Metaphors We Think With: The Role of Metaphor in Reasoning

Citation: Thibodeau PH, Boroditsky L (2011) Metaphors We Think With: The Role of Metaphor in Reasoning. PLoS ONE 6(2): e16782. doi:10.1371/journal.pone.0016782

Editor: Jan Lauwereyns, Kyushu University, Japan

Received: November 3, 2010; Accepted: January 13, 2011; Published: February 23, 2011

Copyright: © 2011 Thibodeau, Boroditsky. This is an open-access article distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited.

Funding: This research was funded by NSF Grant No. 0608514 to LB. The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript.

Competing interests: The authors have declared that no competing interests exist.

Introduction

Both crime, and the criminal justice system designed to deal with crime, impose tremendous costs on society. Over 11 million serious crimes are reported in the United States each year [1], and the US has the highest per capita imprisonment rate of any country [2]. Despite being home to only 5% of the world's population, the United States holds 25% of the world's prisoners, with nearly 1% of the US population living behind bars [3]. Addressing the crime problem is an issue of central importance in social policy. How do people conceptualize crime, and how do they reason about solving the crime problem?

Public discourse about crime is saturated with metaphor. Increases in the prevalence of crime are described as crime waves, surges or sprees. A spreading crime problem is a crime epidemic, plaguing a city or infecting a community. Crimes themselves are attacks in which criminals prey on unsuspecting victims. And criminal investigations are hunts where criminals are tracked and caught. Such metaphorical language pervades not only discourse about crime, but nearly all talk about the abstract and complex [4]–[5]. Are such metaphors just fancy ways of talking, or do they have real consequences for how people reason about complex social problems like crime?

Previous work has demonstrated that using different metaphors can lead people to reason differently about notions like time, emotion, or electricity [6]–[11]. For example, people's reasoning about electricity flow differed systematically depending on the metaphoric frame used to describe electricity (flowing water vs. teeming crowds) [6]. Such findings on metaphorical framing are grounded in a larger body of work that has established the importance of linguistic framing in reasoning [12], and the importance of narrative structure in instantiating meaning [13]. However, questions about the pervasiveness of the role of metaphor in thinking remain. Critics argue that very little work has empirically demonstrated that metaphors in language influence how people think about and solve real-world problems [14].

In this paper we investigate the role of metaphor in reasoning about a domain of societal importance: social policy on crime. Beyond establishing whether metaphors play a role in how people reason about crime, our studies are designed to further illuminate the mechanisms through which metaphors can shape understanding and reasoning. If metaphors in language invite conceptual analogies, then different metaphors should bring to mind different knowledge structures and suggest different analogical inferences. In this paper we ask if metaphors indeed play such a role in reasoning about social policy. That is, do we reason about complex social issues in the same way that we talk about them: through a patchwork of metaphors?

Some observations of crime policy in the real world suggest that people may indeed take metaphors as more than just talk. For example, shifts in metaphors are often accompanied by shifts in policy. In the 1980s Ronald Reagan declared a war on drugs, with smugglers, dealers, and users defined as the enemy to be fought. Policies in line with the war on drugs mandated longer, harsher sentences for drug-related crime. Since then, the incarceration rate has more than quadrupled in the US [15].

Others have taken the crime is a virus metaphor seriously and have implemented programs to treat crime as a contagious disease. For example, a crime-prevention program run by an epidemiologist in Chicago treats crime according to the same regimen used for diseases like AIDS and tuberculoses, focusing on preventing spread from person to person [16].

Some criminal justice scholars have even implicated bad metaphor as the root of failure in crime prevention [17]. In one case described by Kelling, a serial rapist attacked 11 girls over a 15-month period before being captured by the police. During those 15 months, the police had information that (had they shared it with the community) could have prevented some of the attacks. Instead, they opted to keep that information secret to set traps for their suspect. The police, on Kelling's analysis, were entrenched in their metaphorical role of hunting down and catching the criminal, and neglected their responsibility to inoculate the community against further harm. The girls, Kelling writes, “were victims… not only of a rapist, but of a metaphor” (p. 1).

In this paper we empirically investigate whether using different metaphors to talk about crime indeed leads people to reason about crime differently and, in turn, leads them to propose different solutions to the crime problem. We will focus on two contrasting metaphors for crime: crime as a virus and crime as a beast. Do these metaphors subtly encourage people to reason about crime in a way that is consistent with the entailments of the metaphors? For example, might talking about crime as a virus lead people to propose treating the crime problem the same way as one would treat a literal virus epidemic? Might talking about crime as a beast lead people to propose dealing with a crime problem the same way as one would deal with a literal wild animal attack?

To help generate a clear set of predictions, we conducted a norming survey asking 28 participants on Amazon's Mechanical Turk (www.mturk.com; [18]) to describe what should be done to solve a literal virus or beast problem. We asked people to imagine a “virus infecting a city” or a “wild beast preying on a city” and then to describe the best way to solve the problem that they had imagined. Participants who imagined a “virus infecting the city” universally suggested investigating the source of the virus and implementing social reforms and prevention measures to decrease the spread of the virus. That is, they wanted to know where the virus was coming from, whether the city could develop a vaccine and how the virus was spreading. They also wanted to institute educational campaigns to inform residents about how to avoid or deal with the virus and encourage residents to follow better hygiene practices. Participants who imagined a “wild beast preying on a city” universally suggested capturing the beast and then killing or caging it. They wanted to organize a hunting party or hire animal control specialists to track down the beast and stop it from ravaging the city.

Might these schematic representations for solving literal virus or beast problems transfer to people's reasoning about crime if crime is metaphorically framed as a virus or a beast? That is, if crime is talked about as a virus, will people suggest diagnosing the root cause of the problem and enacting social reform to treat and inoculate the community? If crime is a beast, will people suggest catching and jailing criminals in order to fight off the crime attack?

In Experiment 1, we gave people a report about increasing crime rates in the City of Addison and asked them to propose a solution. For half of the participants, crime was metaphorically described as a beast preying on Addison, and for the other half as a virus infecting Addison. The rest of the report contained crime statistics that were identical for the two metaphor conditions. The results revealed that metaphors systematically influenced how people proposed solving Addison's crime problem. When crime was framed metaphorically as a virus, participants proposed investigating the root causes and treating the problem by enacting social reform to inoculate the community, with emphasis on eradicating poverty and improving education. When crime was framed metaphorically as a beast, participants proposed catching and jailing criminals and enacting harsher enforcement laws.

In Experiment 2, we modified the report and repeated the study. Whereas in Experiment 1, the metaphoric frame was established using vivid verbs with rich relational meaning in phrases scattered throughout the report (e.g., crime was said to be either preying & lurking, or infecting & plaguing). In Experiment 2, we used a single word to instantiate the metaphoric frame. Despite this small difference between the virus and beast conditions in the modified report (“Crime is a virus/beast ravaging the city of Addison”), we again found that participants in the two conditions offered different problem solving suggestions. The findings of Experiment 2 demonstrate that these relational elements need not be specified explicitly. People spontaneously extracted the relevant relational inferences even given a single metaphorical noun in Experiment 2.

In Experiment 3 we tested whether the influence of the metaphor observed in the first two studies could have come about through simple spreading activation from lexical associates of the words “beast” and “virus.” Perhaps simply hearing a word like beast, even outside of the context of crime, would activate representations of hunting and caging. These activated lexical associates might then bleed into people's descriptions of how to solve the crime problem. To test for this possibility we dissociated the words “beast” and “virus” from the metaphorical frame in Experiment 3. Before reading the crime report, participants were asked to provide a synonym to the word “beast” or the word “virus” – thereby priming representations for a beast or a virus. They then read the same report about crime as in Experiment 2, but with the metaphorical word omitted (“Crime is ravaging the city of Addison”). This disconnected lexical prime did not yield differences in people's crime-fighting suggestions, revealing that metaphors act as more than just isolated words – their power appears to come from participating in elaborated knowledge structures.

In Experiment 4 we tested whether metaphors can affect not only how people propose solving the problem of crime, but also how they go about gathering information for future problem solving. If participants seek out information that is likely to confirm the initial bias suggested by the metaphor, this may be a mechanism for metaphors to iteratively amass long-term effects on people's reasoning. Indeed, when people were presented with a metaphorically framed crime problem and then given the opportunity to gather further information about the issue, participants chose to look at information that was consistent with the metaphorical frame.

In Experiment 5 we investigated the time-course of how metaphors influence the construal of complex issues. One possibility is that metaphors influence reasoning by providing people a knowledge frame that structures subsequent information. After being exposed to the metaphor, participants assimilate all further information they receive into this knowledge structure, instantiating any ambiguous information in a way that would be consistent with the metaphor. If this is the case, if metaphors actively coerce incoming information, then metaphors should have the most impact when they are presented early. This was the structure of the report in Experiment 4 (and Experiment 2): the metaphoric frame was presented in the first sentence of the report.

Alternatively, if metaphors simply activate a stored package of ideas and do not encourage the kind of active assimilation process described above, then they should be most effective when they are presented late in the narrative, as close to when people are asked to reason about a solution as possible. This way, the memory of the metaphor should be fresh and any knowledge activated by it should have the best chance to influence reasoning. This was the structure of the report in Experiment 5: the metaphoric frame was presented in last sentence of the report. Unlike the results of Experiment 4, this late metaphorical framing had no effect on people's crime-related information foraging. These findings suggest that metaphors can gain power by coercing further incoming information to fit with the relational structure suggested by the metaphor.

One of the most interesting features of the effects of metaphor we find throughout these studies is that its power is covert. When given the opportunity to identify the most influential aspect of the crime report, participants (in all four studies that include a metaphoric frame) ignore the metaphor. Instead, they cite the crime statistics (which are the same in both conditions) as being influential in their reasoning. Together these studies suggest that unbeknownst to us, metaphors powerfully shape how we reason about social issues. Further, the studies help shed light on the mechanisms through which metaphors influence our reasoning.

Methods

Ethics Statement

The experiments reported here were done in accordance with the Declaration of Helsinki. Additionally, they followed the ethical requirements of the Stanford University institutional review board and complied with ethics guidelines set forth by the IRB recommendations. Participants were informed that their data would be treated anonymously and that they could terminate the experiment at any time without providing any reason. We received written informed consent from all participants before they participated in an experiment.

Participants

In Experiment 1, 485 students – 126 from Stanford University and 359 from the University of California, Merced – participated in the study as part of a course requirement. Experiments 2–5 were conducted online with participants recruited from Amazon's mechanical Turk (347, 312, 185, and 190, respectively). In exchange for participation in the study, people were paid $1.60 – consistent with a $10/hour pay rate since the study took 5 to 6 minutes to complete.

Gathering data from these various sub-populations allowed us to sample a broader cross-section of the general population. This is important since people's conceptions of social issues like crime are likely to differ as a function of factors like socioeconomic status and personal experience. This is particularly true of the sample that was recruited online, which was more diverse than that available at Stanford specifically or on college campuses generally [18].

Running Experiments 2–5 online also afforded careful control over our sample population. We used Mechanical Turk's exclusion capabilities and tracked IP addresses to ensure that participants were not repeatedly sampled. We also restricted our study to Turkers with a 95% or better performance record to ensure that we were sampling high quality participants (“Requesters” have the opportunity to publicly give positive or negative feedback to their participants, which can then be used as a criterion for future “Requesters”). At the end of the online version of the study we asked participants to describe their language history, current geographic location, and provide some background information. We then restricted our analysis to residents of the United States who were native English speakers. The characteristics of our samples are detailed in the Results section below.

Materials

In each of the five experiments, participants were presented with a survey that included a short paragraph about crime in the fictional city of Addison and some follow-up questions. The survey differed subtly between experiments, but always contrasted a crime-as-virus framing with a crime-as-beast framing.

It should be noted that there are two somewhat different metaphorical frameworks that treat crime as an illness. In one, the community or population is seen as an organism, and crime is a disease that is developing inside that organism (e.g., “Violent crime is a cancer that eats away at the very heart of society.”). In another, the community is seen as individual agents and crime is a contagious disease that can be passed on from one person to another forming an epidemic. In this paper the stimuli did not strongly distinguish between these different varieties of crime as illness metaphors, but doing so would be an interesting extension of this work, as these metaphors suggest somewhat different implications for treating crime.

Experiment 1.

In the first experiment, participants were presented with one of two versions of the crime paragraph. The two versions of the paragraph differed only in the embedded metaphor: In one, crime was a beast; in the other, crime was a virus. The majority of the paragraph consisted of crime statistics, which were the same in both versions. Half of the participants were given the crime-as-beast version and half the crime-as-virus version. The paragraph read:

---

Crime is a {wild beast preying on/virus infecting} the city of Addison. The crime rate in the once peaceful city has steadily increased over the past three years. In fact, these days it seems that crime is {lurking in/plaguing} every neighborhood. In 2004, 46,177 crimes were reported compared to more than 55,000 reported in 2007. The rise in violent crime is particularly alarming. In 2004, there were 330 murders in the city, in 2007, there were over 500.

This report was followed up with two questions: 1) In your opinion what does Addison need to do to reduce crime? 2) Please underline the part of the report that was most influential in your decision. This question was aimed at discovering if participants explicitly noticed or made use of the metaphor.

Experiment 2.

The crime report used in the second experiment was similar, but not identical to the one used in Experiment 1. Importantly, it instantiated the beast or virus metaphor for crime with a single word. It read as follows:

---

Crime is a {beast/virus} ravaging the city of Addison. Five years ago Addison was in good shape, with no obvious vulnerabilities. Unfortunately, in the past five years the city's defense systems have weakened, and the city has succumbed to crime. Today, there are more than 55,000 criminal incidents a year - up by more than 10,000 per year. There is a worry that if the city does not regain its strength soon, even more serious problems may start to develop.

In Experiment 2, we asked three follow-up questions in the following order: 1) In your opinion what does Addison need to do to reduce crime? 2) What is the role of a police officer in Addison? 3) Please copy the part of the report that was most influential and paste it in the text area below. Questions one and two were free-response. Question three was copy and paste (participants were shown the report adjacent to an open text field and were asked to copy the portion of the report that was most influential in their reasoning and paste it into the open text field).

Experiment 3.

The design of Experiment 3 was similar to that of Experiment 2; however, before participants read the crime report, they were shown the word “beast” or the word “virus” and were asked to “list a synonym” for it. After completing this task, they were presented with the paragraph on crime in Addison on a separate screen. The crime report used in Experiment 3 was the same as the crime report for Experiment 2, except that it did not contain a virus or beast metaphor. The first sentence of the report read: “Crime is ravaging the city of Addison.” It was otherwise identical to the report from Experiment 2.

Experiment 4.

The crime report used in Experiment 4 was the same as the crime report used for Experiment 2. However, instead of asking the follow-up questions from Experiments 2 and 3, we asked participants to select one of four crime-related issues for further investigation – with the knowledge that this information should be used to help them make a more informed crime-reducing suggestion. The instructions read as follows: “Now imagine that Addison has consulted you about the crime problem. You have the resources to investigate one of the following four issues. Please select one from the list below.” The issues included: 1) the education system and availability of youth programs, 2) the economic system including the poverty level and employment rate, 3) the size and charge of the police force, and 4) the correctional facilities including the methods by which convicted criminals are punished.

Experiment 5.

The materials and task in Experiment 5 were identical to those of Experiment 4 except, instead of presenting the metaphor frame at the beginning of the report, we presented the metaphor frame at the end of the report, as shown below. All other aspects of the design were identical to Experiment 4. The paragraphs used were:

---

Five years ago Addison was in good shape, with no obvious vulnerabilities. Unfortunately, in the past five years the city's defense systems have weakened, and the city has succumbed to crime. Today, there are more than 55,000 criminal incidents a year - up by more than 10,000 per year. There is a worry that if the city does not regain its strength soon, even more serious problems may start to develop. Crime is a {beast/virus} ravaging the city of Addison.

Design

In Experiment 1 the survey was included in a larger packet of questionnaires that were unrelated to this study.

In Experiments 2–5, each step of the experiment was presented on a separate screen. That is, the initial crime report was presented on a screen by itself. After participants read the report and clicked a button indicating they had finished reading it, the report disappeared and the first follow-up question appeared on a screen by itself. Similarly, each subsequent question was shown on a separate screen. On the final screen, participants were asked several background questions (e.g., What is the first language you learned to speak?).

Participants in Experiments 2–5 were explicitly instructed not to use the “back” button on their browser. If they did use the “back” button, the experimental session was terminated. This ensured that participants did not reread the crime report when they were later asked questions about it.

Results

Experiment 1

In Experiment 1, we explored whether framing a crime problem with one of two contrasting metaphors for crime could systematically influence how people reasoned about the problem. Participants were presented with one of two versions of the crime paragraph (as detailed above) and asked a set of free response follow-up questions. Of particular interest, participants were asked how they would recommend solving Addison's crime problem.

Coding.

Proposed solutions to the crime problem in Addison were coded into two categories in line with the results of the norming study described in the introduction: 1) diagnose/treat/inoculate, and 2) capture/enforce/punish. Responses were categorized as “diagnose/treat/inoculate” if they suggested investigating the underlying cause of the problem (e.g., “look for the root cause”) or suggested a particular social reform to treat or inoculate the community (e.g., fix the economy, improve education, provide healthcare). Responses were categorized as “capture/enforce/punish” if they focused on the police force or other methods of law enforcement (e.g., calling in the National Guard) or modifying the criminal justice system (e.g., instituting harsher penalties, building more jails). For brevity, we will refer to the “diagnose/treat/inoculate” category as “reform” and the “capture/enforce/punish” category as “enforce.”

Each participant's response was weighted equally – as a single point towards the analysis. For solutions that solely emphasized either reform or enforcement, the respective category was incremented by a point. Responses that exclusively emphasized one approach were the majority. Occasionally, however, participants listed both types of suggestions. In this case, if the response listed a disproportionate number of suggestions that were consistent with one approach (e.g., if the response listed three suggestions in line with reform and only one in line with enforcement, as in “investigate the root cause, institute new educational programs, create jobs, and hire more police”) then it was coded as a full point for the corresponding category. However, if the response equally emphasized both approaches, then the point was split between the categories such that each was incremented by .5.

Thirty of the 485 responses (6%) did not fit into either category. In every case this was because the response lacked a suggestion (e.g., “I don't know”, “I need more information”, “It should be addressed”). These data were omitted from analysis.

Participants' crime reducing suggestions were coded blindly by two coders. Cohen's kappa – a measure of inter-rater reliability – was .75 indicating good agreement between the coders (p<.001). All disagreements between the coders were resolved between them before analyzing the data.

Results.

Overall, participants were more likely to emphasize enforcement strategies (65%) than reform (35%), χ2 = 41.85, p<.001. However, as predicted, the solutions participants proposed to the crime problem in Addison differed systematically as a function of the metaphorical frame encountered in the crime report (see Fig. 1). Participants given the crime-as-beast metaphorical framing were more likely to suggest enforcement (74%) than participants given the crime-as-virus framing (56%), χ2 = 13.94, p<.001. See Table 1 for response frequencies.

Download:

---

• PPT
  PowerPoint slide

---

• PNG
  larger image (33KB)

---

• TIFF
  original image (66KB)

---

Figure 1. Proportion of proposed solutions to crime by metaphor frame.

doi:10.1371/journal.pone.0016782.g001

Interestingly, when asked to identify the most influential aspect of the report, most participants ignored the metaphor. Only 15 participants (3%) identified the metaphoric frame as influential to their problem solving strategy. Removing these participants from the analysis did not affect the results (the proportion of responses that were congruent with the metaphor was not different in the two analyses, χ2 = .0001, p = .991). The vast majority of the participants identified the statistics in the crime report as being most influential in their decision – namely, the final three sentences of the paragraph that state the increasing crime and murder rate.

Download:

---

• PPT
  PowerPoint slide

---

• PNG
  larger image (20KB)

---

• TIFF
  original image (104KB)

---

Table 1. Response frequencies for each of the five experiments by condition and response category.

doi:10.1371/journal.pone.0016782.t001

Discussion.

In this experiment, we found that crime-reducing suggestions differed systematically as a function of the metaphor used to frame the crime problem. Participants who read that crime was a virus were more likely to propose treating the crime problem by investigating the root causes of the issue and instituting social reforms than participants who read that crime was a beast. Participants who read that crime was a beast were more likely to propose fighting back against the crime problem by hiring police officers and building jails – to catch and cage the criminals – than participants who read that crime was a virus.

Further, despite the clear influence of the metaphor, we found that participants generally identified the crime statistics, which were the same for both groups, and not the metaphor, as the most influential aspect of the report. These findings suggest that metaphors can influence how people conceptualize and in turn approach solving an important social issue, even if people don't explicitly perceive the metaphor as being especially influential.

Experiment 2

Experiment 3

Experiment 4

Experiment 5

Discussion