tox 5

How newspapers represent environmental risk: the case of carcinogenic hazards in South Korea Thomas Hove, Hye-Jin Paek *, Moonyoung Yun and Bokyung Jwa Department of Advertising & Public Relations, Hanyang University, Ansan, South Korea (Received 23 April 2013;ﬁnal version received 7 May 2014) This study pursues the following aims: to examine how news stories use frames, emotions, and uncertainty to present environmental risk information; to identify which aspects of risk issues they highlight; and to analyze how these stories’ representations of risk and uncertainty might differ according to the sources they use. Content analysis of 641 news stories in South Korea over the last decade yields threeﬁndings: (1) reassurance was the most frequently used news frame, while uncertainty and emotion were used less often than expected; (2) news sto- ries using government/industry/experts as sources vs. activists/lay people high- lighted different news frames and risk information; and (3) the two most frequently used uncertainty presentation formats were single point estimate and verbal estimate. This study contributes to existing literature on the roles of media in environmental risk communication in two ways. First, it examines the speciﬁc formats journalists use to present uncertainty about risks. Second, it integrates news frames with the emotional characteristics of risk communication and with differences in risk information characteristics according to source. Implications are discussed regarding how a better understanding of news representations of risk could inform and enhance cooperation between experts and journalists, and lead to more effective environmental risk communication. Finally, this content analysis provides a stepping stone for future research that could further investi- gate and test how publics respond to risk messages that have varying permuta- tions of emotional content and risk presentation formats.

Keywords:news framing; uncertainty; environmental risk; emotion; uncertainty presentation format Introduction People today face a variety of environmental risks such as pollution, global warm- ing, radiation exposure, and carcinogenic hazards. But because we usually do not notice or directly experience many of these risks, we rely on the media as our pri- mary source of information about them (Allan 2002 ). To provide this information, news media typically feature experts, ofﬁcials, and professionals (Fischhoff 1995 ; Kitzinger and Reilly 1997 ). However, when journalists cover a risk issue, there are often discrepancies between their own and experts’judgments about what qualiﬁes as important information (Dunwoody and Peters 1992 ; Peters 1995 ).

Previous studies have assessed the degree to which the news media succeed or fail in presenting high-quality risk information (Dudo, Dahlstrom, and Brossard *Corresponding author. Email: [email protected] © 2014 Taylor & Francis Journal of Risk Research, 2015 Vol. 18, No. 10, 1320–1336, http://dx.doi.org/10.1080/13669877.2014.923025 2007 ; Roche and Muskavitch 2003 ). Key assumptions in these studies are that news stories should not only provide accurate and thorough information but also avoid inappropriate sensationalism and undue emphasis on emotional content. Extending this research, the current study pursues three objectives: to examine how news media present environmental risk information; to identify which aspects of risk issues they highlight; and to analyze how journalists’representations of risk might differ according to the sources they use. These three objectives are interrelated because the characteristics of news coverage about a risk issue and the sources used in that coverage could affect the way people perceive and react to the risk at hand (Dudo, Dahlstrom, and Brossard 2007 ; Einsiedel and Coughlan 1993 ; Powell et al. 2007 ; Sandman 1997 ; Viswanath et al. 2008 ; Wahlberg and Sjoberg 2000 ).

This study is guided by theoretical and empirical arguments on media framing, uncertainty, and the media’s role in covering environmental risks. Previous studies have analyzed news frames and differences in framing with respect to which sources journalists use (e.g. Logan, Park, and Shin 2004 ; Oh et al. 2012 ; Viswanath et al. 2008 ; for overview, see Tewksbury and Scheufele 2009 ). In risk communication research, another concept that has been thoroughly discussed and investigated is uncertainty, particularly regarding its role as a key determinant of people’s risk per- ception (e.g. Powell et al. 2007 ; Slovic 1987 ). The way uncertainty is presented has been found to affect the level of uncertainty that people feel about a risk and how they subsequently respond to it, for example by acknowledging it or rejecting it (Johnson and Slovic 1995 ; Kuhn 2000 ; Wardekker et al. 2008 ).

We examine these characteristics of risk framing and uncertainty in Korean newspaper coverage of carcinogenic hazards over the course of ten years, 2003–2012. Recently, scientists, news media, and the general public in South Korea have paid increasingly serious attention to cancers that might be caused by environ- mental hazards such as mobile phone use, diesel engine exhaust fumes, household chemicals, building materials, grilled charcoal, and food products (Lee 2010 ;No 2011 ). Carcinogenic hazards are any substances that can cause cancer. Some well- known examples are asbestos, radiation, and various types of toxins. Cancer is the number one cause of death in Korea, killing 142.8 people per 100,000 (Statistics Korea 2012 ).

This study contributes to existing literature on the roles of media in environmen- tal risk communication in two ways. First, it goes beyond the customary practice of cataloging news frames by also examining the speciﬁc formats journalists use to present uncertainty about risks. Second, it integrates news frames with the emotional characteristics of risk communication and with differences in risk information char- acteristics according to source. Implications are discussed regarding how a better understanding of these media representations of risk could inform and enhance cooperation between experts and journalists, and lead to more effective environmen- tal risk communication.

News frames and environmental risks Decades of research have documented the various ways news media inform the pub- lic about risk issues and information (e.g. Singer 1990 ; Slovic 2000 ; Tankard and Ryan 1974 ). But often, journalists present risk information in ways that diverge from the recommendations of scientists and risk communication experts ( for overview, see Kitzinger 1999 ; Peters 1995 ). One common explanation for this divergence is Journal of Risk Research1321 that journalists tend to focus more on risk issues’newsworthiness than they do on conveying accurate information about them, and that this focus might end up distort- ing or even blocking appropriate public awareness and perceptions of risk (Grifﬁth, Mathias, and Price 1994 ; Miles, Braxton, and Frewer 1999 ). For example, news media may focus more on risks that pose small threats to public safety, such as air- plane accidents, than they do on risks that pose much greater threats to public safety, such as smoking and tobacco use (Greenberg et al. 1989a ,1989b ; Sandman et al. 1987 ).

With regards to how media communicate environmental risks, a key issue has been the degree to which they present the risk with appropriately accurate, thorough, and contextualized information. Some studies, however, have noted that this aim to present the technical aspects of risk issues can be overshadowed by the effort to present the emotional and dramatic dimensions of risks and how people respond to them (Dudo, Dahlstrom, and Brossard 2007 ; Nisbet and Huge 2006 ). For example, when covering risk issues such as epidemics, journalists may highlight people’s feel- ings of doubt, threat, and fear rather than actual scientiﬁc information about the issue. Such tendencies can generate spurious phenomena such as‘media pandemics’ (Gainor and Menefee 2006 ). Journalists might also focus more on human interest topics, exaggerate the magnitude of a risk by focusing on worst-case scenarios, and use sensationalistic and emotionally charged language to describe risks (Dudo, Dahlstrom, and Brossard 2007 ; Shih, Wijaya, and Brossard 2008 ).

Such concerns lie behind previous studies on how the news media succeed or fail at providing high-quality risk information. Several of these studies have addressed this issue by analyzing the speciﬁc frames that have been used for pre- senting that information. A widely used deﬁnition of framing describes it as a pro- cess of‘selecting and highlighting some facets of events or issues, and making connections among them so as to promote a particular interpretation, evaluation, and/or solution’(Entman 2004 , 5; also see Entman 1993 ). Framing may occur either in people’s thoughts or in acts of communication.Media frames(which we here call news frames) occur in communication, and they are deﬁned as‘the words, images, phrases, and presentation styles that a speaker (e.g. a politician, a media outlet) uses when relaying information about an issue or event to an audience’(Chong and Druckman 2007 , 100). News frames have been studied for their potential to inﬂu- ence audience perception of events and social issues by associating them with net- works of information and meaning.

Research has documented various complex relations between the way the public perceives risk issues and the way the mass media frame them (Wahlberg and Sjoberg 2000 ). Many types of media frames have been deﬁned, empirically studied, and evaluated for their appropriateness and effectiveness in presenting risk issues.

Semetko and Valkenburg ( 2000 ) identiﬁedﬁve common news frames that have been studied and used across various types of US and European media coverage: attribu- tion of responsibility, conﬂict, human interest, economic consequences, and morality.

For science topics such as stem cell research and biotechnology, studies have found that frames with dramatic potential, such as strategy/conﬂict, uncertainty, and ethics/ morality, were the most prominently featured (Nisbet, Brossard, and Kroepsch 2003 ; Nisbet and Huge 2006 ). Other studies have found that news stories on risk tend to emphasize who is responsible for causing or solving the risk, what actions people can take to deal with it, and what information should make them feel reassured (An and Gower 2009 ; Oh et al. 2012 ; Shih, Wijaya, and Brossard 2008 ). 1322T. Hoveet al. For the context of carcinogenic hazards, the current study uses the following news frames from Semetko and Valkenburg ( 2000 ): human interest, conﬂict, attribu- tion of responsibility, and consequences in a broader sense than merely economic.

Their ethics/morality frame was not used because, for an environmental risk issue, it was deemed to be contained within the conﬂict frame. However, following previous studies of risk issue coverage, we added the frames of risk magnitude, reassurance, sensationalism, and uncertainty (Dudo, Dahlstrom, and Brossard 2007 ; Oh et al. 2012 ; Shih, Wijaya, and Brossard 2008 ).

In addition to using these news frames, journalists tend to highlight the emo- tional aspects of risk issues (Powell et al. 2007 ; Sandman 1997 ; Sandman et al. 1993 ). For example, a content analysis of the environmental risk news stories that were chosen as best articles by newspaper editors found that 68% of the stories did not include any risk information but instead focused more on conﬂicts and emotion- laden opinions (Sandman et al. 1987 ). Emotions felt by the public with respect to risk issues typically include dread, worry, anger, distrust, and distress (Sandman et al. 1993 ). Because of the important role that emotion conveyed by media has been found to play in public perception of risk in general, it is worthwhile to examine the degree to which emotion is present in the media coverage of carcinogenic hazards in particular.

Based on the above rationales and approaches from previous studies, two research questions are proposed. RQ1. What are the type and frequency of news frames presented in the media cov- erage of carcinogenic hazards?

RQ2. What are the type and frequency of emotional content presented in the news coverage of carcinogenic hazards? Source opinions and differences of emotional content according to source Journalists’use of speciﬁc news frames may differ depending on the sources they use for risk information. For example, they may seek out sources who are likely to have strong opinions, particularly those who are likely to disagree with other types of sources (Sandman 1997 ). Or they may try to provide balanced coverage when theyﬁnd it difﬁcult or impossible to resolve controversies about how the issue should be presented ( for overview, see Kitzinger 1999 ). Within this general practice of balancing out opinions, though, one of the speciﬁc patterns found in previous research is that journalists use government, industry, and expert sources to represent the‘safe’side of risk debates, while they use activists and lay people to represent the‘risky’side (Sandman 1997 ).

Among the sources frequently cited in risk coverage are government ofﬁcials, industry spokespeople, experts, citizens, and advocacy group spokespeople. For sto- ries about environmental risk, the type most frequently relied on is government ofﬁ- cials, particularly when stories cite only a single source. This tendency was conﬁrmed in content analyses of newspapers and network television (Sandman et al. 1993 ). Along with industry sources, government sources are most likely to be used to represent the safe and reassuring perspective on the risk issue. One likely reason is that government and industry stakeholders expect the public to view them as somehow responsible for either causing or controlling the risk, and so they would want to minimize any perceptions of danger. By contrast, activist and lay people Journal of Risk Research1323 sources are likely to be used to represent the alarming perspective so that the people who are responsible for controlling the risk will take action. Activist/lay people sources are therefore likely to emphasize the hazards that the general public faces by highlighting personal experiences and feelings, conﬂicts or dissenting views, and the magnitude of the risks. Theseﬁndings and rationales lead us to formulate the fol- lowing seven-part hypothesis about the use of sources in stories about carcinogenic hazards: H1. Stories from the government/industry/expert source will present reassurance news frames more frequently than those from the activists/lay people source (H1-1), while stories from the activists/lay people source will present human interest (H1-2), conﬂict (H1-3), risk magnitude (H1-4), sensationalism (H1-5), and uncertainty (H1-6) news frames, as well as emotion (H1-7), more fre- quently than the stories from the government/industry/expert source. Uncertainty presentation format in news about risk issues Studies show that news stories about environmental risks often feature the theme of uncertainty (Powell et al. 2007 ). Even when experts communicate about environ- mental risks, uncertainty is inevitable. To deal with it, journalists have developed the strategy of framing risks as scientiﬁc controversies or as phenomena that arouse worry, fear, and anger (Sandman 1997 ; Wardekker et al. 2008 ). For instance, consid- erable uncertainty may exist about such questions as how many people have been exposed to a risk, how severe that exposure was for different people, and how sensi- tive the risk issue is with respect to individual and cultural differences. Even more fundamentally, there may be underlying scientiﬁc uncertainty about whether or not something counts as a signiﬁcant risk (Kuhn 2000 ). It has been observed that the hazards that generate the biggest conﬂict are often those associated with the highest degree of uncertainty, especially when people debate about environmental risk issues such as global warming, radioactive exposure, and carcinogenic hazards. For these types of hazards, experts seem to agree that these questions cannot be answered with absolute certainty (Sellnow et al. 2008 ).

One tendency in risk communication research is to treat uncertainty as a reﬂec- tion of lay people’s probabilistic judgments about a risk event or outcome happening (Powell et al. 2007 ). Previous studies on the uncertainties of risk often focus on lay people’s judgments about how likely a risk event or outcome will occur (Frewer et al. 2003 ; Kahneman and Tversky 1982a ,1982b ). Taking a different approach cen- tered more on communication, the current study focuses on different message for- mats for presenting risk, which are hereafter referred to asuncertainty presentation formats. Studies of such formats should accompany efforts to analyze how people understand and react to the risks they learn about in the media (McCarthy et al. 2008 ; also see Kuhn 2000 ). These formats are important message features that are likely to inﬂuence people’s risk perceptions.

Uncertainty presentation formats can be divided into two basic types, verbal risk estimates and numerical risk estimates (Wardekker et al. 2008 ). Verbal estimates convey risk information through words such as‘likely,’‘unlikely,’‘serious,’‘triv- ial,’‘severe,’‘probable,’and‘improbable’(Wallsten et al. 1986 ). Numerical risk estimates convey information through numbers that represent probabilities, which can be done in several ways (Kuhn 2000 ). Theﬁrst way is called a‘single best estimate,’which is aﬁxed scale that provides the‘best’point estimate without any 1324T. Hoveet al. uncertainty information. The second is called a‘vague point estimate,’which includes a verbal qualiﬁer with the base point estimate. For example, the statement ‘it is likely to have a 10% of leakage occurring over the next 100 years’includes the verbal probability qualiﬁer‘likely’. The third way of representing probabilities is called a‘range estimate,’which provides a range of numbers centered on the base probability value. Examples would be a conﬁdence interval bounding an estimate, or a distribution of estimates, each of which has an associated likelihood that it is the correct number (Kuhn 2000 ; Wardekker et al. 2008 ).

Although researchers have made precise distinctions among uncertainty pre- sentation formats, there is no consensus about how people respond to the differ- ent types and which types make the uncertainties of risk easier to understand.

There are two basic schools of thought, one favoring the verbal and the other the numeric.

Verbal formats can be advantageous when words are used consistently to con- vey speciﬁc meanings. Verbal format advocates argue that words can improve people’s ability to read and remember risk information (Wardekker et al. 2008 ).

However, verbal format detractors argue that representing probabilities with words alone leads to loss of precision, and that words often have different meanings for different people (Wallsten et al. 1986 ). The rejoinder, though, is that words’ potential vagueness, ambiguity, and polysemy (the ability to have multiple mean- ings) can also be advantageous. One positive purpose such characteristics might serve is to correct experts’tendency to be overly precise when they communicate predictions (Slovic et al. 1981 ). When even experts are uncertain about risk events and outcomes, they might avoid this mistake by using broad ranges and wordings to accurately reﬂect their limited knowledge. Another advantage is that when verbal communications are consistent, people are more likely to understand and remember them.

Advocates for presenting uncertainty numerically argue that numbers have a per- suasive advantage over words because they are precise, they allow for calculations, and they have aﬁxed rank order. For instance, 100% is always more than 75%. If a single point estimate is given, along with a brief verbal qualiﬁcation describing why the probability cannot be established with certainty, the association between environ- mental concern and perceived risk is expected to be small. Critics, however, argue that many people are fundamentally unfamiliar with quantitative methods. As a result, the general public might not be able to understand numerically represented risk estimates, and they might focus on the wrong aspects of them (Konheim 1988 ; Viscusi, Magat, and Huber 1991 ).

These considerations suggest that different types of uncertainty presentation for- mat may lead to different consequences regarding people’s perceptions of and responses to risk issues in media. Accordingly, an important task that should precede studies of the effects of uncertainty presentation formats is to examine how fre- quently the types of format are used in news reports on risk. Thus, we ask the fol- lowing question. RQ3. What are the type and frequency of uncertainty presentation format presented in the media coverage of carcinogenic hazards? Journal of Risk Research1325 Method Sample Data consisted of news stories from 2003 to 2012 on carcinogenic hazards drawn from eleven daily South Korean newspapers:Chosun, Dong-A, Joongang, Hankyoreh, Segye, Hankook, Seoul, Kyunghyang, Munwha, Naeil,andKukmin.

These stories were searched online through Naver, the most widely used search engine in Korea (Seol 2011 ). When the term‘carcinogen’(bal-am-mul-jilin Korean) was used to search all news content in Naver across several types of media (daily, specialty, national, and local newspapers; Internet news sites; magazines; and TV), 22,898 cases were found. These cases included various health and risk topics related to cancer (e.g. smoking, drinking, diet, and genetic issues).

To attain the cases for our main inquiry, we used rigorous sampling criteria.

First, we chose only major print news that play a key role in setting the agenda for other types of media and public communication. Second, we did not combine print news and television news because they have different reporting formats, particularly with respect to story length and use of visuals. Third, we selected only news stories that used the key termbal-am-mul-jilin their headlines. This strategy would be more likely to yield stories about environmental hazards and to exclude stories about personal diseases or health problems. The reasons are thatbal-am-mul-jilspeciﬁcally connotes substances in the environment that can cause cancer, and that it is the only term in Korean that is equivalent to‘carcinogen’in English (No 2011 ). Fourth, we chose this generic term over those that refer to more speciﬁc carcinogenic agents such as asbestos and radon because it is very common and familiar to South Kore- ans (Jwa, Yun, and Paek 2013 ; Lee 2010 ), and because we wanted toﬁnd a global pattern for news frames and risk information across a variety of carcinogenic agents.

This rigorous screening process yielded 653 cases. After irrelevant and duplicate news stories were removed, theﬁnal sample size for analysis was 641: 2003 = 24, 2004 = 29, 2005 = 108, 2006 = 34, 2007 = 59, 2008 = 57, 2009 = 84, 2010 = 53, 2011 = 107, 2012 = 86. The news stories appeared in various newspaper sections such as politics (N= 24), society (N= 356), economy (N= 46), health (N= 17), inter- national (N= 37), local (N= 17), culture (N= 28), editorial (N= 22), and information (N= 3). The word count per news story ranged from 63 to 640, with a mean of 174.29 (SD = 84.01).

The stories covered a total of 141 types of carcinogenic agents. Among these, the most frequently found were the following in descending order: formaldehyde (2.5%,N= 80, mainly found in consumer goods), benzene (2.0%,N= 63, mainly found in factories), benzopyrene (1.7%,N= 55, mainly found in food, especially edible oil ), and asbestos (1.3%,N= 42, mainly found in public facilities). Notably, one of the substances most frequently reported as a carcinogenic agent was mala- chite Green (2.3%,N= 74, mainly found in food, especiallyﬁsh ). But according to the International Agency for Research on Cancer, malachite Green is not a carcino- genic agent.

The locations of the carcinogenic agents also varied: food products such as soy sauce, cola, and cooking oil (36.5%,N= 234); consumer goods such as outdoor clothing, chopsticks, and charcoal (31%,N= 199); public facilities such as train stations, government ofﬁces, and elementary schools (10.1%,N= 65); and water supplies such as ground water and tap water (8.1%,N= 51). 1326T. Hoveet al. Coding scheme This study mainly focused on the following characteristics presented in carcinogenic hazard news stories: (1) news frames, (2) types of emotional content, (3) uncertainty presentation formats, and (4) news sources. Based on several studies on the framing of environmental risk issues, we examined the following news frames: human inter- est, conﬂict, attribution of responsibility, economic consequences, risk magnitude, reassurance, uncertainty, and sensationalism (Dudo, Dahlstrom, and Brossard 2007 ; Semetko and Valkenburg 2000 ; Shih, Wijaya, and Brossard 2008 ). The concept of emotional contentis based on the topics covered in Sandman’s study category of‘out- rage’(Sandman 1997 ; Sandman et al. 1993 ), which we deemed to be too narrow a label. Emotional content was coded as present or absent and then categorized further into various types of emotions such as fear, blame, worry, dread, and uneasiness. The uncertainty presentation formats drawn from risk communication literature included single point estimate, vague point estimate, range point estimate, and verbal estimate (e.g. Kuhn 2000 ; Wallsten et al. 1986 ; Wardekker et al. 2008 ). Lastly, news source was coded as exhaustively as possible and then merged into theﬁnal categories of government (government ofﬁcials, regulators), industry (corporate spokespeople), experts, activists (consumer interest groups/NGOs, labor union representatives, politicians, and individual activists), lay people, media organizations, foreign source, multiple sources, and other. Operational deﬁnitions and response options for these categories are provided in Table 1.

Coding procedure Two coders, each of whom was blind to the research questions, independently coded the sample news stories following the standard procedure in mass media studies (Wimmer and Dominick 2005 ). This procedure includes the following steps in chro- nological order: training, pilot coding (187 stories that were not part of the main coding), modifying and reﬁning the coding scheme, main coding, calculating inter- coder reliability, and resolving disagreements through discussion.

Out of 641 news stories, the two coders coded about 20% (N= 129) of the same samples to calculate intercoder reliability, and each coder coded half of the remain- ing samples. This quantity exceeds the rule of thumb that about 10% of the total sample size should be coded to compute intercoder reliability (Wimmer and Dominick 2005 ). Intercoder reliability was computed using Krippendorff’s alpha (Krippendorff 2004 ). We achieved a 0.72 average for intercoder reliability (see Table 1 for reliabilities for all the coding categories), which is acceptable (Krippendorff 2004 ; Lombard, Snyder-Duch, and Bracken 2002 ). All disagreements were resolved through discussions between the two coders.

Results RQ1: type and frequency of news frames Research Question 1 asked the type and frequency of news frames presented in the coverage of carcinogenic hazards. As shown in Table 2, descriptive statistics indicate that 14.5% of the carcinogenic hazard news stories presented the human interest frame and 16.2% presented the conﬂict frame. In addition, 46% presented responsi- bilities attributed to several entities, speciﬁcally to industry (27.3%), government (9.5%), individuals (5.1%), society (1.7%), and multiple responsibilities (1.1%). In Journal of Risk Research1327 Table 1. Operational deﬁnitions and intercoder reliability.

VariablesKrippendorff’s Alpha News frames Human interest(1–Yes; 0–No): The story tells how carcinogen affects one or more persons.0.74 Conﬂict(1–Yes; 0–No): The story emphasizes conﬂict between individuals or groups.0.69 Attribution of responsibility(1–Personal responsibility; 2–Corporate responsibility; 3–Governmental responsibility; 4–Social responsibility; 5–Plural responsibility; 9–Other; 99–None): The story presents an issue or problem and attributes responsibility for a cause or solution to either government, organization, group, or individual.0.75 Consequences(1–Yes; 0–No): The story reports any effects of carcinogen on an individual, society, and/or the economy.0.71 Risk magnitude(1–Yes; 0–No): The story mentions the level or size of a risk effect, for example likelihoods or rates of injury, illness, mortality, or other risk-related consequences.0.85 Reassurance(1–Yes; 0–No): The story addresses public fear about carcinogen by reporting on efforts to reduce or eradicate it, for example product recalls or progress in repair or treatment efforts.0.69 Uncertainty(1–Yes; 0–No): The story emphasizes and exaggerates uncertainties surrounding carcinogen.0.65 Sensationalism(1–Yes; 0–No): The story emphasizes worst case scenario or includes emotionally loaded words and/or phrases0.66 Presence of emotion (1–Anger; 2–Worry; 3–Fear; 4–Ignoring; 5 –Dissatisfaction; 6–Distrust; 7–Surprise; 9–Other; 10–Distress; 99–None): The story refers to speciﬁc emotions and feelings0.65 Types of uncertainty presentation format Single point estimate(1–Yes; 0–No): A single‘best’point estimate, no uncertainty information (e.g.‘Coke made in Korea had 96 micrograms (μg) of the chemical 4 methylimidizole (4-MI), far exceeding amounts in the soda produced in China at 56 ug, and Japan, 72μg’;‘According to the study released Monday, benzene was found in 0.00038 and 0.00990 ppm in two different chip-making lines–both below the permitted 1 ppm.’)0.70 Vague point estimate(1–Yes; 0–No): A verbal qualiﬁer with the base point estimate (e.g.‘About 1.1 particles per million ( ppm) of dioxane and 210 ppm of formaldehyde were detected in the shampoo and 3.0 ppm of dioxane was detected in the baby wash’;‘‘‘In a safety and quality test we conducted on 12 outdoor jackets from nine brands, we found that the inner lining of Kolon’s Active brand contained high traces of carcinogenic arylamine at 595 mg/kg,”said a consumer rights agency. This is roughly 20 times the legal limit.’)0.71 Verbal estimate(1–Yes; 0–No):Verbal expression without numerical probability‘Sesame oil, cooking oil and red pepper oil were found to contain carcinogens exceeding the government-set standard’;‘Kolon Industries, Korea’s foremost manufacturer of textile and chemical materials, is found to have sold outdoor jackets containing an excessive amount of a cancer-causing substance.’)0.74 Range estimate(1– Yes; 0–No):A range centered on the base probability value (e.g.‘The food regulator said the detected amount of benzene ranged from 5.7 to 87.8 ppb ( parts per billion). However, the regulator said the benzene level is not harmful to humans.’)0.83 (Continued) 1328T. Hoveet al. addition, about 33.2% of the news stories presented the consequence frame, which highlights effects of carcinogens on individuals, society, and the economy. More than a half of the stories presented some sort of reassurance (54.4%), such as food recall, repair, and treatment. Lastly, some stories were framed with sensationalism (12.9%) and a relatively small number presented uncertainty (6.7%) and the magnitude of risk (5.1%).

RQ2: frequency and type of emotion Research Question 2 asked the type and frequency of emotion presented in the media coverage of carcinogenic hazards. Only about 6.7% of all the story samples presented types of emotion, which include anger (1.9%), worry (2.3%), fear, sur- prise, uneasiness, and distrust. For example, people‘worry’that tear gas contains carcinogens; people express‘anger’when they learn that carcinogens were found in the importedﬁsh they eat; or a mother‘fears’for her child having been born in an apartment where a carcinogen was found.

H1: differences of news frames and emotional content by source Hypothesis 1 predicted that news stories using the government/industry/expert source would present reassurance news frames more frequently than those using the activists/lay people source (H1-1), while stories using the activists/lay people source would present human interest (H1-2), conﬂict (H1-3), risk magnitude (H1-4), sensa- tionalism (H1-5), and uncertainty (H1-6) news frames, as well as emotion (H1-7), more frequently than the stories using the government/industry/expert source.

First, the news stories about carcinogenic risks included a variety of news sources, including government ofﬁcials (48.2%), experts or expert institutions (10.6%), NGOs or consumer interest groups (8.7%), politicians (5.2%), labor unions (0.9%), lay people (0.6%), and media organizations including social networking sites and foreign and domestic media (10.8%). To test Hypothesis 1, various types of sources were merged to see the difference between,ﬁrst, government/industry/ expert sources and activists/lay people sources (a merger of NGOs or consumer interest groups and labor unions). After cases were removed that could not be cate- gorized into government, industry, activists, or lay people, descriptive statistics show that a total of 506 cases from the entire sample were dichotomized into government/ industry/expert sources (79.8%;N= 404) and activists/lay people (20.1%;N= 102).

Then, cross-tabulation analysis was performed between news frames, presence of emotion ( yes or no), and the dichotomized source. As shown in Table 3, results generally show the predicted patterns. Chi-square statistics indicate that the Table 1. (Continued).

VariablesKrippendorff’s Alpha News sources 1–Government; 2–Expert; 3–Activists; 4–Lay people; 5–Industry; 6–Media organization; 7–Foreign; 8–Politician; 9–Multiple; 10–Other; 99–None.0.76 Journal of Risk Research1329 stories from the government/industry/expert source presented the reassurance news frame signiﬁcantly more than those from the activists/lay people source (59.7% vs. 40.2%; Chi-square (1) = 12.50,p< 0.001). By contrast, the stories from the activists/lay people source presented more frequently than those from the govern- ment/industry/expert source the following: conﬂict frame (21.6% vs. 12.4%; Chi-square (1) = 5.64,p< 0.05), sensationalism frame (16.7% vs. 9.9%; Chi-square (1) = 3.73,p= 0.05), and emotion (10.8% vs. 4.0%; Chi-square (1) = 7.51,p< 0.01). The presences of human interest and risk magnitude frames Table 2. Descriptive statistics of the key categories.

Categories % of presence (N) News Frames Human interest 14.5(93) Conﬂict 16.2(104) Attribution of responsibility Personal responsibility 5.1 (33) Corporate responsibility 27.3 (175) Governmental responsibility 9.5 (61) Social responsibility 1.7 (11) Plural responsibility 1.1 (7) Other 1.2 (8) None 54.0 (346) Consequence 33.2 (213) Risk magnitude 5.1 (33) Reassurance 54.4 (349) Sensationalism 12.9 (83) Uncertainty 6.7 (43) Presence of emotion Anger 1.9 (12) Worry 2.3 (15) Fear 0.5 (3) Dissatisfaction 0.8 (5) Distrust 0.2 (1) Surprise 0.6 (4) Distress 0.2 (1) Other 0.2 (1) None 93.4 (599) Types of uncertainty presentation format Single point estimate 39.6 (254) Vague point estimate 4.1 (26) Verbal estimate 20.6(132) Range estimate 14.8 (95) News source Government 48.2 (309) Experts 10.6(68) NGO 8.7 (56) Labor union 0.9 (6) Lay people 0.6(4) Industry 3.6(23) Media organization 10.8(69) Foreign source 4.2(27) Politician 5.2(36) None 6.1(39) 1330T. Hoveet al. were also more frequently–that is, in the predicted direction–presented in the news stories using the activists/lay people source than those using the govern- ment/industry/expert source; however, for those two frames statistical signiﬁcance did not reach the level of 0.05. In sum, the results support H1-1, H1-3, H1-5, and H1-7, but not H1-2, H1-4, and H1-6.

RQ3: frequency and type of uncertainty presentation format Research Question 3 asked about the frequency and type of uncertainty presentation formats. Descriptive statistics shown in Table 1indicate that, among the four types of uncertainty presentation format, the most prevalent was single point estimate (39.6%), followed by verbal estimate (20.6%), range estimate (14.8%), and vague point estimate (4.1%).

Discussion In the context of government and other institutional efforts to make carcinogenic hazards a publicly salient environmental risk issue in South Korea, this study exam- ined how Korean newspapers use speciﬁc frames to cover those hazards. Speciﬁ- cally, it focused on the extent to which news coverage presents the risk communication characteristics of framing, emotions, and uncertainty. Keyﬁndings can be summarized as follows. First, reassurance was the most frequently used news frame, while the uncertainty frame and emotional content were used less often than existing research has suggested (e.g. Sandman et al. 1987 ). Second, as predicted, news stories using government/industry/expert vs. activists/lay people sources high- lighted noticeably different news frames, emotion, and risk information. Finally, the two most frequently used uncertainty presentation formats were single point estimate and verbal estimate.

Previous research shows that news media tend to focus more on human interest topics, exaggerate the magnitude of a risk by focusing on worst-case scenarios, and use sensationalistic and emotionally charged language to describe risks (Dudo, Dahlstrom, and Brossard 2007 ; Shih, Wijaya, and Brossard 2008 ). By contrast, we found that for carcinogenic hazards the reassurance frame was used much more Table 3. Chi-square tests for journalistic news frames and emotion by source.

Government/industry/ expert source (N= 404)Activists/lay people source (N= 102) χ 2(df )Fisher’s exact test %(n)%(n) News frames Reassurance 59.7 (241) 40.2 (41) 12.497 (1)*** 0.001 Human interest 12.6 (51) 15. 7 (16) 0.665 (1) 0.416 Conﬂict 12.4 (50) 21.6 (22) 5.639 (1)* 0.025 Risk magnitude 5.4 (22) 6.9 (7) 0.303 (1) 0.633 Sensationalism 9.9 (40) 16.7 (17) 3.729 (1)* 0.078 Uncertainty 5.7 (23) 3.9 (4) 0.506 (1) 0.625 Presence of Emotion 4.0 (16) 10.8 (11) 5.508 (1)*** 0.012 Note: Reported in each cell is the percent ofpresence.

*p< 0.05; **p< 0.01; ***p< 0.001. Journal of Risk Research1331 frequently than frames such as sensationalism, human interest, and conﬂict, which previous studies have found to be more commonly used in the context of H5N1ﬂu (Dudo, Dahlstrom, and Brossard 2007 ) and the SARS epidemic (Luther and Zhou 2005 ). Thisﬁnding also contrasts strikingly with that of Oh et al. ( 2012 ), who found that only one news story about H1N1ﬂu employed the reassurance frame. In another result contrasting with that same study, human interest and conﬂict frames appeared more frequently in our sample stories about carcinogenic hazards.

This apparent difference may derive from the fact that carcinogenic hazards and epidemics are two qualitatively different risk topics. Compared to epidemic diseases whose causes were not hotly contested (SARS, H5N1, H1N1), carcinogenic hazards are an environmental risk issue that can involve conﬂicting views between the stake- holder causing the risk and the stakeholder experiencing it. For example, if the issue is radon found in newly built apartments, construction companies would have views that differ from those of the people who live in the contaminated dwellings. Or if the issue is carcinogens found in foods, the companies who produced the foods would have views that differ from the consumers who buy and consume them. This potential for greater conﬂict between different stakeholder views may also be more likely to generate reassurance measures, such as food recalls, statements that assure safety, and promises of immediate policy reactions by industry and government. In any case, with the increased public attention to and concern about carcinogenic hazards found in everyday goods such as household chemicals, mobile phones, and diesel vehicles, public anxiety about them may be heightened, and the media may want to respond more properly by providing more reassuring information to relieve public uneasiness.

Very few news stories ( less than 7%) presented emotional content such as worry, anger, fear, and dissatisfaction. Thisﬁnding diverges from Sandman’s claim that jour- nalists tend to present risk issues more through the emotional frame of‘outrage’than the technical frame of‘hazard’. Although Sandman et al. ( 1987 ) made this claim after performing several content analyses of media coverage on environmental issues, one clear difference between his methods and ours is sampling procedure. His samples were selected from stories that editors judged to be the most important environmental news stories, while our sampling was based on stories that were actually available to audiences across several news outlets. There may be a difference between what edi- tors judge to be the most important stories and what stories are actually available to audiences. In addition, a content analysis of local newspaper stories and nationwide stories that are available online may also lead to different results regarding which top- ics should be more highlighted and which are more newsworthy.

By contrast, consistent with Sandman’s arguments are ourﬁndings on different uses of news frames and presence of emotion by source. Sandman argued that jour- nalists use government, industry, and expert sources to represent the‘safe’side of risk debates, while they use activists and lay people to represent the‘risky’side. We found that,ﬁrst, government is the most frequently cited source ( for a similarﬁnd- ing, see Oh et al. 2012 ). Also, news stories relying on government/industry/expert sources present more reassurance, while those relying on activists/lay people sources focus more on conﬂict, sensationalism, and emotion. This consistentﬁnding may suggest that there are common patterns of journalistic practice in presenting and communicating risks in the news media of various nations (also see Logan, Park, and Shin 2004 ). If such patterns could ultimately be conﬁrmed as consistent, policy-makers and risk communicators might be better able to predict which types 1332T. Hoveet al. of risk messages would be available to the public, and they could plan ahead by pre- paring tailored messages for journalists to relay to the public at the precise moments when they are needed.

Also, risk issue managers and communicators may consider speciﬁc types of risk communication strategies for different situations. Sandman et al. ( 1987 ) and Sandman ( 2003 ) propose several useful strategies for how to handle risk issues in the media. Depending on the magnitude of the hazard and the level of people’s emotion, risk issues can be dealt with, according to stakeholder relations, public rela- tions, outrage management, and crisis management. For example, when there is a low-hazard and low-outrage risk situation, issue managers may work with opinion leaders and/or various stakeholders to make the risk issue more salient through edu- cation and information. On the other hand, if there is a serious high-hazard risk but low outrage, the public audience might be apathetic and inattentive. In such cases, Sandman suggests a communication strategy that entails honing important messages and urging people to take the risk more seriously. Sometimes the news media portray a hazard more seriously than is warranted, as in the case of the avianﬂu‘media pan- demic,’which resulted in unnecessarily high public outrage (Gainor and Menefee 2006 ). For that type of case, Sandman suggests outrage management, which mainly focuses on‘calming down’the public through interpersonal communication or through listening, acknowledging, apologizing, sharing control, and building trust.

Lastly, a high-risk, high-outrage situation may be a crisis in which more strategic communication is needed to avoid over-reassurance, to share dilemmas, to convey empathy, to provide speciﬁc directives, and to acknowledge uncertainty.

With respect to uncertainty presentation format, ourﬁndings suggest that single point estimate and verbal estimate are the types most frequently used. As some researchers claim, verbal estimates may provide a higher level of uncertainty and different meanings to the public (Patt and Schrag 2003 ; Wallsten et al. 1986 ).

Accordingly, risk communicators should determine whether they need to heighten or reduce uncertainty and risk perception, and whether they should use different uncer- tainty presentation formats for their public messages or supply facts directly to jour- nalists. Because there are mixedﬁndings on whether numeric or verbal uncertainty presentation formats would positively or negatively affect risk perception, more studies of these formats’effects are needed.

Lastly, although the issue was not addressed in the research questions, some descriptiveﬁndings suggest that the news media provided somewhat inaccurate risk information for chemicals (e.g. Malachite Green) that are not in fact carcinogenic agents. Because lay people usually know little about the scientiﬁc and technical components of risk, journalists need to be particularly careful in providing the public with accurate and complete information. In turn, experts, ofﬁcials, and other risk communicators should be equally careful about informing journalists.

Some limitations should be noted. First, while carcinogenic hazards are receiving increasing attention as an environmental risk issue in Korea,ﬁndings based on one risk issue in one country may not be generalizable across other risk topics around the world. Promising paths for future research would be to examine media coverage for multiple environmental risk topics and to develop typologies of media formats and message characteristics for communicating risks. Second, the story samples chosen were ones with the word‘carcinogen’(bal-am-mul-jil) in their headlines.

However, some stories may have focused on speciﬁc types of carcinogens (e.g.

radon, benzopyrene, asbestos) without mentioning the general term, and our Journal of Risk Research1333 sampling procedure would have missed those. Third, the content analytic method itself is limited because it focuses only on the topics and terms that are present in media, and it cannot explain how they got there and how the audience would respond. In addition, some of the coding categories had lower-than-ideal intercoder reliabilities ( less than 0.70) and should thus be treated with caution. Nevertheless, a content analysis provides a precise overview of the types of risk information to which the public may be exposed. Accordingly, our study could be used as a step- ping stone for future research that might further investigate and test how publics respond to risk messages that have varying permutations of emotional content and risk presentation formats.

Acknowledgments This research was supported by the research fund of Hanyang University (HY-2013-N) given to the corresponding author.

References Allan, S. 2002.Media, Risk and Science. Philadelphia, PA: Open University Press.

An, S.-K., and K. K. Gower. 2009.“How Do the News Media Frame Crises? A Content Analysis of Crisis News Coverage.”Public Relations Review35: 107–112.

Chong, D., and J. N. Druckman. 2007.“A Theory of Framing and Opinion Formation in Competitive Elite Environments.”Journal of Communication57 (1): 99–118.

Dudo, A. D., M. F. Dahlstrom, and D. Brossard. 2007.“Reporting a Potential Pandemic: A Risk-related Assessment of Avian Inﬂuenza Coverage in US Newspapers.”Science Com- munication28 (4): 429–454.

Dunwoody, S., and H. P. Peters. 1992.“Mass Media Coverage of Technological and Environ- mental Risks: A Survey of Research in the United States and Germany.”Public Under- standing of Science1 (2): 199–230.

Einsiedel, E., and E. Coughlan. 1993.“The Canadian Press and the Environment: Recon- structing a Social Reality.”InThe Mass Media and Environmental Issues, edited by A.

Hansen, 134–149. Leicester: Leicester University Press.

Entman, R. M. 1993.“Framing: Toward Clariﬁcation of a Fractured Paradigm.”Journal of Communication43 (4): 51–58.

Entman, R. M. 2004.Projections of Power: Framing News, Public Opinion, and US Foreign Policy. Chicago, IL: University of Chicago Press.

Fischhoff, B. 1995.“Risk Perception and Communication Unplugged: Twenty Years of Process1.”Risk Analysis15 (2): 137–145.

Frewer, L. J., S. Hunt, M. Brennan, S. Kuznesof, M. Ness, and C. Ritson. 2003.“The Views of Scientiﬁc Experts on How the Public Conceptualize Uncertainty.”Journal of Risk Research6 (1): 75–85.

Gainor, D., and A. Menefee. 2006.“Avian Flu: A Media Pandemic.”Accessed October 19, 2007. http://www.businessandmedia.org/news/2006/news20060308.asp Greenberg, M. R., D. B. Sachsman, P. M. Sandman, and K. L. Salomone. 1989a.“Network Evening News Coverage of Environmental Risk.”Risk Analysis9 (1): 119–126.

Greenberg, M. R., D. B. Sachsman, P. M. Sandman, and K. L. Salomone. 1989b.“Risk, Drama and Geography in Coverage of Environmental Risk by Network TV.”Journalism & Mass Communication Quarterly66 (2): 267–276.

Grifﬁth, C. J., K. A. Mathias, and P. E. Price. 1994.“The Mass Media and Food Hygiene Education.”British Food Journal96 (9): 16–21.

Johnson, B. B., and P. Slovic. 1995.“Presenting Uncertainty in Health Risk Assessment:

Initial Studies of Its Effects on Risk Perception and Trust.”Risk Analysis15 (4):

485–494.

Jwa, B.-K., M.-Y. Yun, and H.-J. Paek. 2013.“Media, Risk Characteristics, and Risk Percep- tions: The Context of Carcinogenic Hazards.”Korean Journal of Public Relations Research17 (4): 72–109.

1334T. Hoveet al. Kahneman, D., and A. Tversky. 1982a.“Variants of Uncertainty.”Cognition11 (2): 143–157.

Kahneman, D., and A. Tversky. 1982b.“The Simulation Heuristic.”InJudgment Under Uncertainty: Heuristics and Biases, edited by D. Kahneman, P. Slovic, and A. Tversky, 201–208. New York: Cambridge University Press.

Kitzinger, J. 1999.“Researching Risk and the Media.”Health, Risk and Society1 (1): 55–69.

Kitzinger, J., and J. Reilly. 1997.“The Rise and Fall of Risk Reporting Media Coverage of Human Genetics Research, False Memory Syndrome and Mad Cow Disease.”European Journal of Communication12 (3): 319–350.

Konheim, C. S. 1988.“Risk Communication in the Real World.”Risk Analysis8 (3):

367–373.

Krippendorff, K. 2004.Content Analysis: An Introduction to Its Methodology. 2nd ed.

Thousand Oaks, CA: Sage.

Kuhn, K. M. 2000.“Message Format and Audience Values: Interactive Effects of Uncertainty Information and Environmental Attitudes on Perceived Risk.”Journal of Environmental Psychology20 (1): 41–51.

Lee, H. J. 2010.“Where is the Carcinogen in Everyday Life.”Accessed December 27, 2013. http://health.chosun.com/site/data/html_dir/2010/04/05/2010040501488.html Logan, R. A., J. Park, and J. Shin. 2004.“Elite Sources, Context, and News Topics: How Two Korean Newspapers Covered a Public Health Crisis.”Science Communication25:

364–398.

Lombard, M., J. Snyder-Duch, and C. C. Bracken. 2002.“Content Analysis in Mass Commu- nication Research: An Assessment and Reporting of Intercoder Reliability.”Human Communication Research28: 587–604.

Luther, C. A., and X. Zhou. 2005.“Within the Boundaries of Politics: News Framing of SARS in China and the United States.”Journalism and Mass Communication Quarterly 82 (4): 857–872.

McCarthy, M., M. Brennan, M. De Boer, and C. Ritson. 2008.“Media Risk Communication– What Was Said by Whom and How Was It Interpreted.”Journal of Risk Research11 (3):

375–394. doi: 10.1080/13669870701566599 .

Miles, S., D. S. Braxton, and L. J. Frewer. 1999.“Public Perceptions about Microbiological Hazards in Food.”British Food Journal101 (10): 744–762.

Nisbet, M. C., D. Brossard, and A. Kroepsch. 2003.“Framing Science: The Stem Cell Controversy in an Age of Press/Politics.”Harvard International Journal of Press/Politics 8 (2): 36–70.

Nisbet, M. C., and M. Huge. 2006.“Attention Cycles and Frames in the Plant Biotechnology Debate: Managing Power and Participation through the Press/Policy Connection.” Harvard International Journal of Press/Politics11 (2): 3–40.

No, Y. H. 2011.Explanation Book for Terms of Hazard. 2nd ed. Chungcheongdo: Korea Food and Drug Administration.

Oh, H. J., T. Hove, H.-J. Paek, B. Lee, H. Lee, and S. K. Song. 2012.“Attention Cycles and the H1N1 Pandemic: A Cross-national Study of US and Korean Newspaper Coverage.” Asian Journal of Communication22 (2): 214–232.

Patt, A. G., and D. P. Schrag. 2003.“Using Speciﬁc Language to Describe Risk and Proba- bility.”Climatic Change61 (1): 17–30.

Peters, H. P. 1995.“The Interaction of Journalists and Scientiﬁc Experts: Co-operation and Conﬂict between Two Professional Cultures.”Media, Culture and Society17 (1): 31–48.

Powell, M., S. Dunwoody, R. Grifﬁn, and K. Neuwirth. 2007.“Exploring Lay Uncertainty about an Environmental Health Risk.”Public Understanding of Science16 (3): 323–343.

Roche, J. P., and M. A. T. Muskavitch. 2003.“Limited Precision in Print Media Communica- tion of West Nile Virus Risk.”Science Communication24 (3): 353–365.

Sandman, P. M. 1997.“Mass Media and Environmental Risk: Seven Principles.”InWhat Risk? Science, Politics and Public Health, edited by R. Bate, 275–284. Oxford: Butter- worth-Heinemann.

Sandman, P. M. 2003.“Four Kinds of Risk Communication.”Accessed June 10, 2012. www.

psandman.com/col/4kind-1.htm Sandman, P. M., P. Miller, B. B. Johnson, and N. D. Weinstein. 1993.“Agency Communica- tion, Community Outrage, and Perception of Risk: Three Simulation Experiments.”Risk Analysis13 (6): 585–598. Journal of Risk Research1335 Sandman, P. M., D. B. Sachsman, M. R. Greenberg, and M. Gochﬁeld. 1987.Environmental Risk and the Press: An Exploratory Assessment. New Brunswick, NJ: Transaction Publishers.

Sellnow, T. L., R. R. Ulmer, M. W. Seeger, and R. Littleﬁeld. 2008.Effective Risk Communi- cation: A Message-centered Approach. New York: Springer.

Semetko, H. A., and P. M. Valkenburg. 2000.“Framing European Politics: A Content Analysis of Press and Television News.”Journal of Communication50 (2): 93–109.

Seol, J.-A. 2011.“A Comparative Study of Internet Search Engines: Naver versus Google.” Journal of Communication Science11 (1): 157–186.

Shih, T.-J., R. Wijaya, and D. Brossard. 2008.“Media Coverage of Public Health Epidemics:

Linking Framing and Issue Attention Cycle toward an Integrated Theory of Print News Coverage of Epidemics.”Mass Communication and Society11: 141–160.

Singer, E. 1990.“A Question of Accuracy: How Journalists and Scientists Report Research on Hazards.”Journal of Communication40: 102–116.

Slovic, P. 1987.“Perception of Risk.”Science236 (4799): 280–285.

Slovic, P. 2000.“Informing and Educating the Public about Risk.”InThe Perception of Risk, edited by P. Slovic, 182–198. Sterling, TX: Earthscan.

Slovic, P., B. Fischhoff, S. Lichtenstein, and F. J. C. Roe. 1981.“Perceived Risk: Psychologi- cal Factors and Social Implications [and Discussion].”Proceedings of the Royal Society of London. A. Mathematical and Physical Sciences376 (1764): 17–34.

Statistics Korea. 2012.“Cause-speciﬁc Death Statistic.”Accessed 2011. http://kostat.go.kr/ portal/korea/kor_nw/2/6/1/index.board?bmode=download&bseq=&aseq=260046&ord=12 Tankard, J. W., and M. Ryan. 1974.“News Sources’Perceptions of Accuracy in Science Coverage.”Journalism & Mass Communication Quarterly51: 219–225.

Tewksbury, D., and D. A. Scheufele. 2009.“News Framing Theory and Research.”Media Effects: Advances in Theory and Research3: 17–33.

Viscusi, W. K., W. A. Magat, and J. Huber. 1991.“Communication of Ambiguous Risk Information.”Theory and Decision31 (2): 159–173.

Viswanath, K., K. D. Blake, H. I. Meissner, N. G. Saiontz, C. Mull, C. S. Freeman, B. Hesse, and R. T. Croyle. 2008.“Occupational Practices and the Making of Health News: A National Survey of US Health and Medical Science Journalists.”Journal of Health Com- munication13: 759–777.

Wahlberg, A. A., and L. Sjoberg. 2000.“Risk Perception and the Media.”Journal of Risk Research3 (1): 31–50.

Wallsten, T. S., D. V. Budescu, A. Rapoport, R. Zwick, and B. Forsyth. 1986.“Measuring the Vague Meanings of Probability Terms.”Journal of Experimental Psychology:

General115 (4): 348–365.

Wardekker, J. A., J. P. van der Sluijs, P. H. M. Janssen, P. Kloprogge, and A. C. Petersen.

2008.“Uncertainty Communication in Environmental Assessments: Views from the Dutch Science-policy Interface.”Environmental Science and Policy11 (7): 627–641.

Wimmer, R. D., and J. R. Dominick. 2005.Mass Media Research. 8th ed. Boston, MA:

Wadsworth.

1336T. Hoveet al. Copyright ofJournal ofRisk Research isthe property ofRoutledge anditscontent maynotbe copied oremailed tomultiple sitesorposted toalistserv without thecopyright holder's express writtenpermission. However,usersmayprint, download, oremail articles for individual use.