What Doesn't Go Without Saying: Communication, Induction, and Exploration

Abstract

Although prior research on the development of causal reasoning has focused on inferential abilities within the individual child, causal learning often occurs in a social and communicative context. In this paper, we review recent research from our laboratory and look at how linguistic communication may influence children's causal reasoning. First, we present a study suggesting that toddlers only treat spontaneously occurring predictive relationships as if they might support intervention if the events are described with causal language. Second, we show that presenting causal hypotheses as contrastive beliefs, rather than neutrally, improves kindergarteners' ability to provide evidence for their causal inferences. Third, we provide a rational analysis suggesting that stronger inductive inferences are licensed when evidence is presented pedagogically than nonpedagogically; preschoolers are sensitive to this with the consequence that, for better and worse, instruction constrains exploration. In each case study, we discuss the implication that language has a unique role in changing children's interpretation of evidence for causal relationships.

ACKNOWLEDGMENT

The authors thank Elizabeth Bonawitz for feedback on this article. This work was supported by the National Science Foundation, the John Templeton Foundation, and the James S. McDonnell Foundation.

Notes

¹The contrast should be qualified however. Although his writing on the topic is somewhat obscure, Piaget was well aware of the ways in which the individual child's play and the communicative context might interact: “We have to attempt to determine the connection between the imitative image, ludic symbolism and representative intelligence, i.e., between cognitive representation and the representation of imitation and play. This very complex problem is still further complicated by the intervention of language, collective verbal signs coming to interfere with the symbols we have already analyzed, in order to make possible the construction of concepts” (Piaget, 1951).

²This is particularly true given that adults seem to have better source memory when the integrity, rather than merely the identity, of the source is at stake (Rahhal, May, & Hasher, 2002). Interestingly, research suggests that this is true of preschoolers as well. Although preschoolers are generally poor at source memory, they encode the difference between knowledgeable and ignorant informants, and remember this information up to a week later in deciding which informant to trust (Birch, Vauthier, & Bloom, 2004; Corriveau & Harris, 2009; Scofield & Behrend, 2008). Importantly, such studies of selective trust involve conflicting claims about the world. Thus consistent with the current proposal, preschoolers seem most likely to check the source of evidence in the face of contrastive beliefs. We are grateful to an anonymous reviewer for bringing this to our attention.

³The children who failed to produce sufficient evidence showed no distinct pattern of behavior. Of the 12 children who failed the task in the Neutral condition, 2/12 produced only a single demonstration. The remaining children produced two demonstrations: 4/12 varied only the ball, 1/12 varied only the height, and 5/12 changed both variables between demonstrations.

⁴Mercier & Sperber (in press) have recently turned this claim on its head. Rather than assuming that arguments from evidence support accurate reasoning, they have suggested that explicit reasoning, in the form of arguments from evidence, serves as a costly signal — advertising to conversational partners the contestants' epistemic reliability. We think this is an interesting possibility. Note however, that if explicit reasoning did not actually improve epistemic reliability it is not clear that it would function effectively as a reliable signal.