Balancing Causes and Consequences: The Magnitude-Matching Principle in Explanations for Complex Social Events

Abstract

We assessed whether people use a magnitude-matching principle in determining causes for complex social events. We hypothesized that individuals tend to favor causal explanations that match the event in terms of size and scope. In Experiment 1, the magnitude of the consequences of events was manipulated, and participants were presented with two potential causes of modest magnitude and two potential causes of high magnitude. Analyses revealed a relative magnitude-matching effect such that participants were more likely to select high magnitude causes for large magnitude events than modest magnitude events and more likely to select modest magnitude causes for modest magnitude events than large magnitude events. Experiment 2 replicated the magnitude-matching effect with a different event and set of causes, and demonstrated that this effect could be reversed by undermining participants' beliefs in the magnitude-matching principle.

ACKNOWLEDGMENTS

Research was funded by grants from the Canadian Institute of Health Research and the Social Science and Humanities Research Council issued to the second and third authors.

Notes

¹A separate pretesting sample of 24 participants was randomly assigned to either the high or low conditions of the predictability/control manipulation. These participants subsequently completed a six-item scale assessing the extent to which they believed that their life circumstances were predictable and controllable. An independent samples t test confirmed that the manipulation was successful in altering beliefs, t(22) = 1.88, p = .04 (one-tailed). The high control condition produced perceptions of greater control/predictability (M = 5.03 on a 7-point scale) than the low-control condition (M = 4.37). The Cohen's d effect size for the manipulation was .86, thereby indicating a large effect (see Cohen, 1988).

²To confirm that the generally symmetrical magnitude-matching effects that we obtained were not artifacts associated with the format of our dependent measures, we computed the correlations between the ratings of each of the possible causes of an event that participants were instructed to evaluate. If people felt constrained to make ratings contingent upon one another, strong negative correlations among ratings should have emerged. However, the four causal ratings were virtually uncorrelated with each other (average absolute value of rs was .09), which indicated that once the participants endorsed one cause, this judgment did not influence endorsement of other causes.

³A separate group of 180 participants were assigned to read either the high- or the moderate-magnitude version of one of the two scenarios. Participants then used 9-point scales to rate the severity and the significance of the event that was described. Subsequent 2 (magnitude of event: high vs. moderate) × 2 (scenario: airline accident vs. contagious illness) between-subjects ANOVAs suggested that our manipulation of magnitude had been successful. High-magnitude events were rated as being more severe (M = 8.07, SD = 1.33) than moderate-magnitude events (M = 6.29, SD = 1.33), F(1, 172) = 80.01, p < .001, and high-magnitude events were rated as being more significant (M = 7.02, SD = 1.65) than moderate-magnitude events (M = 5.57, SD = 1.71), F(1, 172) = 33.58, p < .001.

Note. Degrees of freedom are (1, 110) for all analyses of variance (ANOVAs). Big 1 scenarios were terrorist bomb (airline accident) and biological warfare (contagious illness); Small 1 scenarios were cargo mislabeling (airline accident) and illness from Asia (contagious illness); Small 2 scenarios were electrical malfunction (airline accident) and standard bacterial infection (contagious illness); Big 2 scenarios were poor maintenance policies (airline accident) and superbacteria (contagious illness).

⁴Of importance, keeping the outcome of the scenario used in Experiment 2 constant across conditions also eliminated the possibility that valence might be confounded with magnitude.

⁵The belief-induction procedure that was employed in Experiment 2 also acted as a safeguard against potential confounds. Namely, any confounds inherent in the design of Experiment 2 would not be susceptible to a manipulation that specifically targeted people's beliefs about magnitude matching. Thus, we reasoned that if this belief-induction had an impact on the emergence of magnitude-matching effects, it would constitute further evidence that these effects were indeed based on magnitude-matching beliefs.

⁶We chose to use an actual historical event (the assassination of American President Garfield) as our focal incident in Study 2. As such, one might wonder whether participants were familiar with this event. This possibility was unlikely for several reasons. First, when we questioned a subset of our participants, none of them (all of whom were Canadian undergraduate students) appeared to be aware of his existence (let alone the details of his death). Furthermore, in the unlikely event that a few of our participants might have been knowledgeable about President Garfield and the circumstances surrounding his assassination, random assignment should have ensured that these participants were distributed equally across conditions and the existence of these participants should, if anything, have attenuated the impact of our manipulations.

⁷As in Experiment 1, correlations between participants' ratings of the four causes were computed to ensure that they were relatively independent of each other. The average absolute value of rs was .17, which strongly suggests that this was the case.

Note. Degrees of freedom are (1, 119) for all analyses of variance. For the presidential assassination scenario, Big 1 cause was gunman hired by vice president; Small 1 cause was single, deranged gunman; Small 2 cause was another official was the intended victim; and Big 2 cause was grand conspiracy by confederates.

Note. For the presidential assassination scenario, Big 1 cause was gunman hired by vice president; Small 1 cause was single, deranged gunman; Small 2 cause was another official was the intended victim; and Big 2 cause was grand conspiracy by confederates.

⁸It is important to note that the emergence of the two-way interaction in Experiment 2 also provides strong evidence against the interpretation that the magnitude-matching effect was produced by some other feature of causes and events that was unintentionally confounded with magnitude. Had this alternative been true, the magnitude-mismatching belief induction manipulation should not have altered preferences for the explanations.