On November 29th 2018, Ross Thompson and Heidi Keller were invited to the 2. Wilhelm Wundt Dialog on the topic of “Attachment Theory: Past, Present and Future” at the University of Leipzig. First as speakers and then as discussants, they deliberated about the current state of the field and how it has evolved since its inception in the 1950s. Furthermore, Thompson and Keller reflected on some of the remaining questions and unresolved issues of attachment theory and research, and how the latter could be addressed in the years and decades to come.
The Dialog inspired me to think more about possible future avenues of 21st century attachment theory and research. In this blog post, I would like to suggest that at least some of the remaining questions and unresolved issues could be addressed by embracing a social neuroscience approach – and by fruitfully combining it with already well established methods from the field in multi-method projects. But before elaborating on my suggestion, let us first have a look at some of the remaining questions and unresolved issues of attachment theory and research that were raised during the Dialog.
Remaining questions and unresolved issues
Who qualifies as an attachment figure – conceptual models of attachment. Original attachment theory postulated a strong attachment hierarchy, or even monotropy, in conceptual attachment organization. In the monotropy model, the primary attachment figure (generally the mother) is thought to have an exclusive impact on children’s personality development despite the presence of other attachment figures. Similarly, the hierarchy model still attributes the best predictive value regarding child development to the primary attachment figure (again usually the mother), but acknowledges that relationships with subsidiary attachment figures (like the father) may also contribute, although to a lesser extent. These two initial models have recently been challenged, because the family structure has considerably changed since the 1950s. For example, fathers are nowadays regarded as important attachment figures, and their contribution to child development is increasingly recognized. This development is reflected in the emergence of (at least) two alternative models of conceptual attachment organization. In the independence model, all attachment relationships are assumed equally important for children’s development, but predicted to each contributing in distinct developmental domains. In turn, the integration model suggests that the quality of all attachment relationships taken together as a whole is what optimally predicts children’s developmental outcomes. It is, however, still debated which of the more recent models may more adequately depict conceptual attachment organization, and recent empirical evidence still appears to partially support the hierarchy model as a valid alternative.
How is attachment transmitted from one generation to the next – sensitivity hypothesis. Attachment theory is built upon the assumption that once a certain attachment orientation has been established during early life, this attachment orientation remains relatively stable during childhood, adolescence, and adulthood, and is then transmitted from one generation to the next. As the principle moderating variable of this intergenerational attachment transmission, attachment theory emphasizes parental, and particularly maternal sensitivity. Thus, attachment theory predicts that the strength of the association between the mother’s attachment security (measured by the adult attachment interview) and the child’s attachment security (measured by the strange situation procedure) will be principally mediated by maternal sensitivity. Many studies provide evidence for a role of maternal sensitivity in intergenerational attachment transmission, evidence that is further backed up by several meta-analyses. The persisting problem, however, is that the amount of variance in intergenerational attachment transmission explained by maternal sensitivity alone is relatively modest. It is therefore likely that other variables moderate intergenerational attachment transmission, or the process by which maternal attachment security affects developing security of attachment in the child, and these variables have still to be determined with more confidence. This persistent problem is therefore also referred to as the transmission gap.
Secure attachment is good and insecure attachment is bad – competence hypothesis. Attachment theory also postulates that a secure attachment is prerequisite for healthy social and emotional development, or in other words, necessary to become a competent child, adolescent, and adult. Within this context, competence is equated to good emotion regulation abilities and healthy social functioning. In his original writings, John Bowlby quite strongly implies such competence development in relation to his maternal deprivation hypothesis by saying that “broken attachment leads to delinquency and affectionless psychopathy”. Although it is nowadays still recognized that insecure attachment may represent a risk factor for the development of mental health issues – such as social anxiety or borderline personality disorder –, insecure attachment is not automatically equated with detrimental developmental outcomes anymore. In contrast, the nature versus nurture debate and its investigation using genetic and particularly epigenetic methods has contributed to the view that any attachment, be it secure or insecure, represents a meaningful and thus appropriate adaptation to one’s immediate environment. For example, if a child grows up in a family setting where attachment figures are unavailable and unresponsive in times of need, the emergence of an avoidant attachment orientation in the child represents an adequate developmental adjustment. Furthermore, social defense theory, a more recent derivative of attachment theory, suggests that inter-individual variation in attachment may even have benefits at the group level. The idea is that groups that are composed of people with different attachment orientations will be more effective when dealing with threats and dangers (early detection, rapid response, and effective cooperation) than less heterogeneous groups. Thus, rather than looking at attachment security versus insecurity and the potential implications of this dissociation for emotional and social competence with a black versus white approach, more differentiated consideration is necessary.
Is attachment universal – cross-cultural similarities versus differences. Attachment theory was developed in, and mainly tested on individuals from western, educated, industrialized, rich, and democratic societies – in short, WEIRD cultures. Nonetheless, it claims universality in the sense that attachment should work the same way all around the world. Consequently, measuring attachment anywhere by relying upon, for example, the strange situation procedure, should yield comparable results. Furthermore, if the results from one country do not correspond to “the norm” – usually being mean values of secure versus anxious versus avoidant attachment style distributions from the US and/or Europe and thus WEIRD cultures –, these results are regarded as a deviation. It has been known for quite a while that the core hypotheses of attachment theory, comprising the sensitivity and competence hypotheses mentioned above, are interpreted differently in different cultures, as, for example, very nicely illustrated by Rothbaum and colleagues by a direct comparison between the US versus Japan. There are many implications of cross-cultural differences in attachment on various levels, comprising moral judgments about good and bad parenting. Future attachment research therefore needs more cross-cultural studies and to evaluate cross-cultural differences within the cultural contexts they are observed. In the words of Heidi Keller, “diversity needs to be recognized as the human condition, and the recognition of diversity is an obligation for better science as well as for improving people’s lives”.
Embracing a Social Neuroscience Approach
How can one or more of the above remaining questions and unresolved issues of attachment theory and research be further elucidated during the 21st century? I would like to suggest that one way of doing so is by embracing a social neuroscience approach. I think that investigating the putative biological and neural substrates of inter-individual differences in attachment can be fruitfully combined with already well-established attachment measures in multi-method projects, and that the thereby obtained results are likely to reveal new insights with the potential to inform attachment theory. So far, I have three different avenues in mind.
Understanding the neural substrates of attachment in humans on a single-person level. By relying upon social neuroscience methods such as electroencephalography (EEG), positron emission tomography (PET), and (functional) magnetic resonance imaging ([f]MRI), human brain function can nowadays be investigated with great temporal and/or anatomical detail. In the context of attachment, it therefore is possible to, for example, study how mothers neurally react to images of their own versus unknown children, how adolescents represent positive and negative attributes of themselves versus their best friend, or how adults regulate emotions induced by viewing social versus nonsocial, positive versus negative images and by using either cognitive re-appraisal or expressive suppression. Furthermore, we can nowadays look at brain structure and connectivity either at rest or pre versus post a certain task. Most importantly, all of the above data can be combined with measures of attachment to look at the influence of inter-individual differences in attachment on brain function, anatomy, and connectivity. Derived from findings of my own research and integrating data from other laboratories across the globe, Patrik Vuilleumier and I have proposed a first functional neuro-anatomical model of human attachment in 2012, a model that I am constantly extending and refining. The model postulates that there is no unified brain system devoted to human attachment, but that attachment rather recruits a large network of brain areas also involved in other social and emotional functions. Nonetheless, we specify two core systems either processing information rapidly, in a bottom-up, and sometimes even unconscious manner – termed emotional processing system –, or in a slower, top-down, deliberate, and conscious manner – termed cognitive processing system. We furthermore suggest that the emotional processing system can be divided into a (social) approach versus aversion module, and the cognitive processing system into an emotion regulation versus a mental state representation module. Accumulating evidence suggests that inter-individual differences in attachment security versus insecurity (i.e. anxiety versus avoidance) can be reliably mapped onto brain activation and deactivation, as well as anatomical and connectivity patterns within the above four modules. Most of these data, however, stem from adults and to a lesser degree from adolescents, meaning that data from children is still crucially lacking. Nonetheless, I think that by better understanding how inter-individual differences in attachment affect brain activity, anatomy, and connectivity in single persons across the life span, attachment theory can be further specified. This pertains to all of the above remaining questions and unresolved issues, as such approach will provide fundamental insights into attachment processes on the biological/neural level across domains – for example, in mothers versus fathers, in children of securely versus insecurely attached parents, children from different socioeconomic contexts or with different histories of early life adversities, and individuals from different cultures.
Understanding the neural substrates of attachment in humans on an interpersonal level. Besides measuring brain activity in single participants, recent technological development now allows us to simultaneously measure brain activity in two (or more) individuals. In so doing, a new potential variable empirically describing relationship quality has emerged: brain-to-brain- or inter-brain-synchrony (IBS). According to theory, very close relationships (i.e. parent-child or romantic relationships) should be characterized by high(est) IBS, and IBS should successively decrease as the closeness of the interacting partners decreases (i.e. to friends and strangers). Because relationship quality from an attachment theory perspective not only depends on closeness per se (i.e. how close two individuals are to each other generally speaking), but primarily on how such closeness is individually perceived, we hypothesize that there should be ample variation in IBS even within the above four relationship categories. Furthermore, there should not only be variation as a function of inter-individual differences in attachment in IBS during tasks that necessitate social interaction (i.e. collaboration), but also in the temporal sequence of IBS versus inter-brain-asynchrony (IBA). Such pattern of IBS versus IBA appears particularly relevant for the emergence of attachment at the first place, because children are especially prone to learn about the functioning of their inner and outer environment during times of allostasis deviation through, for example, co-regulation processes with attachment figures. The measurement of IBS and IBA should thus be of ample importance for further elucidating the sensitivity hypothesis, but also clarifying conceptual attachment organization and the competence hypothesis, and revealing potential cross-cultural similarities versus differences on an interpersonal level.
Understanding intergenerational transmission. Attachment orientation emergence is nowadays being understood as a prototypical gene by environment interaction. It therefore appears logical to study this process by using a method that can disentangle the effects of nature versus nurture. A recently developed method offering such opportunity is the analysis of epigenetic DNA modification. As shown in a seminal study in rats that either displayed high versus low maternal behavior (i.e. licking and grooming) towards their pups, maternal behavior in these pups (once they became mothers themselves) could be associated with inter-individual differences in methylation of the glucocorticoid receptor gene (NR3C1) linked to HPA-axis functioning, stress, and anxiety (see also here). Corresponding data in humans has only recently started to emerge, particularly in the context of inter-individual differences in attachment (see also here). Directly testing epigenetic change patterns in children from birth using longitudinal study designs will help determining additional factors important for intergenerational attachment transmission apart from parental (maternal) sensitivity, and likely reveal variance regarding these factors themselves, both within and between cultures.
Conclusions
Attachment theory and research have steadily evolved over the course of the last seventy years. Nonetheless, several questions and issues remain unresolved. I am suggesting that during the 21st century, a combination of well-established attachment measures with new research protocols applying social neuroscience methods promises to yield results that will contribute to important advances of the field. Such approach will require more extensive collaborative efforts not only between different research domains, but also between research groups from different countries and cultural backgrounds. As a 21st century attachment researcher, I am very much looking forward to the next years and decades to come – there are exciting times ahead of us!
What we observed was a specific effect for 
The adolescents’ task in the fMRI scanner consisted of rapidly detecting on which side of a white line there were more dots (see figure, part a). More important, however, was the feedback adolescents obtained after each trial (see figure, part b). The latter feedback was composed of a word (either “won” in case of a correct answer, or “lost” in case of an incorrect answer) and an emotional facial expression (either smiling or angry). While two feedback combinations were congruent (smiling face and word “won,” angry face and word “lost”), two feedback combinations were incongruent (smiling face and word “lost,” angry face and word “won”). This setup allowed us to investigate adolescents’ brain responses while they were integrating objective performance feedback (personal success vs. failure) with the social evaluation of their performance (emotional faces representing social support vs. disapproval). We refer to this integration as social-feedback processing.
