Children's causal reasoning: counterfactual thinking occurs for `negative' outcomes only

Harris, German and Mills (Children's use of counterfactual thinking in causal reasoning. Cognition, 61 (1996), 223± 259), following Mackie, argue that children make explicit use of counterfactual thinking in arriving at causal judgments. They showed that children as young as 3, in explaining simple mishap events, made reference to courses of action that a protagonist had rejected, when that course of action would have prevented the observed outcome. It is hypothesized here that such counterfactual thinking might have been invoked by the `negative' mishaps rather than as part of the causal reasoning process. Although the generation of counterfactuals in explanation was replicated using mishap outcomes such as those used by Harris et al., counterfactual thinking was not evident in children's explanations of `positive' outcomes. These results undermine the view that a counterfactual thinking process, as indexed by reference to possible actions rejected by a protagonist, is necessary for causal reasoning. Alternative characterizations of the relationship between causals and counterfactuals are discussed. The nature of the relationship between causal reasoning and counterfactual thinking has been of interest not only to philosophers and psychologists (e.g. Mackie, 1974; Jackson, 1977; Roese & Olson, 1995a; Harris, German & Mills, 1996; Mandel & Lehman, 1996) but also to areas of enquiry as diverse as history, politics and the law (e.g. Hart & Honore , 1959=1985; Fearon, 1991). In a recent paper exploring children's causal reasoning, Harris et al. (1996) advanced evidence in support of Mackie's (1974) contention that causal reasoning depends on counterfactual thinking. Mackie proposed that in reaching the conclusion that event x caused event y we first consider a counterfactual situation where x did not occur and imagine what outcome would follow. If our simulation reveals that y would also not have occurred, then a causal relationship between the two events is inferred. For example, a lighted match goes out after a window is opened and we consider whether the match would have continued to burn had the window not been opened; if our thought experiment reveals that the match would have continued to burn, we are likely to conclude that opening the window caused the match to be extinguished (see Harris et al., 1996, p. 234). Harris et al. (1996, Experiment 3) presented preschool children with stories modeled on stories presented to adult subjects by Wells and Gavanski (1989), in which a protagonist makes a choice between two courses of action. In experimental stories the choice was important in determining the outcome which followed; in control stories it was irrelevant to the outcome. For example, one story involved a woman dying as a result of an allergic reaction to a dish ordered by her boss at a celebratory meal. In the experimental story the boss wavered between two dishes of which one, unknown to him, would provoke the reaction. In the control story either dish would precipitate the reaction. Subjects rated the choice as more causally important in the experimental condition. Wells and Gavanski interpreted this result in terms of subjects bringing to mind counterfactual situations in reaching their causal conclusions. Harris et al. (1996) assessed such counterfactual thinking in children by requiring them to explain similar simplified stories. For example, an experimental story involved Sally choosing some white chocolate for a snack instead of a sandwich and ending up hungry, while the control story involved an irrelevant choice ± Sally chose white chocolate over brown chocolate and, again, ended # Blackwell Publishers Ltd. 1999, 108 Cowley Road, Oxford OX4 1JF, UK and 350 Main Street, Malden, MA 02148, USA. Address for correspondence: Department of Psychology, University of Essex, Wivenhoe Park, Colchester CO4 3SQ, UK; e-mail: [email protected] up hungry. Children were probed with explanation questions (e.g. `why did Sally get all hungry?') and then prevention questions (e.g. `what should Sally have done instead so she wasn't hungry?') for each of six stories (three experimental, three control). Following Wells and Gavanski (1989), Harris et al. looked at whether children's explanations made reference to the point of choice. In particular, they reasoned that if children made reference, in their explanations, to the rejected course of action, then this would show that they considered events that might have happened, but that did not in fact happen (i.e. counterfactual events) to be causally relevant. For the purposes of the study to be reported here, and following Harris et al., reference to actions that were not chosen is assumed to count as counterfactual thinking. Harris et al. found that children's explanations in experimental stories were indeed more likely to make reference to the rejected course of action (e.g., `she was hungry because she didn't pick the sandwich'). For the control stories, children were more likely to generate alternative counterfactual actions that would have prevented the outcome (e.g., `she should have had more food'). This difference was more pronounced for the answers to the prevention question. Harris et al. interpreted these data as evidence that counterfactual thinking plays a key role in causal reasoning. An issue that arises for Harris et al.'s conclusion that children engage in counterfactual thinking in order to reach a causal conclusion is how well the evidence for references to counterfactual states of affairs generalizes to different events. In particular, Harris et al. made exclusive use of scenarios where the outcome was a minor mishap. In an overview of the counterfactual literature, Roese (1997) defends the thesis that counterfactual thinking, though producing consequences that might be regarded as both beneficial and aversive to an individual, has a net beneficial effect for the individual that is functional. Part of Roese's claim is based on evidence that counterfactual thinking is elicited more by negative events than by positive or neutral events. That is, when events go wrong, thoughts of how things might have been better are provoked (e.g. Johnson, 1986; Boninger, Gleicher & Stratham, 1994). Moreover, these so-called upward counterfactuals, where comparison results in an outcome better than that observed in actuality, are generated spontaneously more frequently than downward counterfactuals, which are generated only very rarely (Roese & Olson, 1995b). In view of this evidence, an alternative explanation is possible for the counterfactual thinking observed by Harris et al. (1996). Rather than counterfactual thinking playing a key role in causal reasoning, this explanation pattern may in fact reflect spontaneous upward counterfactual thinking provoked by negative events; children's explanations might be contaminated by issues of prevention rather than cause. Recall that Harris et al. required children to answer explicit questions about prevention, in alternation with their explanation questions. Perhaps interleaving explanation with prevention questions in this way, across the course of six stories, amplified the tendency to invoke counterfactual thinking in explanation. Analysis of children's answers to the explanation question on the first story, which ought to have been uncontaminated by later prevention questions, showed that only six children (just under 20%) made reference to the rejected option (Harris et al., 1996, p. 250). This result is consistent with the view that thoughts of prevention, rather than causal reasoning, invoke counterfactual thinking. Relatedly, Mandel and Lehman (1996) required adult subjects to judge what caused an outcome, or to judge how an outcome might have been prevented or to imagine counterfactual events (i.e. completions of `if only...' prompts). They showed ascriptions of cause typically made reference to different events than did either judgments of preventability or counterfactual judgments, leading them to suggest that causal ascriptions are based on different criteria than are counterfactual and preventability judgments (see also N'gbala & Branscombe, 1995). The prevention hypothesis is tested in the experiment to be presented here; if mention of rejected options in explanation is evidence that counterfactual judgment forms a critical part of causal reasoning, then evidence of such thinking should emerge when children explain all kinds of events, not just mishaps. However, if the mention of rejected options in explanation reflects counterfactual thinking provoked by negative outcomes, we should find no evidence of such counterfactual thought when the events to be explained are not mishaps.