Metaphor in communication and thought

Metaphor has been regarded as an important tool for persuasion and explanation since classical antiquity. Over the last 50 years, scholars from multiple disciplines have studied the role of metaphor in communication and thought using a broad range of methods. The seminal work by linguist George Lakoff and philosopher Mark Johnson has proven especially influential in this pursuit. They famously proposed that conventional metaphors in language (e.g., “She shot down all my arguments”) are realizations of conceptual metaphors in thought (e.g., argument is war; [11]). In other words, people don’t just speak metaphorically; they think metaphorically.

According to Lakoff and Johnson’s ‘Conceptual Metaphor Theory’ (CMT), we understand a target domain (e.g., argument) by importing knowledge and structure from a source domain (e.g., war). In this way, people can leverage their knowledge of familiar subjects to reason about novel or complex subjects. The choice of source domain influences how the target domain is conceptualized, resulting in potential “framing effects” of metaphors. This is because metaphors have unique entailments; the structure of the source domain facilitates certain inferences—but not others—when it is mapped onto the target domain. For example, if an argument is understood metaphorically as a war, then the goal of interactants is to win—to ensure that one’s own views prevail. If an argument is understood metaphorically as a journey, in contrast, then the goal of interactants may be to meet in the middle—to find a compromise. Lakoff and Johnson argued that human thought is largely organized in this way and that reasoning about abstract domains is virtually impossible without metaphors [11].

Discourse analysts have, in some cases, found Lakoff and Johnson’s notion of the conceptual domain too broad to explain how metaphors are used in discourse, especially in the context of explanation and persuasion. In a series of studies on political discourse, Musolff [12, 13] has employed the sub-domain notion of scenario as a “a set of assumptions made by competent members of a discourse community about ‘typical’ aspects of a source-situation, for example, its participants and their roles, the ‘dramatic’ storylines and outcomes, and conventional evaluations of whether they count as successful or unsuccessful, normal or abnormal, permissible or illegitimate, etc.” [12, p. 28]. In this way, the same overarching domain may generate different conceptual entailments under difference domain-related scenarios. For example, in a study of metaphors in cancer experience, Semino et al. [14] showed how war-related metaphors can have different implications (e.g., for the (dis)empowerment of patients) depending on the specific scenario involved, such as preparing for battle (e.g., “I am sharpening my weapons in case I have to do battle”) vs. outcome of battle (e.g., “I’m not winning this battle”).

While there are ongoing debates over CMT and the cognitive mechanisms that support metaphor processing, there is a general consensus that metaphors play a critical role in communication and cognition [15–18]. Over the past decade, for example, a large body of experimental research has shown that metaphors can enhance the persuasive power of messaging and explanations (for review, see [17, 18]). Using a metaphor to convey a complex issue like climate change [19], crime [20], cancer [21, 22], or the federal budget [23] can shape observer beliefs and attitudes in ways that are compatible with the metaphor’s entailments.

The impact of framing a concept with a metaphor has been studied in a variety of ways. This includes having participants who have received information via metaphor provide or select a solution to a problem [20, 24], express their behavioral intentions [19, 21], and indicate their beliefs, preferences, or emotional states [22, 23, 53]. Table 1 provides a summary of variables that have been shown to moderate metaphor framing effects. This work has provided rich insights into the cognitive, affective, and social-pragmatic factors that influence the potency of metaphorical messages (see [17, 18] for reviews).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 1. Variables that moderate the impact of metaphor framing.

https://doi.org/10.1371/journal.pone.0294739.t001

Explanatory metaphors in vaccine discourse

Health communicators have developed a range of metaphors to address common questions and misconceptions about vaccines. These healthcare metaphors are often extended, elaborated, and used to provide explanations with the goal of persuasion. They contrast with the more subtle metaphoric language studied in most research on metaphor framing described above (although see [23] for an exception).

Recently, for instance, a range of metaphors have been deployed by scientists and healthcare practitioners to explain why people should get the COVID-19 vaccine even though it does not offer complete protection from the virus. For example, the US-based Northeast Georgia Health System published a blog post in 2021 titled “How is the COVID-19 Vaccine like a Seatbelt?” [33]. As the blog post reminds the reader, wearing a seatbelt keeps you safe in the event of an accident, but it is still possible to get hurt. This is also the case with vaccines for COVID-19. A similar idea is conveyed by the metaphor vaccines are raincoats, used, for example, in the following representative X post (formerly called Twitter) by a scientist in July 2021 [34]: “Concerning breakthrough infections. Think of the vaccine as a very effective raincoat. If it’s drizzling, you’ll be protected. If the rain is coming down hard, you might still be fine. But if you are going in and out of rainstorms all the time, you could end up getting wet.”

Another common question about the COVID-19 vaccines is how they were developed so quickly—and whether this means they are not as safe as vaccines that were developed more slowly in the past. Professor Sarah Gilbert and Dr. Catherine Green used a vivid metaphor to address this concern in their book Vaxxers about the development of the Oxford AstraZeneca vaccine [35]. They ask the reader to consider how a baker might speed up the process of producing cakes with personalized messages (p. 66):

“Think about a baker who sells personalised cakes iced with a message like ‘Happy 50^th birthday Joe’ or ‘Congratulations on your engagement Ali and Max’. She might wait until she gets an order, and only then start the process of mixing the ingredients, baking the cake, letting it cool, icing it all over, waiting for the icing to set, and then finally adding the customized message. If she gets the order the day before the cake is needed, that works well. But if she wants to be able to offer a quicker service, she could bake a stock of cakes and put on the base layer of icing every morning. She is taking a financial risk: if no orders come in, the pre-baked cakes will go stale and need to be thrown away. But it may be worth the risk. When a customer comes into the shop, all she has to do is pick up her piping bag and add the custom message while he waits. The cake is then ready to take straight to the party. Only in the case of a vaccine, the party is a pandemic.”

In this metaphor, the cake represents the starting material needed to develop any generic vaccine. The personalized icing represents the pathogen-specific string of DNA that is added to create a pathogen-specific vaccine. As the authors explain, researchers working on COVID-19 vaccines had pre-baked a stock of cakes that could be quickly iced when the novel coronavirus was decoded. Baking each personalized cake from scratch when the order is placed, on the other hand, reflects the older, less efficient method for developing vaccines.

The authors use another food-based metaphor to supplement their explanation. Before the COVID-19 pandemic led governments worldwide to invest enormous sums into vaccine research, funding this work was a slow and arduous process that required a lot of bureaucratic red tape (p. 157): “It is as if you are making a roast dinner and for every ingredient you have to make a separate trip to the shops to buy it, then cook it and demonstrate that it is going to be delicious, before moving on to the next.” When COVID-19 hit and governments infused this sector with substantial funding, on the other hand, “We were allowed to do a big shop and put all the ingredients we needed in the trolley all at once” (p. 158).

There has been limited research on the impact of metaphor framing on vaccine attitudes [36, 37]. In one such study, Scherer and colleagues [36] investigated whether describing influenza metaphorically (as a beast, riot, army, or weed) would increase people’s willingness and interest in being vaccinated. They found that, compared to a comparable literal description, metaphorical descriptions of the flu increased people’s expressed willingness to be vaccinated. This effect was observed more consistently in people who occasionally got the flu vaccine compared with those who never got it. Their study highlights one variable—the strength of prior beliefs or attitudes—that moderated the impact of a metaphoric message (as shown in Table 1, for example). Describing the influenza virus as a beast is quite different from the elaborate explanatory metaphors used in COVID-19 discourse, however. For one thing, beast is a metaphor for a virus rather than a metaphor for some aspect of vaccination. For another, beast is a subtle, one-off metaphor rather than an extended and elaborated explanatory metaphor. In contrast, the cake metaphor described above is explicitly presented as an explanation for how researchers could develop the COVID-19 vaccines so quickly, and it was extended and developed throughout the text. It is this latter type of explanatory metaphor that we aimed to evaluate in the present study.

A recent study by Ervas and colleagues [37] did examine how an explanatory metaphor would impact the efficacy of vaccine communications. Their central message focused on “collective immunity,” emphasizing how all people should be required to receive a vaccination to benefit the larger group. Including an explanatory metaphor—likening people to bees in a beehive who must collaborate—significantly increased a range of assessments of the message, including its emotional impact and perceived convincingness. However, it did not increase intentions to vaccinate. Our study builds on this basic design by sampling a wider range of more elaborate explanatory metaphors in vaccine discourse and using a more comprehensive measure of general vaccine attitudes.

The current study

A critical question concerns the general effectiveness of extended explanatory metaphors in vaccine discourse. That is, do such metaphors lead to positive effects on people’s understanding of and attitudes towards vaccination? To date, this issue has not been systematically assessed in the experimental literature. We addressed this question in the current study by comparing the outcomes of participants of health messages with and without extended explanatory metaphors.

Overview

We used a between-subject, pretest/posttest design. Participants completed a 24-item measure of their attitudes towards vaccines both before and after evaluating health messages that we organized as a series of responses to five common questions about vaccines. We created three responses for each question: two responses included unique explanatory metaphors—adapted from real-world source materials like the ones described above—and a baseline response that provided the same information but did not include an explanatory metaphor. We refer to these baseline responses as ‘literal’ responses, and we operationalized this term as “not containing an extended explanatory metaphor.” Participants were randomly assigned to receive either all metaphor or all literal messages. Those in the metaphor condition randomly received one of the two explanatory metaphors in response to each of the five target questions. After reading each response, metaphor and literal participants rated how understandable, informative, and persuasive it was. This allowed us to measure whether people interpreted metaphor-infused messages as more effective, equally effective, or less effective when compared to literal messages (i.e., those that were not organized around a central explanatory metaphor). We also included an innovative method for assessing the impact of metaphorically framed messages: after rating each message, participants were asked to freely describe how they would answer the associated target question if it were posed to them by a friend. This resulted in a rich natural language dataset that allowed us to conduct novel linguistic analyses to assess the impact of metaphors.

Our findings provide important insights into the effects of explanatory metaphors in vaccine communications. The study also serves as a design model for the systematic investigation of the role of explanatory metaphors in public discourse, as we included a combination of traditional quantitative and linguistic analytic methods.

Participants

Data were collected in June 2022. We recruited 301 participants from Amazon’s Mechanical Turk crowdsourcing platform [38]. We used the CloudResearch interface [39], which includes a set of pre-screening measures shown to improve Mturk participant and data quality [40, 41]. We aimed for a sample size of at least 100 individuals in each condition to be consistent with past research on metaphor framing [e.g., 17, 25, 42, 43]. All participants were at least 18 years of age (M = 38.3, SD = 10.4), residents of the United States, and had a good performance record on previous MTurk tasks (with a minimum of 95% approval rating). The sample included 117 female, 176 male, and four non-binary/no-gender participants, with four additional participants who preferred not to indicate their gender. We did not collect information that could identify individual participants during or after data collection. All participants gave informed consent prior to beginning the study. This study was reviewed and approved under reference FASSLUMS-2021-0576-RECR-2 by Lancaster University Management School Research Ethics Committee’s Faculty of Arts and Social Sciences (FASS-LUMS). Participants were paid $3 (USD) once they completed the study.

Materials

Vaccine Attitude Measure (VAM).

We developed a 24-item questionnaire designed to measure participants’ beliefs and attitudes toward vaccines. Most items were adapted from (a) the Vaccination Attitudes Examination Scale [44], (b) the Vaccine Hesitancy Scale [45], and (c) the Vaccine Conspiracy Beliefs Scale [46]. We also included (d) unique items generated for the purposes of the current study. We drew on multiple sources to create a comprehensive measure that addressed a wide range of vaccine beliefs and attitudes. Items were clustered into eight sub-categories consisting of three items each, as illustrated in Table 2. The instructions required participants to indicate their support for a series of statements on a scale from 1 (strongly disagree) to 7 (strongly agree). Tests of internal consistency showed the measure was reliable (Cronbach’s α = 0.937 for the VAM pretest). In coding, we averaged together responses, reverse-coding as needed, such that higher scores equate to more positive attitudes and more accurate beliefs about vaccines.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 2. Items, coding, and sub-categories in our Vaccine Attitude Measure (VAM).

https://doi.org/10.1371/journal.pone.0294739.t002

Vaccination questions and messages.

We designed the stimulus materials to resemble a public health messaging campaign. To identify common misperceptions, we focused on public health messaging in the U.S., the U.K., and other international organizations. This included, for example, the comprehensive Myths and Facts about COVID-19 Vaccines from the Centers for Diseases Control in the U.S. [47], Vaccine Myths from the British Society for Immunology [48], and Vaccines and immunization: Myths and Misconceptions from the World Health Organization [49]. We focused on five questions aimed at tapping into common questions, misconceptions, and concerns about vaccines, as follows:

1. How do vaccines work?
2. Is ‘natural immunity’ better than the immunity provided by a vaccine?
3. Are vaccines that are developed quickly safe?
4. Why should I take a vaccine if I am personally at low risk for the illness?
5. Why get a vaccine if it isn’t 100% effective?

For each question, we created three corresponding responses: two that used extended explanatory metaphors that were different from each other and a baseline literal response that, as we noted earlier, did not include an extended, explanatory metaphor. Using Musolff’s [12] terms, these extended metaphors exploit the narrative potential of specific source scenarios. Most of the metaphors were adapted from authentic, real-world examples of metaphors in common use in vaccine discourse. These were collected through the authors’ observations from the media and from websites on public health messaging, like those listed above and described in the Introduction. We computed readability consensus scores to ensure comparability of the texts based primarily on the work of Flesch and Flesch-Kincaid and others [50], using a range of measures to gauge the readability or difficulty level of a text in English. These explanatory messages are provided in Table 3.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 3. Vaccination questions and corresponding explanation stimuli.

https://doi.org/10.1371/journal.pone.0294739.t003

After reading one of the three explanations associated with a particular question, participants were asked to consider how the general public would react to the message and to keep this in mind while responding to four questions. Participants were asked to consider the general public in order to reduce any reluctance they might have to go on the record with their personal responses given the highly emotive and politicized nature of vaccine/booster discourse at the time of the study, as well as to reduce any potential audience effect. In other words, the goal was to obtain maximally authentic data. First, participants rated three properties of the message using a 7-point scale: (1) “How understandable was this paragraph?” (1 = very difficult to understand, 7 = very easy to understand); (2) “How informative was this paragraph?” (1 = not informative at all, 7 = very informative), and (3) “How persuasive was this paragraph?” (1 = not persuasive at all, 7 = very persuasive). Finally—advancing a new methodology for metaphor research and mindful of the influence of social networks on vaccine attitudes—we asked participants how they would respond to a friend who asked them one of the five common questions above about vaccines (e.g., Question 5: “Why get a vaccine if it isn’t 100% effective?”), and to provide the response they would give their friend by typing into a blank text box.

Procedure

Participant vaccination data

The average level of support for vaccines, in general, was relatively high among our participants (M = 5.93, SD = 1.46 on our 7-point scale). Most participants reported they had received or planned to receive a COVID-19 vaccination (81.4%) and booster (64.8%). The vaccination rate is in line with the overall U.S. rate of vaccination rate of 78% as of June 29, 2022 (the day after data collection was completed. See https://usafacts.org/visualizations/covid-vaccine-tracker-states). Only 32% of Americans had received a booster as of June 29, 2022, suggesting our sample was more likely to be boosted than the general public. However, our question about booster status also left open the possibility that a participant was planning to receive a booster shot in the future, indicating this gap is likely smaller than it appears. Overall, these data suggest our sample was slightly more accepting of COVID-19 vaccines than the general public.

Explicit ratings of the health messages

All analyses were carried out using jamovi open-source statistics software. We conducted a series of one-way ANOVAs to compare the effect of each of the three stimulus vignettes for each of our five questions on ratings of understandability, informativeness, and persuasiveness. These analyses revealed that participants generally rated the metaphorical and literal responses for each question as similarly understandable, informative, and persuasive. There were, however, several notable significant differences, as revealed by Tukey’s post-hoc t-tests: For Question 1, the bank (M = 5.86, SD = 0.95) and castle (M = 5.89, SD = 1.42) metaphors were rated as more understandable than the literal message (M = 5.35, SD = 1.19); t(298) = 2.96, p = 0.009 and t(298) = 3.18, p = 0.005, respectively. For Question 4, the war metaphor was rated as both less understandable (M = 5.53, SD = 1.27 vs. M = 5.99, SD = 1.13; t(298) = 2.66, p = 0.023) and less informative (M = 5.00, SD = 1.56 vs. M = 5.54, SD = 1.27; and t(298) = 2.63, p = 0.025) than the speed limit metaphor. The war metaphor was also rated as less understandable than the literal message (M = 6.07, SD = 1.25; t(298) = 3.13, p = 0.005). Finally, for Question 5, the raincoat metaphor was rated as more persuasive than the literal message (M = 5.24, SD = 1.69 vs. M = 4.35, SD = 1.93; t(298) = 3.60, p = 0.001). See Table 4.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 4. Explicit ratings of each metaphor.

https://doi.org/10.1371/journal.pone.0294739.t004

This analysis suggests that certain metaphors may help or hinder communications about a particular topic. For example, using a bank or castle metaphor to explain how vaccines work may make a message easier to understand, though it does not appear to impact how informative or persuasive the message seems in relation to a comparable literal message. Using a war metaphor involving mobilization by a country under attack to explain why people should take a vaccine even if they are personally at low risk for the illness appears to be particularly ineffective, eliciting lower understandability and informativeness ratings. We offer additional evidence for the efficacy of certain metaphors in the linguistic analysis section below. Overall, however, the explicit ratings data indicate that people’s evaluations of the messages were only slightly impacted by the presence of a metaphor. For most of the comparisons, participants tended to perceive the metaphor-enriched messages as similarly understandable, persuasive, and informative as the messages that did not include an extended explanatory metaphor.

(How) Did overall vaccine attitudes change from pre- to posttest by condition?

We used a repeated-measures ANOVA comparing attitudes towards vaccines before and after participants read and rated the messages (VAM-pretest vs. VAM-posttest). Whether the participant received explanatory metaphor messages or literal messages was included as a between-subject factor (note that this analysis does not distinguish between which of the two metaphorical messages a participant received if they were in the metaphor condition). Previous research—including our own pilot testing—has shown that a variety of individual differences and demographics are associated with attitudes towards vaccines. Therefore, we analyzed age, self-reported science background, generic conspiracist beliefs, trust in institutions, and political ideology as covariates in the model. Two individuals left certain demographic variables blank and were therefore excluded from this analysis.

Overall vaccine attitudes became significantly more favorable from pretest (M = 4.98, SD = 1.22) to posttest (M = 5.23, SD = 1.24), F(1, 292) = 19.27, p < 0.001. However, there was no interaction between vaccine attitudes (pre- vs. posttest) and condition, F(1, 292) = 0.059, p = 0.81, as illustrated in Fig 1. These findings suggest that simply taking part in a study reading healthcare communications led to more favorable/accurate attitudes towards vaccines, whether or not extended explanatory metaphors were included in the messaging.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 1. Mean pretest and posttest Vaccine Attitude Measures by condition.

Error bars represent standard errors of the mean. Data for all 301 participants is shown.

https://doi.org/10.1371/journal.pone.0294739.g001

Interestingly, a significant within-subject effect was found in the interaction between vaccine attitudes (pre- vs. posttest) and trust in institutions, F(1, 292) = 7.17, p = 0.008. A similar effect was found for generic conspiracy beliefs, though this effect was marginal, F(1, 294) = 5.05, p = 0.05. People whose questionnaire responses indicated they had lower trust in institutions or more generic conspiracy beliefs showed a greater increase in vaccine attitude scores from pre to posttest compared to people with higher trust in institutions or fewer conspiracy beliefs. However, this is likely a result of lower pretest baseline scores for these low-trust and high-conspiracy individuals.

Three covariates independently predicted vaccine attitudes collapsing across pre- and posttest: (a) generic conspiracist beliefs (F(1, 292) = 154.86, p < 0.001), (b) political ideology (F(1, 292) = 118.96, p < 0.001), and (c) trust in institutions (F(1, 292) = 12.75, p < 0.001). Endorsing conspiracy beliefs, conservative political ideology, and lower trust in institutions were all associated with less favorable vaccine attitudes (i.e., lower scores on our measure). This is consistent with previous research on vaccine hesitancy [5–9] and provides additional empirical support for those findings while highlighting the validity of our adapted vaccine attitude measure.

Exploratory analyses by vaccine attitude sub-category

We conducted an identical repeated-measures analysis for each of the eight VAM sub-category scales, including all of the same covariates as before. This enabled us to explore whether the messages led to differential impacts on participants’ attitudes toward certain vaccine attitude items since some of the sub-categories corresponded directly to the content of the messages (e.g., sub-category 2, on “natural immunity,” corresponds to vaccine Question 2 “Is ‘natural immunity’ better than the immunity provided by a vaccine?”).

Overall, most of the same general patterns were obtained for the sub-category scales as for the total VAM analysis, though the findings are noisier due to increased variance in the smaller sub-scales. However, these data do suggest that health messages may have been particularly effective at targeting the specific content included in the messages. VAM scores increased significantly from pre- to posttest for sub-category scales 2 (F(1, 292) = 16.95, p < 0.001), 3 (F(1, 292) = 22.18, p < 0.001), and 5 (F(1, 292) = 12.65, p < 0.001). The effect for subscale 4 was similar, though this test did not quite reach significance (F(1, 292) = 3.88, p = 0.05). Interestingly, VAM scores decreased between pre- and posttest for sub-category scale 1, though not significantly (F(1, 292) = 3.33, p = 0.069), which is a finding we unpack in the General Discussion. Each of these sub-scales was associated with one of our target questions (see Tables 2 and 3). VAM scores did not significantly differ between pre- and posttest for sub-category scales 6, 7, and 8, though the numeric trends were in line with the increase observed in the other sub-scales. Sub-scales 6–8 tapped beliefs and attitudes that were not explicitly addressed in our health messages (i.e., personal attitudes towards vaccines, attitudes towards childhood vaccinations, and vaccine conspiracy beliefs). Future work is needed to develop messages to target these issues directly.

Whether or not the message included an explanatory metaphor was not a significant factor for any of the subscales, with one notable exception: For sub-category scale 2 (‘natural immunity’), scores increased significantly more between pre- and posttest for participants in the metaphor condition than the literal condition, F(1, 292) = 16.95, p = 0.036. See Fig 2. In exploratory research like the current study, the possibility of false positives must always be considered. However, this result may also indicate that metaphors are especially effective for communicating about this particular issue of natural immunity. Current research underway in our labs is further testing this intriguing possibility.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 2. Mean pretest and posttest VAM subcategory 2 scale scores by condition.

Error bars represent standard errors of the mean. Data for all 301 participants is shown.

https://doi.org/10.1371/journal.pone.0294739.g002

Linguistic analyses

As noted above, the current study aimed to advance traditional procedures in metaphor research in that we asked our participants to report how they would respond to a friend asking the target questions (“How would you respond to a friend who asked you…?”). Participants filled in a free-text response box, into which they could type without a time or character limit. This elicitation method, while not uncommon in social science and survey-based research (e.g., [54]), has not been used in prior metaphor framing research (although [22] used a related but more abstract ‘imagine a situation’ measure). This technique was designed to obtain data to triangulate the attitude measures of the efficacy of public health messaging. It also allowed us to gauge the extent to which the different explanatory metaphors participants read may have been memorable or ‘sticky’ for them (in the sense of engagement or involvement). We were interested to see whether metaphors participants had read were reused by them when they answered the vaccine-related concerns of their purported friend. The “explain to a friend” prompt also allowed us to investigate how the different explanatory metaphors were reused. Their free response data allowed us to touch upon social communication in interactions with friends and family in decisions about vaccinations. This dovetails with reports like the UK’s SAGE Working Group on Vaccine Hesitancy [9], where the authors include individual and group influences as one of three groups of determinants of vaccine hesitancy. As we took a first pass through the data, a preliminary posthoc analysis revealed that the answers to the “explain to a friend” prompt also provided novel participant-generated metaphors, which were not a focus of the current study. Although space constraints prevent a full discussion in the current paper, assessment of these data may reveal insightful information about the relationship between the original messages participants read and the new metaphors they constructed themselves to explain vaccines to their friends. We look forward to describing these findings in future reports.

Did the length of free-text responses (in terms of word tokens) differ depending on condition?

To address this component of participant responses, we calculated the length of free-text responses in terms of the number of words (tokens) used. The results are displayed in Table 5 and Fig 3.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 3. Percentage difference in mean word production on the “explain to a friend” free response task for each metaphor relative to the comparable literal condition.

https://doi.org/10.1371/journal.pone.0294739.g003

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 5. Cumulative and mean number of word tokens in free-text responses for each condition and question.

https://doi.org/10.1371/journal.pone.0294739.t005

Overall, participants in the metaphor condition (M = 140.5, SD = 73.6) used significantly more total words on average (~18%) in their “explain to a friend” responses than those in the literal condition (M = 119.2, SD = 62.7), t(299) = 2.51, p = 0.013, d = 0.31. This effect was observed fairly consistently across the questions. Participants used significantly more words in the metaphor condition than the literal condition for Question 2 (t(299) = 3.27, p = 0.001, d = 0.40), Question 3 (t(299) = 2.19, p = 0.029, d = 0.27), Question 4 t(299) = 1.98, p = 0.049, d = 0.24) and Question 5 (t(299) = 2.39, p = 0.018, d = 0.29), though not for Question 1 (t(299) = 0.73, p = 0.466, d = 0.09). See Table 5.

Breaking the word counts down by metaphor is illuminating. As illustrated in Table 5 and Fig 3, compared to participants who read literal passages, those who received the metaphorical messages used more words on average for all their explanations to a friend. This difference was negligible (0.4%) when comparing those who received the bank metaphor to those who read the literal passage for Question 1. Overall, however, the average word count for “explain to a friend” answers in eight out of the ten metaphor conditions was at least 10% higher than for the corresponding literal conditions. And in four of the metaphor conditions (pilot, fire drill, war, and raincoat), participants used more than 25% more words compared to those who received comparable literal messages.

To what extent were the metaphors participants read then reused in their explanations to a friend’s answers?

We also coded the “explain to a friend” responses in terms of whether or not the explanatory metaphor contained in the passage that each participant read was then reused by the participants in the explanations they said they would give to their friends. All data were double-coded by two research assistants who, after a 45-minute training session involving 5% of the data, proceeded with individual coding. Inter-rater reliability (IRR) was calculated using all data across the five metaphor stimuli using Cohen’s Kappa, revealing a high degree of agreement in binary Yes-No metaphor reuse (M Cohen’s Kappa = 0.94 across the five stimuli). The two coders met to resolve coding disagreements through additional discussion until they achieved complete agreement on all codes (M Cohen’s Kappa = 1.00 across the five stimuli). Table 6 and Fig 4 show the outcomes of the coding.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 4. Percentage of reuse of each explanatory metaphor in “explain to a friend” responses.

https://doi.org/10.1371/journal.pone.0294739.g004

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 6. Percentage of reuse of each explanatory metaphor in “explain to a friend” responses.

https://doi.org/10.1371/journal.pone.0294739.t006

Five of the metaphors were reused by more than 10% of the participants—raincoat, bank, castle, seatbelt, and video game—indicating that these metaphors were likely particularly salient for participants as they responded to the “explain to a friend” prompt. Three metaphors were reused to an even greater extent—by more than 20% of participants. These especially salient metaphors were castle (22%), bank (23%), and raincoat (27%).

The metaphors used in response to Questions 1 (“How do vaccines work?”) and 5 (“Why get a vaccine if it isn’t 100% effective?”) were reused most often. This might be because of the topic of the questions (e.g., how and why), or because of an inherent attribute of the metaphors (more memorable or”sticky”), or due to the interaction between the two. The two metaphors for Question 1 were reused to a similar extent, whereas the raincoat metaphor was reused more than the seatbelt metaphor in relation to Question 5. The metaphors for Question 2 were reused less often. When considering which specific metaphors were retained and reused by participants, we also found it interesting to look at how they were used.