It was not long into the theory of mind era before researchers began to speculate about possible relations between attachment and theory of mind (Bretherton, Bates, Benigni, Camaioni, & Volterra, 1979; Fonagy, Steele, Steele, Moran, & Higgitt, 1991; Main, 1991). It was not until 1997, however, that the first study was reported (although see Fonagy, 1996, for an earlier, partial report). Fonagy et al. (1997) used the Separation Anxiety Test to assess differences in attachment in a sample of 3- to 6-year-olds. The theory-of-mind outcome of interest was the emotion-based form of false belief described in Chapter 2—that is, the ability to judge an inappropriate emotion that followed from a false belief. The two measures proved to be related: Children with secure attachments outperformed those whose attachments were less secure, a difference that remained even when chronological age and language ability were controlled. The effect, moreover, was a substantial one: 71% correct responses by the former group, as opposed to 20% for the latter group.
To date, the affective false belief task has appeared in just two other studies in the attachment literature (it was also employed, but dropped because of poor performance, by Repacholi & Trapolini, 2004, and employed, but not analyzed separately, by Ontai & Thompson, 2008). The results are mixed. Like Fonagy et al. (1997), de Rosnay and Harris (2002) used the Separation Anxiety Test to assess attachment in a sample of preschoolers. Their measure of affective false belief understanding included both a typical such story and an attachment-theme story, in which a child was separated from his or her mother. The expectation was that response to the attachment-theme story would relate more strongly to attachment than would response to the typical story. This did not prove to be the case. The basic finding, however, replicated that of Fonagy et al.: a positive relation between security of attachment and affective false belief understanding.
The second study, that of Meins, Fernyhough, Russell, and Clark-Carter (1998), was longitudinal, and I therefore consider it more fully in the next section. As we will see there, security of attachment at age 1, as assessed by the Strange Situation, related to a number of later theory-of-mind outcomes. The only exception was the affective false belief task (although there was a nonsignificant trend toward better performance by the securely attached group).
The affective false belief task taps into both false belief understanding and emotion understanding. As noted earlier, standard forms of these two developments account for most of the relevant literature.
The picture is perhaps somewhat clearer for emotion understanding than for false belief (which, in fact, is the conclusion that Hughes, 2011, reached from the research available at the time of her review). Of the six concurrent examinations of attachment and emotion understanding, five have reported significant relations, that is, better performance by securely attached children (Barone & Lionetti, 2012; Greig & Howe, 2001; Laible & Thompson, 1998; Ontai & Thompson, 2002, Study 2; Repacholi & Trapolini, 2004). The successful demonstrations span a range of attachment measures, including both behavioral (the Attachment Q-Set) and representational (Attachment Story Completion, Manchester Child Attachment Story Task, Separation Anxiety Test) forms. The measures of emotion understanding also span a range of forms, although the most commonly used are the tests developed by Denham (1986; Denham & Auerbach, 1995).
One of the reports with positive results is also the source for the only failure to find a relation between attachment and emotion understanding. Ontai and Thompson’s (2002) Study 1 examined the issue with a sample of 3-year-olds and reported no relation between the measures. The same children and the same measures were reexamined 2 years later, and in this case a significant relation emerged. The link between attachment and understanding of negative emotions was especially strong, something that had also been true in the earlier study by Laible and Thompson (1998). In discussing the reason for the age difference in outcomes, the authors emphasize the child’s developmental level and hence readiness to benefit from potentially helpful experiences. In their words, “the benefits for socioemotional understanding of the more sensitive maternal responsiveness to emotional issues documented by attachment theorists await the greater comprehension of others’ inner, psychological states that develops most significantly after age 3” (p. 447).
Findings are a bit less consistent for false belief than they are for emotion understanding. Three studies have reported significant concurrent relations between attachment and false belief understanding (Arranz et al., 2002; Repacholi & Trapolini, 2004; Symons & Clark, 2000), but two other studies failed to find a relation (Greig & Howe, 2001; Ontai & Thompson, 2008). A report by Meins
(2012) offers findings in both categories: no relation with Attachment Story Completion as the measure of attachment, a positive relation with the Separation Anxiety Test as the measure. There is no obvious difference that divides the studies with positive and negative outcomes. The false belief measures are in every case standard ones, and the attachment measures fall on each side of the positivenegative divide: one positive and one negative outcome with the Attachment Q- Set, one positive and one negative outcome with Attachment Story Completion.
The Repacholi and Trapolini (2004) study adds an interesting wrinkle to the assessment. Ordinarily, the target that must be judged in theory-of-mind research is a generic one—for example, Maxi in the original study of false belief introduced in Chapter 2. Repacholi and Trapolini presented a generic target on half of their false belief and emotion understanding trials; on the other half, however, the child’s mother served as the target. The expectation was that any effects of attachment on performance might be especially pronounced when the attachment object was the target for the child’s reasoning efforts. And such proved to be the case, although only for the false belief trials and not the emotion understanding trials. Children whose attachments fell in the avoidant category (one of the dimensions that can be scored on the Attachment-Q-Set) had special difficulty in judging their own mother’s false beliefs. In explaining this outcome, Repacholi and Trapolini evoke Bowlby’s (1969) concept of a working model. They suggest that children with avoidant attachments have developed negative working models with respect to the mother and the mother’s relationship to the self, and that these working models bias their ability to attend to and reason about the mother’s beliefs.
The Repacholi and Trapolini (2004) study is not the only attempt to explore possible effects of a familiar target on children’s theory-of-mind reasoning; it is, however, the only one to date to show clear effects of attachment. I have already mentioned Symons and Clark’s (2000) use of a caregiver form of the false belief task. The caregiver task was associated with several significant effects—none, however, that had to do with the child’s attachment status. Clark and Symons
- (2009) elicited attributions for behavior for three targets: parent, teacher, and peer. Various effects were associated with target, but none showed any relation to attachment. Ontai and Thompson (2008) used the mother as target on half of their false belief and emotion understanding trials; their analyses, however, were based on composite scores that did not distinguish between targets, and thus any possible effects went unexplored. Finally, Laranjo, Bernier, Meins, and Carlson
- (2010) used the mother as the target for Level 1 perspective taking in 2-year-olds, the rationale being that expected effects of the variables of interest (maternal mind-mindedness as well as security of attachment) would.be greatest with the attachment object as the target. But because no other target was included for comparison, it was not possible to determine whether use of the mother in fact made a difference.
As noted, the target effects in Clark and Symons’s (2009) study of attributions did not vary as a function of attachment status. Attributions in general, however, did relate to attachment: Securely attached children offered more positive attributions than did less securely attached children. Securely attached children also offered more positive self-evaluations than did less securely attached children.
Two other studies provide evidence about possible effects beyond the preschool period. In Humfress et al. (2002), response to the Strange Stories measure by 12 - and 13-year-olds related positively to a concurrent measure of attachment security. Controlling for verbal IQreduced but did not eliminate the relation. In Vanwoerden, Kalpakci, and Sharp (2015), adolescents’ responses to an advanced measure of theory of mind related positively to attachment status—only, however, for girls. Girls with disorganized attachments performed more poorly than did those with other classifications.
As the descriptions in the preceding paragraphs indicate, false belief and emotion understanding are not the only outcomes that are represented in the attachment and theory of mind literature. Two other dependent variables appear in several studies. One is the child’s use of mental state language in conversation— either mental state words in general or emotion words in particular. Two studies under the Concurrent Relations heading provide evidence on this point. In a study by McQuaid and colleagues (McQuaid, Bigelow, McLaughlin, & MacLean, 2008), attachment security was positively related to preschoolers’ use of both mental state terms in general and emotion terms in particular. Bost and colleagues (Bost et al., 2006) also found a significant relation between attachment security and use of emotion words (the only aspect of language assessed); when the mother’s own use of emotion terms was controlled for, however, the relation to attachment disappeared.
The other dependent variable is joint attention in infancy. Joint attention (as I discuss more fully in Chapter 6) is the ability to share the attentional focus of others, an ability that develops gradually across the first year or so of life. Relations between joint attention and attachment are hardly robust in research to date, but some relation has emerged in three of the four examinations of the issue (the exception being a study by Yoon, Kelso, Lock, & Lyons-Ruth, 2014). In two studies (Claussen, Mundy, Mallik, & Willoughby, 2002; Scholmerich, Lamb, Leyendecker, & Fracasso, 1997) the disorganized pattern of attachment, as assessed by the Strange Situation, was associated with relatively poor joint attention, in one case (Claussen et al.) in interaction with the experimenter and in one case (Scholmerich et al.) in interaction with the mother. In a third study (Meins et al., 2011) it was the avoidant pattern that diverged from the other classifications. Compared to the other attachment groups, infants with avoidant attachment showed both more joint attention with the experimenter and less joint attention, especially of higher level forms (i.e., pointing and showing), with the mother. The lower response to the mother was interpreted as evidence that the avoidant response shown under conditions of stress (which is what defines the attachment relationship) is a more general characteristic of the infant-mother relationship. The greater response to the experimenter was interpreted as evidence that infants who tend to avoid the mother compensate by increased responding to other social partners.