Você está na página 1de 11

Risk Analysis, Vol. 31, No.

2, 2011

DOI: 10.1111/j.1539-6924.2010.01516.x

Interpersonal Amplication of Risk? Citizen Discussions and Their Impact on Perceptions of Risks and Benets of a Biological Research Facility
Andrew R. Binder,1, Dietram A. Scheufele,2 Dominique Brossard,2 and Albert C. Gunther2

Much risk communication research has demonstrated how mass media can inuence individual risk perceptions, but lacks a comprehensive conceptual understanding of another key channel of communication: interpersonal discussion. Using the social amplication of risk as a theoretical framework, we consider the potential for discussions to function as amplication stations. We explore this possibility using data from a public opinion survey of residents living in potential locations for a new biological research facility in the United States. Controlling for a variety of key information variables, our results show that two dimensions of discussionfrequency and valencehave impacts on residents perceptions of the facilitys benets and its risks. We also explore the possibility that an individuals overall attitude moderates the effect of discussion on their perceptions of risks and benets. Our results demonstrate the potential for discussions to operate as ampliers or attenuators of perceptions of both risks and benets.
KEY WORDS: Interpersonal discussion; risk communication; social amplication of risk

1. INTRODUCTION In spite of an abundance of empirical research exploring how communication channels inuence individual perceptions of risk, our understanding of mass media inuences far outweighs our understanding of interpersonal inuences. In fact, we know little about how discussions among citizens can impact perceptions directly or how they work in tandem with other information and communication channels to
1

Department of Communication, North Carolina State University, Raleigh, NC, USA. 2 Department of Life Sciences Communication, University of WisconsinMadison, Madison, WI, USA. Address correspondence to Andrew R. Binder, Department of Communication, North Carolina State University, Campus Box 8104, Raleigh, NC 27695-8104, USA; tel: 919-513-2430; fax: 919515-9456; arbinder@ncsu.edu.

shape various perceptual and cognitive outcomes. In this study, we focus on such inuences as potential amplication stations within the social amplication of risk framework. Originally, the framework was proposed to address why and how certain risks capture public attention and become either intensied (through an amplication process) or ignored (through an attenuation process).(1) The model concentrated on two components of risk research: the difference between technical denition and social experience of risk,(2) and the social mechanisms underlying the communication and reception of risk messages. Information about risks is transmitted through various amplication stations, such as social groups, institutions, media outlets, and, ultimately, the individual.(3) Communication is therefore at the heart of the social amplication of risk, since people are exposed

324

0272-4332/11/0100-0324$22.00/1

2010 Society for Risk Analysis

Interpersonal Amplication of Risk to information about risks most commonly through their use of mass media or through discussions with others. Similar to most of the risk communication literature, interpersonal discussion has received relatively little emphasis in research on the social amplication of risk. A number of elds in the social sciences, such as political communication, science communication, and social psychology, in contrast, have long research traditions dealing with the interpersonal antecedents of perceptual and cognitive variables. This research can directly inform our thinking about the social dynamics linking mass-mediated communication, interpersonal discussion, and perceptions of emerging technologies, and allow us to make concrete predictions about the role that interpersonal discussion can play as an agent of risk amplication and attenuation. Our analyses rely on primary survey data from ve communities that were potential sites for a new biological research facility, which would study deadly animal diseases such as hoof-and-mouth disease and African swine fever. This context, which we describe in greater detail below, proves ideal for exploring the inuence of interpersonal discussion because citizens in all ve communities were actively encouraged to voice their opinions and take part in the decisionmaking process. In conceptualizing talk as an amplication station, we emphasize two dimensions that may be particularly relevant to evaluations of the risks and benets of this research facility, as well as how the direct effects of those dimensions may be moderated by an individuals own attitude toward the facility. 1.1. Dimensions of Interpersonal Talk The communication literature linking mass media with interpersonal inuences has focused on a handful of key dimensions of interpersonal communication. In order to situate these dimensions within the context of risk, it is also important to understand the underlying functions of interpersonal discussion. A useful typology provides at least ve purposes: to inform and explain, to argue and persuade, to provide emotional support, to tell stories, and to interpret information.(4) Much normative democratic theorizing emphasizes the informational and persuasive functions and their relevance for public opinion and decision making at the societal level.(58) Interpersonal discussion has also been proposed as a key factor in interpreting messages made salient through the mass media,(9,10) particularly in the context of risk.(11)

325 While they are not the primary focus of this study, these functions underlie at least three dimensions of talk that need to be distinguished when examining social dynamics involving interpersonal communication: content, frequency, and valence. 1.1.1. Content of Discussion Scholars of public deliberation have long distinguished conceptually between goal-directed, mutually informed discussions and casual conversations without a particular topical focus(7,12) and argued that deliberative societal decision making depends on the former. This conceptual distinction and the particular importance of issue-specic, goal-oriented discussion among citizens for informed decision making was conrmed empirically in subsequent research.(8) More recent research has also highlighted the need to develop more granular distinctions of the particular issues that citizens may discuss and how the content-specic focus of discussions is critical for understanding their effects.(13) 1.1.2. Discussion Frequency A number of existing studies in the risk communication literature also propose that the amount of exposure to various communication situations in addition to the content of discussionscan have an impact on individual risk perceptions. In these studies, time is treated as the crucial factor: communication is proposed to have an inuence on attitudes through the number of hours exposed to media or the frequency of talk. The conceptualization of communication channels as competing inuences on personal and/or societal-level risk judgments (e.g., the impersonal impact hypothesis)(14) dominate this literature. As Lehmkuhl(15) argues, this focus may follow from much political communication research in the same vein, including the two-step ow of communication(16) and the so-called limited-effects paradigm of political communication research.(17) The ndings from this line of risk research are rather equivocal regarding the overall impact of discussion frequency. In some cases, frequency of discussion has a signicant and more powerful impact on risk perceptions than do mass media, for both personal and societal risks(18) or subdimensions of personal risks.(11) In another study, interpersonal discussion frequency mediated the inuence of mass media on similar outcomes.(19) In some analyses,

326 researchers operationalize frequency of discussion with different discussants as separate variables, which makes it difcult to evaluate unique relationships between overall frequency of discussion and perception of risk. This is not to say that the type of discussion partner is a negligible inuence, but rather that it is important to distinguish the overall frequency of discussion from other dimensions of talk. Our conceptualization of potential discussion effects differs from these studies in two important ways. First, we are less concerned with the impact of competing communication channels than we are with the net impact of discussion frequency on judgments of both risks and benets. Thus, our focus has more in common with recent studies examining discussion effects on attitudes related to science and technology more generally, where evidence suggests discussion frequency can have a direct impact on various attitudinal outcomes.(13,20,21) Second, the studies reviewed above examine the relative inuence of discussion frequency on personal versus societal-level health risks. This distinction seems less useful for perceptions of a biological research facility, which are relevant on both levels. At the same time, studies examining public opinion toward hazardous waste sites and research laboratories, a number of which also use the social amplication of risk as a theoretical framework,(22,23) have not explored the possible role of interpersonal communication. Since, to our knowledge, there are no extant studies evaluating the effect of frequency of discussion on both risk and benet perceptions, the nature of such a relationship (in terms of magnitude and direction) remains relatively unclear. We therefore pose the following research question: RQ1: What is the relationship between an individuals discussion frequency and his or her perceptions of (a) risks and (b) benets of a biological research facility? 1.1.3. Most Recent Discussion Valence As reviewed above, many studies exploring the interpersonal inuence on risk judgments and perceptions have focused solely on the frequency of discussion. Of course, knowing how often one discusses an issue tells us little about the substance of such conversations. We have therefore chosen to examine another important dimension of interpersonal discussion: the valence of an individuals most recent discussion. That is, if an individual recently discussed a risk-related issue, to what degree was his or her dis-

Binder et al. cussant concerned about it and how does the valence of that discussion color his or her evaluations of risks and benets? Discussion valence is likely to have a direct impact on perceptions of risks and benets for two reasons. First, as studies in political communication have shown in a variety of contexts, it is relatively rare for people to nd themselves in nonlikeminded discussion networks.(24) It follows that people who perceive a higher level of risk may be more likely a priori to discuss an issue more often with others who also perceive a higher level of risk. The same relationship may hold true for perceived benets. Second, the valence of previous discussions has received much attention in research on public opinion, and in particular spiral of silence theory.(25) It is unclear from this research whether we should expect valence to have a direct impact on evaluations of risks and benets, although recent conversations can have a direct causal impact on perceptions of the surrounding opinion climate,(26) which may in turn inuence ones own perception of risks and benets. Regardless of the underlying theoretical mechanismeither self-selection into likeminded discussion groups or an inuence of the climate of opinionit seems likely for there to be some correspondence between ones most recent discussion and evaluations of risks and benets. We formalize this expectation with a second research question: RQ2: What is the relationship between the valence of an individuals most recent discussion and his or her perceptions of (a) risks and (b) benets of a biological research facility? 1.2. Issue-Specic Attitude as a Moderator of Discussion Effects We have focused thus far on intrinsic factors of conversation, that is, content, frequency, and valence, and their potential direct effects on perceptions of risks and benets. However, in addition to drawing direct connections between these dimensions of discussion and attitudinal or cognitive outcomes, the literature also suggests that it is possible that extrinsic factors, such as preexisting characteristics of discussion participants, may shape outcomes of discussion. In particular, an individuals overall attitude toward an object could moderate the relationship between discussion and more specic evaluations of the attributes of that object, and the linear relationship between both frequency and valence of discussion and perceptions of risks and benets may

Interpersonal Amplication of Risk therefore differ among different subgroups in a population. Much of the research cited above, in fact, explores the inuence of interpersonal discussion as contingent on a variety of variables in this way.(24,26) Moreover, as we stated at the outset, the social amplication of risk suggests dynamic processes whereby individual judgments of risk have subsequent impacts originating at the microlevel, continuing through the meso- and macrolevels, resulting in secondary and tertiary ripple effects.(27) We therefore address several reasons why we may reasonably expect interpersonal discussion to function as an amplication station and inuence such dynamic processes. 1.2.1. Discussion and Persuasive Argumentation A large body of research in social psychology has explored small-group decision-making scenarios that feature two of the main functions of interpersonal discussion: explanation and persuasion. This research has focused mainly on explaining the phenomenon of group polarizationwhere a small group comes to a decision that is more extreme than the average initial position of its members(28) in terms of individual thought(29) and repeated expressions.(30) Thus, the resulting shift in the average attitude of the group can be explained, in part, as a function of individuals engaging in discussion about a topic, which increases the number of times they think about and express their attitude toward an object, which, in turn, strengthens their initial position. In the context of risk amplication and attenuation, such a process suggests that engaging more frequently in discussions about a topic will reinforce ones own initial attitude (of support or opposition) and have an impact on perceptions of both risks and benets. We formalize this expected moderation effect with the following hypothesis: H1: An individuals overall attitude toward a biological research facility moderates the relationship between frequency of discussion and perception of the facilitys (a) risks and (b) benets. 1.2.2. Discussion Congruency and the Potential for Disagreement While the valence of discussion might be directly related to evaluations of risks and benets, it is also possible that this relationship differs depending on how that valence corresponds to the individuals overall attitude. In exploring this possibility, we fol-

327 low a long line of research in political communication and a rare example of risk research(15) emphasizing a second dimension: congruency of discussion, or the degree to which discussants agree or disagree with one another. In political communication, congruency has mainly been explored in terms of its effects on learning. Discussion can foster information-seeking(31) and elaboration,(32) although disagreement within ones discussion network may hamper knowledge gain.(33) Regarding political attitudes, research suggests heterogeneous discussion networks can result in ambivalent attitudes(34,35) while homogeneous discussion may polarize attitudes.(21) The theme emerging from all of these studies is that the presence or absence of disagreement with discussion partners can differentially inuence cognitive complexity and thought regarding a given topic. Similarly, we might expect an individuals own attitude to interact with the tenor of his or her more recent discussion (i.e., congruency or incongruency), resulting in differences in perceptions of risks and benets. For example, a supporter discussing an issue with another supporter may perceive less risk than if they discussed the same issue with an opponent. With this reasoning, we propose a second hypothesis: H2: An individuals overall attitude toward a biological research facility moderates the relationship between the congruency of that individuals last discussion and his or her perception of the facilitys (1) risks and (2) benets.

1.3. Context of Inquiry: The National Bio- and Agro-Defense Facility As we mentioned briey above, this study is part of a larger project focusing on public opinion and media coverage surrounding the site-selection process by the U.S. Department of Homeland Security for the National Bio- and Agro-Defense Facility (NBAF). The facility is intended to be a state-of-the art biosafety level 4 facility that will conduct research on highly contagious foreign animal pathogens.(36) Five communities were short-listed as sites for housing the facility in diverse locations throughout the United States: Athens, Georgia; Butner, North Carolina; Flora, Mississippi; Manhattan, Kansas; and San Antonio, Texas. The process was the subject of intense debate, and elected ofcials and policymakers encouraged citizens to attend numerous town hall meetings and debate the merits of hosting the

328 facility in their community. The site-selection process for the NBAF therefore provided an ideal context for exploring the role of interpersonal discussion in the social amplication of risk framework. 2. METHODS 2.1. Data Public opinion surveys were carried out by the University of Wisconsin Survey Center between November 2008 and January 2009 according to the tailored design method.(37) An identical mail questionnaireadjusted for location-specic wordingwas sent to a probability sample of people residing within 25 miles from each of the ve proposed sites for the NBAF.3 Response rates, based on the American Association for Public Opinion Research (AAPOR) Standard Denition RR1,(38) varied by community but were in an acceptable range for a mail survey with the general public: Athens, Georgia (28.5%); Butner, North Carolina (28.7%); Flora, Mississippi (26.0%); Manhattan, Kansas (34.4%); and San Antonio, Texas (25.1%). For our analyses the separate samples were pooled (N = 1,716).4 2.2. Measures First, to control for potential betweencommunities differences in our dependent variables, dummy variables were created for four of the ve communities: Athens (20.4%), Butner (20.6%), Flora (18.3%), and Manhattan (24.1%). San Antonio served as the excluded category in our regression analyses. We measured respondents age (M = 51.57, SD = 16.47) by asking them for their year of birth and subtracting it from 2008, the year the survey was conducted. Respondents also indicated their sex (40.6% female). Educational attainment was measured with
3

Binder et al. the question: What is the highest level of education you have completed? (1 = Some high school, 5 = Completed a graduate or professional degree; Mdn = 4, Completed four-year college with a bachelors degree). We measured household income with the question: Please estimate your total 2007 household income before taxes, combining income from all household members, from all sources (1 = $20,000 or less, 6 = $100,001 or more; Mdn = 3, $40,001 $60,000). These two indicators (r = 0.45, p < 0.001) were standardized and averaged into a two-item index of socioeconomic status.5 Predispositions were each constructed as multiple item indices. Religiosity (M = 3.61, SD = 1.28; r = 0.71, p < 0.001) was measured with two questions: (1) How much guidance does religion give you in your everyday life? (1 = No guidance at all, 5 = A lot of guidance) and (2) How often do you attend religious services? (1 = Every week, 5 = Never; reverse-coded). Political ideology (M = 3.26, SD = 0.98; r = 0.62, p < 0.001) was also measured with two questions: (1) In terms of economic issues, would you say you are . . . and In terms of social issues, would you say you are . . . (1 = Very liberal, 5 = Very conservative). Respondents attention to news media was operationalized as separate variables along two content distinctions. The rst focused on a four-item index of overall attention to public-affairs news (M = 3.62, SD = 0.82; Cronbachs = 0.80), within two content domains (i.e., national government & politics and local affairs & politics) and two types of media (i.e., on television and in newspapers, either in print or online). All four items were measured on a vepoint scale (1 = None, 5 = A lot). We also evaluated respondents specic attention to NBAF news (M = 2.32, SD = 1.10; r = 0.70, p <0.001). The two indicators, measuring attention to both television and newspaper content (both in print and online), were averaged as a two-item index. Both were assessed on a 5-point scale (1 = None, 5 = A lot). As with news media use, we included a multiitem index of how often respondents discussed: (1) local issues and (2) national issues with other people (1 = Never, 5 = Very often). The nal variable interpersonal public-affairs discussionwas an
5

A total of 1,400 individuals were contacted in each of the communities. Names and addresses were purchased from GENESYS Sampling Systems, a subsidiary of Marketing Systems Group. To encourage responses, a $2 bill was included as an incentive in the initial contact mailing. 4 We report AAPOR Response Rate 1 because the address-based sampling procedure (vs. a named sampling procedure) we employed essentially makes ours a random survey, which is not specically discussed in the AAPOR standards. With this design, it is difcult, if not impossible, for us to dene eligible versus ineligible households, which would be required to report more robust, less restrictive response rates. We therefore report RR1, with the caveat that it is very likely the most conservative estimate of responses to our survey.

This was done primarily to minimize the inuence of missing cases, which were particularly numerous on the measure of household income. The mean index allows us to maximize the use of available information in the data set without running the risk of systematically losing cases in our regression model.

Interpersonal Amplication of Risk average of the two (M = 3.47, SD = 0.84; r = 0.61, p < 0.001). Two questions tapped dimensions of interpersonal discussion about the NBAF. The rst indicator focused on issue-specic NBAF discussion frequency: How often do you have discussions with other people about the National Bio- and AgroDefense Facility (1 = Never, 5 = Very often; M = 1.79, SD = 0.86). The second focused on NBAF discussion valence: If you have talked about the NBAF with other people, was your last discussion with . . . . Responses were in one of three categories: Someone who was in favor of the facility (23.2%), Someone who was opposed to the facility (16.9%), and I have not discussed the facility with anyone (59.8%). For the analysis, this valence variable was recoded on a three-point ordinal scale with values 1 (discussion with opponent), 0 (no discussion), and 1 (discussion with supporter). To test our moderation hypotheses, we also included an indicator of overall support for NBAF : Different people have different opinions when it comes to the new facility. How about you? Do you favor or oppose the new facility? (M = 3.39, SD = 1.05). Responses were measured on a vepoint scale (1 = Strongly oppose, 5 = Strongly support). Our two dependent variables tapped respondents evaluations of different dimensions of the facility. Perception of risks (M = 2.87, SD = 0.89) was measured with four items: (1) The NBAF will have a negative impact on the safety of my communitys water supply, (2) Because of the type of research going on at the facility, my community may be the target of a terrorist attack, (3) The facility will increase the risk of environmental contamination in this community, and (4) People in this community may have a higher risk of contracting dangerous diseases because of the NBAF. The four items, which demonstrated high reliability (Cronbachs = 0.86), were averaged to form the nal variable. Perception of benets (M = 3.50, SD = 0.77) was also measured with four items: (1) The NBAF will provide more jobs for my community, (2) If the facility is built in this community, it will have a positive impact on our economy, (3) The construction of the NBAF will have positive effects on the infrastructure of my community (such as road conditions), and (4) The research conducted at the facility will benet local farmers. Based on their reliability (Cronbachs = 0.83), the four items were averaged into a single variable. 2.3. Analytic Framework

329

Our analyses were conducted using hierarchical ordinary least-squares (OLS) regression. The analysis proceeded in two steps. First, each block of variables was entered into the equation according to their assumed causal order (e.g., rst the block of dummy variables controlling for community differences, then demographics and predispositions, etc.). The nal block contains a single variable (risk or benet perception), which controls for the interrelationships between perceptions of risks and benets. The block-by-block approach allows us to evaluate the variance explained by each set of variables as they are entered as predictors. The before-entry standardized betas allow us to evaluate the main effects of each variable controlling for those already entered into the equation,(39) and are especially useful for evaluating the independent effects of variables that may be highly correlated with one another (in this case, for example, the effects of public-affairs vs. issue-specic news attention and discussion). Second, because our hypotheses focus on the interactive effects of two dimensions of interpersonal discussion and respondents own attitude toward the facility, we tested two two-way interaction terms. These multiplicative terms were calculated after standardizing the original variables, a transformation that helps reduce the impact of multicollinearity in estimating the interaction effect.(40,41) Each interaction term was entered separately in the nal block of the regression equation. 3. RESULTS Prior to reporting the results of our hypothesis tests, we rst provide an overview of the main effects of several control variables. All main-effect regression estimates are reported in Table I. 3.1. Main Effects In the nal equation, only one of the demographic and predispositional variables was a signicant predictor of either risks or benets: higher levels of socioeconomic status were associated with slightly higher perceptions of benets. Overall, demographic variables explained more variance than predispositions. Issue-specic attention to news media was signicantly and negatively related to perceptions of risks, but positively related to perceptions of benets, before they were entered into the regression model.

330

Binder et al.
Table I. Ordinary Least-Squares (OLS) Regression Predicting Perceived Risks and Perceived Benets of NBAF Perceived Risks Before-Entry Final 0.04 0.05 0.01 0.04 1.1 0.02 0.02 0.02 3.0 0.03 0.00 0.6 0.05 0.05 1.4 0.03 0.04 0.13 0.52 34.3 0.07 0.3 40.7 Perceived Benets Before-Entry Final 0.05 0.04 0.05 0.10 3.2 0.01 0.01 0.06 4.1 0.02 0.03 1.3 0.03 0.02 3.1 0.07 0.08 0.10 0.45 27.2 0.3 39.2

Block 1: Community dummies Athens, Georgia Butner, North Carolina Flora, Mississippi Manhattan, Kansas Incremental R2 (%) Block 2: Demographics Age Sex (female) Socioeconomic status Incremental R2 (%) Block 3: Predispositions Religiosity Ideology Incremental R2 (%) Block 4: News media attention Public-affairs news attention NBAF news attention Incremental R2 (%) Block 5: Interpersonal discussion and issue attitude Frequency of public-affairs discussion Frequency of NBAF discussion Valence of last NBAF discussion Own support for NBAF Incremental R2 (%) Block 6: Risk/benet perception Perceived risks Perceived benets Incremental R2 (%) Final Equation R2 (%)

0.00 0.04 0.07 0.06

0.03 0.02 0.16 0.14

0.03 0.13 0.13 0.02 0.08

0.01 0.12 0.18 0.08 0.11 0.10 0.18 0.12 0.06 0.33 0.55 0.07

0.01 0.11 0.01 0.07 0.38 0.62

Note: p < 0.05, p < 0.01, p < 0.001. N = 1,716. All coefcients are standardized betas. Block 1 includes dummy variable representing four of the communities, with San Antonio excluded. For this block, the left-hand column contains upon-entry standardized betas.

Not surprisingly, there was a strong relationship between respondents overall support for the facility and their perceptions of its risks and benets. The more respondents supported the facility, the less likely they were to perceive a high level of risk and the more likely they were to perceive a high level of benets. Finally, perceptions of risk and benet were signicantly (at the 0.05 level) related to each other when entered as predictors in the last block of their respective equations. The coefcient, while negative, is relatively small in magnitude. Regarding Research Question 1, frequency of issue-specic discussion was positively related to both perception of risks and benets before-entry (i.e., controlling for all other variables entered prior). For each of these variables, greater levels of discussion corresponded to a slight increase in levels of perceived risks and benets, providing some insight into

the overall inuence of discussion frequency on these outcome variables. Research Question 2 focused on the valence of respondents most recent discussion. Valence was signicantly related to perceptions of both risks and benets. The direction of this relationship was such that, if respondents last spoke with a supporter, they were more likely to report lower levels of risk but higher levels of benet perceptions. The reverse was true if they last spoke to an opponent.

3.2. Interaction Effects Our hypotheses focused on the moderating inuence of ones own support for the facility on the effects of interpersonal discussion. The interaction tests provide evidence supporting a

Interpersonal Amplication of Risk


Table II. Results of Two-Way Interactions Testing Hypothesized Moderation Effect of Own Support on Effects of Last Discussion Valence and Discussion Frequency Perceived Risks 0.04 0.08 Perceived Benets 0.02 0.08

331

Block 7: Two-Way Interactions Last discussion valence Own support Discussion frequency Own support

Note: p < 0.05, p < 0.01, p < 0.001. N = 1,716. Coefcients are upon-entry betas.

frequency-amplication hypothesis but not a congruency-amplication hypothesis (see Table II). In order to interpret the signicant interaction for frequency of NBAF-related discussion, we divided the sample into subgroups according to the response categories and plotted the observed means on the dependent variable for each. As shown in Fig. 1, respondents who opposed the facility were more likely to perceive a higher level of risk; as they discussed the facility more often, however, their perception of risk was likely to increase signicantly. Similarly, those who reported feeling neutral (i.e., neither supporting nor opposing the facility) were also likely to feel neutral about risks related to the facility if they never discussed it. Among respondents in this group, as their frequency of discussion increased, their neutrality also shifted toward a perception of increased risk. Finally, supporters perceived fewer risks than the other two groups, and as their frequency of discussion increased their level of perceived risks decreased. In the equation predicting perceived benets, a similar pattern emerged (see Fig. 2). As with per-

Fig. 2. Two-way interaction effect of frequency of interpersonal discussion and individuals overall attitudes toward the National Bio- and Agro-Defense Facility (NBAF) on perceived benets.

ceived risks, supporters and opponents of the facility fall unambiguously on either side of the neutral point (3 on the y-axis). As their frequency of discussion increases, their levels of perceived benets shift positively and negatively, respectively. Interestingly, people who reported feeling neutral about the facility initially reported higher than neutral levels of benet perceptions, and perceived benets increased as their level of discussion increased. 4. DISCUSSION This study explored interpersonal inuences on individual perceptions of risks and benets. Conceptualizing interpersonal discussion as a potential amplication station, we found that frequency of discussion functions to amplify and attenuate both risk and benet perceptions, although this effect is contingent on an individuals overall level of support and canin factpolarize perceptions between groups with different initial perceptions. Our second hypothesis, that discussion valence would operate in a similar manner, found no support. Before we explore how these results inform social theories of risk in greater detail, we outline some limitations of our study. 4.1. Limitations First, our analyses relied on cross-sectional data, which cannot account for potential reverse causation as an explanation for our results. As part of our larger study, we collected data in a two-wave panel design in three of these ve communities. Unfortunately, however, it was beyond the scope of this study to

Fig. 1. Two-way interaction effect of frequency of interpersonal discussion and individuals overall attitudes toward the National Bio- and Agro-Defense Facility (NBAF) on perceived risks.

332 replicate the measures for the analyses presented here across both waves, even among the panel communities. We therefore do not formally test our theoretically derived assumptions about causal inuences among variables, and our results should be interpreted in light of this constraint. Nonetheless, studies that have shown that discussion frequency can have a greater inuence on attitudes than attitudes have on discussion frequency(21) increase our condence that the causal arrow runs in a similar direction. Second, we have tested our hypothesized relationships on a single issue at a single point in time. In particular, we were faced with a tradeoff between conducting a detailed, contextualized study of a single issue or a more supercial inquiry across different issues. We opted in favor of the former, and this raises questions about the ecological validity of our ndings. We would argue, however, that the complexity of our study design in terms of data collection and also the ability to develop an in-depth understanding of the context of inquiry far outweigh potential concerns about focusing on one particular issue. Finally, as one anonymous reviewer observed, there are many other features of interpersonal discussion about risk that we have not taken into account with this study. Among these are the size and composition of discussion networks, which have played a large role in the study of political discussion among citizens.(42,43) In order to appreciate fully the role that these other dimensions may play in amplifying or attenuating risk perceptions in different segments of the populace, future research should focus on appropriate methodologies and measures for assessing those inuences. Our own effort to tap the valence of discussion is one promising step in this direction, although more attention needs to be paid in future research to developing good measures for use in public opinion surveys. Addressing more negrained aspects of discussionfor example, the quality of discussion or the process of negotiating meaning based upon available news coveragemay be better suited to methodologies employing ethnography or in-depth interviews.(44) 4.2. Contributions to Risk Communication Overall, the results presented here offer novel and more nuanced ways of understanding the underexplored inuence of interpersonal discussion on perceptions of risks. Since past research has focused

Binder et al. on interpersonal discussion as one of the many competing communication channels,(15) this study rst evaluated the overall impact of frequency of discussion on both risk and benet perceptions. Our results suggested a small positive inuence in both cases; however, we further hypothesized that such a direct inuence may hide differential effects for different subgroups within a population. This moderation effect was supported, and we discuss its implications in more detail below. Similarly, we explored the inuence of an individuals most recent discussion. For this dimension of interpersonal talk, the direct impact was unambiguous: respondents most recent discussion had a strong impactsecond in magnitude only to their overall attitudein predicting risk and benet perceptions of the biological research facility. Moreover, this relationship did not differ depending on whether or not their own support was congruent or incongruent with their discussant. There are two possible explanations, both of which merit attention in future research, for why this might be the case. First, it seems that when people discussed the facilityand roughly half our sample did not discuss it at allpeople mostly conned themselves to like-minded environments. It was beyond the scope of this study to examine potential motivations or constraints that made it more or less likely for citizens to discuss the facility and with whom, and answering these questions will be the logical next step for research in this area. Second, the circumstances did not differentially impact peoples overall evaluations of risks and benets, even for respondents within an incongruent discussion environment. That is, discussions neither increased nor decreased (i.e., amplied or attenuated) levels of these evaluations. Both of these explanations have normative implications regarding democratic decision-making processes surrounding risk-related facilities and the inuence of public opinion on policymakers.(22) Going beyond direct main effects of discussion, our interaction tests suggest that, depending on ones own overall attitude, frequency of discussion amplies or attenuates judgments of risk. In this sense, engaging in discussion more frequently does appear to work as an amplication station within the social amplication of risk framework. Notably, the mechanism applies equally well to perceptions of benets. Perhaps the most intriguing aspect of our results is the inuence of frequency of discussion on neutrals, or those respondents who reported

Interpersonal Amplication of Risk neither supporting nor opposing the new research facility. For these individuals, perceptions of both risks and benets were amplied through more frequent discussion. One explanation for this phenomenon could be the amount of exposure to different viewpoints within these discussions. For example, if an individual is neutral but discusses the facility in an opinion climate of majority supporters, one would expect his or her evaluation of risks to decrease and benets to increase. This explanation does not t our results, however, because upon making judgments of both risks and benets associated with the research facility, neutrals were pulled in a slightly positive direction. That is, in the present case, discussion frequency appears to have amplied both risk and benet perceptions in strikingly similar ways. Thus, rather than functioning to shore them up as supporters and opponents, as shown in experimental and survey research,(21,28) discussion seems to amplify their already existing ambivalence or neutrality toward the NBAF. To the degree that heightened attitudinal ambivalence is related to decreased levels of participation,(35) such a phenomenon also has consequences for democratic decision making regarding the NBAF specically, as well as for the implementation of risk-related policies more generally. Even with the contributions of this study to our understanding of how interpersonal discussion can impact on individual judgments of risks and benets, it seems evident that much more conceptual and empirical work needs to be done. Among other things, researchers must expand risk communication research beyond the notion that mass media compete with discussion. The current study and other recent research(15) demonstrate that widening the conceptual scope of risk communication can lead to fruitful new insights. ACKNOWLEDGMENTS A previous version of this article was awarded rst place in the Risk Communication Specialty Group ExxonMobil student paper competition at the 2009 annual meeting of the Society for Risk Analysis. This material is based upon work supported by the University of WisconsinMadison Graduate School (Grant No. 090012) and the National Science Foundation (Grant No. SES-0820474). Any opinions, ndings, and conclusions or recommendations expressed in this material are those of the authors and do not necessarily reect the views of the National Science Foundation. REFERENCES

333

1. Kasperson RE, Renn O, Slovic P, Brown HS, Emel J, Goble R, Kasperson JX, Ratick S. The social amplication of risk: A conceptual framework. Risk Analysis, 1988; 8(2):177187. 2. Renn O, Burns WJ, Kasperson JX, Kasperson RE, Slovic P. The social amplication of risk: Theoretical foundations and empirical applications. Journal of Social Issues, 1992; 48(4):137160. 3. Pidgeon NF, Kasperson RE, Kasperson JX. The Social Amplication of Risk. New York: Cambridge University Press, 2003. 4. Greene JO, Burleson BR. Handbook of Communication and Social Interaction Skills. Mahwah, NJ: Erlbaum, 2003. 5. Delli Carpini MX, Cook FL, Jacobs LR. Public deliberation, discursive participation, and citizen engagement: A review of the empirical literature. Annual Review of Political Science, 2004; 7:315344. 6. Fishkin JS. Democracy and Deliberation. New Haven: Yale University Press, 1991. 7. Schudson M. Why conversation is not the soul of democracy. Critical Studies in Mass Communication, 1997; 14(4):297309. 8. Scheufele DA. Talk or conversation? Dimensions of interpersonal discussion and their implications for participatory democracy. Journalism and Mass Communication Quarterly, 2000; 77(4):727743. 9. Scheufele DA. Examining differential gains from mass media and their implications for participatory behavior. Communication Research, 2002; 29(1):4665. 10. Scheufele DA. Democracy for some? How political talk both informs and polarizes the electorate. Pp. 1932 in Hart RP, Shaw D (eds). Communication and U.S. Elections: New Agendas. Lanham, MD: Rowman and Littleeld Publishers, 2001. 11. Dunwoody S, Neuwirth K. Coming to terms with the impact of communication on scientic and technological risk judgments. Pp. 1130 in Wilkins L, Patterson P (eds). Risky Business: Communicating Issues of Science, Risk, and Public Policy. Westport, CT: Greenwood, 1991. 12. Gutmann JO, Thompson D. Democracy and Disagreement. Cambridge, MA: Harvard University Press, 1996. 13. Binder AR. Routes to attention or shortcuts to apathy? Exploring domain-specic communication pathways and their implications for public perceptions of controversial science. Science Communication, in press. DOI: 10.1177/1075547009345471. 14. Tyler TR, Cook FL. The mass media and judgments of risk: Distinguishing impact on personal and societal-level judgments. Journal of Personality and Social Psychology, 1984; 47(4):693708. 15. Lehmkuhl MJ. Congruency within rural social networks as an indicator of interpersonal inuence on risk judgments: The great stir caused by BSE in a village in northern Germany. Public Understanding of Science, 2008; 17(4):485502. 16. Katz E, Lazarsfeld PF. Personal Inuence. Glencoe, IL: Free Press, 1955. 17. Klapper JT. The Effects of Mass Communications. Glencoe, IL: Free Press, 1960. 18. Coleman CL. The inuence of mass media and interpersonal communication on societal and personal risk judgments. Communication Research, 1993; 20(4):611628. 19. Morton TA, Duck JM. Communication and health beliefs: Mass and interpersonal inuences on perceptions of risk to self and others. Communication Research, 2001; 28(5):602 626. 20. Southwell BG, Torres A. Connecting interpersonal and mass communication: Science news exposure, perceived ability to understand science, and conversation. Communication Monographs, 2006; 73(3):334350. 21. Binder AR, Dalrymple KE, Brossard D, Scheufele DA. The soul of a polarized democracy: Testing theoretical linkages

334
between talk and attitude extremity during the 2004 presidential election. Communication Research, 2009; 36(3):315340. Flynn J, Peters E, Mertz CK, Slovic P. Risk, media, and stigma at Rocky Flats. Risk Analysis, 1998; 18(6):715727. Metz WC. Historical application of a social amplication of risk model: Economic impacts of risk events at nuclear weapons facilities. Risk Analysis, 1996; 16(2):185193. Huckfeldt R, Sprague JD. Networks in context: The social ow of political information. American Political Science Review, 1987; 81(4):11971216. Noelle-Neumann E. The spiral of silence: A theory of public opinion. Journal of Communication, 1974; 24(2):4351. Scheufele DA, Shanahan J, Lee E. Real talk: Manipulating the dependent variable in spiral of silence research. Communication Research, 2001; 28(3):304324. Renn O. Risk communication and the social amplication of risk. Pp. 287324 in Kasperson RE, Stallen PJ (eds). Communicating Risks to the Public. Dordrecht, The Netherlands: Kluwer, 1991. Moscovici S, Zavalloni M. The group as a polarizer of attitudes. Journal of Personality and Social Psychology, 1969; 12:125135. Tesser A, Conlee MC. Some effects of time and thought on attitude polarization. Journal of Personality and Social Psychology, 1975; 31(2):262270. Brauer M, Gliner MD, Judd CM. The effects of repeated expressions on attitude polarization during group discussions. Journal of Personality and Social Psychology, 1995; 68(6):10141029. Scheufele DA, Nisbet MC, Brossard D, Nisbet EC. Social structure and citizenship: Examining the impacts of social setting, network heterogeneity, and informational variables on political participation. Political Communication, 2004; 21(3):315338. Eveland WP. The effects of political discussion in producing informed citizens: The roles of information, motivation, and elaboration. Political Communication, 2004; 21(2):177193.

Binder et al.
33. Feldman L, Price V. Confusion or enlightenment? How exposure to disagreement moderates the effects of political discussion and media use on candidate knowledge. Communication Research, 2008; 35(1):6187. 34. Huckfeldt R, Mendez JM, Osborn T. Disagreement, ambivalence, and engagement: The political consequences of heterogeneous networks. Political Psychology, 2004; 25(1):6595. 35. Mutz DC. The consequences of cross-cutting networks for political participation. American Journal of Political Science, 2002;46(4):838855. 36. Johnson JV. National Bio- and Agro-Defense Facility: Final Environmental Impact Statement. Washington, DC: U.S. Department of Homeland Security, 2008. Available at: http://www.dhs.gov/les/labs/gc 11877346776.shtm#1, Accessed on August 3, 2009. 37. Dillman DA, Smyth JD, Christian LM. Mail and Internet Surveys: The Tailored Design Method, 3rd ed. New York: Wiley, 2008. 38. AAPOR. Standard Denitions: Final Dispositions of Case Codes and Outcome Rates for Surveys, 6th ed. Lenexa, KS: AAPOR, 2009. 39. Cohen J, Cohen P. Applied Multiple Regression/Correlation Analysis for the Behavioral Sciences, 2nd ed. Hillsdale, NJ: Erlbaum, 1983. 40. Eveland WP. Interactions and nonlinearity in mass communication: Connecting theory and methodology. Journalism and Mass Communication Quarterly, 1997; 74(2):400416. 41. Aiken LS, West SG. Multiple Regression: Testing and Interpreting Interactions. Thousand Oaks, CA: Sage, 1991. 42. Huckfeldt R, Sprague JD. Citizens, Politics, and Social Communication: Information and Inuence in an Election Campaign. New York: Cambridge University Press, 1995. 43. Eliasoph N. Avoiding Politics: How Americans Produce Apathy in Everyday Life. New York: Cambridge University Press, 1998. 44. Babbie E. The Practice of Social Research, 11th ed. Belmont, CA: Thomson Wadsworth, 2007.

22. 23. 24. 25. 26. 27.

28. 29. 30.

31.

32.

Você também pode gostar