www.psychspace.com心理学空间网
The attachment system as a bio-social mechanism of homeostatic regulation
Attachment theory, developed by John Bowlby (Bowlby, 1969; Bowlby, 1973; Bowlby, 1980), postulates a universal human need to form close affectional bonds. At its core is the reciprocity of early relationships, which is a precondition of normal development probably in all mammals, including humans (Hofer, 1995). The attachment behaviours of the human infant (e.g. proximity seeking, smiling, clinging) are reciprocated by adult attachment behaviours (touching, holding, soothing) and these responses strengthen the attachment behaviour of the infant toward that particular adult. The activation of attachment behaviours depends on the infant’s evaluation of a range of environmental signals which results in the subjective experience of security or insecurity. The experience of security is the goal of the attachment system, which is thus first and foremost a regulator of emotional experience (Sroufe, 1996). In this sense it lies at the heart of many forms of mental disorder and the entire psychotherapeutic enterprise.
None of us is born with the capacity to regulate our own emotional reactions. A dyadic regulatory system evolves where the infants’ signals of moment to moment changes in their state that are understood and responded to by the caregiver thereby achieving their regulation. The infant learns that arousal in the presence of the caregiver will not lead to disorganisation beyond his coping capabilities. The caregiver will be there to re-establish equilibrium. In states of uncontrollable arousal, the infant will come to seek physical proximity to the caregiver in the hope of soothing and the recovery of homeostasis. The infant’s behaviour by the end of the first year is purposeful, and apparently based on specific expectations. His past experiences with the caregiver are aggregated into representational systems which Bowlby (1973) termed ‘internal working models’. Thus, the attachment system is an open bio-social homeostatic regulatory system.
Patterns of attachment in infancy
The second great pioneer of attachment theory, Mary Ainsworth (1969; 1985; Ainsworth, Blehar, Waters, & Wall, 1978), developed the well-known laboratory based procedure for observing infant’s internal working models in action. Infants, briefly separated from their caregiver in a situation unfamiliar to them, show one of four patterns of behaviour. Infants classified as Secure explore readily in the presence of the primary caregiver, are anxious in the presence of the stranger and avoid her, are distressed by their caregivers’ brief absence, rapidly seek contact with the caregiver afterwards, and are reassured by this. The infant returns to exploration. Some infants, who appear to be made less anxious by separation, may not seek proximity with the caregiver following separation, and may not prefer the caregiver over the stranger; these infants are designated ‘Anxious/Avoidant’. A third category, ‘Anxious/Resistant’ infants show limited exploration and play, tend to be highly distressed by the separation, but have great difficulty in settling afterwards, showing struggling, stiffness, continued crying, or fuss in a passive way. The caregiver’s presence or attempts at comforting fail to reassure, and the infant’s anxiety and anger appear to prevent them from deriving comfort from proximity.
Secure infants’ behaviour is based on the experience of well co-ordinated, sensitive interactions where the caregiver is rarely over-arousing and is able to restabilise the child’s disorganising emotional responses. Therefore, they remain relatively organised in stressful situations. Negative emotions feel less threatening, and can be experienced as meaningful and communicative (Grossman, Grossmann, & Schwan, 1986; Sroufe, 1979; Sroufe, 1996).
Anxious/Avoidantly attached children are presumed to have had experiences where their emotional arousal was not restabilised by the caregiver, or where they were over aroused through intrusive parenting; therefore they over-regulatetheir affect and avoid situations that are likely to be distressing. Anxious/Resistantly attached children under-regulate, heightening their expression of distress possibly in an effort to elicit the expectable response of the caregiver. There is a low threshold for threat, and the child becomes preoccupied with having contact with the caregiver, but frustrated even when it is available (Sroufe, 1996).
A fourth group of infants exhibits seemingly undirected behaviour, giving the impression of disorganisation and disorientation (Main & Solomon, 1990). Infants who manifest freezing, hand clapping, head-banging, the wish to escape the situation even in the presence of the caregiver, are referred to as ‘Disorganised/Disoriented’. It is generally held that for such infants the caregiver has served as a source of both fear and reassurance, thus arousal of the attachment behavioural system produces strong conflicting motivations. Not surprisingly, a history of severe neglect or physical or sexual abuse is often associated with this pattern (Cicchetti & Beeghly, 1987; Main & Hesse, 1990). I would like to consider this group in much greater detail, this afternoon.
The continuity of patterns of attachment
Bowlby proposed that internal working models of the self and others provide prototypes for all later relationships. Such models are relatively stable across the lifespan (Collins & Read, 1994). Early experiences of flexible access to feelings are regarded as formative by attachment theorists. The autonomous sense of self emerges fully from secure parent-infant relationships (Emde & Buchsbaum, 1990; Fonagy et al., 1995a; Lieberman & Pawl, 1990). Most importantly the increased control of the secure child permits him to move toward the ownership of inner experience, and toward understanding self and others as intentional beings whose behaviour is organised by mental states, thoughts, feelings, beliefs and desires (Fonagy et al., 1995a; Sroufe, 1990). Consistent with this, prospective longitudinal research has demonstrated that children with a history of secure attachment are independently rated as more resilient, self-reliant, socially oriented (Sroufe, 1983; Waters, Wippman, & Sroufe, 1979), empathic to distress (Kestenbaum, Farber, & Sroufe, 1989), with deeper relationships (Sroufe, 1983; Sroufe, Egeland, & Kreutzer, 1990).
Prediction from adult attachment measures
The stability of attachment is demonstrated by longitudinal studies of infants assessed with the Strange Situation and followed up in adolescence or young adulthood with the Adult Attachment Interview (AAI) (George, Kaplan, & Main, 1996). I assume most of you are familiar with this wonderful structured clinical instrument which elicits narrative histories of childhood attachment relationships – the characteristics of early relationships, experiences of separation, illness, punishment, loss, maltreatment or abuse. The AAI scoring system (Main & Goldwyn, 1994) classifies individuals intoSecure/Autonomous, Insecure/Dismissing, Insecure/Preoccupied or Unresolved with respect to loss or trauma, categories based on the structural qualities of narratives of early experiences. While autonomous individuals value attachment relationships, coherently integrate memories into a meaningful narrative and regard these as formative, insecure individuals are poor at integrating memories of experience with the meaning of that experience. Those dismissing of attachment show avoidance in denying memories, idealizing or devaluing (or both idealizing and devaluing) early relationships. Preoccupied individuals tend to be confused, angry or passive in relation to attachment figures, often still complaining of childhood slights, echoing the protests of the resistant infant. Unresolved individuals give indications of significant disorganisation in their attachment relationship representation in semantic or syntactic confusions in their narratives concerning childhood trauma or a recent loss. Again, we shall revisit this clinically most important group this afternoon.
Three major longitudinal studies (Hamilton, 1994; Main, 1997; Waters, Merrick, Albersheim, Treboux, & Crowell, 1995) have shown a 68-75% correspondence between attachment classifications in infancy and classifications in adulthood. This is an unparalleled level of consistency between behaviour observed in infancy and outcomes in adulthood. Obviously, such individual differences may well be maintained by consistent environments as well as by patterns laid down in the first year of life.
Attachment relationships play a key role in the transgenerational transmission of deprivation. Secure adults are 3 or 4 times more likely to have children who are securely attached to them (van IJzendoorn, 1995). This is true even where parental attachment is assessed before the birth of the child (Benoit & Parker, 1994; Fonagy, Steele, & Steele, 1991b; Radojevic, 1992; Steele, Steele, & Fonagy, 1996; Ward & Carlson, 1995). Parental attachment patterns predict variance in addition to temperament measures or contextual factors, such as life events, social support and psychopathology (Steele, Steele, & Fonagy, in preparation). How is such transgenerational transmission mediated? Genetics may appear to provide an obvious explanation. The early findings of an ongoing twin study in our laboratory have yielded no evidence of differential levels of concordance of attachment classification between identical and non-identical twins (Fearon, 1998). Attachment theorists have assumed that securely attached adults are more sensitive to their children’s needs thus fostering an expectation in the infant that dysregulation will be rapidly and effectively met (Belsky, Rosenberger, & Crnic, 1995; De Wolff & van IJzendoorn, 1997). Disappointingly, standard measures of caregiver sensitivity do not appear to explain at all well transgenerational consistencies in attachment classification (van IJzendoorn, 1995).
Attachment and mentalising: The move from a teleological to a mentalistic construal of attachment relationships in development
To do so, we have to explore a further aspect of the determinants of attachment. Mary Main and Inge Bretherton independently drew attention to what the philosopher Dennett called the "intentional stance". Dennett (1987) stressed that human beings try to understand each other in terms of mental states: thoughts and feelings, beliefs and desires, in order to make sense of and, even more important, to anticipate each others’ actions. If the child is able to attribute an unresponsive mother’s apparently rejecting behaviour to her sadness about a loss, rather than simply feeling helpless in the face of it, the child is protected from confusion and a negative view of himself. The hallmark of the intentional stance is the child’s recognition at around 3-4 years that behaviour may be based on a mistaken belief. Developmentalists have designed numerous tests of the quality of understanding false beliefs and tend to refer to this capacity as ‘a theory of mind’. We prefer the term mentalisation or reflective function which denotes the understanding of one’s own as well as others’ behaviour in mental state terms.
Say a three-year-old sees his friend, Maxi, hiding a piece of chocolate in a box, saying that he has to leave now but will come back to eat it later (Perner, 1991). After Maxi leaves, the child sees the experimenter move the chocolate to a basket. The child is asked: "Where will Maxi look for the chocolate, when he comes back?" The three year olds tend to predict that Maxi will look in the basket where the chocolate actually is, rather than in the box where he left it. Four and five year olds are already able to predict Maxi’s behaviour on the basis of what one might expect to be his belief, that the chocolate will still be where he left it. The four year old is said to have "a theory of mind", which is indicated by his ability to attribute false beliefs (Wimmer & Perner, 1983). He adopts an intentional stance and reasons in terms of the beliefs that may be attributed to Maxi. The three year old, however, is basing his prediction on his own representation of reality, and not on the other’s mind state.
One way to interpret this finding is that the three-year-old’s expectations are based on a nonmentalistic, "teleological" model of behaviour, rather than a mentalisitic intentional one (Gergely & Csibra, 1997). Within this simpler, teleological, model, the behavior of human objects is interpreted in terms of visible outcomes rather than inferred desires, and constraints of physical reality rather than attributed beliefs about reality. The three-year-old child in the Maxi task bases his/her prediction on an assumption of rational action presupposing that the other will do whatever is most rational to bring about the future goal state (to eat the chocolate), given the current state of external reality (that the chocolate is in the basket).
In fact, the capacity to discriminate between rational and non-rational actions has been demonstrated as early as 9 months of age (Gergely, Nadasdy, Csibra, & Biro, 1995). Infants were shown a computer-animated display depicting a small circle repeatedly jumping over a wall and making contact with a large circle on the other side. With repeated presentations, infants become "habituated", losing interest in the display. At this point, the infant is presented with one of two new displays. In both cases the wall is removed: in one case the small circle now approaches the large one directly (in a straight line), in the other it performs exactly the same jumping approach as before. Infants show surprise at the latter of the displays but not the former, indicating that they expected the circle to act ‘rationally’, i.e. to approach the target object by the shortest available route. We argue that the young child’s teleological interpretation of action is transformed into a mentalizing one during the second and third years. By 18 months (but not yet at 14) infants show a mentalistic understanding of desire (Meltzoff, 1995; Repacholi & Gopnik, 1997) being already able to understand that another person’s actions may be driven by desires other than the child’s own. At this stage the young child also becomes able to infer the other’s intention when the person uses new words to label objects unfamiliar to the child (Baldwin & Moses, 1996; Tomasello, Strosberg, & Akhtar, 1996). During the second year children already talk about desire states of self and other, and in the third year talk about beliefs also emerges (Bartsch & Wellman, 1995). A full-fledged mentalizing ability as indicated by the capacity to attribute false beliefs in theory of mind tasks is achieved towards the end of the third year (Perner, 1991).
The acquisition of "a theory of mind" is far from the end-point of this developmental process. In fact, it might be argued that reflective function is never fully achieved. At moments of high arousal, in the context of intimate relationships, we all find it hard to construct accurate representations of the mental word of the other. We reason about the behaviour of those close to us on the basis of what seems obvious, what is visible, the physical rather than the mental world. If we use mental states, these tend to be stereotypic, distorted, or confused overly analytical and inaccurate. We have attempted to operationalise individual differences in adults’ mentalising capacities. Our operationalisation was relatively simple, based on the presence of unequivocal descriptions of mental states (e.g. false beliefs) in the narrative. To score high on these ratings attachment narratives had to show awareness of mental states, manifest explicit efforts to tease out the internal reasons behind behaviour, show awareness that a child’s thoughts and feelings are likely to differ from those of an adult, and reflect a sensitivity to the mental states of the interviewer (Fonagy, Target, Steele, & Steele, 1998). The measure correlates only negligibly with I.Q. and educational background. We were curious to know if the extent of reflective observations about the mental states of self and others in AAI narratives could predict infant security. Reflectiveness ratings made before the child’s birth powerfully predicted the child’s attachment security in the 2nd year of life. Both fathers and mothers who were rated high in this capacity were three or four times more likely to have secure children than parents whose reflective capacity was poor (Fonagy, Steele, Moran, Steele, & Higgitt, 1991a).
The capacity for understanding the mental states that lie behind the parent’s behaviour may be particularly important when the child is exposed to unfavourable experiences, in the extreme, abuse or trauma. We divided our sample into those who had reported significant deprivation (overcrowding, parental mental illness) and those who had not. Our prediction was that mothers in the deprived group (childhood separations, would be far more likely to have children securely attached to them if their reflective function rating was high. All of the mothers in the deprived group with high reflectiveness ratings had children who were secure with them, whereas only 1 out of 17 of deprived mothers with low ratings did so. Our findings imply that this cycle of disadvantage may be interrupted if the caregiver has acquired a capacity to reflect productively on mental experience (Fonagy, Steele, Steele, Higgitt, & Target, 1994).
Mentalising and the development of the self
Not only are parents high in reflective capacity more likely to promote secure attachment in the child, particularly if their own childhood experiences were adverse, but also secure attachment may be a key facilitator of reflective capacity (Fonagy et al., 1995a). In our longitudinal study of 92 children, the proportion of secure children was twice as high in the group which passed a false belief task, compared to the group which failed. Mother’s reflective function was also associated with the child’s success. 80% of children whose mothers were above the median in reflective function passed, whereas only 56% of those whose mothers were below did so. The caregiver’s reflective function predicted attachment security which, in its turn, predicted the precocious acquisition of a theory of mind. Attachment to father also appeared to contribute to this developmental achievement. On a test of second-order mentalising skills, where the child is required to predict the behaviour of a person on the basis of that person’s beliefs about a third person’s false beliefs, those secure with both parents were most likely to succeed whereas those secure with neither parent were least likely to do so.
These results suggest that the parents’ capacity to observe the child’s mind facilitates the child’s general understanding of minds through the mediation of secure attachment. A reflective caregiver increases the likelihood of the child’s secure attachment which, in turn, facilitates the development of mentalisation. We assume that a secure attachment relationship provides a congenial context for the child to explore the mind of the caregiver, and in this way to learn about minds. The philosopher Hegel (1807) suggested that it is only through exploring the mind of the other that the child develops full appreciation of the nature of mental states. Reflectiveness in the child is facilitated by secure attachment. The process is intersubjective: the child gets to know the caregiver’s mind as the caregiver endeavours to understand and contain the mental state of the child.
In elucidating this process I would like to suggest three critical components. These are: (1) the role of mirroring, (2) the move to the interpretation of the caregiver’s behaviour in intentional rather than teleological terms, and (3) the integration of a primitive dual form of psychic reality into a singular mentalising representation of the mind.
1. The Role of Mirroring
Second-order (or symbolic) representations of mental states in our view evolve in the context of attachment relationships. The child’s concept of emotions is arrived at by introspection (Gergely & Watson, 1996; Target & Fonagy, 1996). Anxiety for the infant, for example, is associated with a confusing mixture of physiological experiences, behaviours and visual images. Once these become symbolically bound, the corresponding experience at a mentalised or symbolic secondary level will be one of fear or of anxiety. This symbolic binding process is essential for the child to be able to label the experience as one of a specific emotion. This knowledge is not inherent. Assume that the child’s constitutional or physical self is in a state of arousal. Associated with this are signals (non-verbal expressions, facial as well as vocal). The caregiver resonates with these and ideally reflects on her internal experience and generates an appropriate responsive expression. Such mirroring displays are innate and generated non-consciously by the caregiver (Meltzoff, 1993). The mother’s representation of the infant’s affect is represented by the child and "mapped on to" the infant’s constitutional self-state (Rogers & Pennington, 1991). The discrepancy between the child’s original experience and the internalisation of the caregiver’s mirroring representation is helpful insofar as it allows this somewhat modified representation (which is the same yet not the same) to become a higher order representation of the infant’s experience.
Within this model mirroring would be expected to fail if it is either too close to the infant’s experience or too remote from it. If the mirroring is too accurate, the perception itself can become a source of fear and it loses its symbolic potential. If it is unavailable, or is contaminated with the mother’s own preoccupation, the process of self-development is profoundly compromised. We may presume that individuals for whom the symptoms of anxiety signify catastrophes (e.g. heart attack, imminent death etc.) have second-order representations of their emotional responses which cannot be limited in intensity through symbolization, perhaps because the original mirroring by the primary caregiver exaggerated the infant’s emotions.
Admittedly this is a speculative model, but it is empirically testable. It might help answer the thorny question of why individuals with panic disorders attribute immense significance to physiologically relatively mild levels of disequilibrium. The suggestion here is that the secondary representation, or symbolic representation, of affect in these cases contains too much of the primary experience; hence, instead of labelling the experience having the potential to attenuate it, it tends to stimulate and exacerbate symptoms of the affect state, which in turn accentuates the secondary expression, in a cycle of escalating panic. In a recent study (Fonagy et al., 1995b), we have confirmed that mothers who soothe their distressed 8 month olds most effectively following an injection rapidly reflect the child’s emotion, but this mirroring is mixed with other affects (smiling, questioning, mocking display and the like). In displaying such "complex affect" (Fónagy & Fónagy, 1987) they ensure that the infant recognizes their emotion as analogous to, but not isomorphic with, their experience and thus the process of symbol formation may begin. In this way, the representational mapping between affect of self and emotions of others, the exchange of affect between young child and caregiver, provides a unique source of information to the child about his own internal states.
The child who looks for a way of managing his distress identifies in the response of the caregiver a representation of his mental state which he may internalize and use as part of a higher order strategy of affect regulation. The secure caregiver soothes by combining a "mirror" with a display incompatible with the child’s affect (thus perhaps implying coping). This formulation of sensitivity has much in common with the British psychoanalyst, Wilfred Bion’s (1962) notion of the role of the mother’s capacity to mentally "contain" the affect state intolerable for the baby, and respond in terms of physical care in a manner that acknowledges the child’s mental state yet serves to modulate unmanageable feelings. The finding that the clarity and coherence of the mother’s representation of the child mediates between her attachment status and her behavior is certainly consistent with this model (Slade, Belsky, Aber, & Phelps, in press).
We suggest that the meaning or sense of affect develops out of the integrated representation of the affect in self and other. The combination of the representation of self experience and the representation of the reaction of the caregiver elaborates the child’s teleological model of the mind, and ultimately enables him to interpret and understand affective displays in others as well as arriving at the regulation and control of his own emotions. The representational mapping of emotion displays and self experience is seen here as a prototypical instance of caregiver sensitivity, which, as we shall attempt to demonstrate, is likely to be an important component of the development of mentalizing. The reflective function of the caregiver prompts the child to begin organizing self-experience according to clusters of responses which will eventually come to be verbally labelled as specific emotions (or desires). The high contingent response is the means by which this mapping can take place. The child’s affective experiences are given further meaning by becoming associated with clusters of reality constraints within the parent-infant interaction (leading to rudimentary beliefs about the causes and consequences of his emotional state).
2. Reflective parenting and the move from teleological to intentional mental models
We take the view that the acquisition of mentalising is part of an intersubjective process between the infant and caregiver (see Gopnik, 1993, for a highly elegant elaboration of such a model). In our view, the caregiver facilitates the creation of mentalizing models through complex linguistic and quasi-linguistic processes, primarily through behaving towards the child in such a way that leads him eventually to see that his own behavior may be best understood by assuming that he has ideas and feelings which determine his actions, and the reactions of others to him, which can then be generalized to other similar beings. The caregiver approaches the crying infant with a question in her mind: "Do you want your nappy changed?" "Do you need a cuddle?" The sensitive caregiver is unlikely to address the situation teleologically, without having the person in mind, so is unlikely to say to herself, "Are you wet around your bottom?" or "Have you been standing alone too long?" The sensitive caregiver can cover the gap between a focus on physical reality and internally-directed attention, sufficiently for the child to identify contingencies between internal and external experience. Ultimately, the child arrives at the conclusion that the caregiver’s reaction to him may be understood as rational given the assumption of an internal state of belief or desire within himself. Unconsciously and pervasively, the caregiver ascribes a mental state to the child with her behavior, treats the child as a mental agent, which is perceived by the child and used in the elaboration of teleological models, and then in the development of a core sense of mental selfhood. We assume that this, by and large, is a mundane process, happening routinely throughout early life, not reflected on, and so rarely modified. Caregivers, however, differ in their ways of carrying out this natural human function. Some may be particularly alert to the earliest indications of intentionality, others may need stronger indications before perceiving the child’s mental state and modifying their behavior accordingly. Others, as we described in the context of early infancy, may systematically misperceive the child’s states of mind, with resulting deformation of the child’s sense of himself.
The parent’s capacity to observe the moment to moment changes in the child’s mental state, then, lies at the root of sensitive caregiving, which is viewed by attachment theorists as the cornerstone of secure attachment (e.g. Ainsworth et al., 1978; Grossmann, Grossmann, Spangler, Suess, & Unzner, 1985; Isabella & Belsky, 1991). Secure attachment in its turn provides the psychosocial basis for acquiring an understanding of mind. The secure infant feels safe in making attributions of mental states to account for the behavior of the caregiver. By contrast the avoidant child to some degree shuns the mental state of the other, while the resistant child focuses on his own state of distress to the exclusion of close intersubjective exchanges. Disorganized infants may represent a special category; hypervigilant of the caregiver’s behavior they use all cues available for prediction and may be acutely sensitized to intentional states, and thus may be more ready to construct a mentalized account of the caregiver’s behavior. We would argue (see below) that in such children mentalization may be evident but it does not have the central and effective role in self-organization which characterizes securely attached children.
We believe that most important for the development of mentalizing self-organization is that exploration of the mental state of the sensitive caregiver enables the child to find in her mind an image of himself as motivated by beliefs, feelings and intentions, in other words, as mentalizing. There is considerable evidence to support the view that secure attachment enhances the development of inner security, self worth and autonomy (e.g. Londerville & Main, 1981). Disorganized infants, even if they acquire the skill of mentalization, fail to integrate this with their self-organization. There may be a number of linked reasons for this: a) the caregiver of the disorganized infant is less likely to be reliably contingent in responding to the infant’s self-state, and further to show systematic biases in her perception and reflection of his state; b) the mental state of the caregiver evokes intense anxiety through either frightening behavior suggesting malevolence towards the child, or behavior suggesting fear, which may include fear of the child himself; c) the child needs to use disproportionate resources to understand the parent’s behavior, at the expense of reflecting on self-states. These factors combine, perhaps, to make disorganized infants become keen readers of the caregiver’s mind under certain circumstances, but (we suggest) poor readers of their own mental states.
3. The move from dual to singular psychic reality
In two previous papers we have used both clinical and research evidence to show that the normal experience of psychic reality is not an inherent property of the mind, but rather a developmental achievement (Fonagy & Target, 1996; Target & Fonagy, 1996). It is the consequence of the successful integration of two distinct modes of differentiating internal from external. We see the child’s development as normally moving from an experience of psychic reality in which mental states are not depicted as representations, to an increasingly complex view of the internal world, which has as its hallmark the capacity to mentalise, to think flexibly about thoughts and feelings in others and in oneself. Initially, the child’s experience of the mind is as if it were a recording device, with exact correspondence between internal state and external reality. We use the term "psychic equivalence" to denote this mode of functioning, to emphasise that for the young child mental events are equivalent in terms of power, causality and implications, to events in the physical world. Equating internal and external is inevitably a two-way process. Not only will the small child feel compelled to equate appearance with reality (how it seems is how it is), but also internal representations, distorted by phantasy, will be projected onto external reality in a manner unmodulated by awareness that the experience of the external world might be misconstrued in this way.
Perhaps because it can be terrifying for thoughts and feelings to be experienced as concretely ‘real’, the infant develops an alternative way of construing mental states. In "pretend mode", the child experiences feelings and ideas as totally representational, or symbolic, as having no implication for the world outside. Even though the child of two years knows that his pretence to be a policeman is not real, this is not because he understands that he is being a ‘pretend policeman’, but rather because the mode of psychic reality which prevails at that time presupposes strict separation from external reality (Gopnik & Slaughter, 1991). Thus his play by itself it can form no bridge between inner and outer reality. Only gradually, and through the close participation of another mind which can simultaneously hold together the child’s pretend and serious perspectives, does the integration of these two modes give rise to a psychic reality in which feelings and ideas are known as internal, yet in close relationship with what is outside (Dunn, 1996).
Normally, the child then integrates these alternative modes to arrive at mentalization, or reflective mode, in which mental states can be experienced as representations. Inner and outer reality can then be seen as linked, yet they are accepted as differing in important ways, and no longer have to be either equated or dissociated from each other (e.g. Gopnik, 1993). Mentalization comes about through the child’s experience of his mental states being reflected on, for instance through secure play with a parent or older child. In playfulness, the caregiver gives the child’s ideas and feelings (when he is "only pretending") a link with reality, by indicating the existence of an alternative perspective, which exists outside the child’s mind. The parent or older child also shows that reality may be distorted by acting upon it in playful ways, and through this playfulness a pretend but real mental experience may be introduced.
The child’s development and perception of mental states in himself and others thus depends on his observation of the mental world of his caregiver. He is able to perceive mental states when the caregiver is in a shared pretend mode of playing with the child (hence the association between pretend and early mentalization), and many ordinary interactions (such as physical care and comforting, conversations with peers) will also involve such shared mentation. This is what makes mental state concepts such as thinking inherently intersubjective; shared experience is part of the very logic of mental state concepts.
To pre-empt this afternoon’s topic somewhat, in traumatized children, intense emotion and conflict lead to a partial failure of this integration, so that aspects of the pretend mode of functioning become part of a psychic equivalence manner of experiencing reality. This may be because where maltreatment or trauma has occurred within the family, the atmosphere tends to be incompatible with the caregiver "playing with" the most pressing aspects of the child's thoughts; these are often disturbing and unacceptable to the adult, just as they are to the child. The rigid, controlling behavior of the pre-school child with a history of disorganized attachment is thus seen as arising out of a partial failure on the part of the child to move beyond the mode of psychic equivalence in relation to specific ideas or feelings, so that he experiences them with the intensity that might be expected had they been current, external events.
In the insecure relationship, what is internalised by the child will tend to be of the caregiver’s defences, and the infant’s distress is repeatedly avoided. Most importantly, the process of self-development has been compromised. Insecure parent-child relationships may be regarded as laying the groundwork for subsequent distortions of personality development in one of two ways. These correspond to the two modes of experiencing psychic reality. The mother may echo the child’s state without modulation, as in the mode of psychic equivalence, concretising or panicking at the child’s distress. Alternatively she may avoid reflection on the child’s affect through a process akin to dissociation, which effectively places the mother in a pretend mode, unrelated to external reality, including the child. The mother may then ignore the child’s distress, or translate it into illness, tiredness, and so on. Both strip the child’s communication of the potential for meaning which he can recognise and use. It may also lead to a currency between mother and child of interpretation of feelings in physical terms, so that the physical state is the "real" thing. Lynne Murray (1997), in her work with mothers suffering from puerperal depression, has provided some vivid illustrations of such mothers offering an alternative reality, marked by the exaggeration associated with pretence, but not related to the infant’s experience. Psychoanalytic observers of this type of interaction would rapidly identify the operation of massive denial and even manic defence. The infant has not been able to find a recognisable version of his mental states in another person’s mind, and the opportunity to acquire a symbolic representation of those states has been lost and a disorganised pattern of attachment may follow.
Conclusion
In summary, the securely attached child perceives in the caregiver’s reflective stance an image of himself as desiring and believing. He sees that the caregiver represents him as an intentional being, and this representation is internalised to form the self. "I think therefore I am" will not do as a psychological model of the birth of the self; "She thinks of me as thinking and therefore I exist as a thinker" perhaps comes closer to the truth. If the caregiver’s reflective capacity has enabled her accurately to picture the child’s intentional stance, then he will have the opportunity to "find himself in the other" as a mentalising individual. At the core of our selves is the representation of how we were seen. Our reflective capacity is thus a transgenerational acquisition. We think of others in terms of desires and beliefs because, and to the extent that, we were thought of as intentional beings. Only following this process of internalisation, can the development of awareness of mental states in oneself be generalised to others including the caregiver.
The theory of a transgenerational reflective function has these components: 1. We assume that the internalisation of second order representations of internal states depends upon the sensitive reflection of the caregiver and it offers the building blocks with which a reflective internal working model is constructed. 2. The gradual move from a teleological to an intentional stance is intrinsically linked to the child’s experience of safety in exploring the caregiver’s mind to ferret out the feelings and thoughts that might account for her behaviour. Needless to say, this is easiest and safest to do in the context of a secure attachment relationship. 3. The caregiver makes a further important contribution, perhaps most important at a somewhat later stage. Prototypically, while engaging in pretend play with the child, the caregiver simultaneously engages the child’s internal world while retaining an external reality-based perspective. This is analogous to psychoanalytic discussions of the cognitive impact of the oedipal triad, where the shared reality of two people is suddenly experienced from the point of view of the third. The parents’ engagement in the child’s internal world moves the child beyond the conception of their mind as a replica of the external world.
These three components (the second-order representation of affect, the intentional representation of the caregiver and ultimately the intentional representation of the self) equip the child to confront a sometimes unduly harsh social reality. I shall go on to argue that the robust establishment of reflective function has a protective effect and, by contrast, its relatively fragile status indexes a vulnerability to later trauma. Secure attachment and reflective function are, I believe, overlapping constructs and the vulnerability associated with insecure attachment lies primarily in the child’s diffidence in conceiving of the world in terms of psychic rather than physical reality. Given trauma of sufficient intensity, even a secure bond may sometimes crumble and in the absence of psychosocial pressures, reflective function may offer only marginal developmental advantage. To understand severe personality disorder, as I hope we shall see, it is important we are attuned to our patient’s capacity to use the language of mental states for self organisation as well as social understanding.
Go to Reference Page
Peter Fonagy, PhD, FBA
Freud Memorial Professor of Psychoanalysis, UCL
Director of Research, The Anna Freud Centre
Co-ordinating Director, Child and Family Center and Center for Outcomes Research and Effectiveness, Menninger Foundation
Address for correspondence:
Sub-Department of Clinical Health Psychology
University College London
Gower Street
London WC1E 6BT
E-mail: p.fonagy@ucl.ac.uk
Return to Home Page
Return to Papers
Peter Fonagy, PhD, FBA
Paper to the Developmental and Psychoanalytic Discussion Group, American Psychoanalytic Association Meeting, Washington DC 13 May 1999
The attachment system as a bio-social mechanism of homeostatic regulation
Attachment theory, developed by John Bowlby (Bowlby, 1969; Bowlby, 1973; Bowlby, 1980), postulates a universal human need to form close affectional bonds. At its core is the reciprocity of early relationships, which is a precondition of normal development probably in all mammals, including humans (Hofer, 1995). The attachment behaviours of the human infant (e.g. proximity seeking, smiling, clinging) are reciprocated by adult attachment behaviours (touching, holding, soothing) and these responses strengthen the attachment behaviour of the infant toward that particular adult. The activation of attachment behaviours depends on the infant’s evaluation of a range of environmental signals which results in the subjective experience of security or insecurity. The experience of security is the goal of the attachment system, which is thus first and foremost a regulator of emotional experience (Sroufe, 1996). In this sense it lies at the heart of many forms of mental disorder and the entire psychotherapeutic enterprise.
None of us is born with the capacity to regulate our own emotional reactions. A dyadic regulatory system evolves where the infants’ signals of moment to moment changes in their state that are understood and responded to by the caregiver thereby achieving their regulation. The infant learns that arousal in the presence of the caregiver will not lead to disorganisation beyond his coping capabilities. The caregiver will be there to re-establish equilibrium. In states of uncontrollable arousal, the infant will come to seek physical proximity to the caregiver in the hope of soothing and the recovery of homeostasis. The infant’s behaviour by the end of the first year is purposeful, and apparently based on specific expectations. His past experiences with the caregiver are aggregated into representational systems which Bowlby (1973) termed ‘internal working models’. Thus, the attachment system is an open bio-social homeostatic regulatory system.
Patterns of attachment in infancy
The second great pioneer of attachment theory, Mary Ainsworth (1969; 1985; Ainsworth, Blehar, Waters, & Wall, 1978), developed the well-known laboratory based procedure for observing infant’s internal working models in action. Infants, briefly separated from their caregiver in a situation unfamiliar to them, show one of four patterns of behaviour. Infants classified as Secure explore readily in the presence of the primary caregiver, are anxious in the presence of the stranger and avoid her, are distressed by their caregivers’ brief absence, rapidly seek contact with the caregiver afterwards, and are reassured by this. The infant returns to exploration. Some infants, who appear to be made less anxious by separation, may not seek proximity with the caregiver following separation, and may not prefer the caregiver over the stranger; these infants are designated ‘Anxious/Avoidant’. A third category, ‘Anxious/Resistant’ infants show limited exploration and play, tend to be highly distressed by the separation, but have great difficulty in settling afterwards, showing struggling, stiffness, continued crying, or fuss in a passive way. The caregiver’s presence or attempts at comforting fail to reassure, and the infant’s anxiety and anger appear to prevent them from deriving comfort from proximity.
Secure infants’ behaviour is based on the experience of well co-ordinated, sensitive interactions where the caregiver is rarely over-arousing and is able to restabilise the child’s disorganising emotional responses. Therefore, they remain relatively organised in stressful situations. Negative emotions feel less threatening, and can be experienced as meaningful and communicative (Grossman, Grossmann, & Schwan, 1986; Sroufe, 1979; Sroufe, 1996).
Anxious/Avoidantly attached children are presumed to have had experiences where their emotional arousal was not restabilised by the caregiver, or where they were over aroused through intrusive parenting; therefore they over-regulatetheir affect and avoid situations that are likely to be distressing. Anxious/Resistantly attached children under-regulate, heightening their expression of distress possibly in an effort to elicit the expectable response of the caregiver. There is a low threshold for threat, and the child becomes preoccupied with having contact with the caregiver, but frustrated even when it is available (Sroufe, 1996).
A fourth group of infants exhibits seemingly undirected behaviour, giving the impression of disorganisation and disorientation (Main & Solomon, 1990). Infants who manifest freezing, hand clapping, head-banging, the wish to escape the situation even in the presence of the caregiver, are referred to as ‘Disorganised/Disoriented’. It is generally held that for such infants the caregiver has served as a source of both fear and reassurance, thus arousal of the attachment behavioural system produces strong conflicting motivations. Not surprisingly, a history of severe neglect or physical or sexual abuse is often associated with this pattern (Cicchetti & Beeghly, 1987; Main & Hesse, 1990). I would like to consider this group in much greater detail, this afternoon.
The continuity of patterns of attachment
Bowlby proposed that internal working models of the self and others provide prototypes for all later relationships. Such models are relatively stable across the lifespan (Collins & Read, 1994). Early experiences of flexible access to feelings are regarded as formative by attachment theorists. The autonomous sense of self emerges fully from secure parent-infant relationships (Emde & Buchsbaum, 1990; Fonagy et al., 1995a; Lieberman & Pawl, 1990). Most importantly the increased control of the secure child permits him to move toward the ownership of inner experience, and toward understanding self and others as intentional beings whose behaviour is organised by mental states, thoughts, feelings, beliefs and desires (Fonagy et al., 1995a; Sroufe, 1990). Consistent with this, prospective longitudinal research has demonstrated that children with a history of secure attachment are independently rated as more resilient, self-reliant, socially oriented (Sroufe, 1983; Waters, Wippman, & Sroufe, 1979), empathic to distress (Kestenbaum, Farber, & Sroufe, 1989), with deeper relationships (Sroufe, 1983; Sroufe, Egeland, & Kreutzer, 1990).
Prediction from adult attachment measures
The stability of attachment is demonstrated by longitudinal studies of infants assessed with the Strange Situation and followed up in adolescence or young adulthood with the Adult Attachment Interview (AAI) (George, Kaplan, & Main, 1996). I assume most of you are familiar with this wonderful structured clinical instrument which elicits narrative histories of childhood attachment relationships – the characteristics of early relationships, experiences of separation, illness, punishment, loss, maltreatment or abuse. The AAI scoring system (Main & Goldwyn, 1994) classifies individuals intoSecure/Autonomous, Insecure/Dismissing, Insecure/Preoccupied or Unresolved with respect to loss or trauma, categories based on the structural qualities of narratives of early experiences. While autonomous individuals value attachment relationships, coherently integrate memories into a meaningful narrative and regard these as formative, insecure individuals are poor at integrating memories of experience with the meaning of that experience. Those dismissing of attachment show avoidance in denying memories, idealizing or devaluing (or both idealizing and devaluing) early relationships. Preoccupied individuals tend to be confused, angry or passive in relation to attachment figures, often still complaining of childhood slights, echoing the protests of the resistant infant. Unresolved individuals give indications of significant disorganisation in their attachment relationship representation in semantic or syntactic confusions in their narratives concerning childhood trauma or a recent loss. Again, we shall revisit this clinically most important group this afternoon.
Three major longitudinal studies (Hamilton, 1994; Main, 1997; Waters, Merrick, Albersheim, Treboux, & Crowell, 1995) have shown a 68-75% correspondence between attachment classifications in infancy and classifications in adulthood. This is an unparalleled level of consistency between behaviour observed in infancy and outcomes in adulthood. Obviously, such individual differences may well be maintained by consistent environments as well as by patterns laid down in the first year of life.
Attachment relationships play a key role in the transgenerational transmission of deprivation. Secure adults are 3 or 4 times more likely to have children who are securely attached to them (van IJzendoorn, 1995). This is true even where parental attachment is assessed before the birth of the child (Benoit & Parker, 1994; Fonagy, Steele, & Steele, 1991b; Radojevic, 1992; Steele, Steele, & Fonagy, 1996; Ward & Carlson, 1995). Parental attachment patterns predict variance in addition to temperament measures or contextual factors, such as life events, social support and psychopathology (Steele, Steele, & Fonagy, in preparation). How is such transgenerational transmission mediated? Genetics may appear to provide an obvious explanation. The early findings of an ongoing twin study in our laboratory have yielded no evidence of differential levels of concordance of attachment classification between identical and non-identical twins (Fearon, 1998). Attachment theorists have assumed that securely attached adults are more sensitive to their children’s needs thus fostering an expectation in the infant that dysregulation will be rapidly and effectively met (Belsky, Rosenberger, & Crnic, 1995; De Wolff & van IJzendoorn, 1997). Disappointingly, standard measures of caregiver sensitivity do not appear to explain at all well transgenerational consistencies in attachment classification (van IJzendoorn, 1995).
Attachment and mentalising: The move from a teleological to a mentalistic construal of attachment relationships in development
To do so, we have to explore a further aspect of the determinants of attachment. Mary Main and Inge Bretherton independently drew attention to what the philosopher Dennett called the "intentional stance". Dennett (1987) stressed that human beings try to understand each other in terms of mental states: thoughts and feelings, beliefs and desires, in order to make sense of and, even more important, to anticipate each others’ actions. If the child is able to attribute an unresponsive mother’s apparently rejecting behaviour to her sadness about a loss, rather than simply feeling helpless in the face of it, the child is protected from confusion and a negative view of himself. The hallmark of the intentional stance is the child’s recognition at around 3-4 years that behaviour may be based on a mistaken belief. Developmentalists have designed numerous tests of the quality of understanding false beliefs and tend to refer to this capacity as ‘a theory of mind’. We prefer the term mentalisation or reflective function which denotes the understanding of one’s own as well as others’ behaviour in mental state terms.
Say a three-year-old sees his friend, Maxi, hiding a piece of chocolate in a box, saying that he has to leave now but will come back to eat it later (Perner, 1991). After Maxi leaves, the child sees the experimenter move the chocolate to a basket. The child is asked: "Where will Maxi look for the chocolate, when he comes back?" The three year olds tend to predict that Maxi will look in the basket where the chocolate actually is, rather than in the box where he left it. Four and five year olds are already able to predict Maxi’s behaviour on the basis of what one might expect to be his belief, that the chocolate will still be where he left it. The four year old is said to have "a theory of mind", which is indicated by his ability to attribute false beliefs (Wimmer & Perner, 1983). He adopts an intentional stance and reasons in terms of the beliefs that may be attributed to Maxi. The three year old, however, is basing his prediction on his own representation of reality, and not on the other’s mind state.
One way to interpret this finding is that the three-year-old’s expectations are based on a nonmentalistic, "teleological" model of behaviour, rather than a mentalisitic intentional one (Gergely & Csibra, 1997). Within this simpler, teleological, model, the behavior of human objects is interpreted in terms of visible outcomes rather than inferred desires, and constraints of physical reality rather than attributed beliefs about reality. The three-year-old child in the Maxi task bases his/her prediction on an assumption of rational action presupposing that the other will do whatever is most rational to bring about the future goal state (to eat the chocolate), given the current state of external reality (that the chocolate is in the basket).
In fact, the capacity to discriminate between rational and non-rational actions has been demonstrated as early as 9 months of age (Gergely, Nadasdy, Csibra, & Biro, 1995). Infants were shown a computer-animated display depicting a small circle repeatedly jumping over a wall and making contact with a large circle on the other side. With repeated presentations, infants become "habituated", losing interest in the display. At this point, the infant is presented with one of two new displays. In both cases the wall is removed: in one case the small circle now approaches the large one directly (in a straight line), in the other it performs exactly the same jumping approach as before. Infants show surprise at the latter of the displays but not the former, indicating that they expected the circle to act ‘rationally’, i.e. to approach the target object by the shortest available route. We argue that the young child’s teleological interpretation of action is transformed into a mentalizing one during the second and third years. By 18 months (but not yet at 14) infants show a mentalistic understanding of desire (Meltzoff, 1995; Repacholi & Gopnik, 1997) being already able to understand that another person’s actions may be driven by desires other than the child’s own. At this stage the young child also becomes able to infer the other’s intention when the person uses new words to label objects unfamiliar to the child (Baldwin & Moses, 1996; Tomasello, Strosberg, & Akhtar, 1996). During the second year children already talk about desire states of self and other, and in the third year talk about beliefs also emerges (Bartsch & Wellman, 1995). A full-fledged mentalizing ability as indicated by the capacity to attribute false beliefs in theory of mind tasks is achieved towards the end of the third year (Perner, 1991).
The acquisition of "a theory of mind" is far from the end-point of this developmental process. In fact, it might be argued that reflective function is never fully achieved. At moments of high arousal, in the context of intimate relationships, we all find it hard to construct accurate representations of the mental word of the other. We reason about the behaviour of those close to us on the basis of what seems obvious, what is visible, the physical rather than the mental world. If we use mental states, these tend to be stereotypic, distorted, or confused overly analytical and inaccurate. We have attempted to operationalise individual differences in adults’ mentalising capacities. Our operationalisation was relatively simple, based on the presence of unequivocal descriptions of mental states (e.g. false beliefs) in the narrative. To score high on these ratings attachment narratives had to show awareness of mental states, manifest explicit efforts to tease out the internal reasons behind behaviour, show awareness that a child’s thoughts and feelings are likely to differ from those of an adult, and reflect a sensitivity to the mental states of the interviewer (Fonagy, Target, Steele, & Steele, 1998). The measure correlates only negligibly with I.Q. and educational background. We were curious to know if the extent of reflective observations about the mental states of self and others in AAI narratives could predict infant security. Reflectiveness ratings made before the child’s birth powerfully predicted the child’s attachment security in the 2nd year of life. Both fathers and mothers who were rated high in this capacity were three or four times more likely to have secure children than parents whose reflective capacity was poor (Fonagy, Steele, Moran, Steele, & Higgitt, 1991a).
The capacity for understanding the mental states that lie behind the parent’s behaviour may be particularly important when the child is exposed to unfavourable experiences, in the extreme, abuse or trauma. We divided our sample into those who had reported significant deprivation (overcrowding, parental mental illness) and those who had not. Our prediction was that mothers in the deprived group (childhood separations, would be far more likely to have children securely attached to them if their reflective function rating was high. All of the mothers in the deprived group with high reflectiveness ratings had children who were secure with them, whereas only 1 out of 17 of deprived mothers with low ratings did so. Our findings imply that this cycle of disadvantage may be interrupted if the caregiver has acquired a capacity to reflect productively on mental experience (Fonagy, Steele, Steele, Higgitt, & Target, 1994).
Mentalising and the development of the self
Not only are parents high in reflective capacity more likely to promote secure attachment in the child, particularly if their own childhood experiences were adverse, but also secure attachment may be a key facilitator of reflective capacity (Fonagy et al., 1995a). In our longitudinal study of 92 children, the proportion of secure children was twice as high in the group which passed a false belief task, compared to the group which failed. Mother’s reflective function was also associated with the child’s success. 80% of children whose mothers were above the median in reflective function passed, whereas only 56% of those whose mothers were below did so. The caregiver’s reflective function predicted attachment security which, in its turn, predicted the precocious acquisition of a theory of mind. Attachment to father also appeared to contribute to this developmental achievement. On a test of second-order mentalising skills, where the child is required to predict the behaviour of a person on the basis of that person’s beliefs about a third person’s false beliefs, those secure with both parents were most likely to succeed whereas those secure with neither parent were least likely to do so.
These results suggest that the parents’ capacity to observe the child’s mind facilitates the child’s general understanding of minds through the mediation of secure attachment. A reflective caregiver increases the likelihood of the child’s secure attachment which, in turn, facilitates the development of mentalisation. We assume that a secure attachment relationship provides a congenial context for the child to explore the mind of the caregiver, and in this way to learn about minds. The philosopher Hegel (1807) suggested that it is only through exploring the mind of the other that the child develops full appreciation of the nature of mental states. Reflectiveness in the child is facilitated by secure attachment. The process is intersubjective: the child gets to know the caregiver’s mind as the caregiver endeavours to understand and contain the mental state of the child.
In elucidating this process I would like to suggest three critical components. These are: (1) the role of mirroring, (2) the move to the interpretation of the caregiver’s behaviour in intentional rather than teleological terms, and (3) the integration of a primitive dual form of psychic reality into a singular mentalising representation of the mind.
1. The Role of Mirroring
Second-order (or symbolic) representations of mental states in our view evolve in the context of attachment relationships. The child’s concept of emotions is arrived at by introspection (Gergely & Watson, 1996; Target & Fonagy, 1996). Anxiety for the infant, for example, is associated with a confusing mixture of physiological experiences, behaviours and visual images. Once these become symbolically bound, the corresponding experience at a mentalised or symbolic secondary level will be one of fear or of anxiety. This symbolic binding process is essential for the child to be able to label the experience as one of a specific emotion. This knowledge is not inherent. Assume that the child’s constitutional or physical self is in a state of arousal. Associated with this are signals (non-verbal expressions, facial as well as vocal). The caregiver resonates with these and ideally reflects on her internal experience and generates an appropriate responsive expression. Such mirroring displays are innate and generated non-consciously by the caregiver (Meltzoff, 1993). The mother’s representation of the infant’s affect is represented by the child and "mapped on to" the infant’s constitutional self-state (Rogers & Pennington, 1991). The discrepancy between the child’s original experience and the internalisation of the caregiver’s mirroring representation is helpful insofar as it allows this somewhat modified representation (which is the same yet not the same) to become a higher order representation of the infant’s experience.
Within this model mirroring would be expected to fail if it is either too close to the infant’s experience or too remote from it. If the mirroring is too accurate, the perception itself can become a source of fear and it loses its symbolic potential. If it is unavailable, or is contaminated with the mother’s own preoccupation, the process of self-development is profoundly compromised. We may presume that individuals for whom the symptoms of anxiety signify catastrophes (e.g. heart attack, imminent death etc.) have second-order representations of their emotional responses which cannot be limited in intensity through symbolization, perhaps because the original mirroring by the primary caregiver exaggerated the infant’s emotions.
Admittedly this is a speculative model, but it is empirically testable. It might help answer the thorny question of why individuals with panic disorders attribute immense significance to physiologically relatively mild levels of disequilibrium. The suggestion here is that the secondary representation, or symbolic representation, of affect in these cases contains too much of the primary experience; hence, instead of labelling the experience having the potential to attenuate it, it tends to stimulate and exacerbate symptoms of the affect state, which in turn accentuates the secondary expression, in a cycle of escalating panic. In a recent study (Fonagy et al., 1995b), we have confirmed that mothers who soothe their distressed 8 month olds most effectively following an injection rapidly reflect the child’s emotion, but this mirroring is mixed with other affects (smiling, questioning, mocking display and the like). In displaying such "complex affect" (Fónagy & Fónagy, 1987) they ensure that the infant recognizes their emotion as analogous to, but not isomorphic with, their experience and thus the process of symbol formation may begin. In this way, the representational mapping between affect of self and emotions of others, the exchange of affect between young child and caregiver, provides a unique source of information to the child about his own internal states.
The child who looks for a way of managing his distress identifies in the response of the caregiver a representation of his mental state which he may internalize and use as part of a higher order strategy of affect regulation. The secure caregiver soothes by combining a "mirror" with a display incompatible with the child’s affect (thus perhaps implying coping). This formulation of sensitivity has much in common with the British psychoanalyst, Wilfred Bion’s (1962) notion of the role of the mother’s capacity to mentally "contain" the affect state intolerable for the baby, and respond in terms of physical care in a manner that acknowledges the child’s mental state yet serves to modulate unmanageable feelings. The finding that the clarity and coherence of the mother’s representation of the child mediates between her attachment status and her behavior is certainly consistent with this model (Slade, Belsky, Aber, & Phelps, in press).
We suggest that the meaning or sense of affect develops out of the integrated representation of the affect in self and other. The combination of the representation of self experience and the representation of the reaction of the caregiver elaborates the child’s teleological model of the mind, and ultimately enables him to interpret and understand affective displays in others as well as arriving at the regulation and control of his own emotions. The representational mapping of emotion displays and self experience is seen here as a prototypical instance of caregiver sensitivity, which, as we shall attempt to demonstrate, is likely to be an important component of the development of mentalizing. The reflective function of the caregiver prompts the child to begin organizing self-experience according to clusters of responses which will eventually come to be verbally labelled as specific emotions (or desires). The high contingent response is the means by which this mapping can take place. The child’s affective experiences are given further meaning by becoming associated with clusters of reality constraints within the parent-infant interaction (leading to rudimentary beliefs about the causes and consequences of his emotional state).
2. Reflective parenting and the move from teleological to intentional mental models
We take the view that the acquisition of mentalising is part of an intersubjective process between the infant and caregiver (see Gopnik, 1993, for a highly elegant elaboration of such a model). In our view, the caregiver facilitates the creation of mentalizing models through complex linguistic and quasi-linguistic processes, primarily through behaving towards the child in such a way that leads him eventually to see that his own behavior may be best understood by assuming that he has ideas and feelings which determine his actions, and the reactions of others to him, which can then be generalized to other similar beings. The caregiver approaches the crying infant with a question in her mind: "Do you want your nappy changed?" "Do you need a cuddle?" The sensitive caregiver is unlikely to address the situation teleologically, without having the person in mind, so is unlikely to say to herself, "Are you wet around your bottom?" or "Have you been standing alone too long?" The sensitive caregiver can cover the gap between a focus on physical reality and internally-directed attention, sufficiently for the child to identify contingencies between internal and external experience. Ultimately, the child arrives at the conclusion that the caregiver’s reaction to him may be understood as rational given the assumption of an internal state of belief or desire within himself. Unconsciously and pervasively, the caregiver ascribes a mental state to the child with her behavior, treats the child as a mental agent, which is perceived by the child and used in the elaboration of teleological models, and then in the development of a core sense of mental selfhood. We assume that this, by and large, is a mundane process, happening routinely throughout early life, not reflected on, and so rarely modified. Caregivers, however, differ in their ways of carrying out this natural human function. Some may be particularly alert to the earliest indications of intentionality, others may need stronger indications before perceiving the child’s mental state and modifying their behavior accordingly. Others, as we described in the context of early infancy, may systematically misperceive the child’s states of mind, with resulting deformation of the child’s sense of himself.
The parent’s capacity to observe the moment to moment changes in the child’s mental state, then, lies at the root of sensitive caregiving, which is viewed by attachment theorists as the cornerstone of secure attachment (e.g. Ainsworth et al., 1978; Grossmann, Grossmann, Spangler, Suess, & Unzner, 1985; Isabella & Belsky, 1991). Secure attachment in its turn provides the psychosocial basis for acquiring an understanding of mind. The secure infant feels safe in making attributions of mental states to account for the behavior of the caregiver. By contrast the avoidant child to some degree shuns the mental state of the other, while the resistant child focuses on his own state of distress to the exclusion of close intersubjective exchanges. Disorganized infants may represent a special category; hypervigilant of the caregiver’s behavior they use all cues available for prediction and may be acutely sensitized to intentional states, and thus may be more ready to construct a mentalized account of the caregiver’s behavior. We would argue (see below) that in such children mentalization may be evident but it does not have the central and effective role in self-organization which characterizes securely attached children.
We believe that most important for the development of mentalizing self-organization is that exploration of the mental state of the sensitive caregiver enables the child to find in her mind an image of himself as motivated by beliefs, feelings and intentions, in other words, as mentalizing. There is considerable evidence to support the view that secure attachment enhances the development of inner security, self worth and autonomy (e.g. Londerville & Main, 1981). Disorganized infants, even if they acquire the skill of mentalization, fail to integrate this with their self-organization. There may be a number of linked reasons for this: a) the caregiver of the disorganized infant is less likely to be reliably contingent in responding to the infant’s self-state, and further to show systematic biases in her perception and reflection of his state; b) the mental state of the caregiver evokes intense anxiety through either frightening behavior suggesting malevolence towards the child, or behavior suggesting fear, which may include fear of the child himself; c) the child needs to use disproportionate resources to understand the parent’s behavior, at the expense of reflecting on self-states. These factors combine, perhaps, to make disorganized infants become keen readers of the caregiver’s mind under certain circumstances, but (we suggest) poor readers of their own mental states.
3. The move from dual to singular psychic reality
In two previous papers we have used both clinical and research evidence to show that the normal experience of psychic reality is not an inherent property of the mind, but rather a developmental achievement (Fonagy & Target, 1996; Target & Fonagy, 1996). It is the consequence of the successful integration of two distinct modes of differentiating internal from external. We see the child’s development as normally moving from an experience of psychic reality in which mental states are not depicted as representations, to an increasingly complex view of the internal world, which has as its hallmark the capacity to mentalise, to think flexibly about thoughts and feelings in others and in oneself. Initially, the child’s experience of the mind is as if it were a recording device, with exact correspondence between internal state and external reality. We use the term "psychic equivalence" to denote this mode of functioning, to emphasise that for the young child mental events are equivalent in terms of power, causality and implications, to events in the physical world. Equating internal and external is inevitably a two-way process. Not only will the small child feel compelled to equate appearance with reality (how it seems is how it is), but also internal representations, distorted by phantasy, will be projected onto external reality in a manner unmodulated by awareness that the experience of the external world might be misconstrued in this way.
Perhaps because it can be terrifying for thoughts and feelings to be experienced as concretely ‘real’, the infant develops an alternative way of construing mental states. In "pretend mode", the child experiences feelings and ideas as totally representational, or symbolic, as having no implication for the world outside. Even though the child of two years knows that his pretence to be a policeman is not real, this is not because he understands that he is being a ‘pretend policeman’, but rather because the mode of psychic reality which prevails at that time presupposes strict separation from external reality (Gopnik & Slaughter, 1991). Thus his play by itself it can form no bridge between inner and outer reality. Only gradually, and through the close participation of another mind which can simultaneously hold together the child’s pretend and serious perspectives, does the integration of these two modes give rise to a psychic reality in which feelings and ideas are known as internal, yet in close relationship with what is outside (Dunn, 1996).
Normally, the child then integrates these alternative modes to arrive at mentalization, or reflective mode, in which mental states can be experienced as representations. Inner and outer reality can then be seen as linked, yet they are accepted as differing in important ways, and no longer have to be either equated or dissociated from each other (e.g. Gopnik, 1993). Mentalization comes about through the child’s experience of his mental states being reflected on, for instance through secure play with a parent or older child. In playfulness, the caregiver gives the child’s ideas and feelings (when he is "only pretending") a link with reality, by indicating the existence of an alternative perspective, which exists outside the child’s mind. The parent or older child also shows that reality may be distorted by acting upon it in playful ways, and through this playfulness a pretend but real mental experience may be introduced.
The child’s development and perception of mental states in himself and others thus depends on his observation of the mental world of his caregiver. He is able to perceive mental states when the caregiver is in a shared pretend mode of playing with the child (hence the association between pretend and early mentalization), and many ordinary interactions (such as physical care and comforting, conversations with peers) will also involve such shared mentation. This is what makes mental state concepts such as thinking inherently intersubjective; shared experience is part of the very logic of mental state concepts.
To pre-empt this afternoon’s topic somewhat, in traumatized children, intense emotion and conflict lead to a partial failure of this integration, so that aspects of the pretend mode of functioning become part of a psychic equivalence manner of experiencing reality. This may be because where maltreatment or trauma has occurred within the family, the atmosphere tends to be incompatible with the caregiver "playing with" the most pressing aspects of the child's thoughts; these are often disturbing and unacceptable to the adult, just as they are to the child. The rigid, controlling behavior of the pre-school child with a history of disorganized attachment is thus seen as arising out of a partial failure on the part of the child to move beyond the mode of psychic equivalence in relation to specific ideas or feelings, so that he experiences them with the intensity that might be expected had they been current, external events.
In the insecure relationship, what is internalised by the child will tend to be of the caregiver’s defences, and the infant’s distress is repeatedly avoided. Most importantly, the process of self-development has been compromised. Insecure parent-child relationships may be regarded as laying the groundwork for subsequent distortions of personality development in one of two ways. These correspond to the two modes of experiencing psychic reality. The mother may echo the child’s state without modulation, as in the mode of psychic equivalence, concretising or panicking at the child’s distress. Alternatively she may avoid reflection on the child’s affect through a process akin to dissociation, which effectively places the mother in a pretend mode, unrelated to external reality, including the child. The mother may then ignore the child’s distress, or translate it into illness, tiredness, and so on. Both strip the child’s communication of the potential for meaning which he can recognise and use. It may also lead to a currency between mother and child of interpretation of feelings in physical terms, so that the physical state is the "real" thing. Lynne Murray (1997), in her work with mothers suffering from puerperal depression, has provided some vivid illustrations of such mothers offering an alternative reality, marked by the exaggeration associated with pretence, but not related to the infant’s experience. Psychoanalytic observers of this type of interaction would rapidly identify the operation of massive denial and even manic defence. The infant has not been able to find a recognisable version of his mental states in another person’s mind, and the opportunity to acquire a symbolic representation of those states has been lost and a disorganised pattern of attachment may follow.
Conclusion
In summary, the securely attached child perceives in the caregiver’s reflective stance an image of himself as desiring and believing. He sees that the caregiver represents him as an intentional being, and this representation is internalised to form the self. "I think therefore I am" will not do as a psychological model of the birth of the self; "She thinks of me as thinking and therefore I exist as a thinker" perhaps comes closer to the truth. If the caregiver’s reflective capacity has enabled her accurately to picture the child’s intentional stance, then he will have the opportunity to "find himself in the other" as a mentalising individual. At the core of our selves is the representation of how we were seen. Our reflective capacity is thus a transgenerational acquisition. We think of others in terms of desires and beliefs because, and to the extent that, we were thought of as intentional beings. Only following this process of internalisation, can the development of awareness of mental states in oneself be generalised to others including the caregiver.
The theory of a transgenerational reflective function has these components: 1. We assume that the internalisation of second order representations of internal states depends upon the sensitive reflection of the caregiver and it offers the building blocks with which a reflective internal working model is constructed. 2. The gradual move from a teleological to an intentional stance is intrinsically linked to the child’s experience of safety in exploring the caregiver’s mind to ferret out the feelings and thoughts that might account for her behaviour. Needless to say, this is easiest and safest to do in the context of a secure attachment relationship. 3. The caregiver makes a further important contribution, perhaps most important at a somewhat later stage. Prototypically, while engaging in pretend play with the child, the caregiver simultaneously engages the child’s internal world while retaining an external reality-based perspective. This is analogous to psychoanalytic discussions of the cognitive impact of the oedipal triad, where the shared reality of two people is suddenly experienced from the point of view of the third. The parents’ engagement in the child’s internal world moves the child beyond the conception of their mind as a replica of the external world.
These three components (the second-order representation of affect, the intentional representation of the caregiver and ultimately the intentional representation of the self) equip the child to confront a sometimes unduly harsh social reality. I shall go on to argue that the robust establishment of reflective function has a protective effect and, by contrast, its relatively fragile status indexes a vulnerability to later trauma. Secure attachment and reflective function are, I believe, overlapping constructs and the vulnerability associated with insecure attachment lies primarily in the child’s diffidence in conceiving of the world in terms of psychic rather than physical reality. Given trauma of sufficient intensity, even a secure bond may sometimes crumble and in the absence of psychosocial pressures, reflective function may offer only marginal developmental advantage. To understand severe personality disorder, as I hope we shall see, it is important we are attuned to our patient’s capacity to use the language of mental states for self organisation as well as social understanding.
Go to Reference Page
Peter Fonagy, PhD, FBA
Freud Memorial Professor of Psychoanalysis, UCL
Director of Research, The Anna Freud Centre
Co-ordinating Director, Child and Family Center and Center for Outcomes Research and Effectiveness, Menninger Foundation
Address for correspondence:
Sub-Department of Clinical Health Psychology
University College London
Gower Street
London WC1E 6BT
E-mail: p.fonagy@ucl.ac.uk
Return to Home Page
Return to Papers
To comment on this paper, contact Peter Fonagy, PhD, FBA p.fonagy@ucl.ac.uk
*DO NOT REPRODUCE WITHOUT PERMISSION*
