Assessing attachment, a work in progress: to look, to listen or both?

The title chosen for this discursive editorial in the journal aims to reflect the two main approaches that have been used in the assessment of ‘attachment’, one involving behavioural observation, and the other utilising self-report methodology.

To look

Ainsworth’s Strange Situation epitomises the observational approach (Ainsworth, 1964). It involves a separation–reunion paradigm in which the infant’s attachment is classified into one of several categories based on its response to reunion with (almost invariably) the mother. The strengths of this approach are substantial. First, the infant is unlikely to be influenced by the demand characteristics of the situation, although the responses of the mother might well be. Second, the procedure is standardised, and can utilise a number of independent raters to determine the categorisation. Third, considerable research has been conducted on both the causes and consequences of these infant attachment styles, providing some support for construct validity. However, the weaknesses of this approach are also substantial. First, it is a resource-intensive assessment whose time and cost often limit its application to one-off assessments in relatively small samples. Second, the commitment required of participants risks introducing sample bias. Third, a number of extraneous factors potentially influence the infant’s behaviour (tiredness, hunger, minor illness, etc.). Fourth, the focus is on the infant’s behaviour, at the expense of that of the mother. Fifth, it often involves causing distress to the infant.

To listen

Our group’s work over the last three decades epitomises the alternative approach (Condon & Corkindale, 1998; Condon, Corkindale, & Boyce, 2008). We have sought to identify parents’ subjective cognitive and emotional experiences in relation to their offspring, and to quantify these experiences in terms of frequency, intensity or both. This work began in the early 1980s, in the context of providing grief therapy to men and women who had suffered perinatal bereavement, and subsequently to mothers who had relinquished their neonate (often unseen) for adoption. That almost all these women and men had developed, during pregnancy, a profound ‘love’ for their unborn baby was undeniable, as was the agonising quality of the grief they experienced as a result of the loss of their love object. The notion that love develops to the unborn child during pregnancy began to emerge in the literature as early as the 1940s in the writings of Helena Deutsch (1944) and Donald Winnicot (1958), and empirical explorations began in the late 1970s. Our work has led us to conclude that both parent-to-foetal attachment and parent-to-infant attachment involve a core experience of ‘love’, which language limitations render difficult to articulate. Vaillant (1985) noted that adult humans require the use of metaphor to express their love experiences.

We postulate that this core experience gives rise to a number of needs, desires or dispositions which can be more readily articulated, but may or may not find expression in overt behaviour. Additionally, ‘love’ gives rise to a range of cognitive and emotional experiences which can also be articulated. Antenatal examples could include the desire ‘to know’ and understand the foetus, the need to protect it, the disposition to gratify its needs, the cognitions and emotions associated with fantasised loss, etc. Postnatal examples could include the disposition to gratify the infant’s needs and give it pleasure, the desire to understand the infant’s inner world and parents’ emotions and cognitions associated with interactions, separations and reunions.

The philosopher Kaplan (1946) discussed at length the notion of a ‘scientific construct’. Within his framework, these dispositions would be regarded as ‘indicators of the probable presence’ of attachment. Kaplan believed that for a construct to have meaning, there must be relatedness and linkage between its constituents. Such linkages constitute the essential essence of ‘construct’, i.e. the construct represents more than just the sum of its parts. Thus, our self-report instruments attempt to assess the indicators of the presence of this construct. Many attachment theorists were, and remain, uncomfortable with equating this construct with ‘attachment’. However, in support of our conclusion I will simply quote several excerpts from four of the principal founders of attachment theory, which highlight the subjective, experiential dimension of attachment.

John Bowlby (1971) wrote:

No form of behaviour is accompanied by a stronger feeling than attachment behaviour. The figures towards whom it is directed are loved, and their advent is greeted with joy.

Mary Ainsworth (1969), who in this same paper suggested that attachment was ‘synonymous with love’, wrote:

Attachment is something inside the organism and is distinct from attachment behaviours which may be increased or decreased without any implication that the attachment per se has altered in intensity.

Robert Hinde (1976), whom Bowlby acknowledged as the source of many of his own ideas, wrote:

Objective and behavioural data can be misleading if devoid of meaning ... and the quickest (and sometimes only) way to meaning may be through the use of introspective evidence. The student of interpersonal relationships must thus walk along a knife-edge: objective criteria are essential for purposes of description and communication, but this need must not lead to neglect of the complexity and intersubjectivity inherent in relationships

And in 1978 wrote

To describe a relationship fully we must describe also the affective and cognitive components which accompany and transcend the behavioural ones.

Marris (1982) concluded that:

The ability to make sense of experience depends on connecting feeling to action by way of purpose, the acknowledgement and placing of emotion is therefore fundamental.

If we choose to access the subjective experiences of parents, both antenatally and postnatally, in relation to their offspring, we need to ask several questions. What are our interviews or self-report questionnaires actually assessing? Is this entity worthy of study?

In the psychological and social sciences, almost all the entities with which we deal are constructs. In contrast to the biological sciences, these constructs have no physical existence. They are abstractions about whose nature and usefulness we have achieved some degree of consensus, but which we cannot directly measure. We must infer their presence from various ‘indicators’ of that presence. Our constructs are ‘scientific’ to the extent that we can detect their indicators in a reliable manner, and to the extent to which these are valid indicators of that construct. The American statistician Judd Nunnally (1959, 1978) has addressed these issues, from the viewpoint of construct validity, with outstanding insight and lucidity in a number of his books.

‘Attachment’ is a construct, whose nature and measurement, I would suggest, remain a work in progress, so it may be timely to re-examine some fundamental issues. Does attachment reside in an individual or can it only be construed as a product of the interaction between two individuals? Is the subjective feeling state of a parent towards their child (born or unborn) grounded in some kind of relationship (reality-based or fantasised or idealised), or does it reflect a personality domain akin to a ‘capacity to love’ which may, in turn, be a legacy of the parenting they themselves received, i.e. their internal working model of attachment? To what extent, if any, does this subjective feeling state towards the infant predict a parent’s style of relating to their infant in terms of sensitivity, responsivity, warmth, etc. These are difficult questions, but we should not avoid trying to address them.

The resources required ‘to listen’ to parents are not onerous, facilitating the use of larger, less-biased samples. Well-established techniques of quantitative research can be used to refine our measuring instruments. For example, Nunnally’s approach to item analysis can maximise internal consistency; psychometric properties such as reliability and validity can be evaluated; techniques such as factor analysis can explore the underlying dimensions of our construct, etc. However, akin to behavioural approaches, there are limitations to this methodology. For example, parents’ may be reluctant to acknowledge less-positive experiences towards their infant which are deemed socially unacceptable.

To both look and listen

Whether ‘to look’ versus ‘to listen’ appears to have created a split in parent/infant attachment research, namely: infant observation versus parental self-report. After half a century of attachment research, is it really necessary for us to water down a parent’s subjective feeling state of attachment to their infant by referring to it by some euphemism such as ‘bonding’, in order to preserve the purity of ‘attachment’ exclusively for the infant side of the relationship? Perhaps our efforts over future decades could be more productively directed towards healing this split, restoring a balance, and correcting the under-utilisation of the ‘listen’ approach. Both approaches have much to contribute to research and our understanding of the fascinating complexities of human fathering and mothering from conception throughout childhood. This understanding can, in turn, potentially contribute to useful clinical approaches and insights.