Environmental Risk in the Media: Modeling the Reactions of the Audience

 

Josef Nerb, Hans Spada, and Katja Lay

University of Freiburg

Germany

Abstract

This article is about media reports on environmental risks and about the question how recipients react to such media reports. We first discuss the mechanisms of news selection by journalists followed by the results of a media analysis of reports on environmental risks. We then will present a psychological theory of risk appraisal that includes emotional and behavioral reactions to typical media reports. The theory is implemented as a computer simulation (ITERA). After introducing the ITERA model and its assumptions, we will provide various empirical support for the model. We will review experiments dealing with one newspaper report and present in more detail an experiment that scrutinizes the effects of a series of two media reports. The paper concludes with a discussion of the implications for risk perception and risk communication research.

 

1. Introduction and Overview

For many environmental risks, mass media are the most important source of information for the public. By selecting and presenting risk information, the media can shape public perceptions of environmental problems. As a further consequence the media influence — at least indirectly — what efforts are taken at personal as well as at a societal level to tackle environmental problems (Koné & Mullet, 1994; McCombs, 1992; Nerb, 2000; Slovic, Fischhoff, & Lichtenstein, 1982).

This article scrutinizes the role of the media in shaping the public perception of environmental risks from two perspectives. We will show how journalists select information about environmental risks and how this information is processed by the audience influencing cognition, emotion, and behavior.

After explaining the mechanisms for journalistic news selection, we will present the results of a content media analysis of environmental problems. We then apply psychological theories to the field of environmental risk perception and communication. The theoretical core assumptions of our approach are formally specified and implemented in the computer model ITERA (Intuitive Thinking in Environmental Risk Appraisal) using a constraint satisfaction network. We will review several experimental studies that were conducted to test the empirical validity of the model. ITERA makes predictions about the cognitive evaluation of environmental problems and provides an explanation for the development of distinct emotions (anger and sadness), and the resulting action tendencies (e.g., boycotting groups seen as being responsible for environmental damages). Furthermore, the model describes how cognition and emotions interact in making judgments entailing coherence biases. The framework of ITERA will be also applied to explain lingering effects of thoroughly refuted information on people’s judgments. Finally, we will draw conclusions from our theoretical and empirical results for the empirical validity of appropriate risk-communication and crisis-response strategies.

2. Risk and Media

A substantial part of our experiences with many environmental risks comes indirectly, through various forms of media exposure. News media are important in forming public risk judgments and are often accused for their biased coverage and selection of risk issues. Many industry and government officials have complained for years, that biased news coverage about real or alleged environmental risks raises illegitimate public concern and encourages government spending on selected hazards while other perhaps more pressing hazards are ignored (Mazur, 1994).

Before we examine the effects of media on the audience it is necessary to describe the content and quality of information people are most often confronted with in the media. For that purpose we will analyze what is reported about environmental problems (i.e., news selection), and how it is reported (i.e., news quality).

2.1 News Selection

At first glance, one might presume that the reality that is presented in news reports about risk topics might somehow reflect the criteria used by risk analysis experts. This would have two implications: first, the frequency of reports on a topic should correspond with the risk associated with that topic; second, coverage over time should also reflect the development of a risk over time. As we will see, both implications turn out to be empirically invalid indicating that journalists use different criteria for selecting news.

2.1.1 Quantitative Criteria for Selection

Ample findings in the literature show that media coverage of environmental topics can only partly be explained by the frequency of their occurrence or by the perils that are coupled with them (e.g., Combs & Slovic, 1979, for an early analysis; Kepplinger, 1994). Greenberg, Sachsman, Sandman and Salomone (1989) analyzed reports about anthropogenic environmental risks in the evening news of three American television-broadcasting companies and found very frequent and elaborate coverage of accidents and singular events. Technical risk proved to be a poor predictor for coverage in the news. To illustrate this point, the authors report that there were seven times as many reports on airplane crashes than on risks associated with smoking, although the latter contributed 26.5 times more to the number of fatalities in the same period of time. Kepplinger (1994) found that in Germany over the time period of 1965 to 1986, the frequency of reports about risks of a technology does by no means correlate with the real situation; indeed he found more negative than positive correlations between severity and frequency of coverage. One reason why the severity of an issue only insufficiently explains its coverage can be attributed to so-called “key-events” (Kepplinger & Hartung, 1995). Key-events are extraordinary and spectacular single events that have an especially great influence on the “media-reality”. After such a key-event, the number of reports about similar, otherwise presumably unimportant, events increases drastically, although the number of such events has not changed in reality.

Contrary to those results, Freudenburg, Coleman, Gonzales and Helgeland (1996) found surprisingly little support for exaggerated media coverage on risk. Analyzing the media coverage of 128 diverse hazard events, the authors showed that in a broad sample of hazard events the amount of space devoted to each event in the print media was indeed well predicted by variables reflecting the severity of the event. This finding thus, contradicts the common belief that media display an “anti-technology” bias.

Put together, the findings reported above call for extreme caution when attempting to extrapolate any single analysis. Quantitative criteria such as severity and likelihood of negative outcomes sometimes serve as good predictors for media coverage; sometimes those determinants of technical risk analysis, however, prove to be very poor predictors for news selection by journalists. It its because of this latter mismatch that the mass media are often blamed for being poor at conveying appropriate notions of risk to general audiences (Dunwoody, 1992).

2.1.2 Qualitative Criteria for Selection

The study by Greenberg et al. (1989) also gives insight into qualitative aspects of news selection. Journalists more frequently cover acute single events than chronic risks or abstract risk issues. This is in line with the results of a media analysis about risk coverage by Singer and Endreny (1987):

A rare hazard is more newsworthy than a common one, other things being equal; a new hazard is more newsworthy than an old one; and a dramatic hazard – one that kills many people at once, suddenly or mysteriously – is more newsworthy than a long-familiar illness. (p. 13)

“Bad news is good news; good news is no news” is a journalistic maxim that is especially valid in media coverage of risk (Cohn, 1989, p. 7). Such negativism is a consequence of an investigative journalistic style that has evolved in our modern Western societies. According to their role in our society, journalists are eager to control, uncover, and recognize negative developments at an early stage – or even better right in advance before a more serious problem arises. Thus, there is a reasonable motive behind journalistic inclination towards reporting on negative events. It should be noted, however, that in genuine emergency situations, such as after a major accident, journalists usually are not exaggerating risks, but instead are helping to prevent unwarranted panic in the public (see for instance, the content analysis of media coverage of the Three-Mile Island accident by Stephens & Edison, 1982).

Another important criterion that is related to negativity is paradoxically the entertainment value of a story. McCartney (1997) showed that the people in charge of news selection base their decision, at least in part, on the entertainment value of their material. As cynical as it might sound, however, a story about a catastrophe is considered more entertaining and exciting than a report that shows how such a catastrophe can be avoided. For example, a story of a flooded metropolis has higher entertainment value than a story of a dam built to prevent such flooding, because it is simply not exciting to see a dam holding back a flood (Aronson, 1999). This entertainment function of journalism seems to be increasingly accepted by younger journalists  (Ehmig, 2000, for a recent analysis among German journalists).

Finally, information selection also is culturally specific. The media more frequently cover topics that are geographically or culturally relevant to the recipient. To exemplify the criteria for news selection in the media we will next present an empirical content analysis about media coverage of the topic “sea”.

2. 2 Media Analysis about Environmental Maritime Risks

For the period of 1990–1998, we analyzed 520 reports about the topic “sea” that where published in a local daily German newspaper (Badische Zeitung ; circulation number: 200, 000). The analysis is presented in more detail in Wahl et al. (2000). The findings can be summarized as follows:

Prevalence of single events. Reports about single events such as tanker accidents far outweigh reports about persistent hazards such as the ongoing pollution of the sea. Thirty-one percent of all articles were about tanker accidents. However, chronic problems such as daily industrial and agricultural emissions or the negative consequences of tourism were rarely covered (< 3 %). The Exxon Valdez and Braer tanker accidents were among the most often reported events and may have served as key-events for selecting subsequent topics. Note that the Exxon Valdez disaster occurred in 1989, which was before the analyzed period of time.

Extent of coverage ¹risk. Reports about the oil-industry make up 44 percent of all reports. This does not correctly reflect the damage done to the sea by this industry. Other industries and agriculture do much more harm to the maritime systems (cf. Clark, 1992). Although tanker accidents contribute no more than 5 percent to the overall oil emission, such accidents were the most mentioned topic among all reports (26%).

Affinity towards negativity.  Almost all reports (91%) were of negative content and covered environmental hazards. A more detailed analysis showed clearly an affinity to reports that convey environmental hazards to be anthropogenic and caused by negligence or even criminal behavior.

Geographic and cultural factors. The reported events had mostly happened around German or European coasts. The emphasis was on the North Sea (43%); only 16 percent covered the Pacific, 13 percent the Atlantic, five percent the Mediterranean, four percent the Baltic Sea; twelve percent were undefined.

2.3 News Quality

Of course, the quality of news is partially determined by journalistic criteria for selection. To reveal other factors that help determining the quality of environmental news we will first examine the content of typical risk reports in the media and we will also ask what journalists themselves consider a high quality news report about environmental risks. Finally, we deal with a specific aspect of news quality, namely its precision according to scientific criteria.

2.3.1 What is the Content of Risk Reports?

For analyzing and describing risk communication in general, Sandman and his coworkers (Sandman, Sachsman, Greenberg, & Gochfeld, 1987) differentiate between outrage- and hazard-related content of a message. They categorize non-technical aspects such as blame, political background, or vividness as outrage-related and scientific and technical aspects as hazard-related. The authors analyzed 248 newspaper articles that were pre-selected by journalists themselves for high journalistic quality, and found that only 17 percent of these reports contained hazard-related information. Singer and Endreny (1993) found that between 1964 and 1984, assignments of blame have increased in reports about environmental risks, industry and economy being the primary objectives of the blame.

2.3.2 What Do Journalists Consider as Being a High Quality News Story?

Further evidence for journalistic preference for outrage-related content comes from a study conducted by Salomone, Greenberg, Sandman and Sachsman (1990). The authors presented over 200 media reports about four environmental risks to scientists, politicians, members of special interest groups, and to journalists. While scientists, politicians and members of special interest groups primarily demanded precision of coverage, for journalists the most important criteria for a good article on environmental risks turned out to be vividness of presentation, emphasis on the catastrophic potential and in particular the attribution of causation and the assignment of blame.

2.3.3 The Scientific Precision of News Reports

As numeric information (epidemiological figures, dose-response rates, statistical information etc.) sometimes is very crucial in environmental debates, public understanding of numbers is important for an appropriate appraisal of risk. However, the relation of “News & Numbers” seems to be at odds (Friedman, 1994). The media convey scientific information not only insufficiently but often even in a misleading or wrong way. The meaningfulness and value of reported scientific results is severely restricted by the absence of information about the reliability of data and about the plausibility of underlying scientific models from which the data are derived.

Of course, one has to take into account that the role of the media is not primarily to convey technical information. Journalistic work is constrained by the demands of the recipients who often show only scant interest in technical details and hair-splitting discussions, but rather want to know about causes of hazards and crisis. In particular, the audience wants to know what happened, how it happened, who is to blame, and what the authorities are going to do after an accident. Additionally, journalists have to work under a tight deadline pressures and may be confronted by sources with competing assertions whose validity is difficult to determine. Instead giving difficult validity assessments, journalists very often present competing, sometimes extreme and rather speculative claims so that the recipients are left alone to judge what is right and what is wrong (Dunwoody & Peters, 1992; Sandman, 1994).

As Fong, Rempel, and Hall (1999) have noted, however, it is unrealistic and even self-righteous to only blame the media for doing poor in conveying scientific information appropriately. Given the severe structural constraints imposed on journalists, it might be more feasible, if researchers change and learn how to better communicate with the media. The authors encourage researchers to increase the awareness of the broader implications of their research and to work together with journalists by describing their findings in professional press releases. Well-crafted press releases than can negotiate between the opposing goals of precision and simplicity.

3 Modeling the Effects of Typical Media Reports

As has been shown, people typically receive information in the form of short, specific, and highly vivid media reports consisting of single event descriptions (e.g., accidents) rather than in depth treatments of issues and chronic problems. Thus, lay people’s evaluation of environmental risks is often based on mass media descriptions of single, often disastrous events rather than on long-term probability information about possible damages. Especially in environmental issues, the media often do not report on risks, but describe harm (Singer & Endreny, 1987). These single case reports may be crucial, however, because of their profound influence on the audience. According to the framework of social amplification of risk single events may entail societal or economic consequences that extend far beyond the directly caused harm, including indirect impacts such as liability, insurance costs, loss of trust in institutions, social group mobilization, or alienation from community affairs (Kasperson et al., 1980). So, a better understanding of the psychological processes that occur at the very beginning of such an amplification process is essential.

As an aid for understanding how typical news reports about environmental problems and its presentation influence cognition, emotion, and behavior, the computational model ITERA (Intuitive Thinking in Environmental Risk Appraisal) was developed (Nerb, Spada & Wahl, 1998; Nerb, 2000; Nerb & Spada, 2000). The model is implemented by means of a parallel constraint satisfaction network (Thagard & Verbeurgt, 1998). ITERA explains and predicts the cognitive evaluation of environmental problems, the development of distinct emotions (anger and sadness), and the resulting action tendencies (e.g., boycotting groups seen as being responsible for environmental damage). In addition, the model predicts how cognitions and emotions cohere in making judgments.

3. 1. The Model

The assumptions of the model rest on solid theoretical ground gained from psychological research. The model is based upon the IMP model (IMPression formation) of Kunda and Thagard (1996). Before we discuss each of these assumptions in more detail, we first give an overview of the technical and methodological background of the model.

3.1.1 Parallel constraint satisfaction

ITERA is a so-called parallel constraint satisfaction network (Rumelhart, Smolensky, McClelland, & Hinton, 1986). Such networks consist of a set of nodes and positive (excitatory) and negative (inhibitory) links connecting those nodes. For example, nodes may represent beliefs, concepts, or hypotheses and the bidirectional links between nodes represent the extent to which those entities are consistent or inconsistent with one another. Every node has an activation, and in the network, activation spreads between nodes according to an updating rule until the activations of all nodes settle into a stable state. If not otherwise specified, all nodes start this updating process with an initial activation value of zero. When the network has reached such a stable state, the overall consistency of the network is maximized (see Nerb, 2000; Thagard & Verbeurgt, 1998, for more details). Because the links are bidirectional, each node can influence and be influenced by nodes to which it is directly or indirectly connected. In addition, weights on the links indicate the strength of the relationship between nodes. Since we make no assumptions about the strength of a relation between constructs in ITERA, all weights are held constant, and accordingly all links are symmetrical.

In ITERA, the entities of interest are the emotions sadness and anger, their hypothesized cognitive determinants, and the behavioral intentions help and boycott. Those entities are represented by nodes in the network; the assumed interrelationship between these entities are represented by links connecting the nodes (see Figure 1). This network of nodes and connecting links between them form the theoretical assumptions on which the model rests upon. The role of these entities in the evaluation of environmental risks and the relationship between these entities will be described next.

3.1.2 Appraisal of Environmental Problems

When people are confronted with an environmental problem that was caused by an accident, what kind of cognitive processes are likely to occur? Generally, people are eager to find out the causes of that accident, and this desire is indeed well reflected by the journalistic preference for assigning responsibility and blame when they report about such an accident. People want to know the agent responsible for the damage, because this enables preventing the agent from repeating further transgressions. Assigning responsibility for a particular situation is important not only to understand what has happened, but also to develop a sense of control over what might happen in the future. Random effects can affect anyone, whereas effects of human actions can often be avoided if brought under control.

People infer different degrees of responsibility for an event from information about its causation (Weiner 1995). Because natural causes for a negative event are beyond anyone’s control, they do not require any further assignment of responsibility. However, if the event is attributable to human agency, some factors are influencing the degree of ascribed responsibility. Foremost, the agent is judged more responsible the more controllable the cause. Put differently, agents are seen more responsible when the cause for a problem could have been avoided or prevented. In addition, some mitigating circumstances alleviate, or even eliminate assigned responsibility. According to Weiner, mitigating circumstances are given, if (a) an act serves a higher goal, or if (b) the agent is lacking insight or knowledge and thus is not able to comprehend the “wrongfulness” of the act. Applied to environmental problems, a higher goal may be given if an action tries to increases the overall benefit for a society, or tries to attain a societal highly valued objective. Knowledge or insight reflects, for examples, whether agents knew in advance that there is a possible contingency between their action and threats to the environment

ITERA represents the damage and the determinants of responsibility (human agency, controllability, higher goal, and knowledge) as input nodes within a connectionist network. These input nodes can be set to have initial activations, reflecting whether the corresponding determinant is given (+1), not given (-1), or as unknown (0). By manipulating these input nodes, ITERA allows us to model different scenarios of environmental accidents.

3.1.3 Emotions: Anger and Sadness

Having specified how people cognitively appraise environmental accidents, we now turn to the question what emotional reactions are likely to occur after such accidents. As others (e.g., Lerner & Keltner, 2000a; Forgas, 1995) we suggest that appraisal theories present a rich and often unexploited source of hypotheses about the relationship between cognitive appraisals and emotional reactions. According to many appraisal theories of emotion (cf. Ellsworth & Smith, 1988; Roseman, Antoniou, & Jose, 1996; Scherer, 1999), the appraisals outlined in the previous section should elicit and differentiate between sadness and anger.

According to these theories, sadness and anger should both arise when an event is appraised as unpleasant and thus contrary to ones motives. A damage done to the environmental fulfils this criterion, because most people in Western societies hold pro-environmental attitudes and appraise nature as highly valuable. Furthermore, the perceived height of the environmental damage should affect the intensity of both emotions equally (Ortony, Clore, & Collins, 1988). The determinants of responsibility finally allow differentiating the relative intensity of anger and sadness: when there is an agent who is seen highly responsible for the event, anger should arise; in contrast, sadness should be felt stronger when the event is caused by situational forces (for more details see Nerb & Spada, 2001). Note, however, that this does not preclude that both emotions are felt simultaneously.

3.1.4 Cognition and Emotion Cohere

Not only do cognitions influence emotions, emotions have effects on cognition in various ways, too (see Forgas, 1995, for a review). Most interestingly in our context is a finding by Keltner, Ellsworth and Edwards (1993). The authors showed that inducing anger or sadness influences causal judgments differently. Sad participants perceived situational forces to be more responsible for an ambiguous event than angry participants. Angry participants, in contrast, judged people as more responsible for the same ambiguous event. Thus, an emotion influences judgments in a way that reflects the appraisal pattern of the emotion. Lerner and Keltner (2000a, 2000b) present further evidence for this specific influence of emotions on judgments. ITERA captures this bidirectional relationship between emotion and cognition by using symmetrical bidirectional links between the nodes in the network (Figure 1).

3.1.5 Behavioral Reactions: Boycott and Help

Emotions also have a motivational role. Qualitative different emotions carry specific behavioral intentions or “action tendencies” (see Frijda, 1986). According to Roseman and his colleagues (Roseman, Wiest & Swartz, 1994) anger is likely to bring about a punitive goal (e.g., boycotting the responsible agent), whereas sadness should lead to the goal of replacing or recovering a loss (e.g., helping the victim(s), or helping repair the damage). This would explain the findings that natural hazards motivate people more to help mitigate the negative consequences whereas human-made accidents often bring about blame, sometimes even outrage directed at the responsible agents (Böhm & Pfister, 2000; Brun, 1992; Hallman & Wandersman, 1992).

3.1.6 Coherence Effects

To represent an accident within ITERA, the appraisals (technically, input nodes) damage, human agency, controllability, higher goal, and knowledge will be set as given, not given, or as unknown according to the accident description (see Figure 1). By manipulating these input values, different experimental settings can be realized as a simulation. The results of those simulations can subsequently be compared to empirical data. Because the appraisal-emotion connection is realized bidirectional within ITERA, these appraisal-nodes may interact. As described above, a parallel constraint satisfaction network settles into a state where the overall coherence of the whole network is maximized. This very feature of the network, however, might cause that the final activation of a non-manipulated appraisal-node changes across simulation experiments, in which the input to other nodes were varied. Thus, the model may predict that the very same aspect of a situation will be construed differently depending on other aspects of the situation. We call such a model prediction a coherence effect or a coherence bias.

3. 2. Empirical Validation of the Model

A series of experiments was conducted to test the specific assumptions of ITERA empirically. To this end, results of model simulations that represent different circumstances of environmental accidents were compared with results from corresponding experimental studies.

The first experiment compared the effects of the variation of controllability for the model’s and participants’ results. In the second study, the existence of a higher goal was manipulated. The third experiment contrasted model predictions for the variable knowledge with empirical results. We only give a brief summary of the results of these studies; for illustrating the procedure we describe the first experiment in more detail, however. All three experiments are fully described in Nerb and Spada (2001).

For all experiments, the material was constructed uniformly: Each participant evaluated a fictitious but realistic newspaper report about an environmental problem. The reported damage was held constant while information about the cause and the circumstances of the cause were manipulated. The questionnaire administered after reading the report is introduced next.

3.2.1 Questionnaire

Table 1 shows the wording of the questions in the order in which they appeared in the questionnaire, the scale endpoints, and the associated variables. Questions 1 and 2 assessed the felt intensity of the emotions anger and sadness after reading the newspaper report. Questions 3 to 7 assessed the cognitive evaluation (appraisal) of the reported case. Questions 8 and 9 requested ratings of the behavioral intentions help and boycott. Finally, Question 10 asked for a typicality rating of the described event. The appraisal questions assessed the estimated damage that was described in the newspaper report, whether the damage was caused by a human agent (human agency), whether the cause of the damage could have been avoided (controllability), whether the agent acted in service of a higher goal, and whether the agent had insight into the possible contingency of action and damage (knowledge).

The typicality question has no correspondence in the ITERA model. Typicality ratings allow inferences about how often the media presumable provide information about the type of accident described in the experiments: The more typical participants rate an accident, the more frequently and detailed the media presumably have reported about that type of accident

3.2.2 Experiment: Controllability

In this experiment we examined the effects of the variation of controllability on participants’ judgments. The sample consisted of 78 paid volunteers, mainly students from the University of Freiburg (Germany). Participants were randomly assigned to three experimental conditions in a between subjects design.

The scenario used in this experiment was a report about a tanker accident: a tanker running aground in a severe storm and spilling oil into the North Sea. The material was constructed according to the following principle: besides reporting the damage and the cause of the accident, the circumstances of the cause (manipulation) were described. Three experimental conditions were realized: (a) the tanker did not fulfill the safety guidelines; the damage could have been avoided (high-controllability); (b) the tanker did fulfill the safety guidelines; the damage could not have been avoided (low-controllability); and (c) in a neutral condition no information about the fulfillment of the safety guidelines was given.

Model predictions. To represent the constant part of the three newspaper reports in ITERA, the DAMAGE and HUMAN AGENCY nodes were set as given and the HIGHER GOAL and KNOWLEDGE nodes were set as unknown. In correspondence with the empirical manipulation, model predictions were generated by manipulating the CONTROLLABILITY node. Three conditions were realized by setting CONTROLLABILITY to given (high controllability), to not given (low controllability), or by setting no input to the node (neutral condition). Given these inputs, the model predicts differences for the three conditions as follows: the higher controllability, the higher the empirical ratings for controllability, anger, human agency, knowledge, and boycott and the lower the ratings for sadness, higher goal, and help. The model predicts no differences across conditions for damage.

Results. Statistical analysis revealed a significant effect for the controllability ratings across conditions. Thus the intended manipulation was successful and in accordance with the predictions of ITERA. In line with ITERA, anger and boycott values were higher when participants were confronted with a highly controllable cause. Contrary to model predictions, however, there were no effects for sadness and help ratings. For the variables knowledge and higher goal, the experimental manipulation led to coherence effects. Though both variables were not manipulated in the experiment, participants gave ratings for knowledge and higher goal as if maximizing the overall consistency of their impression of the reported event. In the high-controllability condition ratings for knowledge were highest and ratings for higher goal lowest, whereas the low-controllability condition revealed the opposite pattern for these variables. The findings for human agency are also in line with a coherence interpretation. These coherence effects are in accordance with model predictions. Furthermore, as predicted by the model, the reported damage was not affected by the experimental manipulation of the between subjects variable controllability. For typicality, we found significant differences between experimental conditions. The report in which controllability was manipulated as being lowest was seen as least typical, whereas the highly controllable cause was estimated as most typical.

To summarize, while data showed that a highly controllable cause induces more anger and a higher willingness to boycott the responsible agent than a low controllable cause, the presumed reciprocal effects for sadness and willingness to help did not occur. A possible explanation for this finding might be that people integrate in their sadness ratings not only the case at hand but also the likelihood that similar events might happen in the future. This interpretation is inline with the typicality ratings that were highest in the condition in which the controllability of the cause was high. Participants in this condition might have felt increased sadness, because this negative report about a typical instance of an environmental accident might have confirmed their suspicions that this is a common and often recurrent event. Finally, the model disconfirming results for help may be ascribed to the lack of effects for sadness, because within ITERA help is seen as being dependent on sadness.

3.2.3 Summary of Comparisons of Model and Data

Table 2 gives an overview of the results from three independent experiments conducted to test the model. Generally, we found a good correspondence between the empirical results and the model’s predictions for anger and the intention to boycott the transgressor. Results for help and sadness did not conform as well with model predictions, indicating where the model could be improved.

Perhaps the most interesting result is the observation of “coherence biases” in which non-manipulated aspects of an event are evaluated (differently across experimental conditions) so that the overall coherence of the event representation is maintained. . A similar finding is reported by Finucane, Alhakami, Slovic, and Johnson (2000) in studies investigating the relationship between judgments of risks and judgments of benefits. In one of their studies, the authors provided information that changed the perceived risk of a stimulus item (manipulated attribute) and found an affectively congruent but inverse effect for the perceived benefit of this item (non-manipulated attribute). The data of Finucane et al. support the view that risk and benefit judgments are causally determined, at least in part, by the overall affective evaluation. Our data complement and extend these findings. In all three studies, we found that some non-manipulated attributes of the reported event were evaluated so as to maximize the overall consistency of judgments. In particular, we observed that participants have construed judgments that “go beyond the information given” (Bruner, 1957). The ITERA model provides a formal, rigorous account of the kinds of` ”construal” effects that are often documented in social psychology (e.g., Kunda, & Sherman-Williams, 1993). Overall, the empirical validity of this coherence bias strongly corroborates the assumption of a bidirectional appraisal-emotion connection.

3.3 How the Model could be improved and extended

In our studies, ratings for anger and sadness were not inversely related as predicted by the model. In order to account for that finding, an additional node might be added to ITERA that has an excitatory (or likewise an inhibitory) connection to both anger and sadness. The appraisal of the likelihood that agents will repeat their transgressions in the future might be a candidate for this node. The imagination that an agent will continue to commit transgressions may make people sad as well as angry. Thus, such a node for stability of human caused negative events could have excitatory links to both anger and sadness. Adding a node for stability to ITERA would be consistent with emotion research based on attribution theory (cf. Weiner, 1995).

In addition, we regard two further extensions as appropriate to improve the ITERA model. First, we suggest differentiating between qualitative and quantitative aspects of a damage; second we consider to incorporate a mechanism for integrating multiple and sequential information into ITERA.

3.3.1 Qualitative and Quantitative Aspects of Damage

An additional experiment (Nerb, 2000; unpublished raw data) shows how the model might be extended. In this experiment, we found that qualitative and quantitative aspects of an environmental damage influence the emotional reactions of participants differently. In case where the damage is about endangering a rare animal species, participants feel increased sadness (“qualitative high damage”); whereas quantitative aspects of a damage only affect the ratings of anger. This is inline with findings reported by Böhm & Pfister (2000). The authors also showed that the type of potential consequences of a risk determines which emotional reaction and subsequently, which action tendency will result. In particular, they found that the tendency to help to prevent environmental problems are strongest if consequences are involved that may also be harmful to humans, and if the risk is geographic distant. Altogether, these findings suggest differentiating within ITERA between qualitative different aspects of an environmental problem.

3.3.2 Integrating Multiple Information Sources within the Model

The ITERA model as it is described here only deals with the evaluation of single descriptions of environmental accidents. As we have shown such single case descriptions are typical sources of information for appraising environmental problems. Thus, ITERA in its current version is a valid approach and an important first step to model potential reactions of the audience. However, very often the audience learns about environmental problems in a sequential, piece by piece fashion. Thus, lay people have to integrate in their judgments information that is not only incomplete and controversial, but often inconsistent or even wrong. In the next section we will present results of an experimental study that showed how lay people integrate new information into their judgments that contradicts previously given information. This form of knowledge updating is not covered in ITERA. However, we have recently developed an extension of ITERA that represents hindsight effects by integrating schematic knowledge, current and previously presented information. This model extension is called HIBERIA for HIndsight Bias in Environmental RIsk Appraisal (Lay, Nerb, & Spada, 2000).

3.4 How misinformation affects environmental risk appraisal

Media coverage on environmental accidents often consists of a series of reports. Within a few hours after the occurrence of the accident, headlines appear that soon got augmented with more and more details about the incident. If the accident further unfolds in a newsworthy way, the media disseminate even more information after several days or even weeks and months. Thus, information available to the public is steadily complemented and updated. In this process, one may also learn facts that later turn out to be false or inconsistent. Following the news about a topic therefore implies continuously updating and even correcting of mental representations and repeated revision of previously formed evaluations of the accident.

Research from cognitive psychology and social cognition suggests, that the integration of new information into existing knowledge structures is mainly achieved by spontaneous and effortless processes of knowledge updating. Although the constructive processing of incoming information is usually very effective and yields appropriate results, there are several pitfalls to consider. For example, once an impression is formed, it turns out to be very resistant to change, even in the face of disconfirming information. This is especially true for evaluation of an event consistent with generally held assumptions such as stereotypes and schemata (Kunda & Oleson, 1995). Moreover, judgment and decision processes often turn out to be continuously influenced by information marked as unfounded or irrelevant (Gilbert, Tafarodi, & Malone, 1993; Johnson & Seifert, 1994). This continued influence effect and related phenomena such as the so called hindsight bias are interpreted as manifestations of fundamental difficulties to ignore certain pieces of information for judgment once activated in working memory (cf. Hawkins & Hastie, 1990; Hoffrage & Hertwig, 1999; Lay, 2000).

Johnson & Seifert (1994) have empirically shown the effects of misinformation on later inferences and judgments. In a series of experiments, participants read a story about a warehouse fire. Some participants were first given information about a closet in the warehouse containing cans of flammable material. Later, they were told that this information had turned out to be wrong and that the closet had been found empty. Although the participants reported to have recognized the correction, their answers on different questions about the incident showed significantly more inferences based on the misinformation than control participants that had never learned about any flammable material in the closet.

3.5 Experiment: Continued Influence Effects after Exposure to Misinformation

.To scrutinize the effect of misinformation on the evaluation of environmental accidents an experimental study was conducted (see Lay, 2000, Study 4). In the experiment, we examined the effect of presenting and discrediting information concerning the role of a company involved in an environmental accident. Thus, we compared evaluations of participants who receive a causal explanation (reference condition) and judgments of participants who receive the same explanation but are later informed that this information had turned out to be wrong (misinformation condition) to judgments of participants in a neutral condition with no causal explanation presented at all (control condition). In line with research on the continued influence effect, it was hypothesized that evaluations of participants in the misinformation condition will differ significantly from controls, their judgments being biased towards the evaluation of participants in the reference condition, thus showing a continued influence effect.

It was also hypothesized that the occurrence of the continued influence effect will be influenced by stereotypic beliefs, as was shown to be the case for related phenomena such as the hindsight bias (see Brown, Williams & Lees-Haley, 1994; Lay, 2000). A continued influence effect as reported by Johnson & Seifert (1994) is only predicted to occur after the correction of schema-consistent causal information.

To test this additional hypothesis, participants in the reference and misinformation condition were provided with either one of two alternative explanations, stating that the accident resulted from negligence on part of the company or from an act of nature beyond the control of humans. Based on the media analysis reported in this article and the well-documented tendency to attribute negative events to personal failure rather than to situational factors (cf. Nerb, Bender & Spada, in press), it was assumed that the negligence explanation is consistent with such an event-schema, whereas the act of nature explanation is not. A systematic shift in judgments in the same direction as in the respective reference condition is therefore only expected for the presentation and correction of the negligence explanation, not for the act of nature explanation.

3.5.1 Methods

The sample consisted of 80 paid volunteers, all students of a local high school in Freiburg (Germany). The median age was 17 (range 16 to 21). Participants were randomly assigned to five experimental conditions in a between subjects design. Participants were asked to evaluate an environmental accident based on two fictitious but realistic newspaper reports containing different information about the circumstances that caused the incident

The scenario used in this experiment described the pollution of a stretch of running water due to toxic substances that escaped from a mining company. It was stated that a leakage in the dam wall of a reservoir containing the industrial effluents of the company caused the pollution. As in the other studies described in this article, the reported damage was held constant while information about the circumstances that caused the leakage was manipulated.

There were three versions for the first report: In one version, participants were informed that the leak in the dam wall was caused by the company having discharged illegally large amounts of effluents in the reservoir (negligence explanation); in the framework of ITERA, this amounts to a situation where there is a human agency with high controllability. In another version, the report stated that the leak was caused by continued heavy storms and unforeseeable rainfalls in the foregoing weeks (act of nature explanation); in the framework of ITERA, this amounts to a situation where the cause of the problem is beyond the control of the agent. In the third version, no information concerning the occurrence of the leakage was provided (neutral version).

All participants then read a second report on the same incident. For the participants who got the negligence explanation or the act of nature explanation as the first report, this second report consisted either of an official denial of the first report (misinformation groups) or of some general information (reference groups). The denying reports stated that the prior explanation had turned out to be unfounded, but gave no alternative explanation. For the reference groups, the causal explanation given before was not mentioned again in the second report. For these participants, the second report only contained some general information about the accident that also was available to the participants in the misinformation groups. Thus, for some of the participants the explanation given in the first report remained unchallenged after reading the second report, whereas for the others, this explanation was discredited.

The participants who read the neutral version with no information about the cause of the leakage in the first report received in the second report some general information about the accident that was equally available to participants in the misinformation and reference groups.

Table 3 summarizes the design of the experiment.

After reading the two reports about the pollution, all participants rated the accident, and their reactions towards the event. The questionnaire consisted of a subset of the variables of ITERA (damage, controllability, typicality, boycott, and help; see Table 1). In addition, the item responsibility assessed the extent to which the company was seen responsible for the damage caused by the leakage. Like the other items in the questionnaire, this item was also measured on a nine-point scale. Finally, subjects in the misinformation and reference conditions answered questions concerning the announcement or correction of the causal explanations.

3.5.2 Results and Discussion

Table 4 shows the empirical means together with the standard deviations for all variables in all five conditions of the experiment. Statistical analyses revealed significant effects of the experimental manipulation on all variables, with the exception of help (Table 4; right columns). To illustrate the effect of the two alternative explanations on judgment processes in this study, we will first focus on differences concerning the judgments of the reference groups and the no explanation control group. We will then report the effect of discrediting these explanations by comparing the evaluative judgments of the misinformation groups with the neutral control condition.

Although held constant, the damage caused by the leakage was rated more severe when it was attributed to negligent and supposedly illegal behavior on part of the company compared to the bad weather conditions. Furthermore, participants in the negligence reference condition assigned more responsibility to the company and show more willingness to engage in a boycott against the company than control participants. They also rated the accident as easier to control and as more typical than participants in the neutral control condition. Participants in the act of nature reference condition assigned less responsibility to the company and reported to be less willing to engage in a boycott than control participants. If compared to the control condition, they also rated the accident as harder to control and as less typical. These results are consistent with the predictions of ITERA and replicate the findings described above. Again, there was no effect of the experimental manipulation on the variable help. But unlike the other studies reported in this article, the results of this experiment reveal an additional coherence effect with damage being rated more severe when attributed to human failure (see Böhm & Pfister, 2000; Brown et al., 1994; Nerb et al, in press, for complementary findings).

Analyzing the judgments of the misinformation groups adds another important aspect to these findings. Both misinformation groups differed significantly from the neutral control condition even though, like in the Johnson and Seifert (1994) study, all misinformation participants reported to have recognized the correction information. Participants confronted with the denial of the negligence explanation showed a significantly higher tendency to blame the company and to support a boycott when compared to controls. They also rated the accident as easier to control and more typical, just like participants for whom the negligence explanation was not discredited (negligence reference group). The shift of judgments in the negligence misinformation group towards judgments in the negligence reference group can be interpreted as a manifestation of the continued influence effect after exposure to misinformation.

Even more interestingly, when compared to controls, participants confronted with the denial of the act of nature explanation also assigned more responsibility to the company, rated the damage as easier to prevent and showed enhanced willingness to engage in a boycott. Their judgments did only marginally differ from judgments in the negligence misinformation group. If we recall that the respective reference group displayed significantly lower ratings on all these variables if compared to control participants, the different judgments between the control and the act of nature misinformation condition can be interpreted as a contrast effect of the discredited information.

Taken together, if compared to the neutral control condition, judgments in both misinformation groups are biased towards judgments in the reference group provided with the explanation that the reported damage resulted from negligent behavior of the company involved in the accident. To illustrate this general finding, Figure 2 shows the mean values for the variable responsibility. In other words, any misinformation concerning the company’s involvement in the accident is leads to more negative evaluations and a enhanced tendency to engage in actions directed against the company, no matter whether the misinformation originally accused the company or not. This result is consistent with the notion that the audiences’ trust in industrial and public representatives concerning environmental damage and risk perception in general is low and therefore hard to obtain and easily lost (Slovic, 1993, 1999).

Our findings concerning the continued influence of misinformation lead to two important conclusions: (1) Discrediting causal information concerning an environmental accident is psychologically not equivalent to re-creating a neutral information stand. This finding is consistent with research on the so-called perseverance effect (Ross, Lepper, & Hubbard, 1975) or continued influence effect reported in social cognition research. (2) Discrediting causal information — at least when no alternative information is provided — leads to judgments that appear more schema-driven and therefore more negative for the party presumably or actually responsible for causing an environmental damage. This finding is consistent with the hypothesis that event schemata and previously formed attitudes provide default values that will be relied on in situations in which information crucial for understanding and evaluating is not reliable or contradictory.

Evaluative memory processes on actual judgments are also assumed to play a role in the effect of time on risk perception and attributions of responsibility and blame (Eiser, Podpadec, Reicher & Stevenage, 1998). Eiser et al. (1998) provide empirical evidence for the dynamic nature of environmental risk perception. In our study, however, reactions towards an environmental damage were measured immediately after presenting the relevant information. Further research with repeated measures over time might directly address the temporal dimension of evaluative judgments in order to explain in more detail the interaction of judgment and memory processes in the formation and maintenance of attitudes after receiving valid and invalid information.

In sum, the results of this experiment provide strong evidence for the potentially detrimental effects of any kind of misinformation in media reports on environmental accidents. This opens a dilemma for professionals in risk communication: Whereas refusing to inform the public provides the ground for speculations that are hard to overcome, jumping to conclusions that later have to be denied and corrected bears the risk of continued influence of misinformation (see also Kepplinger & Hartung, 1995). Both situations might enhance the influence of undifferentiated, stereotypic beliefs on the public opinion and related behavioral responses.

4. Conclusions

At the end of this article, we will recapitulate in what respect the model supplements classical contributions to risk research (see Slovic, 1999, for an overview). Finally, we will draw tentative conclusions and suggestions from our framework for risk communication.

The approach taken here differs from prior work in environmental risk research in two important respects. First, we take into account different qualities of negative emotions, sadness and anger, by utilize appraisal theories. Most theories of affective influences on judgment and choice, however, are exploiting a valence-based approach and thus are only contrasting the effects of positive versus negative feeling states. According to the valence-based approach, distinct emotions of the same valence, such as sadness, anger, and fear, should exert similar influences on judgment and choice (see Elster, 1998; Forgas, 1995, for overviews). Lerner and Keltner (2000a, 2000b; see also Keltner et al, 1993) point to an obvious shortcoming of valence-based approaches and have started to specify whether and how different emotions of the same valence differentially influence judgments. In line with the approach taken within ITERA, Lerner and Keltner found that specific emotional states or dispositions influence judgments and decisions in a way that reflects the appraisal pattern underlying those emotions. For example, the authors demonstrate, that two negative emotions, fear and anger, guide judgments of risk in opposite ways: Whereas fearful individuals made pessimistic judgments about future events, angry individuals made optimistic judgments (Lerner & Keltner, 2000a, 2000b).

Second, we look at the most typical form of information that lay people receive about environmental problems, namely, media reports about single events. Traditionally, risk research focuses on classes of hazards, technologies, or human activities. However, single instances of such classes often are crucial because of their profound influence on judgments concerning the significance of an issue. For example in communication research, a number of recent studies have demonstrated the predominant role of single exemplars in shaping judgments about the seriousness of a problem (Brosius & Bathelt, 1994; Daschmann & Brosius, 1999; Zillmann & Brosius, 2000).

Our studies show how details about a single case (different levels of attributed responsibility) may have effects on social group mobilization, as was indicated by participants’ expressed willingness to boycott. Similar reactions were even observed when the information about an agent’s responsibility had been discredited, even though participants had been well aware of this denial. This is remarkable, because in each experiment, the reported consequences of the accidents were held constant across conditions and – the study on misinformation effects being the exception – produced no reliable differences in participants’ judgments of the damage. Thus, a nonconsequential factor has caused the difference in participants’ ratings (see also Baron, 1994; Böhm & Pfister, 2000). In psychometric studies that deal with classes of risk sources, these nonconsequential factors, however, typically are not taken into account appropriately. The importance of single cases in risk research, was recently also put forward by Ritov and Kahneman (1997). They propose the evaluation by exemplar hypothesis stating that representative exemplars might determine the evaluation of a whole class of events. Findings for typicality suggest that highly negative events (such as damage caused by a highly responsible agent) are seen as more representative for a class than positive or less negative events and thus should have more weight in shaping attitudes and opinions. For risk communication, this hypothesis seriously underpins the significance of the old proverb “Bad news travels fast” and calls to awareness that “trust is easily lost but hard to create” (see Slovic, 1999, for an overview).

The framework of  ITERA rests on solid theoretical and empirical basis that easily allows transferring basic research to the applied field of risk communication that often suffers from being a-theoretical and ephemeral. Eventually, the framework presented here helps to understand the root causes of social conflict and provides means to create better risk management. There is no doubt that a better understanding of the complex psychological, social, cultural, and political forces in the end will be crucial of the successes and failures of all kind of risk management.

 

Acknowledgements

This research was supported by grant no. Sp 251/10-x from the German National Research Foundation (DFG) to the second author. We are grateful to Susanne Frings, Fabian Hermann, Miriam Tonne, and Stefan Wahl for their assistance. We thank Joseph Arvai, Gisela Böhm and Derek Koehler for their valuable comments on earlier versions of this manuscript.

Correspondence concerning this article should be addressed to

Josef Nerb, Department of Psychology, University of Freiburg,

D-79085 Freiburg, Germany.

Email: nerb@psychologie.uni-freiburg.de.

 

References

Aronson, E. (1999). The social animal (8th Edition). New York: Worth Publishers.

Baron, J. (1994). Nonconsequentialist decisions. Behavioral and Brain Sciences, 17, 1-42.

Böhm, G., & Pfister, H.-R. (2000). Action tendencies and characteristics of environmental risks. Acta Psychologica, 104, 317-337.

Brosius, H.-B., & Bathelt, A. (1994). The utility of exemplars in persuasive communications. Communication Research, 21, 48-78.

Brown, R.S., Williams, C.W., and Lees-Haley, P.R. (1994). The effects of hindsight bias and causal attribution on human response to environmental events. Journal of Applied Social Psychology, 24, 661-674.

Brun, W. (1992). Cognitive components in risk perception: Natural versus manmade risks. Journal of Behavioral Decision Making, 5, 117-132.

Bruner, J.S. (1957). Going beyond the information given. In H.E. Gruber, K.R. Hammond, and R. Jessor (Eds). Contemporary approaches to cognition (pp. 41-69). Cambridge, MA: Harvard University Press.

Cohn, V. (1989). News & Numbers. Ames, IA: Iowa State University Press.

Combs, B., and Slovic, P. (1979). Newspaper coverage of causes of death. Journalism Quarterly, 56, 837-843,849.

Coombs, W. T. (1995). Choosing the right words: The development of guidelines for the selection of the “appropriate” crisis-response strategies. Management Communication Quarterly, 8, 447-476.

Daschmann, G., & Brosius, H.-B. (1999). Can a single event create an issue? Journalism and Mass Communication Quarterly, 76, 35-51.

Dunwoody, S. (1992). The media and public perceptions of risk: How journalists frame risk stories. In D. W. Bromley and K. Segerson (Eds.), The social response to environmental risk: Policy formulation in an age of uncertainty (pp. 75-100). Dordrecht: Kluwer Academic Publishers.

Dunwoody, S., and Peters, H. P. (1992). Mass media coverage of technological and environmental risks: a survey of research in the United States and Germany. Public Understanding of Science, 1, 199–230.

Ehmig, S. (2000). Generationswechsel im deutschen Journalismus: Zum Einfluß historischer Ereignisse auf das journalische Selbstverständnis. Freiburg: Alber.

Eiser, J. R., Podpadec, T. J., Reicher, S. D., and Stevenage, S. V. (1998). Muddy waters and heavy metal: Time and attitudes guide judgements of pollution. Journal of Environmental Psychology, 18, 199-208.

Ellsworth, P. C., and Smith, C. A. (1988). From appraisal to emotion: Differences among unpleasant feelings. Motivation and Emotion, 12, 271-302.

Elster, J. (1998). Emotions and economic theory. Journal of Economic Literature, 36, 47-74.

Finucane, M. L., Alhakami, A. S., Slovic, P., and Johnson, S. M. (2000). The affect heuristic in judgements of risks and benefits. Journal of Behavioral Decision Making, 13, 1-17.

Fong, G.T., Rempel, L.A., & Hall, P.A. (1999). Challenges to improving health risk communication in the 21^st Century: A discussion. Journal of the National Cancer Institute Monographs, 25, 173-176.

Forgas, J. P. (1995). Mood and judgement: The affect infusion model (AIM). Psychological Bulletin, 117, 39-66.

Freudenburg, W.R, Coleman, C.-L., Gonzales, J., & Helgeland, C. (1996) Media coverage of hazard events: Analyzing the assumptions. Risk Analysis, 16, 31-42.

Friedman, S. M. (1994). The media, risk assessment and numbers: They don’t add up. Risk: Health, Safety and Environment, 5, 203-212.

Frijda, N. H. (1986). The emotions. Cambridge, UK: Cambridge University Press.

Gilbert, D. T., Tafarodi, R. W., and Malone, P. S. (1993). You can’t not believe everything you read. Journal of Personality and Social Psychology, 65, 221-233.

Greenberg, R. M., Sachsman, B. D., Sandman, M. P., & Salomone, L. K. (1989). Network evening news coverage of environmental risk. Risk Analysis, 9, 119-126.

Hallman, W. K., and Wandersman, A. (1992). Attribution of responsibility and individual and collective coping with environmental threats. Journal of Social Issues, 48(4), 101-118.

Hawkins, S. A., and Hastie, R. (1990). Hindsight: Biased judgements of past events after the outcomes are known. Psychological Bulletin, 107, 311-327.

Hearit, K. M. (1994). Apologies and public relations crises at Chrysler, Toshiba, and Volvo. Public Relations Review, 20, 113-125.

Hoffrage, U., and Hertwig, R. (1999). Hindsight bias: A price worth paying for fast and frugal memory. In G. Gigerenzer, P. M. Todd, and the ABC Research Group (Eds.), Simple heuristics that make us smart (p. 191-208). New York, NY: Oxford University Press.

Johnson, H. M., and Seifert, C. M. (1994). Sources of the continued influence effect. When misinformation in memory affects later inferences. Journal of Experimental Psychology: Learning, Memory, and Cognition, 20, 1420-1436.

Kasperson, R. E., Renn, O., Slovic, P., Brown, S. H., Emel, J., Goble, R., Kasperson, J. X., and Ratick, S. (1988). The social amplification of risk: A conceptual framework. Risk Analysis, 8, 177-187.

Keltner, D., Ellsworth, P. C., and Edwards, K. (1993). Beyond simple pessimism: Effects of sadness and anger on social perception. Journal of Personality and Social Psychology, 64, 740-752.

Kepplinger, H. M. (1994). Historical Notes on German press coverage of technology. Risk: Health, Safety and Environment, 5, 213-222.

Kepplinger, H. M., and Hartung, U. (1995). Störfall-Fieber: Wie ein Unfall zum Schlüsselereignis einer Unfallserie wird. Freiburg: Alber.

Koné, D., and Mullet, E. (1994). Societal risk perception and media coverage. Risk Analysis, 14, 21-24.

Kunda, Z., and Oleson, K. C. (1995). Maintaining stereotypes in the face of disconfirmation: Constructing grounds for subtyping deviants. Journal of Personality and Social Psychology, 68, 565-579.

Kunda, Z., & Sherman-Williams, B. (1993). Stereotypes and the construal of individuating information. Personality and Social Psychology Bulletin, 19, 90-99.

Lay, K. (2000). Urteils- und Gedächtnisverzerrungen bei der Bewertung von Umweltschadensfällen. (Biased judgment and memory in environmental risk appraisal) Unpublished doctoral dissertation, University of Freiburg.

Lay, K., Nerb, J., and Spada, H. (2000). Die Bewertung von Umweltrisiken in der Rückschau (Hindsight Judgments of Environmental Risks). 42. Kongress der Deutschen Gesellschaft für Psychologie, Jena.

Lerner, S. L., & Keltner, D. (2000a). Beyond valence: Toward a model of emotion-specific influences on judgement and choice. Cognition and Emotion, 14, 473-493.

Lerner, S. L., & Keltner, D. (2000b). Fear, anger, and risk. Submitted manuscript. Carnegie Mellon University.

Mazur, A. (1994). Technical risk in mass media. Risk: Health, Safety and Environment, 5, 189-192.

McCartney, J. (1997). News lite. The tendency of network newscasts to favor entertainment and tabloid stories. American Journalism Review, 19, 18–26.

McCombs, M. E. (1992). Explorers and surveyors: Expanding strategies for agenda-setting research. Journalism Quarterly, 69, 813-824.

Nerb, J. (2000). Die Bewertung von Umweltschäden: Kognitive und emotionale Folgen von Medienmeldungen (The evaluation of environmental damages: Cognitive and emotional reactions of media reports). Bern: Huber.

Nerb, J., Bender, A., and Spada, H. (in press). Attributed causes of environmental problems – Cross-cultural studies of coping strategies. In M. Casimir and U. Stahl (Eds.), Culture and the changing environment: Uncertainty, cognition, and risk management in cross-cultural perspective. Oxford: Berghahn.

Nerb, J., and Spada, H. (2001). Evaluation of environmental problems: A coherence model of cognition and emotion. Cognition and Emotion, 15(4), 521-551.

Nerb, J., Spada, H., and Wahl, S. (1998). Kognitive Determinanten der emotionalen Bewertung von Umweltschadensfällen: Modellierung und Empirie (Cognition and emotion in the evaluation of environmental accidents: modeling and empirical studies). Zeitschrift für Experimentelle Psychologie, 45, 251-269.

Ortony, A., Clore, G. L., and Collins, A. (1988). The cognitive structure of emotions. Cambridge: Cambridge University Press.

Ritov, I., and Kahneman, D. (1997). How people value the environment. In M. Bazerman and D. Messick and A. Tenbrunsel and K. Wade-Benzoni (Eds.), Environment, ethics, and behavior (pp. 33--51). San Francisco: The New Lexington Press.

Roseman, I. J., Antoniou, A. A., and Jose, P. E. (1996). Appraisal determinants of emotions: Constructing a more accurate and comprehensive theory. Cognition and Emotion, 10, 241-277.

Roseman, I. J., Wiest, C., and Swartz, T. S. (1994). Phenomenology, behaviors, and goals differentiate discrete emotions. Journal of Personality and Social Psychology, 67, 206-221.

Ross, L., Lepper, M. R., & Hubbard, M. (1975). Perseverance in self perception and social perception: Biased attributional processes in the debriefing paradigm. Journal of Personality and Social Psychology, 32, 880-892.

Rumelhart, D. E., Smolensky, P., McClelland, J. L., and Hinton, G. E. (1986). Schemata and sequential thought processes in PDP models. In D. E. Rumelhart and J. L. McClelland (Eds.), Parallel distributed processing: Explorations in the microstructure of cognition, Vol. 2 (pp. 7-57). Cambridge, MA: MIT Press.

Salomone, K. L., Greenberg, M. R., Sandman, P. M., and Sachsman, D. B. (1990). A question of quality: How journalists and news sources evaluate coverage of environmental risk. Journal of Communication, 40, 117-130.

,Sandman, P. M. (1994). Mass media and environmental risk: Seven principles. Risk: Health, Safety and Environment, 5, 251–260

Sandman, P. M., Sachsman, D. B., Greenberg, M. R., and Gochfeld, M. (1987). Environmental risk and the press: An explanatory assessment. New Brunswick, NY: Transaction Books.

Scherer, K. R. (1999). Appraisal theory. In T. Dalgleish and M. J. Power (Eds.), Handbook of cognition and emotion (pp. 637-664). Chichester: Wiley.

Singer, E., & Endreny, P. M. (1987). Reporting hazards: Their benefits and costs. Journal of Communication, 37, 10-26.

Singer, E., & Endreny, P. M. (1993). Reporting on risk: How the mass media portray accidents, diseases, disasters and other hazards. New York, NY: Russel Sage Foundation.

Slovic, P. (1993). Perceived risk, trust, and democracy. Risk Analysis, 13, 675-682.

Slovic, P. (1999). Trust, emotion, sex, politics, and science: Surveying the risk-assessment battlefield. Risk Analysis, 19, 689-701.

Slovic, P., Fischhoff, B., and Lichtenstein, S. (1982). Facts versus fears: Understanding perceived risk. In D. Kahneman and A. Tversky and P. Slovic (Eds.), Judgment and uncertainty: Heuristics and biases (pp. 463-489). Cambridge: Cambridge University Press.

Stephens, M., & Edison, N. G. (1982). News media coverage of issues during the accident at Three-Mile Island. Journalism Quarterly, 59, 99-204,259.

Thagard, P., and Verbeurgt, K. (1998). Coherence as constraint satisfaction. Cognitive Science, 22, 1-24.

Wahl, S., Frings, S., Hermann, F., Nerb, J., and Spada, H. (2000). So ein Ärger! – Die Rezeption von Zeitungsmeldungen über Umweltprobleme (How annoying! – The reception of newspaper reports about environmental problems). Medienpsychologie, 12, 223-241.

Weiner, B. (1995). Judgements of responsibility: A foundation for a theory of social conduct. New York, NY: Guilford Press.

Zillmann, D., & Brosius, H.-B. (2000). Exemplification in communication: the influence of case reports on the perception of issues. Mahwah, N.J.: Lawrence Erlbaum.

 

---

Figure 1: Graphical illustration of ITERA. Network showing nodes and links representing relationships between nodes. Solid lines are positive or excitatory links; dashed lines are negative or inhibitory links. Input nodes can be set to given (+1), not given (-1), or to unknown (0). More details in text.

Figure 2. Means for the variable responsibility.Ratings were made on a scale ranging from 1(low) to 9 (high).

Table 1: Questions administered to participants after reading the newspaper report.

 

Note: Questionnaire used by Nerb and Spada (2001). Questions are listed in the order in which they appeared in the questionnaire. Endpoints of response scales are in parentheses.

Table 2: Overview of the match between empirical and simulated data from three  Experiments by Nerb and Spada (2001).

Variable	Experiment 1 Controllability	Experiment 2 Higher Goal	Experiment 3 Knowledge
anger	+	+	+
sadness	—	+	—
damage	+	+	+
human Agency	+	+	—
controllability	+a	+	—
higher goal	+	+a	+
knowledge	+	—	+a
boycott	+	+	+
help	—	—	+

 

Note: “+”: empirical  data are in accordance with model predictions.

          “—”: empirical data differ from model.

          ^a variable that was manipulated in respective experiment.

Table 3. Experimental design for  continued influence study.

condition			first report	second report
reference conditions	human negligence		human negligence	neutral information
	act of nature		act of nature	neutral information
misinformation conditions	human negligence		human negligence	neutral information negligence denied
	act of nature		act of nature	neutral information act of nature denied
control/neutral condition			no causal information	neutral information

 

Table 4: Continued influence effects. Empirical results.

 

Note: Variables were measured on nine-point scales, ranging from 1 (low) to 9 (high). Data are mean values (with standard deviations). Significant mean differences in comparison to the neutral control condition as yielded by Tukey´s HSD are indicated as follows: * p < .05; ** p < .01.