Table 1: Questions after the report.
Click here to get a more recent paper about this line of research.
Nerb, J. & Spada H.
Department of Psychology,
University of Freiburg
D-79085 Freiburg, Germany
{nerb,spada}@psychologie.uni-freiburg.de
Lay people's evaluations of environmental risks seldom rely on probabilities for potential damages. One reason for this neglect simply is the absence of these data--even for experts; another reason is that information about global environmental risks is communicated--printed or on TV--by short, specific, and highly vivid reports consisting of single case descriptions (e.g., accidents and hazards) or of worst case scenarios. This kind of information often leads to negative emotional evaluations of environmental risks. McDaniels, Axelrod and Slovic (1995) found that the emotionality of a risk source almost predicts its acceptance. Though the importance of emotional reactions towards environmental risks is widely acknowledged, the cognitive mechanisms underlying the evaluation process were not investigated sufficiently and the specific emotions that those risks elicit were not differentiated appropriately.
Evaluative judgments about a risk source require the determining of responsibility for the (potential) damage. In line with Weiner (1995), it is claimed that responsibility is inferred from information about the cause of an event. Foremost, a natural--and therefore uncontrollable--cause of an event requires no assignment of responsibility at all. If a human agent caused an event, then he/she is judged to be more responsible the more controllable the cause was. While controllability increases responsibility, mitigating circumstances alleviate, or even totally eliminate assigned responsibility. Mitigating circumstances are given if (a) an act serves a higher goal, or if (b) the agent is lacking knowledge and thus is not able to comprehend the ``wrongfulness'' of the action.
A judgment of responsibility is intertwined with distinct emotional reactions. Based on so called appraisal theories of emotions (e.g., Roseman, 1991), we focus on two emotions that depend on responsibility: anger and sadness. High responsibility judgments should go hand in hand with high feelings of anger and low feelings of sadness and vice versa. Unlike McDaniels et al. (1995) we thus differentiate between (negative) emotions towards environmental risks. For risk research, this distinction is crucial, because different emotions imply different cognitive and behavioral consequences. Roseman and his colleagues (Roseman, Wiest & Swartz, 1994) showed that anger is likely to bring about a ``want to hurt someone''-goal (e.g., boycotting the responsible agent), whereas sadness should lead to a ``want to recover something''-goal (e.g., helping the victim(s), or helping repair the damage).
In this paper, we examine the influence of controllability of a cause on the cognitive and emotional evaluation of risks. Moreover, possible behavioral consequences (boycott, donation) were investigated. To manipulate this variable, we constructed appropriate risk stories in the form of typical news reports about an oil spill caused by a tanker. Combining these ideas about the influence of controllability on judgments of responsibility, distinct emotions, and behavioral consequences, yields two experimental hypotheses: (1) People will judge an agent more (less) responsible for an outcome, if there was a high (low) controllable cause. (2) People confronted with a high controllable cause will react more angry, less sad, and will be more inclined to boycott the responsible agent as when confronted with a low controllable risk. (3) People confronted with a low controllable cause will react more sad, less angry, and will be more inclined to make a donation for repairing the damage as when confronted with a high controllable risk.
Subjects. 79 paid volunteers (35 female, 44 male), mainly students from the University of Freiburg (Germany). Psychology students were not allowed to participate. The median age was 23 (range, 20 to 46). After a brief introduction by the experimenter, all subjects read a newspaper report and then completed a questionnaire.
Material. The material consisted of fictitious, but realistic newspaper reports about an environmental problem: a tanker running aground in a tempest and spilling oil in the North Sea. The material was constructed uniformly: besides reporting the damage (constant part), information about the cause and the circumstances of the cause (manipulation) was described. Three experimental conditions were realized: (1) the tanker did not fulfill the safety guidelines (controllable cause high); (2) the tanker did fulfill the safety guidelines (controllable cause low); (3) no information about safety guidelines was presented (no information; control condition).
Questionnaire. The questionnaire contained 17 items; only twelve items are of interest for the present paper. These items demanded judgments about emotional and cognitive evaluation of the reported case. In addition, a general question asked to what extent a protected value was at risk in the described case. Eleven judgments were made on scales which ranged from 1 (low) to 9 (high); one question about whether or not there was a human cause had a binary response format (Table 1).
Table 1: Questions after the report.
Some General Findings.
Aggregation of data across all experimental conditions revealed that
females showed higher values on the protected values scale than
males (M 
= 6.2, M 
= 5.6,
t(77) = 2.78, p < .007). No other gender differences
reached significance at p < .05. Five subjects rated the
oil spill as not being caused by a human agent and therefore only 74
subjects completed the controllability, higher goal,
knowledge, responsibility, and boycott questions.
All five subjects were in the experimental condition where the
controllability of the cause was low.
Figure 1: The evaluation of the news report in three conditions: (a)
controllability of the cause was low (N=28; N=23 for
variables marked with an asterisk), (b) no
information about controllability of the cause (N=24), and (c)
controllability of the cause was high (N=27). Data represent
averaged values of scales ranging from 1 to 9.
Manipulation Check. The manipulation was successful. An ANOVA of the reported controllability revealed a highly significant effect of the experimental manipulation (F(2,71) = 12.82, p < .0000). Figure 1 shows the mean values for the variable controllability and all other variables.
Effects of Experimental Variations. For all variables, ANOVAs with the between subject variable controllability of cause were computed. As predicted, subjects judged an agent as more (less) responsible for an outcome, if the cause was more (less) controllable. The ANOVA of the responsibility judgments revealed a highly significant effect of the experimental manipulation (F(2,71) = 9.73, p < .0001) in the predicted direction. Furthermore, correlational analyses (computed separately for each experimental condition) of the relation between responsibility and controllability, higher goal, and knowledge confirmed that responsibility is contingent on those variables. The correlation between responsibility and controllability varied between .45 and .56, with a mean of .49; the correlation between responsibility and higher goal was lower and varied between -.03 and -.47, with a mean of -.27; and the correlation between responsibility and knowledge varied between .41 and .59, with a mean of .49.
Subjects confronted with a highly controllable cause also reported significantly higher values of anger (F(2,76) = 7.08, p < .0015) and boycott (F(2,71) = 9.73, p < .0001). There was, however, no effect of the experimental variation on sadness and donation ratings (both Fs(2,76) < 1, n.s.).
The reported harm was not affected by the experimental manipulation (F(2,76) < 1, n.s.). For the variables knowledge, typicality, and protected value the data showed a consistent pattern: in the controllable cause high condition, subjects showed higher values for those variables than in the no information and controllable cause low condition (all Fs(2,71/76) > 3.1, all ps < .05). For the variable higher goal, the experimental manipulation revealed the inverse pattern; but here the differences failed significance (F(2,71) = 1.6, p < .21).
The results clearly indicate that subjects` judgments of responsibility depend on whether a cause of environmental damage is more or less controllable. Confirming Hypothesis 1, high controllability leads to higher responsibility ratings and low controllability leads to lower responsibility ratings with the control condition lying in between. Furthermore, correlational measures of associations between responsibility, controllability, higher goal and knowledge support the validity of the proposed determinants of responsibility.
The study only provides partial support for hypotheses 2 and 3. While data showed that a high controllable cause induces more anger and a higher willingness to boycott the responsible agent than a low controllable cause, the presumed reciprocal effect for sadness and willingness to help did not occur. The lack of support for predictions concerning sadness may be explained by the rather subtle manipulation that was not able to induce differences between the experimental conditions on this variable. It may also be speculated that people integrate in their sadness ratings not only the case at hand but in addition the likelihood that similar events might happen in the future. Since typicality ratings were highest in the condition where the controllability of the cause was high, subjects in this condition might feel increased sadness, because they expect a worse future development than subjects in the condition where the controllability of the cause was low. Thus, the intended effect of the manipulation on sadness ratings may be neutralized. Overall, the differentiated findings for the emotions anger and sadness underline the importance to further investigate the origins and the effects of distinct negative emotions in risk research.
The effect of the experimental manipulation on typicality may be best characterized by the old proverb: ``Bad news travels fast''. Risk communication studies often found that highly negative events (such as damage caused by a highly responsible agent) are more visible than positive or less negative events and thus have more weight in shaping attitudes and opinions--and presumably typicality ratings (compare Slovic, 1997, for this argument) .
Finally, the results for the variables knowledge, higher goal, and protected value showed that these variables were interrelated with responsibility. This may be interpreted as a form of coherent impression making (see, Kunda & Thagard, 1996). Although not manipulated in the experiment, people rated these variables in a way that maximized the coherence of their global impression. Here, a coherent judgment is seen as a matter of building an impression in a way that maximizes compliance with given positive and negative information.
Beside its implications for risk perception, studies like the presented one are crucial to shed light on the empirical validity of appropriate risk-communication and crisis-response strategies as put forward by Coombs (1995).
This research was supported by grant no. Sp 251/10-1 from the German National Research Foundation (DFG) to the second author. We would like to thank Stefan Wichmann, Susanne Frings, and Fabian Hermann for assistance.