation?
Be as lean as possible without becoming underweight.	Engage
Be physically active for at least 30 min every day.	Engage
Avoid sugary drinks. Limit consumption of energy-dense foods.	Limit
Eat more of a variety of vegetables, fruits, whole grains, and legumes such as beans.	Engage
Limit consumption of red meats (such as beef, pork, and lamb) and avoid processed meats.	Limit
If consumed at all, limit alcoholic drinks to 2 for men and 1 for women a day.	Limit
Limit consumption of salty foods and foods processed with salt (sodium).	Limit
Do not use supplements to protect against cancer.	Limit
It is best for mothers to breastfeed exclusively for up to 6 months and then add other liquids and foods.	—
After treatment, cancer survivors should follow the recommendations for cancer prevention.	—
And always remember, do not smoke or chew tobacco.	Limit

As the prevention recommendations in Table 1 illustrate, cancer prevention is only occasionally characterized by active engagement in behaviors that also come to mind while thinking of fighting a battle against an aggressor. Instead, many of the more efficient prevention behaviors amount to self-limitation (avoid smoking, avoid alcohol, avoid fatty foods, avoid red meats, and so on). Unfortunately, self-limitation is a class of behaviors that is unlikely to figure prominently in people’s associations with fighting enemies.

A pilot study using the Corpus of Contemporary American English (COCA; Davies, 2008) confirmed this intuition. Linguistically, limitation is less associated with attacking than it is with prevention. In natural language, there is less co-occurrence of limitation-related words with the word ATTACK than with the word PREVENT; for instance, the odds of seeing the word INHIBIT within nine words before or after the target word ATTACK in natural language are one-twentieth that of seeing it within nine words before or after the target word PREVENT. Given that word co-occurrence corresponds with semantic association (Landauer & Dumais, 1997), these data highlight that limitation is less associated with an enemy concept than with a prevention concept. This suggests that people do not think of limiting the self when they think of fighting enemies. If so, framing cancer in terms of an enemy metaphor may lessen people’s intention to engage in some of the more effective prevention behaviors available to them.

In the following series of studies, we test this possibility. In Study 1, we examine if the enemy representation affects the accessibility of limitation-related prevention behaviors. In Studies 2 and 3, we explore if the enemy representation affects intention for various prevention, monitoring, and treatment behaviors. In addition, Study 3 explores whether the predicted adverse effects of enemy framing are attenuated when prevention behaviors are presented in a metaphorical language that matches the metaphor used in the message.

Study 1

Assuming that limiting the self is not closely associated with the concept of fighting an enemy, framing cancer as an enemy should impair the accessibility of prevention behaviors that involve self-limitation. Study 1 tests this hypothesis.

Results and Discussion

We counted the number of self-bolstering and self-limiting behaviors listed and conducted a 2 (metaphor: enemy, neutral) × 2 (behavior type: self-bolstering, self-limiting) mixed analysis of variance. Overall, participants listed more self-bolstering than self-limiting behaviors, F(1, 62) = 12.18, p = .001, η p 2 = .164, for the main effect of behavior type. More important, this observation was qualified by the predicted interaction of behavior type and metaphor, F(1, 62) = 5.68, p = .020, η p 2 = .084. As shown in Figure 1, exposure to the enemy metaphor reduced the listing of self-limiting behaviors relative to the neutral frame condition, F(1, 62) = 6.96, p = .011, d = .66, 95% confidence interval [CI] = [0.17, 1.29] for the simple effect, without increasing the listing of self-bolstering prevention behaviors, F(1, 62) = 1.30, p = .26 for the simple effect. This indicates that an enemy metaphor reduces the accessibility of self-limiting prevention behaviors without increasing the accessibility of self-bolstering ones. Finally, the main effect of metaphor was not significant, F < 1, suggesting that, counter to the prevailing wisdom, enemy framing does not simply increase thoughts of beneficial behaviors.

OPEN IN VIEWER

Figure 1.

Message metaphor by behavior type on the mean number of behaviors listed.

These results parallel those observed in the linguistic analysis. When thinking of cancer as an enemy, limitation-related prevention behaviors do not come to mind. However, there was no beneficial effect of enemy framing on self-bolstering behavior accessibility as one would intuitively expect. One reason for this may be differences in chronic accessibility of the behavior types; participants listed more self-bolstering than self-limiting behaviors, suggesting that self-bolstering behaviors may be more chronically accessible than self-limiting ones. We explore these differences in chronic accessibility and ambiguity in the next study.

While these effects on accessibility provide preliminary support for our hypotheses, additional framing manipulations are necessary to rule out alternate explanations stemming from the use of one specific method of framing. Therefore, Study 2 employs a different framing manipulation in a conceptual replication and extension of Study 1.

Study 2

Building on the observation that metaphorical framing influences what comes to mind when people are asked to list potential prevention measures (Study 1), Study 2 tests whether the observed effects extend to behavioral intentions. The prevailing wisdom would suggest that enemy framing would increase intentions for all behaviors that would help someone “fight” cancer (i.e., prevention, monitoring, and treatment). However, in line with conceptual metaphor theory, we hypothesized that metaphorically framing cancer as an enemy in a message would lessen intention for self-limiting prevention behaviors.

Furthermore, many have offered the conjecture that an enemy representation might boost motivation for active, agentic behaviors against cancer. For instance, because one must be active to fight enemies, one must also take active steps to fight cancer (Harrington, 2012; Reisfield & Wilson, 2004). In addition, because one fights enemies aggressively, one must also fight cancer by opting for aggressive treatments (Aktipis et al., 2010). To our knowledge, Study 2 provides the first experimental test of these conjectures by examining the effect of enemy cancer framing on a range of preventive intentions (i.e., self-limiting and self-bolstering prevention, monitoring, and treatment). If the enemy representation boosts motivation for active agentic behaviors, enemy framing should boost intention for self-bolstering, monitoring, and treatment behaviors, while undermining intentions to engage in self-limiting behaviors.

Finally, Study 2 added another metaphoric framing condition to the design—that of cancer as imbalance. Recent research has shown that the use of any applicable metaphor in health information about vaccination can increase a reader’s intention to get vaccinated (Scherer, Scherer, & Fagerlin, 2014). Therefore, the imbalance metaphor condition serves as a control that tests whether the predicted effect of enemy framing is unique to this metaphor (as we expect) or also observed for other metaphors that have different substantive implications. Furthermore, the imbalance conceptualization was once the dominant conceptual metaphor for disease (Goatly, 2007; Mukherjee, 2010) and is still the dominant conceptual metaphor for most diseases in Chinese cultures (Stibbe, 1996). In fact, some medical scholars suggest that illness and treatment may better fit an imbalance metaphor than an enemy metaphor. For example, ecological balance metaphors emphasize population-level prevention and sustainable treatment practices (Annas, 1995; Nesse & Williams, 1996), unlike bellicose metaphors that emphasize defeating diseases at all costs. For these reasons, the remaining studies include an imbalance metaphor framing group for comparison purposes.

Results and Discussion

We created indices for intention to self-limit (α = .77), self-bolster (α = .91), monitor (α = .87), and treat (α = .90) by averaging the ratings of the respective items. A 3 (message metaphor: enemy, imbalance, neutral) × 4 (intention: self-limiting, self-bolstering, monitoring, treatment) × 2 (gender: male, female) mixed analysis of variance revealed the predicted significant omnibus two-way interaction between message metaphor and intention, F(6, 921) = 2.48, p = .022, η p 2 = .016.

Self-limiting behavior

As predicted, enemy framing lowered intentions for self-limiting behaviors compared with the neutral representation, t(307) = −2.21, p = .028, d = .29, 95% CI = [–0.96, –0.03], and compared with the imbalance representation, t(307) = −2.54, p = .012, d = .28, 95% CI = [–0.80, –0.14] (Table 2). These differences are reflected in an omnibus simple main effect of message frame, F(2, 307) = 3.91, p = .021, η p 2 = .025, which was not moderated by gender, F < 1 for the Metaphor × Gender interaction. Thus, framing cancer as an enemy hurt intentions to limit risky behaviors, consistent with the accessibility effects observed in Study 1.

OPEN IN VIEWER

Table 2.

Message Metaphor by Intention Index on Mean (SD) Behavioral Intention.

	Message metaphor
Intention Index	Enemy	Imbalance	Neutral
Self-limiting	3.69 (1.70)	4.16 (1.66)	4.17 (1.62)
Self-bolstering	4.92 (1.42)	5.33 (1.44)	4.89 (1.48)
Monitoring	2.65 (1.58)	2.95 (1.52)	2.98 (1.70)
Treatment	5.78 (1.25)	5.80 (1.21)	5.48 (1.43)

Self-bolstering behaviors

An omnibus simple main effect of message metaphor for the self-bolstering index, F(2, 307) = 4.13, p = .017, η p 2 = .026, further indicated that imbalance framing increased bolstering intentions. Specifically, participants exposed to imbalance framing intended to self-bolster more than both enemy framing, simple effect t(307) = 2.33, p = .020, d = .29, and the neutral representation, simple effect t(307) = 2.63, p = .009, d = .30. In contrast, enemy framing did not increase intention to self-bolster relative to the neutral representation, t(307) = 0.40, p = .686, indicating that enemy framing does not boost intentions to engage in risk-reducing behaviors. This also replicates the non-effect of enemy framing on self-bolstering behaviors observed in Study 1. The effect of message framing was not moderated by an interaction with gender, F(2, 307) = 1.53, p = .218, for the Metaphor × Gender interaction.

Monitoring and treatment

Finally, the framing manipulations did not affect the monitoring or treatment intentions. For the monitoring index, the simple main effects of message metaphor, F(2, 307) = 1.61, p = .201; gender, F < 1; and the message Metaphor × Gender simple interaction, F < 1, were not significant. Similarly, for the treatment index, the simple main effects of message metaphor, F(2, 307) = 1.12, p = .327; gender, F < 1; and the message Metaphor × Gender simple interaction, F(2, 307) = 1.04, p = .356, were not significant.

Gender effects

In the overall model, there was an additional significant two-way interaction between gender and intention, F(3, 921) = 8.26, p < .001, η p 2 = .026. Although gender did not moderate the effect of message framing on intentions, F(6, 921) = 1.24, p = .285, for the three-way interaction, the observation is worth noting. Specifically, women intended to engage in prevention behaviors more than men: simple main effect of gender on self-limiting intentions, F(1, 307) = 23.37, p < .001, η p 2 = .071, and simple main effect of gender on self-bolstering intentions, F(1, 307) = 6.28, p = .013, η p 2 = .020. However, both males and females had equal intentions for monitoring (F < 1) and treatment (F < 1) behaviors.

Study 3

The preceding two studies consistently showed that framing cancer messages in terms of an enemy metaphor has adverse consequences for prevention behaviors that involve self-limitation, from avoiding overexposure to the sun to avoiding fatty foods. The theoretical rationale of metaphor framing implies that these adverse effects can be attenuated when the target behavior is framed in a way that matches the metaphor (Lakoff & Johnson, 1980; Landau et al., 2010; Lee & Schwarz, 2014b). Because aptness facilitates metaphoric processing (Jones & Estes, 2006), metaphorically describing the behaviors in a way that makes them a better fit with a metaphoric conceptualization should facilitate fluent processing (Thibodeau & Durgin, 2008, 2011), which may attenuate the adverse effects of the enemy metaphor. Study 3 tests this implication.

Results

We computed intention indices for each type of behavior (self-limit α = .73; self-bolster α = .93; monitor α = .89; treatment α = .90) by averaging the respective behavior ratings. We first address whether Study 3 replicates the effects of message framing observed in Study 2; subsequently, we turn to the new predictions and address the influence of metaphor matching.

Do the effects of Study 2 replicate?

The materials of the neutral behavior framing condition of the present study were identical to the materials of Study 2. This allows us to assess the robustness of our results with a new sample. To do so, we assessed the influence of message framing on participants’ prevention intentions under neutral behavior framing.

Message framing again affected participants’ intention to engage in self-limiting prevention behaviors, F(2, 57) = 2.68, p = .077, η p 2 = .068. Replicating Study 2, participants reported lower intention under enemy framing than under neutral framing conditions, t(57) = 1.95, p = .056, d = .60. In contrast to Study 2, however, intentions under imbalance framing did not differ from those under enemy framing, t(57) < 1, and intentions under imbalance framing were significantly less than those under neutral framing, t(57) = 2.03, p = .047, d = .62.

There were no effects of message framing on the remaining behavioral intention indices; omnibus main effect of message framing on the self-bolstering index, F(2, 57) = 1.59, p = .212; on the monitoring index, F < 1; and on the treatment index, F(2, 57) = 1.07, p = .351.

In a meta-analytic analysis of Study 2 and the replication conditions of Study 3, enemy-framed messages undermine recipients’ intention to engage in self-limiting prevention behaviors, relative to neutrally framed messages; Z = 4.11, p = .004, following the procedures of the Stouffer method as described by Rosenthal (1978). Further, enemy-framed messages do not affect intention to engage in self-bolstering (Z = .25, p = .80), monitoring (Z = −.74, p = .46), and treatment behaviors (Z = 1.51, p = .13) in the individual studies or the meta-analytic analyses. Thus, across studies, enemy-framed messages lessen intentions for self-limiting behaviors while remaining ineffective on altering intentions for self-bolstering prevention, monitoring, and treatment intentions.

In addition, the effects of imbalance-framed messages in Study 2 did not replicate in our new sample. In contrast to Study 2, imbalance-framed messages did not increase intention for self-bolstering behaviors. Further, imbalance-framed messages lessened intentions for self-limiting behaviors. Thus, it appears as if the effects of this once dominant metaphor for disease have as-of-yet unknown moderating conditions which preclude us from drawing firm conclusions. In addition, populations differed between Study 2 (MTurk) and Study 3 (subject pool) and differences between populations in chronic accessibility, aptness, or conventionality of the imbalance metaphor may also account for the inconsistent effects.

Does metaphor matching improve intentions?

A 3 (message metaphor: enemy, imbalance, neutral) × 3 (behavior frame: enemy, imbalance, neutral) × 4 (intention: self-limit, self-bolster, monitor, treat) × 2 (gender: male, female) mixed analysis of variance revealed the predicted omnibus three-way interaction between message metaphor, behavior frame, and intention, F(12, 474) = 1.90, p = .033, η p 2 = .046. ¹

To diagnose this three-way interaction, we examined the simple two-way interactions of message metaphor and behavior frame for each set of behaviors. There was no simple two-way interaction of message metaphor and behavior framing for treatment intentions, F(4, 158) = 1.19, p = .315; monitoring intentions, F < 1; or self-bolstering intentions, F(4, 158) = 1.63, p = .170, indicating that metaphor matching did not influence intentions for these behaviors. This was expected and replicates Study 2, which found no effect of initial metaphoric framing in the message on intentions for these behaviors.

In contrast, the intention to engage in self-limiting prevention behaviors was affected by the experimental variables, F(4, 158) = 3.07, p = .018, η p 2 = .072, for the simple interaction of message and behavior framing. Next, we turn to a more detailed analysis of this simple interaction.

Intention to engage in self-limiting behaviors

Enemy frame

Previous research has shown that a neutral, non-metaphoric sentence about a target concept can be considered to “mismatch” a metaphoric representation of the target concept; after reading a metaphoric sentence, reading times for a neutral non-metaphoric sentence about the same topic are delayed to the same extent as reading times for a sentence utilizing a different source metaphor for the topic (e.g., Thibodeau & Durgin, 2008, Study 3). Our planned contrasts take this observation into account. A planned contrast pertaining to the enemy message metaphor, which compared enemy behavioral framing (match) to neutral and imbalance behavioral framing (mismatch), revealed a marginally significant metaphor matching effect (Figure 2), t(167) = 1.88, p = .062, d = .52. Follow-up comparisons showed that participants who had received an enemy-framed message reported higher intentions to engage in self-limiting behaviors when these behaviors were framed as apt for fighting enemies than when they were framed neutrally, t(167) = 1.82, p = .070, d = .66, 95% CI = [−0.01, 1.79], or as apt for restoring imbalance, t(167) = 1.41, p = .159, d = .43, 95% CI = [−0.36, 1.72]. Thus, metaphor matching can eliminate the otherwise observed adverse effects of the enemy metaphor—for participants with an enemy conceptualization of cancer, describing self-limiting behaviors in enemy-consistent terms increased intention relative to describing those behaviors in enemy-inconsistent terms (either described literally or in terms of imbalance).

OPEN IN VIEWER

Figure 2.

Message metaphor by behavior frame on self-limiting intention.

Of theoretical and applied interest is a comparison of the above conditions with the neutral control conditions. As metaphor matching effects typically only compare metaphor matches against metaphor mismatches (see Thibodeau & Durgin, 2008), it is currently unknown if metaphor matching effects actually facilitate comprehension and intention over basic literal statements. Further, this comparison has applied interest, as it might potentially demonstrate that enemy framing has a persuasive advantage over non-framing when behaviors are described as apt for the enemy conceptualization.

The comparison with the neutral control condition shows that metaphor matching did not endow the enemy metaphor with any persuasive advantage. When the enemy frame of the message matched the enemy frame of the target behavior, participants’ intentions merely equaled, but did not exceed, the intentions under neutral control conditions, where neither the message nor the behavior was described in metaphorical terms, t(167) = 0.05, p = .963.

Imbalance frame

The planned contrast looking at imbalance message framing and comparing imbalance behavior framing (match) to neutral and enemy behavior framing (mismatch) also revealed a marginally significant metaphor matching effect (Figure 2), t(167) = −1.69, p = .092, d = .46. Follow-up comparisons showed that participants who had received an imbalance-framed message reported higher intentions to engage in self-limiting behaviors when they were framed as apt for restoring balance than when they were framed neutrally, t(167) = 2.52, p = .013, d = .80, 95% CI = [0.24, 2.07], but not when they were framed as apt for fighting enemies, t(167) = 0.41, p = .679. This indicates a conditional metaphor matching effect—relative to literal behavioral framing, imbalance behavioral framing enhanced intention for self-limiting behaviors when participants had an imbalance conceptualization of the disease.

Neutral frame

Finally, the planned contrast looking at neutral message framing and comparing neutral behavior framing to enemy and imbalance behavior framing showed no effect of matching on intentions in the neutral control conditions, t(167) = 1.31, p = .192.

Gender effects

As in Study 2, there were significant effects of gender in the overall model that, while not bearing on the proposed hypotheses, still deserve discussion. Also, as in Study 2, gender did not moderate the effects described earlier as the four-way interaction of gender, message metaphor, behavior frame, and intention was not significant, F(12, 474) = 1.38, p = .17.

There was a significant main effect of gender, F(1, 158) = 20.04, p < .001, η p 2 = .113. Females had higher intentions for the behaviors (M = 4.5) than males (M = 3.9). There was also a marginal interaction of gender and behavior frame, F(2, 158) = 2.51, p = .085, η p 2 = .031. Females had much more intentions than males for the behaviors when they were neutrally framed, F(1, 158) = 8.28, p = .005, η p 2 = .050 for the simple effect; framed as an enemy, F(1, 158) = 6.92, p = .009, η p 2 = .042 for the simple effect; and slightly more when they were framed as imbalance, F(1, 158) = 5.04, p = .026, η p 2 = .031 for the simple effect.

Last, there was an interaction of gender and intention type that was similar to the one observed in Study 2, F(3, 474) = 3.46, p = .016, η p 2 = .021. Females had more intention than males for self-limiting behaviors, F(1, 158) = 20.84, p < .001, η p 2 = .117 for the simple effect, and self-bolstering behaviors, F(1, 158) = 13.95, p < .001, η p 2 = .081 for the simple effect. Further, males and females did not differ in monitoring intentions, F < 1 for the simple effect. But unlike Study 2, females had more intention than males for hypothetical treatment intentions, F(1, 158) = 8.97, p = .003, η p 2 = .054 for the simple effect.

General Discussion

Counter to the prevailing wisdom’s expected benefits of framing cancer as a feared enemy, our findings suggest that framing cancer in terms of bellicose enemy metaphors has unintended side-effects that may impair efficient prevention strategies. Many behaviors that reduce the risk of cancer require one to limit enjoyable activities, from sunbathing to drinking alcohol and eating red meats. Yet, limiting and constraining oneself is not a concept closely associated with fighting enemies. Hence, a bellicose message frame that emphasizes fighting an enemy may render these protective behaviors less compelling than they might otherwise be. Three studies provided consistent support for this prediction. First, enemy framing reduced the likelihood that self-limiting behaviors were listed when participants described prevention options in a free response format (Study 1). Second, enemy framing reduced participants’ intention to engage in self-limiting prevention behaviors when a list of prevention options was presented to them (Studies 2 and 3). Third, counter to the prevailing wisdom, this negative impact of enemy framing on prevention intentions was not offset by the increased intentions to engage in other preventive behaviors (Studies 2 and 3), most notably self-bolstering behaviors, such as eating more fruits or more high fiber foods. Fourth, also in contrast to the prevailing wisdom, enemy framing did not increase participants’ intention to engage in effective monitoring procedures (Studies 2 and 3), nor did it affect their preference for different treatment options (Studies 2 and 3). Most notably, it did not increase their intention to seek aggressive treatments, in contrast to conjectures offered in the literature (Aktipis et al., 2010; Reisfield & Wilson, 2004). Finally, framing self-limiting behaviors as particularly apt in “fighting” cancer eliminated the negative impact of enemy framing, but did not provide any advantage over a neutral frame (Study 3). In combination, these findings cast doubt on the benefits of a metaphorical frame that has come to dominate public discourse about cancer: cancer as an enemy against whom we ought to wage a war “as aggressive as the war cancer wages against us” (Obama, cited in Lennon, 2009).

The findings are consistent with conceptual metaphor theory (Lakoff & Johnson, 1980) and extend earlier work that showed metaphorical framing effects on sociopolitical attitudes (Landau et al., 2009), relationship perception (Lee & Schwarz, 2014a), and reasoning about fictional cities (Thibodeau & Boroditsky, 2011). However, not all predictions of conceptual metaphor theory were fully supported. Most notably, the expected beneficial effect of metaphor matching (Landau et al., 2010; Lee & Schwarz, 2014b) was only partially observed. On one hand, describing preventive behaviors as apt to fight cancer eliminated the otherwise observed disadvantage of enemy framing; on the other hand, an imbalance-framed message not only increased prevention intentions when the behavior was described as apt at restoring balance (as predicted), but also when it was described as apt at fighting enemies (Study 3). One might conjecture that fighting enemies is one way of restoring balance on a battle field, but this possibility merely highlights an ambiguity that frequently plagues metaphor research: as the meaning of any other utterance, the meaning of metaphors is highly context sensitive and open to unanticipated interpretations.

Although the current research makes no predictions about the time-course of the effect of metaphorically framing cancer as an enemy, prior research suggests the effect would be relatively short lived. Metaphoric framing effects are often thought to operate similarly to conceptual priming effects (Lee & Schwarz, 2014b), so the metaphoric framing effect might be quickly replaced by the next activated concept. However, our messages with very minor wording differences were able to affect behavioral intentions, which are reliable predictors of behavior (at r = .47 in meta-analyses; Armitage & Conner, 2001). Hence, even though the influence of a single metaphoric framing manipulation may be short lived, it may have long-term consequences if people can be induced to form behavioral intentions, preferably in ways that facilitate the later implantation of the intention (see Gollwitzer, 1999). More important for the present issue, enemy metaphoric framing of cancer is pervasive in public discourse, and the influence of this continuous exposure is likely to far exceed the observed impact of a single additional exposure in an experiment. To illustrate the ubiquity of enemy framing in discussions about cancer, we conducted a collocation analysis of sources of contemporary American English using the COCA database (Davies, 2008). This analysis revealed that two verbs related to the enemy metaphor, FIGHT and BATTLE, are among the top 10 verbs found up to two words prior to the word CANCER. This high frequency of use may ultimately make the enemy metaphor for cancer a powerful influence on public health.

Our findings carry implications for public health messages, which now follow the view that enemy framing of cancer, through evoking fear, would increase public adherence to beneficial health behaviors. The enemy metaphor has pervaded media portrayals of cancer (Camus, 2009) and information created by cancer funding organizations. Yet, our studies suggest that enemy metaphoric language for cancer undermines intention for limitation-related prevention behaviors. Further, it does not increase motivation for active, agentic behaviors to fight the disease among a lay audience. As such, the evidence suggests that the use of enemy metaphors for cancer in public health information does not boost intention for active agentic behaviors as intended. Rather, it seems more likely that it hurts intention for self-limiting prevention behaviors. Hence, the continued use of the enemy metaphor in public information campaigns on cancer may not be warranted and may ultimately be hurting more than helping public health.