Methodological Issues

Measuring Parental Talk

Measuring how parents talk to their children requires two methodological decisions. The first concerns the context within which to record parental talk. The second concerns the aspects of parental talk on which to concentrate.

Table 7.1 provides an overview of the possibilities with respect to the first decision. The entries in the table reflect a familiar divide that applies to the study of many topics: whether to carry out the measurements in the natural setting, which in this case is the home (the first entry in the Table 7.1), or whether to measure the outcomes of interest in some specially constructed laboratory environment (the remaining entries).

Observations in the home account for close to a third of the relevant studies. Among other challenges, such studies face the fact that mental state utterances do not occur very frequently in most parent-child conversations. One study, for example (Jenkins, Turrell, Kogushi, Lollis, & Ross, 2003), reported that mental state terms occurred at a rate of 2.79 per hour in talk with 3-year-olds and 5.13 per

Table 7.1 Methods for Collecting Samples of Parent Talk

hour in talk with 4-year-olds. Although such utterances become more common as children grow older, the base rate remains low.

Researchers have employed various strategies in an attempt to maximize the yield from home observations. It is common to schedule visits at times when family talk is most likely; the time period surrounding and including family meals is a frequent focus. It is also common to request that family members temporarily forgo activities (e.g., TV, video games) that may cut down on the possibilities for conversation. Finally, the main strategy is one that is common to naturalistic observational research in general: namely, to sample as broadly as possible. In one (admittedly exceptional) example, parents and children were observed across six 90 minute home visits (Jenkins et al., 2003).

One of the virtues of laboratory study is that situations can be contrived to maximize the occurrence of the outcomes of interest, and thus the yield per time expended is typically greater than is true with naturalistic observation. The most popular method of eliciting parental talk in the laboratory has been the storybook approach, in which talk is recorded as the parent and child look together at a storybook. In some instances commercial storybooks have been used, including books with written text (e.g., Ensor, Devine, Marks, & Hughes, 2014). More common, however, is the use of wordless picture books constructed for the study. The absence of a written text clearly maximizes the opportunity for the parent to provide her own interpretation of the story, including references to the mental states of the characters (and many books are designed to encourage such references). The storybook approach has been the most common method of measuring parental talk; indeed, this approach accounts for close to half of the studies to be reviewed.

Given the challenges of any of the possible forms of measurement, it is perhaps understandable that most studies have included only a single method of measuring parental talk. There are, however, some exceptions (e.g., Beeghly, Bretherton, & Mervis, 1986; Ensor et al., 2014; Howe, Rinaldi, & Recchia, 2010; Kucirkova & Tompkins, 2014; Laible & Song, 2006). As we will see, these studies do show some variations in outcomes across different methods of assessment. To date, however, no study has provided a clear test of whether the effects of parental talk vary across the two most often used forms of measurement: home observation and storybook reading. It has been argued, based on across-study comparisons, that the more extensive data from home observations make possible more precise conclusions about links between parental talk and theory of mind than does the laboratory approach—specifically, not just effects of general mental state talk but differential effects of talk about different mental states (Hughes, 2011). As we will see, however, many applications of the storybook approach also identify specific and not just general links between parental talk and theory of mind. On my reading, the various methods of study are a good deal more similar than different in the conclusions they provide.

Identifying differential effects of talk about different mental states requires that several mental states be examined in the same study. Not all studies provide such comparative data. As I have already noted, many studies focus solely on emotion talk. A smaller number consider cognitive states only (e.g., Furrow, Moore, Davidge, & Chiasson, 1992; Howard, Mayeux, & Naigles, 2008).

When multiple states are included, the most frequently examined are those shown in Table 7.2. Of the various categories, perception is the least often studied. Perception was included, however, in the infancy studies discussed in Chapter 6 (Roberts et al., 2013; Slaughter et al. 2008, 2009), and it appears as well in some studies of postinfant development (e.g., Adrian, Clemente, Villanueva, & Rieffe, 2005). Other categories not listed in the table (e.g., preferences, goals) are found

Table 7.2 Categories and Examples ofMental State Words

in a handful of studies, although in at least some instances the category label may be just a different term for a concept typically addressed under another heading (e.g., “volition” instead of “desire”).

As the examples in the table suggest, assigning terms to particular categories is generally straightforward. It is not always straightforward, however. Symons (2004) offers one example. Depending on the context, “dream” or “dreamy” might refer to a cognitive state, a desire, or an emotion. Similarly, the phrase “was hurt” might refer to an emotional reaction (and thus would need to be included) or a physical injury (and thus would need to be ignored). Because of such complexities, all studies report interobserver reliability, and the reliabilities are generally quite good.

In addition to deciding about particular items, researchers must make two general decisions about their coding system. One is whether to apply so-called exclusionary rules. This issue arises most obviously in the coding of cognitive terms. In some instances a cognitive term seems to function more as a conversational device than as a genuine mental state reference (e.g., “You know?” “Guess what”). Some researchers (e.g., Ensor & Hughes, 2008) exclude such conversational uses from their cognitive tally; others (e.g., Jenkins et al., 2003) do not. This decision may have some effect on conclusions about children’s mental state understanding, especially young children (Brown, Donelan-McCall, & Dunn, 1996). There is no evidence that it affects conclusions about parental talk.

The second decision about coding concerns whether to correct for verbosity. Some parents talk more than others, and they may produce more mental state words simply because they are producing more words in general. For this reason, some researchers (e.g., Slaughter et al., 2008) use proportion scores rather than totals in their analyses. Others (e.g., Ruffman, Slade, & Crowe, 2002) work with raw totals, the argument being that more exposure means more information about mental states, even if the exposure comes in the midst of a good deal of less relevant input. As with exclusionary rules, this decision may affect the descriptive information from a study, but there is no evidence that it affects conclusions about the effects of parental talk.

A final point about coding is a more substantive one. In many studies, tabulation of mental state words is the only goal of the observational system, and the total number of such words is the only aspect of parental talk examined in the analyses. Some researchers, however, have gone beyond a simple tally of words in an attempt to identify variations in the quality of the linguistic input that children receive. The various ways of defining quality, as well as possible effects of differences in the quality of parental talk, will be an important theme in the review of studies.