Establishing Cause-and-Effect Relations
Like parenting research in general, studies of parental talk and theory of mind are correlational, in that both constructs are measured rather than experimentally controlled. Because such research is correlational, it cannot establish cause- and-effect relations with certainty. As with parenting research in general, any significant parent-child correlation has three possible explanations. One is that the causal locus is with the parent—that parental talk contributes to children’s theory-of-mind development. A second possibility is that the causal locus is with the child—that children with relatively advanced theory-of-mind skills elicit mental state talk from their parents. Finally, the third possibility is that no causal relation exists at all—that some third factor or set of factors is responsible for both outcomes.
The ways to attempt to pull apart these possibilities were discussed in Chapter 3. The main way to address the direction-of-effect issue is through longitudinal study. If parental talk is the causal factor in the correlation, then variations in talk at time 1 should relate to variations in theory of mind at time 2; the reverse, theory of mind to talk direction should be either weaker or nonexistent. As we will see, longitudinal studies are frequent in this literature, and most report the pattern just described: namely, that early talk relates to later theory of mind.
As P. L. Harris (2006) notes, another way to show that variations in how parents talk are not determined solely by variations in their children’s level of development is to measure parental talk independently of the child’s presence. The last two methods described in Table 7.1 do precisely this. Only a handful of studies have adopted these methods, and thus the evidence is limited. Research to date, however, suggests that the same variations in parental talk and the same effects of those variations occur when the talk that is measured is independent of the child’s immediate presence.
The main way to rule out third factor explanations is to control for potentially important third factors, either in advance through selection ofparticipants or after the fact through statistical controls. Such controls are common in the research to be reviewed. At the child level, age and language ability are the variables that are most often controlled; at the adult level, education is the variable that is most often controlled. As we will see, the great majority of parent-child relations remain significant even when third factor alternatives are ruled out.
The discussion of methods in Chapter 3 identified another way to address both the directionality and third factor issues: namely, experimental manipulation of the hypothesized causal variable. A small such literature now exists, for providing extra experience with mental state terms is one way in which researchers have attempted to train theory-of-mind understanding. I consider this research in Chapter 9.