2. 680 P. Fonagy and M. Target
ingful and predictable. As children learn to has great generalizability and explanatory
understand people’s behavior, they can flexi- value. Recent philosophers of mind (Hopkins,
bly activate, from multiple sets of self–other 1992; Wollheim, 1995) have extended Den-
representations organized on the basis of prior nett’s approach to unconscious processes.
experience, the one(s) best suited to respond They illustrated that one of Freud’s most sub-
adaptively to particular interpersonal transac- stantive contributions was to extend folk psy-
tions. chology to unconscious mental states, a the-
The interdependence of reflective function ory of unconscious mind, thus making those
as it applies to others and to the self was high- aspects of behavior meaningful which—using
lighted by the second pioneer of psychologi- the ordinary constructs of intentionality—
cal self theory, Cooley (1902/1964): “The make little sense (e.g., dreams, neurotic symp-
thing that moves us to pride and shame is not toms, humor). These behaviors may be under-
the mechanical reflection of ourselves, but an stood if we add unconscious beliefs, thoughts,
imputed sentiment, the imagined effect of this and feelings to our everyday model of the
reflection upon another’s mind” (p. 153). De- mind.
velopmentally, this may be thought to imply Extending these ideas, we consider reflec-
that a mental operation is required in early tive function to be the mental function which
childhood, to derive the self-state from the ap- organizes the experience of one’s own and
perception of the mental state of the other. others’ behavior in terms of mental state con-
Exploring the meaning of others’ actions is structs. Reflective function concerns knowl-
then a precursor of children’s ability to label edge of the nature of experiences which give
and find meaningful their own psychological rise to certain beliefs and emotions, of likely
experiences. This ability arguably underlies behaviors given knowledge of beliefs and de-
the capacities for affect regulation, impulse sires, of the expectable transactional relation-
control, self monitoring, and the experience of ships between beliefs and emotions, and of
self-agency, the building blocks of the organi- feelings and beliefs characteristic of particular
zation of the self. In previous papers (e.g., Fo- developmental phases or relationships. Its es-
nagy, Steele, Moran, Steele, & Higgitt, 1991a) sence is not that the individual should be able
we have labeled the predisposition to under- to articulate this theoretically, and this is clear
stand behavior in mental state terms reflective in our operationalization (Fonagy, Steele,
self-function, or more simply reflective func- Steele, & Target, 1997). Individuals differ in
tion. the extent to which they go beyond observ-
The notion of reflective function is rooted able phenomena to explain their own or oth-
in Dennett’s (1978, 1987) proposal that three ers’ actions in terms of beliefs, desires, plans,
stances are available in the prediction of be- and so on. This undoubtedly high level cogni-
havior: the physical stance, the design stance, tive capacity is, we believe, an important de-
and the intentional stance. Dennett’s thesis is terminant of individual differences in self-or-
that explanation in terms of beliefs and de- ganization, intimately involved with many
sires, so-called intentional states, provides defining features of selfhood such as self-con-
good grounds for predicting human behav- sciousness, autonomy, freedom, and responsi-
ior—the only grounds accessible to all of us; bility (Bolton & Hill, 1996; Cassam, 1994).
this knowledge is embodied in the theory of Intentional stance, in the broad sense consid-
mind of folk psychology (see Churchland, ered here (i.e., including apparently irrational
1986; Mele, 1992).2 unconscious motives), explains one’s own be-
Theory of mind is an interconnected set of havior and therefore creates the continuity of
beliefs and desires, attributed to explain a per- self-experience which is the underpinning of
son’s behavior. The theory of mind concept a coherent self-structure.
It is important that reflective function is
2. Dennett’s formulation is unnecessarily restrictive (Bol-
not conflated with introspection. Bolton and
ton & Hill, 1996). It does not address predicting the Hill (1996) note that the weakness of intro-
behavior of systems which do not function rationally. spection is to define mental states in terms of
3. Attachment and reflective function 681
conscious motivation rather than, as here, in do not ask what the child feels about the men-
terms of their capacity to regulate behavior. tal states he encounters in others. Yet, in this
Introspection or self-reflection is quite differ- context at least, the question of knowledge
ent from reflective function as the latter is an and that of emotional investment are evi-
automatic procedure, unconsciously invoked dently closely related. The child may know
in interpreting human action. We see it as an what the other feels but care little or not at all
overlearned skill, which may be systemati- about this; alternatively this information, for
cally misleading in ways much more difficult some youngsters, may be an issue of survival.
to detect and correct than mistakes in con- The emotional significance of mental states
scious attributions would be. Reflective func- determines the evolution of the capacity or
tion similarly lends a shape and coherence to structure available for processing, but this is
self-organization which is outside awareness, not usually addressed. In current models of
in contrast to introspection, which has a clear theory of mind development the child tends to
impact on experience of oneself. be seen as an isolated processor of informa-
Our central concern here is the acquisition tion, engaged in the construction of a theory
of reflective function and the light this might of mind using biological mechanisms which
cast on the development of self-organization. may fail if the child’s endowment is poor.
Baron–Cohen and Swettenham (1996) appro- This, from the viewpoint of developmental
priately ask “ . . . how on earth can young psychopathology and its psychosocial treat-
children master such abstract concepts as be- ment, is a barren picture which ignores the
lief (and false belief) with such ease, and central role of the child’s emotional relation-
roughly at the same time the world over?” (p. ship with the parents or other caregivers in
158). Their answer is that of modularity theo- fostering the capacity to understand interac-
rists, along the lines of Chomsky’s solution to tions in terms of mental states. The develop-
the problem of the acquisition of a knowledge ment of children’s understanding of mental
of syntax. They postulate an innate (learning) states is embedded within the social world of
mechanism with a specific location in the the family, with its interactive network of
brain (see also Leslie, 1994; Segal, 1996). complex and at times intensely emotionally
Other current psychological theories stress the charged relationships, which, after all, consti-
cognitive precursors of theory of mind. Some tute the primary content of early reflection.
favor the folk psychology, theory–theory, ap- Therefore it should not surprise us that the na-
proach assuming that the child evolves a sci- ture of family interactions, the quality of pa-
entific theory-like network of interdependent rental control (Dunn, Brown, Somkowski,
propositions about the mind on the basis of Telsa, & Youngblade, 1991b), parental talk
experience (e.g., Botterill, 1996; Gopnik, about emotions (Denham, Zoller, & Cou-
1996). Others assume that theory of mind is choud, 1994), and the depth of parental dis-
acquired via simulation of the mental state of cussion involving affect (Dunn, Brown, &
the other, either through making inferences Beardsall, 1991a) are all strongly associated
from what we ourselves would do in the with the acquisition of the intentional stance
imagined circumstances (e.g., Goldman, 1993; in observational studies. The involvement of
Harris, 1992) or an even more radical assump- the family in the child’s acquisition of a the-
tion of imagined transformation into the other ory of mind is further highlighted by the ro-
which does not involve introspection or infer- bust finding that the presence of siblings in
ence (Gordon, 1995). the family appears to improve the child’s per-
Both simulation and theory–theory models formance on a range of false-belief tasks (Jen-
may appear to emphasize social learning as- kins & Astington, 1996; Perner, Ruffman, &
pects of the development of mentalization but, Leekman, 1994; Ruffman, Perner, Naito, Par-
upon closer scrutiny, their focus is at the level kin, & Clements, in press).
of mechanism rather than content. They ques- Modular accounts of theory of mind devel-
tion how and when the child acquires knowl- opment have some difficulty with such data.
edge of other minds in an abstract sense and Both the theory–theory and the simulation ac-
4. 682 P. Fonagy and M. Target
counts also fall short of adequately addressing Jaffe (1997) have shown that interaction be-
the social origins of this critical aspect of self- tween baby and mother shows both self-regu-
organization. In the theory–theory account lation and sensitivity to the state of the other.
mental concepts are thought to develop within Their facial expressions showed rapid fluctua-
a network of interdependent concepts on the tions: affect, space, and degree of contact in
basis of data from the social world, but the each 1⁄ 12 s time period responded to the ex-
social world does not generally “give” con- pression of the other in the previous period,
cepts to the child—it provides him with data presumably on the basis of schemata of antici-
for concept building. In the simulation model pated reactions. High coordination predicted
mental state concepts are thought to arise later good cognitive performance, whereas
from introspection, but this begs the question lower levels of coordination were optimal for
of how children come to think of their own secure attachment and easy temperament. In-
mental states in terms of feelings, beliefs, terestingly, security with the mother in the
wishes, and so on. This paper attempts to ex- Strange Situation at 1 year was better pre-
plore the role of parent–child relationships in dicted by coordination with a stranger than
the transformation of prereflective experience with the mother at 4 months.
of mental states into reflective understanding Interactions at this stage may be argued to
of them. Within this social context both social be presymbolic, in the sense that they are non-
models of mentalization may have their place; mentalistic; the infant is not required to repre-
the predominance of one or other route to un- sent the thoughts or feelings of the caregiver.
derstanding the mind may be a function of in- However, they involve reference to future
dividual differences between children and be- states such as goals as explanatory constructs
tween environments, but, in our view, a in the interpretation of the behavior of the
satisfactory model must have the child’s rela- other. Thus they can be used to predict behav-
tionships with attachment figures as its start- ior although these structures would be limited
ing point. in their capacity to modify behavior. Recent
experimental work by Gergely and Csibra
(e.g., Gergely, Nadasdy, Csibra, & Biro, 1995)
Developmental Roots of Reflective suggests that the infant’s perception of social
Function in Infancy contingencies by the second half of the 1st
“Teleological” stance year is teleological in that they make refer-
ence to future states (goals) as explanatory
There is general agreement that self-organiza- entities in the interpretation of behavior based
tion initially entails the integration of body- on the principle of “rational action.” The tele-
related experiences, defining the physical ological stance is applied by infants to human
boundaries of self and world (e.g., Brown- and nonhuman objects alike. Studies by Ger-
ell & Kopp, 1991). Once the physical self is gely and Csibra (1997) demonstrated that in-
established, social exchanges, the identifica- fants express surprise when nonhuman but
tions of social boundaries and, somewhat moving objects (such as various sized discs in
later, the identification of social causality be- a computer-generated animated display) ap-
come central self-functions. The caregiver’s pear to act “irrationally”—not choosing the
recognition of the child’s intentional stance, optimal action given specific goals and reality
however, is communicated nonverbally, be- constraints. The infant is assumed to extend
ginning at birth. Between birth and 5 months, teleological models beyond the prediction of
face-to-face exchanges of affective signals be- human behavior. Teleological models, how-
tween infant and caregiver (Beebe, Lach- ever, evolve into mentalizing ones in the re-
mann, & Jaffe, 1997; Tronick, 1989) play a stricted domain of human action. They be-
key role in the development of the child’s rep- come fully mentalizing once representations
resentation of affect. of goal states come to be thought of as desires,
For example, using a microanalytic obser- and constraints come to be thought of in terms
vational paradigm, Beebe, Lachmann and of the agent’s beliefs about physical reality.
5. Attachment and reflective function 683
The infant’s behaviors in dyadic interac- contingent actions of others. Taking Stern’s
tions are underpinned by an evolving model (1985) and Neisser’s formulation of the inter-
of rational action by the caregiver. We would personal self together, we can identify three
argue that the development from teleological aspects of the intersubjective development of
to mentalizing models will depend upon the the self, which Mundy and Hogan (1994) term
quality of interpersonal interactions between instrumental action states, sensory or percep-
the infant and the caregiving adult. It should tual action states, and affective action states.
be noted that although such models may Rogers and Pennington (1991) offered a
merely represent rational action, it is the per- model of the cognitive underpinnings for such
ceived rather than actual rationality of an act an intersubjective process in their concept of
which defines the teleological model. Thus representational mapping (the process of co-
misapprehension of reality constraints (e.g., ordinating representations of self and other)
assumed dangerousness) will provide and cre- which is thought to underlie the sharing of af-
ate a model where action which is clearly irra- fect, attention, and higher order aspects of
tional from an external standpoint is neverthe- cognition such as beliefs. The existence of im-
less seen as based on the principle of rational itation skills from the neonatal stage repre-
action. The predictive significance of the in- sents strong evidence for the model (Meltzoff,
fant’s response to a stranger in the Beebe 1993). The acquisition of an appreciation of
study suggests that representations (working mental states, however, goes beyond mir-
models) of self–other relations even when not roring.
yet mentalized begin to vary in quality in the The development of an understanding of
1st year, and this quality is related to infant– affect in self and other may be a good illustra-
caregiver interactions, as observed in the lab tion of the role of representational mapping in
situation. If sufficiently coherent to be gener- the development of reflective abilities (Ger-
alized to other relationships in characteristic gely & Watson, 1996; Target & Fonagy,
ways, they may index processes crucial to the 1996). Anxiety, for example, is for the infant
creation of a secure mother–infant bond. a confusing mixture of physiological changes,
ideas, and behaviors. When the mother re-
flects, or mirrors, the child’s anxiety, this per-
Representational mapping
ception organizes the child’s experience, and
Representational mapping is likely to under- he now “knows” what he is feeling. The
pin the gradual move in infancy from teleo- mother’s representation of the infant’s affect
logical to mentalizing models of mind. Be- is represented by the child and is mapped on
tween 6 and 18 months the child becomes to the representation of his self-state. The dis-
increasingly able to match his mental state crepancy between these is helpful insofar as it
`
with that of the caregiver vis-a-vis a third ob- provides organization for the self-state and
ject or person, as, for example, in requesting thus the caregiver’s mirroring can become the
or joint attention (Bretherton, 1991). The higher order representation of the child’s ex-
communication is evidently deliberate, since perience. Within this model mirroring would
children at this phase try to repair failed com- be expected to fail if it is either too close to
municative bids and thus show some recogni- the infant’s experience or too remote from it.
tion of awareness and agency in self and other If the mirroring is too accurate, the perception
including affective states, perceptions, and in- itself can become a source of fear, and it loses
tentions (Stern, 1985; Wellman, 1993). Neis- its symbolic potential. If it is unavailable, or
ser (1991) suggested that based on perceptual is contaminated with the mother’s own preoc-
processes, two preconceptual aspects of the cupation, the process of self-development is
self emerge: the ecological and the interper- profoundly compromised. We may presume
sonal. While the former involves self-aware- that individuals for whom the symptoms of
ness in reference to perception of nonsocial anxiety signify catastrophes (e.g., heart attack,
surroundings, the latter is generated via the imminent death, etc.) have metarepresenta-
coperception of actions of the self and related tions of their emotional responses which can-
6. 684 P. Fonagy and M. Target
not be limited in intensity through symboliza- plays in others as well as arriving at the regu-
tion, perhaps because the original mirroring lation and control of his own emotions. The
by the primary caregiver exaggerated the in- representational mapping of emotion displays
fant’s emotions. and self-experience is seen here as a prototyp-
Admittedly this is a speculative model, but ical instance of caregiver sensitivity, which,
it is empirically testable. It might help answer as we shall attempt to demonstrate, is likely
the thorny question of why individuals with to be an important component of the develop-
panic disorders attribute immense significance ment of mentalizing. The sensitivity of the
to physiologically relatively mild levels of caregiver prompts the child to begin organiz-
disequilibrium. The suggestion here is that the ing self-experience according to clusters of re-
metarepresentation, or symbolic representa- sponses which will eventually come to be ver-
tion, of affect in these cases contains too bally labeled as specific emotions (or desires).
much of the primary experience; hence, in- The high contingent response is the means by
stead of labeling the experience having the which this mapping can take place. The
potential to attenuate it, it tends to stimulate child’s affective experiences are given further
and exacerbate symptoms of the affect state, meaning by becoming associated with clusters
which in turn accentuates the secondary ex- of reality constraints within the parent–infant
pression, in a cycle of escalating panic.3 In interaction (leading to rudimentary beliefs
collaboration with George Gergely, we are about the causes and consequences of his
designing a series of studies of the infant’s emotional state).
emotional understanding which will more di-
rectly test these ideas. In a recent study (Fo-
Transmission of attachment security
nagy et al., 1995), we have confirmed that
mothers who soothe their distressed 8-month- The attachment system (Bowlby, 1969, 1973,
olds most effectively following an injection 1980) is intimately connected with the pro-
rapidly reflect the child’s emotion, but this cess of representational mapping and the de-
mirroring is mixed with other affects (smiling, velopment of the reflective function of the self.
questioning, mocking display, and the like). There is general agreement that, as the self
In displaying such “complex affect” (Fo- ´ exists only in the context of the other, the de-
nagy & Fonagy, 1987) they ensure that the
´ velopment of the self is tantamount to the ag-
infant recognizes their emotion as analogous gregation of experiences of self in relation-
to, but not isomorphic with, their experience ships (e.g., Crittenden, 1994; Sroufe, 1990).
and thus the process of symbol formation may Psychoanalytic object relations (Kernberg,
begin. In this way, the representational map- 1982; Winnicott, 1965) and attachment theo-
ping between affect of self and emotions of rists (Bowlby, 1980) are in agreement that
others, the exchange of affect between young repeated, invariant aspects of self–other rela-
child and caregiver, provides a unique source tions are abstracted into internal representa-
of information to the child about his own in- tional mental models and structured, to use
ternal states. Kernberg’s term, into self–other–affect triads,
We suggest that the meaning or sense of or internal working models, according to
affect develops out of the integrated represen- Bowlby. Although in its original formulation
tation of the affect in self and other. The com- the concept of internal working model lacked
bination of the representation of self-experi- specificity (Dunn, 1996), more recent empiri-
ence and the representation of the reaction of cal work by psychoanalysts has greatly im-
the caregiver elaborates the child’s teleologi- proved this (Horowitz, 1995; Luborsky & Lu-
cal model of the mind, and ultimately enables borsky, 1995).
him to interpret and understand affective dis- At the same time, cognitive scientists have
elaborated the notion of procedural memories
3. In terms of linguistic theory, one may say that the sig-
based on the nonconscious implicit use of past
nifier is not sufficiently “demotivated”; in other words experience (e.g., Johnson & Multhaup, 1992;
it resembles the signified too closely. Schacter, 1992). There is general agreement
7. Attachment and reflective function 685
that the memory system is at least of a dual Steele, & Fonagy, 1996). The small overlap
nature with two relatively independent, neuro- between the two sets of classifications could
logically and psychologically homogeneous, be equally well accounted for by assuming a
systems underpinning it. In addition to the au- temperament factor or by the generalization
tobiographical memory, which is at least in of the child’s behavior with the mother (re-
part accessible to awareness, an important flecting her attachment classification) to his
component to memory is a nonvoluntary sys- behavior with the father. The results suggest
tem which is implicit, principally perceptual, that the infant develops independent models
nondeclarative, and nonreflective. It is possi- (self–other schemata) for his major attach-
ble that it is, at least in certain respects, more ment relations based on his past history of in-
dominated by emotional and impressionistic teractions with each of those individuals. In
information than its autobiographical counter- turn, these interaction experiences are indexed
part (e.g., van der Kolk, 1994). It stores the by the caregiver’s representation of her or his
“how” of executing sequences of actions, mo- attachment history.
tor skills being prototypical instances. The There has been considerable research on
procedural knowledge that it contains is ac- the manner in which representations of attach-
cessible only through performance. It mani- ment might influence the caregiver’s behavior
fests itself only when the individual engages with the child. Van IJzendoorn’s (1995) com-
in the skills and operations into which knowl- prehensive meta-analysis identifies a “trans-
edge is embedded. Given these features, it mission gap,” to the extent that the variability
seems likely that the schematic representa- which AAI narratives and SSn classifications
tions postulated by attachment and object re- share is not accounted for by observational
lations theorists are most usefully construed data concerning the sensitivity of caregiver
as procedural memories, the function of behavior. Indeed, studies of the AAI-SSn as-
which is to adapt social behavior to specific sociation, which concurrently measured the
interpersonal contexts. sensitivity of caregiver–infant interaction, have
The classification of patterns of attachment yielded negative (Ward & Carlson, 1995) or
in infancy (Ainsworth, Blehar, Waters, & inconclusive results (van IJzendoorn, Kranen-
Wall, 1978) taps into procedural memory burg, Zwart–Woudstra, Van Busschbach, &
(Crittenden, 1990; Fonagy, 1995). The strength Lambermon, 1991). Previously, we have sug-
of the Strange Situation (SSn) as a method gested that the transmission gap may be a
of psychological assessment is to provide a consequence of the limitations of measures of
powerful analogue of past situational contexts sensitivity (Fonagy et al., 1995). Sensitivity is
within which knowledge concerning the a generic construct covering a wide range of
“how” of behavior with a specific caregiver is parental behaviors (Belsky, Rosenberger, &
accrued. In this sense attachment is a skill, Crnic, 1995). Not all of these may be equally
one which is acquired in relation to a specific powerful in engendering secure attachment. If
caregiver encoded into a teleological model of secure attachment is conceived of as the ac-
behavior. In the London Parent–Child Study, quisition of procedures of goal oriented ra-
the Adult Attachment Interview (AAI), ad- tional action for the regulation of aversive
ministered before the birth of the first child to states of arousal within relationships (Cas-
100 predominantly middle class primiparous sidy, 1994; Sroufe, 1996), it is argued that
parents, was tested as a predictor of attach- these would be most consistently acquired
ment classification at 1 year to mother and at and coherently represented when the child’s
18 months to father (Fonagy, Steele, & Steele, acute affective state is accurately, but not
1991b). There was only a marginally signifi- overwhelmingly, reflected back to the child.
cant association between the attachment clas- The child who looks for a way of manag-
sification with mother and that with father. ing his distress identifies in the response of
However, both SSn results were powerfully the caregiver a representation of his mental
predicted by the attachment classification of state which he may internalize and use as part
the respective parent on the AAI (Steele, of a higher order strategy of affect regulation.
8. 686 P. Fonagy and M. Target
The secure caregiver soothes by combining a tenden (1988; Crittenden & DiLalla, 1988) re-
“mirror” with a display incompatible with the ports that maltreated toddlers display falsely
child’s affect (thus perhaps implying coping). positive affect which does not match their true
This formulation of sensitivity has much in feelings. At an extreme, the internalization of
common with the British psychoanalyst, Wil- the caregiver’s defenses can not only lead to
fred Bion’s (1962) notion of the role of the a failure to adequately represent and display
mother’s capacity to mentally “contain” the actual emotional experience, but also to the
affect state intolerable for the baby, and re- construction of an experience of self around
spond in terms of physical care in a manner this false internalization (Winnicott, 1960).
that acknowledges the child’s mental state yet While the experience of “putting on an
serves to modulate unmanageable feelings. act” may be common, particularly in adoles-
The finding that the clarity and coherence of cence (Harter, Marold, Whitesell, & Cobbs,
the mother’s representation of the child medi- 1996), here we are referring to the highly dis-
ates between her attachment status and her be- tressing experience of personality disordered
havior is certainly consistent with this model children who feel a sense of alienation from
(Slade, Belsky, Aber, & Phelps, in press). their core self (Bleiberg, 1994). A strategy
Ratings of the quality of the reflective func- many such children adopt later in develop-
tion of each caregiver were found indepen- ment is to attempt to externalize this false part
dently to predict the child’s security of attach- of their self-representation, and manipulate
ment in the London Parent–Child Project the behaviors of others around them so these
(Fonagy et al., 1991a). match the incongruent self-representation. We
If secure attachment is the outcome of suc- would argue that this explains the strangely
cessful containment, insecure attachment may coercive behavior with the caregiver of pre-
be seen as the infant’s identification with the school children whose attachment in infancy
caregiver’s defensive behavior. Proximity to was classified as disorganized (Crittenden,
the caregiver is maintained at the cost of a 1992; Main & Cassidy, 1988). In a desperate
compromise to reflective function. A dismiss- way they try, we suggest, to provoke behavior
ing (Ds) caregiver may altogether fail to mir- in another person which expresses part of
ror the child’s distress because of the painful their self-representation experienced as
experiences this evokes for her or because she “alien,” they can then experience a more co-
lacks the capacity to create a coherent image herent residual self (Fonagy & Target, 1995).
of the child’s mental state. By contrast, the
preoccupied (E) caregiver may represent the
Secure infant becomes mentalizing child
infant’s state with amplification and insuffi-
cient marking, or complicated by responses to There is general agreement that the “harmoni-
the parent’s ambivalent preoccupation with ousness of the mother–child relationship con-
her own experience, so much so that the sym- tributes to the emergence of symbolic
bolic potential of the exchange is lost. In both thought” (Bretherton, Bates, Benigni, Camai-
cases the infant internalizes the caregiver’s at- oni, & Volterra, 1979, p. 224). Bowlby recog-
titude and “this dysynchrony becomes the nized the significance of the emergence of
content of the experience of the self” (Critten- “the child’s capacity both to conceive of his
den, 1994, p. 89). mother as having her own goals and interests
separate from his own and to take them into
account” (1969, p. 368). Moss, Parent, and
Infant attachment and developing self
Gosselin (1995) reported that attachment se-
We may speculate about the impact of this on curity with mother was a good concurrent pre-
the development of the child’s sense of self. dictor of metacognitive capacity in the child
We know that avoidant infants respond to in the domains of memory, comprehension
separation with minimal displays of distress and communication. The Separation Anxiety
while experiencing considerable physiological Test, a projective test of attachment security,
arousal (Spangler & Grossman, 1993). Crit- has been shown to be a good predictor of be-
9. Attachment and reflective function 687
lief–desire reasoning capacity in 3- to 6-year- of secure attachment and false belief under-
old children when age, verbal mental age, and standing was due to an as yet unknown and
social maturity were all controlled for (Fo- unmeasured third factor, such as tempera-
nagy, Redfern, & Charman, 1997). ment. More plausibly, it could be argued that
We have recently completed a prospective the facilitative effect of secure attachment is
study of the relationship of attachment secu- due to a more relaxed, task-oriented attitude,
rity to mother (1 year) and father (18 months) a general facility to engage in a cognitively
and children’s performance on three tests of demanding task, to relate to an adult experi-
theory of mind at 5 years (Fonagy, Steele, menter in a playful, exploratory way, and so
Steele, & Holder, submitted). Ninety-two of on: that it reflects performance, rather than
96 children tested in the SSn at 12 and 18 competence. This suggestion could be tested
months were seen. Eighty-two percent of using a false-belief task where implicit and
those classified as secure at 12 months with explicit knowledge of false belief is sepa-
mother passed the belief–desire reasoning rately assessed. If attachment security relates
task, whereas 46% of those who had been to performance, then securely attached chil-
classified as insecure failed. Infant–father at- dren would be expected to do better only on
tachment (at 18 months) also predicted the the explicit (verbal/pointing) task. Implicit,
child’s performance, with 77% of infants clas- procedural false belief reasoning would be ex-
sified as secure passing the test compared to pected to be facilitated by secure attachment
55% of children classified as insecure. There only if this was associated with superior re-
was some indication of an additive relation- flective capacity. This study remains to be
ship, in that 87% of children with two secure performed, and is planned in our laboratory.
relationships passed the belief–desire task, In what follows we shall, however hazard-
63% of those with only one secure relation- ously, assume that the relationship between
ship and only 50% of those insecure with both false belief reasoning and security of attach-
did so. A similar but somewhat weaker pat- ment is nontrivial.
tern could be observed with the second-order We then envisage two alternative sets of
false-belief task. Thirty-six percent of those models to explain this relationship: (a) Secu-
secure with both parents passed compared rity of attachment in infancy predisposes chil-
with 23% who were secure with one and 9% dren to benefit from social processes directly
who were insecure with both. facilitating reflective abilities and social un-
In a somewhat smaller but careful longitu- derstanding (mediational models), and (b) se-
dinal study of mother–infant dyads, Meins curity of attachment is an indicator of that
and colleagues (Meins, Fernyhough, Russel, & quality of infant–caregiver relationship which
Clark–Carter, in press) reported that 83% of generates psychological understanding. In this
children who were securely attached in infancy second model, the social processes which ac-
passed a false-belief task at age 4, in compari- celerate the mentalizing quality of self-organi-
son with 33% of insecurely attached peers. At zation are the very same as those which en-
age 5, 85% of securely attached children and sure security of attachment.
50% of insecurely attached ones passed a task Mediational models would require that
requiring an understanding of information ac- specific social processes are shown to be in-
cess. Although, probably because of its small volved in this aspect of the development of
sample, the study was not able to replicate our self-organization, and such social processes
results on the false belief and emotion task, are enhanced in securely attached individuals.
the general trend of the findings confirms that There are at least three candidates which meet
security of attachment is significantly linked these criteria.
to symbolic abilities in general and precocious The first is pretense. There is evidence that
mentalizing in particular. children in their 3rd year who engage more
There are both trivial and substantive ex- readily in cooperative interaction (Dunn et al.,
planations which could account for these find- 1991b), and specifically in joint pretend play
ings. They would be trivial if the association (Astington & Jenkins, 1995; Youngblade &
10. 688 P. Fonagy and M. Target
Dunn, 1995), show superior mentalization and presumes a degree of trust, in so far as the
emotion understanding performance. There is child relies on the other’s version or percep-
a separate body of observations from longitu- tion of reality.
dinal studies of attachment that preschool The second is talking. There is evidence
children securely attached to their mother in that conversations about feelings and about
infancy engage more strongly in fantasy play the reasons behind people’s actions are linked
than avoidant children, whose engagement is to the relatively early achievement of reflec-
low and whose pretend play is impoverished tive function (Dunn & Brown, 1993). Three-
(e.g., Belsky, Garduque, & Hrncir, 1984; year-olds whose mothers spontaneously ex-
Bretherton et al., 1979; Main, Kaplan, & Cas- plained their emotions in a lab task showed
sidy, 1985). There is also evidence that se- enhanced emotion understanding over the
curely attached young children can more eas- subsequent 15 months (Denham et al., 1994).
ily use help from adults to elaborate their play Patterns of mother–child interaction char-
(Meins et al., in press; Slade, 1987). acteristic of secure dyads—shared play, com-
It is highly plausible that joint pretend play forting, or joking—are also contexts within
or playfulness fosters the understanding of which mothers’ explanations of mental states
mental states. Deliberate role-taking is seen as are particularly found to facilitate reflective
integral to the off-line simulation model of the function (Dunn, 1996). Secure attachment may
performance of mentalization tasks (Gold- then engender patterns of verbal interaction be-
man, 1989). Within other models pretend play tween child and caregiver which in turn sup-
is an early manifestation of the theory of mind port thinking about feelings and intentions.
mechanisms (Leslie, 1987). It is an important The central role of language in the acquisi-
puzzle why 3-year-olds can understand that tion of mentalizing capacity was forcefully
someone is entertaining a pretend representa- advanced by Smith (1996), using primate evi-
tion but not a false belief (Harris, Kava- dence. Even more pertinent is Harris’s (1996)
naugh, & Meredith, 1994), a pretend/real dis- proposal that the experience of engaging in
tinction but not an appearance/reality one conversations per se shows children that peo-
(Flavell, Flavell, & Green, 1987). In the case ple are receivers and providers of information,
of pretend, the representations, while they are whether or not the conversation refers to men-
different from reality, are shared by those en- tal states. The structure of informative conver-
gaged in the pretend game. As Astington sations (e.g., being told about an event one
(1996) put it, “they are intermental, not intra- has not witnessed, dissent and denial, filling
mental” (p. 193). The sharing of representa- in information gaps) implies that partners in
tions different from reality may help in under- a conversation differ in what they know and
standing situations where representations are believe about a shared topic. Effective con-
not only different from reality but are not versation requires that gaps in shared knowl-
shared in a social pretend domain. In joint edge and belief are acknowledged and ad-
pretend play or playfulness the adult adopts dressed. The measurement of attachment in
the child’s mental stance and re-presents it to adults (Main, Kaplan, & Cassidy, 1985) em-
the child in relation to a third object which is phasizes that secure attachment involves
symbolically held in mind by both (Target & greater sensitivity to the rules of conversation.
Fonagy, 1996). The scaffolding provided by The third potential mediator is peer group
the child’s playmate in pretend play (Vygot- interaction. We have already noted that inter-
sky, 1967) not only promotes earlier success action with siblings enhances theory of mind
but is also the mechanism whereby the devel- performance. There is an independent body of
opment of reflection comes about. Lillard evidence which supports a strong link be-
(1993) argued that symbolic play may offer a tween secure attachment in infancy and rat-
“zone of proximal development” for the skills ings of peer competence: social orientation,
which subserve mentalization ability. Chil- reciprocity, popularity, and empathy (e.g.,
dren with a secure attachment history may be Elicker, Englund, & Sroufe, 1992; Park &
more likely to engage in an activity which Waters, 1989).
11. Attachment and reflective function 689
Both simulation theory and theory–theory from Dunn’s work suggests that these differ-
explanations of the development of mentaliza- ent contexts correlate poorly with one another
tion offer good explanations of the facilitative (Dunn, 1996). For example, observational
effect of more intense peer group interaction data shows that individual differences found
(Ruffman et al., in press). Peer group interac- in pretend play, management of conflict, and
tion should increase the opportunities the talking about mental states are not correlated
child has for simulation, imagining what he between social situations (mothers, siblings,
would see, think, feel etc., if he were in an- close friend) although each correlates with
other person’s situation. Equally, interaction sociocognitive assessments (e.g., Young-
with peers or older sibs could be seen from a blade & Dunn, 1995). These findings could
theory–theory perspective as a rich source of suggest that there are a number of indepen-
ideas about how the mind works. An alterna- dent, simultaneous pathways between attach-
tive view may be that enculturation is itself ment, social situations, and social cognition.
the source of the child’s mental state concepts Alternatively, there is the second possibil-
(Astington, 1996). Bruner (1983) proposed ity, that the suggested mediating variables are
that parents’ tendency to treat the infant’s not on the causal path at all, that their correla-
spontaneous gestures as if they were inten- tion with the rate of acquisition of mentaliza-
tional communications leads to infants seeing tion is spurious, that this facility is directly
themselves as having intentions and starting related to the child’s attachment status. Early
to communicate intentionally. The social experience with the caregivers in the 1st year
world (in the first instance, the parent) fosters of life may create a bedrock of theory of mind
the child’s sense of his mental self through competence, helping the child to move from
complex linguistic and interactional pro- a teleological to a mentalizing model of be-
cesses, behaving towards the infant in a way havior. What evidence do we have to support
that leads him eventually to share the assump- such a contention?
tion that his own behavior and (by simulation First, recall that Fonagy, Steele, Steele, and
or the observation of similar interactions be- Holder (1997) found that a mother’s attach-
tween the caregiver and others) that of others ment classification before the birth of the
may be best understood in terms of mental child was a powerful predictor of the child’s
states (Fonagy & Target, 1996; Target & Fo- theory of mind competence at 5 years. Al-
nagy, 1996). Through participation in activi- though, on the face of it, this can be ac-
ties of their culture they come to share their counted for by the mediational models, we be-
culture’s way of understanding people’s ac- lieve that there is now evidence that the
tions. If there is a process of “apprenticeship” caregiver brings something to the parent–
in which peers and caregivers encourage the child relationship, evident even before the
child’s use of mentalizing concepts (Asting- birth, which may be critical in the child’s es-
ton, 1996), then secure attachment may be a tablishment of both secure attachment and
catalyst to this learning process. The greater mentalization.
readiness with which secure children are will- What is this capacity? It is well established
ing to explore and engage with the social that in infancy, mothers of securely attached
world could then account for their mentalizing children are more sensitive to their children’s
skill. needs (Ainsworth, Bell, & Stayton, 1971; Isa-
There is nothing exclusive about these bella, 1993). We have already mentioned that
three mediational models. Pretense often the parent’s capacity to envision the mental
involves the use of mental state language. In- states of her or his own parents is predictive
teraction with peers often involves both lan- of the infant’s security of attachment to each
guage and pretense. In general, social engage- parent (Fonagy et al., 1991a). In a follow-up
ment tends to enhance social understanding, of the same group, the same capacity also pre-
and such engagement is more accessible to se- dicted superior performance on a false belief
curely attached children. There is, however, a task at 5 years, controlling for verbal fluency
problem with a singular model. Evidence in the child. (However, this result was not
12. 690 P. Fonagy and M. Target
found for all tasks which could be thought to ideas and feelings which determine his ac-
index mentalization). tions, and the reactions of others to him,
A path analysis of the above data showed which can then be generalized to other similar
that not all the variance predicted was medi- beings. The caregiver approaches the crying
ated by mother–infant attachment status at 1 infant with a question in her mind: “Do you
year. Mother’s mentalizing ability seemed to want your nappy changed?” “Do you need a
have a direct as well as an indirect relation- cuddle?” The sensitive caregiver is unlikely to
ship with the child’s theory of mind. Thus, the address the situation without having the per-
child’s attachment security was not the only son in mind, so is unlikely to say to herself,
predictor; the mother’s tendency to envision “Are you wet around your bottom?” or “Have
people (including the child) as mental entities you been standing alone too long?” The sensi-
also seemed to be important. tive caregiver can bridge the focus on physi-
The above data suggest that common cal reality and internally directed attention,
mechanisms underpin attachment organiza- sufficiently for the child to identify contingen-
tion in caregiver and infant, and the preco- cies between internal and external experience.
cious emergence of mentalizing in the child. Ultimately, the child arrives at the conclusion
It should be remembered that no clear causal that the caregiver’s reaction to him may be
path was identified among mediational mod- understood as rational given the assumption
els. The relative importance of various po- of an internal state of belief or desire within
tential mediational mechanisms for the at- himself. Unconsciously and pervasively, the
tachment–theory of mind relationship varies caregiver ascribes a mental state to the child
according to context but intergenerational with her behavior, treats the child as a mental
data may be consistent with at least two of the agent, which is perceived by the child and
models (pretense, language). Further experi- used in the elaboration of teleological models,
mental research which manipulates parental and then in the development of a core sense
behavior and explores attachment and theory of mental selfhood. We assume that this, by
of mind task performance (van IJzendoorn, and large, is a mundane process, happening
Juffer, & Duyvesteyn, 1995) will be necessary routinely throughout early life, not reflected
to show whether specific behaviors which en- on, and so rarely modified. Caregivers, how-
gender secure attachment simultaneously en- ever, differ in their ways of carrying out this
hance mentalizing. For such a study to be fea- natural human function. Some may be partic-
sible, we need a model of how attachment ularly alert to the earliest indications of inten-
may directly relate to theory of mind perfor- tionality; others may need stronger indications
mance. Next we outline a tentative model of before perceiving the child’s mental state and
how such a mechanism may operate. modifying their behavior accordingly. Others,
as we described in the context of early in-
fancy, may systematically misperceive the
Reflective parenting and development child’s states of mind, with resulting deforma-
of mentalization tion of the child’s sense of himself.
The child’s development and perception of
We take the view that the acquisition of the mental states in himself and others thus de-
theory of mind is part of an intersubjective pends on his observation of the mental world
process between the infant and caregiver (see of his caregiver. He is able to perceive mental
Gopnik, 1993, for a highly elegant elaboration states, to the extent that his caregiver’s behav-
of such a model). In our view, the caregiver ior implied such states. This he does when the
facilitates the creation of mentalizing models caregiver is in a shared pretend mode of play-
through complex linguistic and quasilinguistic ing with the child (hence the association be-
processes, primarily through behaving to- tween pretend and early mentalization), and
wards the child in such a way that leads him many ordinary interactions (such as physical
eventually to see that his own behavior may care and comforting, conversations with
be best understood by assuming that he has peers) will also involve such shared menta-
13. Attachment and reflective function 691
tion. This is what makes mental state concepts the mental state of the caregiver evokes in-
such as thinking inherently intersubjective; tense anxiety through either frightening be-
shared experience is part of the very logic of havior suggesting malevolence towards the
mental state concepts. child, or behavior suggesting fear, which may
The parent’s capacity to observe the mo- include fear of the child himself; and (c) the
ment to moment changes in the child’s mental child needs to use disproportionate resources
state, then, lies at the root of sensitive care- to understand the parent’s behavior, at the ex-
giving, which is viewed by attachment theo- pense of reflecting on self-states.
rists as the cornerstone of secure attachment These factors combine, perhaps, to make
(e.g., Ainsworth et al., 1978; Grossmann, disorganized infants become keen readers of
Grossmann, Spangler, Suess, & Unzner, 1985; the caregiver’s mind under certain circum-
Isabella & Belsky, 1991). Secure attachment stances, but (we suggest) poor readers of their
in its turn provides the psychosocial basis for own mental states. Thus, in terms of the rival
acquiring an understanding of mind. The se- models of theory of mind development, such
cure infant feels safe in making attributions of children may acquire a theory–theory of
mental states to account for the behavior of mind, but cannot use simulation of mentaliz-
the caregiver. By contrast the avoidant child ing with the same confidence as children
to some degree shuns the mental state of the whose attachment (albeit insecure) is orga-
other, while the resistant child focuses on his nized. The alternative models may be more
own state of distress to the exclusion of close usefully thought of as alternative routes to
intersubjective exchanges. Disorganized in- mentalization, the first (theory–theory) acces-
fants may represent a special category; hyper- sible to all, the second (simulation) more
vigilant of the caregiver’s behavior they use readily available to children whose early at-
all cues available for prediction and may be tachment relationships made such a strategy
acutely sensitized to intentional states, and safely possible.
thus may be more ready to construct a mental-
ized account of the caregiver’s behavior. We
Theoretical model of development
would argue (see below) that in such children
of mentalization
mentalization may be evident but it does not
have the central and effective role in self-or- In previous papers (Fonagy & Target, 1996;
ganization which characterizes securely attached Target & Fonagy, 1996), we have attempted
children. to describe the normal development of reflec-
We believe that most important for the de- tive function in the child of 2–5 years. We
velopment of mentalizing self-organization is suggested that there is a transition from a dual
that exploration of the mental state of the sen- mode of experience to mentalization. Primar-
sitive caregiver enables the child to find in ily from a clinical perspective, we advanced a
her mind an image of himself as motivated by number of propositions concerning the devel-
beliefs, feelings, and intentions, in other opment of the psychological part of the self.
words, as mentalizing. There is considerable These were
evidence to support the view that secure at- 1. Until 3 or 4 years of age, reflective func-
tachment enhances the development of inner tion is characterized by two modes of relating
security, self-worth, and autonomy (e.g., Lon- internal experiences to the external situation:
derville & Main, 1981). Disorganized infants, (a) In a serious frame of mind, the child ex-
even if they acquire the skill of mentalization, pects the internal world in himself and others
fail to integrate this with their self-organiza- to correspond to external reality, and subjec-
tion. There may be a number of linked rea- tive experience will often be distorted to
sons for this: (a) The caregiver of the disorga- match information coming from outside
nized infant is less likely to be reliably (“psychic equivalence mode”), (e.g., Gop-
contingent in responding to the infant’s self- nik & Astington, 1988; Perner, Leekam, &
state, and further to show systematic biases in Wimmer, 1987); and (b) while involved in
her perception and reflection of his state; (b) pretend play, the child knows that internal ex-
14. 692 P. Fonagy and M. Target
perience does not reflect external reality (e.g., Reflective function and self-development
Bartsch & Wellman, 1989; Dias & Harris,
1990), but then the internal state is thought While mentalization may not be an unequivo-
to have no relationship to the outside world cally positive experience, Dunn’s work shows
(“pretend mode”). us at any rate that the understanding of emo-
2. Normally, the child then integrates these tion at 3.5 years predicts a positive perception
alternative modes to arrive at mentalization, of social relations, mature moral sensibility,
or reflective mode, in which mental states can and the understanding of complex emotions
be experienced as representations. Inner and (Dunn, 1996). Stern (1985) pointed out that a
outer reality can then be seen as linked, yet sense of ownership of one’s actions, whether
they are accepted as differing in important derived from the experience of forming plans,
ways, and no longer have to be either equated proprioceptive feedback, or the objective con-
or dissociated from each other (e.g., Gopnik, sequences of physical actions on the environ-
1993). Mentalization comes about through the ment, contributes to the sense of self-agency.
child’s experience of his mental states being In our view, such agency also crucially de-
reflected on, for instance through secure play pends on the quality and reliability of reflec-
with a parent or older child, which facilitates tive function, as ownership of action is inti-
integration of the pretend and psychic equiva- mately tied to the mental state (belief or
lence modes, through a process which may be desire) which initiated it. It is impossible to
an elaboration of the complex mirroring of the conceive of self-agency as fully established
infant by the caregiver. In playfulness, the by the physical actions of the child, as such a
caregiver gives the child’s ideas and feelings large proportion of these will fail to achieve
(when he is “only pretending”) a link with re- their intended objective, because of the
ality, by indicating the existence of an alterna- child’s immature physical and cognitive ca-
tive perspective, which exists outside the pacities. The recognition of the child’s inten-
child’s mind. The parent or older child also tional stance by others must then be critical in
shows that reality may be distorted by acting making the thought “real” for the child. We
upon it in playful ways, and through this play- believe that interaction which links percep-
fulness a pretend but real mental experience tions, thoughts, and emotions as causes and
may be introduced. consequences of action, and the contempla-
3. In traumatized children, intense emotion tion of mental states without fear, contribute
and conflict lead to a partial failure of this significantly to self-agency. The earliest foun-
integration, so that aspects of the pretend dation is presumably the baby’s sense that he
mode of functioning become part of a psychic brings about the caregiver’s mirroring behav-
equivalence manner of experiencing reality. ior (Gergely & Watson, 1996).
This may be because where maltreatment or Of course, the core of self-agency must
trauma has occurred within the family, the at- originally lie with the body, where the in-
mosphere tends to be incompatible with the fant’s attempts to exercise control frequently
caregiver “playing with” the most pressing as- succeed after early infancy. Higher level, more
pects of the child’s thoughts; these are often complex actions, particularly those which in-
disturbing and unacceptable to the adult, just volve others in the child’s life, often require
as they are to the child. The rigid, controlling the reflective caregiver to make sense of the
behavior of the preschool child with a history young child’s wishes and translate these into
of disorganized attachment is thus seen as action sequences for the links between mental
arising out of a partial failure on the part of states and action to be established. It is to be
the child to move beyond the mode of psychic expected then that individuals who have expe-
equivalence in relation to specific ideas or rienced severe neglect or coercive, rigid,
feelings, so that he experiences them with the frightening, and, at an extreme, abusive par-
intensity that might be expected had they been enting will frequently experience their sense
current, external events. of self-agency as massively curtailed, and
15. Attachment and reflective function 693
limited to the more firmly established bodily adoxically drive them physically closer to a
(physical) domain. potential abuser. Their ability to adapt to,
modify, or avoid the perpetrator’s behavior is
also constrained by limited mentalizing skills.
Reflective function and pathological
There are several reasons why the family
self-development
environment of maltreatment is likely to un-
The model of the development of mentalizing dermine the development of reflective func-
capacity which we propose has considerable tion.
clinical implications, a few of which we will First, in abusive families the public world
mention here. of school and community—where reflective
function is common and desirable—is often
Impact of maltreatment on reflective function. kept very separate from the world of home,
Maltreated children, perhaps even more than where the inhumane behavior of an adult
insecure ones, are at risk of failing to find makes recognition of the mental state of the
their own intentional being within the mind other dangerous to the developing self. Even
of the caregiver, and are thus at risk of poor where a maltreated child benefits from sensi-
development of mentalization. There is accu- tivity and reflectiveness in his public world,
mulating evidence that maltreatment does im- so developing an alternative model of relating
pair the child’s reflective capacities and sense and experiencing himself, the models derived
of self. Schneider–Rosen and Cicchetti (1984, from public and family experiences are likely
1991) noted that abused toddlers showed neu- to be kept insulated from each other, and rigid
tral or negative affect on recognizing them- in their application to the separate contexts.
selves in the mirror, unlike their nonabused Second, the child may have specific prob-
peers. Beeghly and Cicchetti (1994) showed lems in dealing with different experience. In
that toddlers with a history of maltreatment abusive families the meaning of intentional
were not retarded in receptive language but states may be denied or distorted. Abusive
were significantly behind in productive lan- parents may claim beliefs and feelings at odds
guage, reflecting a withdrawal from social in- with their behavior. Abuse, particularly within
teractions. Their specific deficit was in the the family, prevents the child testing and
relative absence of internal state words and modifying representations of mental states.
the context-bound (concrete) nature of their Thus, the mental representation of ideas tends
internal state language. They also showed less to become rigid, maladaptive, and inappropri-
differentiation in attributions. Their internal ate, and consequently may be partially or
state language was particularly sparse in terms largely abandoned.
of words pertaining to cognition and belief A third possibility is that the maltreated
states, but was richer for perception and de- child is forced to construct a model of the
sire. Cicchetti and Beeghly (1987) found that caregiver’s mind based on an awareness of
young school-age children who had been mal- analogous mental states in himself. It may be
treated used proportionally fewer words about argued, on the basis of the simulation model,
internal states, attributed their internal states that simulation is compromised by both the
to fewer social agents, and were more context dissimilarity between the child’s mental expe-
bound than their counterparts who were not rience and that of the abuser, and the threat
maltreated. They appeared to control their that such simulation inevitably brings with it.
anxiety by modifying their language to ex- If understanding the behavior of his caregiv-
clude certain aspects and contexts associated ers requires the maltreated child to try to gen-
with maltreatment. This pattern of results sug- erate their probable thoughts and feelings,
gests that maltreatment may cause children to then he will be confronted with attitudes to-
withdraw from the mental world. For mal- wards himself which are extremely painful to
treated children, physical experiences proba- recognize: hatred, cruelty, indifference. Abuse
bly become more important, and this may par- could destroy the child’s belief that one can
16. 694 P. Fonagy and M. Target
understand others through one’s own feelings sionals and family members. These anomalies
(Herman, 1992), and the child would be likely can be clarified by more sophisticated devel-
to inhibit his capacity for simulation in in- opmental theory.
tense attachment relationships. Our chosen framework is provided by “dy-
A fourth possibility is that the difficulty is namic skills theory” (Fischer & Farrar, 1987;
not a result of the maltreatment itself, but of Fischer, Kenny, & Pipp, 1990) which depicts
the family atmosphere surrounding it (which development as the elaboration of increas-
may well also occur where maltreatment does ingly complex control systems (skills). Re-
not). Social constructivist ideas concerning the flective function may be readily conceived of
development of mentalization (e.g., Asting- as one such control system, critical to the or-
ton, 1996) are pertinent here. Authoritarian ganization of the self. Within dynamic skills
punishment of bad behavior and demanding theory, reflective function would be seen as
of obedience is clearly less facilitating of the not simply a property of the person, but of the
child’s development of mentalization than are person and situation together, because all
equivalent interactions with authoritative par- skills are composed of both the person’s ac-
ents, who reason with the child and explain tivities and the contexts within which these
decisions and rules with reference to people’s occur. Particular tasks, specific events, other
different points of view (Baumrind, 1971). people, as well as culture are seen as part of
There is some evidence that authoritarian par- the skill. Further, the development of a skill is
enting is associated with delayed false belief not seen as progression along a singular path,
task performance (Holmes, Roldan, & Miller, determined by maturation. Rather, reflective
1994, cited in Astington, 1996). As, in a Vy- function, as a skill, evolves through varied
gotskian framework, the individual’s compe- pathways, molded by many dynamically inter-
tence originates in their social interactions and acting influences, such as the individual’s
is then internalized, we would expect the ab- emotions, social interaction, family relation-
normal patterns of parent–child relations in ships and environment, important social
the families of maltreated children to lead to groups, the reactions of the wider social world,
a distorted experience of minds. Alessandri etc. (Fischer, Knight, & Van Parys, 1993).
(1991, 1992) noted that the incompetence of Reflective function is a strand within the
maltreated youngsters in pretend symbolic developmental web, one of the many distinct
play was mirrored by their mothers’ difficulty control systems that are neither strongly con-
in taking a playful stance with their child, nected with each other, nor coordinated or
directing their attention, and engaging in posi- integrated (Fischer & Pipp, 1984). The “frac-
tive interactions. This pattern of results is con- tionation” or splitting of all abilities as a func-
sistent with the model that the lack of appro- tion of tasks and domains is well demon-
priate social scaffolding may undermine the strated, and we might expect reflective
normal development of mentalizing in mal- function to be subject to the same kind of de-
treated children. ´
velopmental decalage (unevenness) which
characterizes the rest of cognitive develop-
Developmental framework for abnormal re- ment (Flavell, 1982). Fractionation refers to
flective function. It is tempting to argue that the tendency for a person not to coordinate
disorders of conduct and borderline states can skills or experiences that are naturally sepa-
be explained as dismissive and preoccupied rate, but may be thought of as belonging to-
forms of nonmentalizing self-organizations gether by some external criterion (Fischer &
respectively, but this would be simplistic. In Ayoub, 1994). Just as the understanding of
both instances, there are often variations conservation of liquid does not generalize to
across situations, or types of relationship. The conservation of area, reflective capacity in
delinquent adolescent is aware of the mental one domain of interpersonal interaction should
states of other gang members and the border- not be expected to generalize to others. Re-
line individual is at times hypersensitive to flective function does not begin as a general
the affective states of mental health profes- capacity, but is a particular skill tied to the
17. Attachment and reflective function 695
task and domain where it is learned, a specific naturally move toward integration. The family
category of relationship. Reflective function might of course, as we mentioned, support
as a skill may be more or less present in situa- such splits with sharp dissociations between
tions as a function of contextual support and their public, proper world and their private,
emotional state, which push an individual up tyrannical one. The split is context and affect
or down a developmental strand. We have dependent; within an attachment theory
noted above that the child’s observed use and framework we might say that the self is orga-
experience of mental state language can differ nized so that certain internal working models
markedly across social contexts. It is clearly include considerable reflective components—
possible for task-based skills such as reflec- expectations incorporating the mental states
tive function to come to be coordinated, but of self and other—while other working mod-
this should not be seen as automatic. Uneven- els of relationships appear impoverished, indi-
ness across situations is likely to remain prev- cating only minimal mentalizing skills. In the
alent even in adults, especially when they are latter contexts the subject will offer only ster-
emotional (Fischer & Ayoub, 1994). eotyped, concrete, low level descriptions.
Normal development is from fractionation This does not imply developmental delay or
towards integration, which involves the coor- regression; rather it suggests a remarkably
dination of previously separate skills and pro- complex ability to coordinate two distinct lev-
vides the foundation for more complex, so- els of functioning. The abusive or emotionally
phisticated control systems (Bidell & Fischer, depriving world within which they developed
1994). Abnormalities of reflective function, has engendered in them the sophisticated
the continued use of a teleological rather than skills that were required for adaptation. Thus
a mentalizing model for predicting behavior, to talk of deficit or absence of a capacity in
should not then be seen as either a conse- such individuals is an oversimplification.
quence of arrest and fixation at an early stage, Measures of global abilities may not yield a
or a regression to that stage. Pathologies in difference between these individuals and
the reflective function of the maltreated child other groups. Efforts at going beyond clinical
may be expected to develop increased com- impression in terms of measurement have to
plexity with age and time, in a manner similar take on board the situational specificity of the
to other skills. The skill for limited reflective- failure of reflective function.
ness developed by the child to anticipate and We will return to the example of conduct
forestall maltreatment and its painful physical disordered children, for whom we suggest that
and psychological impact would be adaptive nonreflective internal working models may
in their original world, but would be expected dominate behavior when an element of con-
to produce sophisticated forms of difficulty flict is present within a relationship. Conflict,
rather than straightforward adaptations in or rather its adaptive resolution, particularly
other contexts (Noam, 1990). The ability to calls for the perception both of the self and of
be reflective in general, but to show only min- the other in relation to the self, requiring the
imal reflectiveness in the context of one’s individual to reconcile his own legitimate
own childhood and parents, or in specific rela- claims with concern for the other (Killen &
tionships which reactivate the same schemata, Nucci, 1995). The abnormality of the early
could be a result of natural fractionation. Un- family environment of individuals with severe
evenness or splitting of reflective ability could problems of conduct has been clearest in the
also be the consequence of an active (pur- context of normal conflicts (Patterson, 1982;
poseful, conscious, or unconscious) attempt Perry, Perry, & Kennedy, 1992). Here the
on the part of the individual not to coordinate child with a vulnerable capacity for mentali-
or generalize reflective function to specific re- zation finds no affirmation of his intentional
lationship domains. Here the unevenness is a stance and fails to acquire the sense of owner-
developmental achievement, in that the per- ship or inner endorsement of his actions es-
son must create a coordination in order ac- sential for a sense of self-agency. Consequently
tively to keep separate contexts which would his sense of autonomy becomes vulnerable
18. 696 P. Fonagy and M. Target
and the importance of his original intention predominant response to emotional situations
is exaggerated. The characteristics of opposi- will be a nonreflective one, readily disowned
tional defiant disorder (e.g., negativity, dis- by the self. Naturally the absence of reflective
obedience, aggression) may in part be seen as function in such situations will give the ap-
attempts at reasserting self-agency in a rela- pearance of rigidity to the person’s behavior
tionship where the connection between mental as if only a singular pattern of response were
state and action within the self has been un- accessible. Furthermore, the response may
dermined by insensitive and coercive par- frequently be in conflict with social norms be-
enting. cause the tendency to take the perspective of
Abnormalities of parenting represent but others has been abandoned in that context
one route to limitations on reflective function. and, consequently, the moral emotions used
The child’s biological vulnerabilities such to make judgments about the consequences of
as hyperactivity, attention problems, low im- actions and regulate behavior are absent. The
pulse control, are all likely to obstruct the absence of reflective function may further ex-
opportunity the child has for evolving a men- aggerate an antisocial response by forcing the
talized reflective model of conflict-related in- individual to see the other not as another in-
terpersonal situations. Within a dialectic or tentional agent, but in nonhuman terms, as a
transactional model there is a bidirectional body, as representing a social position or
causality inherent to such biological vulnera- agency, or as a faceless member of a group.
bilities: They both provoke situations of con- Maltreatment, or more broadly trauma, is
flict and place grave limitations on the child’s seen as interacting with the domain- and situ-
capacity to acquire the flexibility needed for ation-specific restrictions upon reflective func-
their adaptive handling. This may bear on the tion at two levels. First, as we have argued,
well demonstrated comorbidity between con- maltreatment presents the young child with a
duct disorders and hyperactivity or attention powerful emotional disincentive for taking the
deficit disorder (Kazdin, 1995). Similarly, perspective of others, because of the actual
factors associated with early behavioral prob- hostility of the intentional stance of the
lems, such as poor parental adjustment (ma- abuser, as well as the constraints upon the self
ternal aggression, suspiciousness, and mood which an older person’s failure to understand
disorder) (Shaw, Owens, Vondra, Keenan, the child’s budding intentionality imposes.
& Winslow, 1996; Zahn–Waxler, Ianotti, Second, the child misses a protective factor,
Cummings, & Denham, 1990), and resources the capacity to understand traumatic interper-
(marital dissatisfaction, parental conflict) sonal situations, which would be likely to
(Block, Block, & Gjerde, 1986; Campbell, limit their impact (Fonagy, Steele, Steele,
Pierce, Moore, Marakovitz, & Newby, 1996) Higgitt, & Target, 1994). Thus, individuals
may limit the parents’ capacity to respond to traumatized by their family environment are
the child in ways which promote a mentaliz- vulnerable both in terms of the long-term mal-
ing model of self–other relationships. adaptive effect of their reaction to the trauma
The separation of action from intention un- and in terms of their reduced resilience in the
dermines the emotional reaction an individual face of it. The predominantly nonmentalizing
may have to the consequences of their actions stance adopted in such situations therefore
since, as Hart and Killen (1995) pointed out, handicaps the individual and, if the vicious
the acquisition of moral emotions requires circle is unbroken, may come to dominate all
that individuals are “active contributors to interpersonal relationships. We believe that at
their own development, interpreting their this stage severe developmental psychopathol-
world and making judgments that determine ogy, in the adult entrenched personality disor-
their actions in it” (p. 7). Subsequently, the der, is the likely consequence.
References
Ainsworth, M. D. S., Bell, S. M., & Stayton, D. J. (1971). olds. In H. R. Schaffer (Ed.), The origins of human
Attachment and exploratory behaviour of one year social relations. New York: Academy Press.
19. Attachment and reflective function 697
Ainsworth, M. D. S., Blehar, M. C., Waters, E., & Wall, Botterill, G. (1996). Folk psychology and theoretical sta-
S. (1978). Patterns of attachment: A psychological tus. In P. Carruthers & P. K. Smith (Eds.), Theories
study of the strange situation. Hillsdale, NJ: Erlbaum. of theories of mind (pp. 105–118). Cambridge: Cam-
Alessandri, S. M. (1991). Play and social behaviours in bridge University Press.
maltreated preschoolers. Development and Psycho- Bowlby, J. (1969). Attachment and loss, Vol. 1. Attach-
pathology, 3, 191–206. ment. London: Hogarth Press and the Institute of Psy-
Alessandri, S. M. (1992). Mother–child interactional cor- cho-Analysis.
relates of maltreated and nonmaltreated children’s Bowlby, J. (1973). Attachment and loss, Vol. 2. Separa-
play behavior. Development and Psychopathology, 4, tion: Anxiety and anger. London: Hogarth Press and
257–270. Institute of Psycho-Analysis.
Astington, J. (1996). What is theoretical about the child’s Bowlby, J. (1980). Attachment and loss, Vol. 3. Loss:
theory of mind?: A Vygotskian view of its develop- Sadness and depression. London: Hogarth Press and
ment. In P. Carruthers & P. K. Smith (Eds.), Theories Institute of Psycho-Analysis.
of theories of mind (pp. 184–199). Cambridge: Cam- Bracken, B. A. (Ed.). (1996). Handbook of self-concept:
bridge University Press. Developmental, social and clinical considerations.
Astington, J., & Jenkins, J. M. (1995). Theory of mind New York: Wiley.
development and social understanding. Cognition and Bretherton, I. (1991). Intentional communication and the
Emotion, 9, 151–165. development of an understanding of mind. In D.
Baron–Cohen, S., & Swettenham, J. (1996). The relation- Frye & C. Moore (Eds.), Children’s theories of mind:
ship between SAM and ToMM: Two hypotheses. In Mental states and social understanding (pp. 271–
P. Carruthers & P. K. Smith (Eds.), Theories of theo- 289). Hilldale, NJ: Erlbaum.
ries of mind (pp. 158–168). Cambridge: Cambridge Bretherton, I., Bates, E., Benigni, L., Camaioni, L., &
University Press. Volterra, V. (1979). Relationships between cognition,
Baron–Cohen, S., Tager–Flusberg, H., & Cohen, D. J. communication, and quality of attachment. In E.
(1993). Understanding other minds: Perspectives Bates, L. Benigni, I. Bretherton, L. Camaioni, & V.
from autism. Oxford: Oxford University Press. Volterra (Eds.), The emergence of symbols. New
Bartsch, K., & Wellman, H. M. (1989). Young children’s York: Academic Press.
attribution of action to beliefs and desires. Child De- Brownell, C. A., & Kopp, C. B. (1991). Common threads,
velopment, 60, 946–964. diverse solutions: Concluding commentary. Develop-
Baumrind, D. (1971). Current patterns of parental author- mental Review, 11, 288–303.
ity. Developmental Psychology Monographs, 4(1, Bruner, J. (1983). Child’s talk: Learning to use language.
Part 2). Oxford: Oxford University Press.
Beebe, B., Lachmann, F., & Jaffe, J. (1997). Mother–in- Campbell, S. B., Pierce, E. W., Moore, G., Marakovitz,
fant interaction structures and presymbolic self and S., & Newby, K. (1996). Boys’ externalizing problems
object representations. Psychoanalytic Dialogues, 7, at elementary school age: Pathways from early behav-
133–182. ior problems, maternal control and family stress. De-
Beeghly, M., & Cicchetti, D. (1994). Child maltreatment, velopment and Psychopathology, 8, 701–719.
attachment, and the self system: Emergence of an in- Carlson, E., & Sroufe, L. A. (1995). Contribution of at-
ternal state lexicon in toddlers at high social risk. De- tachment theory to developmental psychopathology.
velopment and Psychopathology, 6, 5–30. In D. Cicchetti & D. J. Cohen (Eds.), Developmental
Belsky, J., Garduque, L., & Hrncir, E. (1984). Assessing Psychopathology. Vol. 1. Theory and methods (pp.
performance, competence and executive capacity in 581–617). New York: Wiley.
infant play: Relations to home environment and secu- Cassam, Q. (Ed.). (1994). Self-knowledge. Oxford: Ox-
rity of attachment. Developmental Psychology, 20, ford University Press.
406–417. Cassidy, J. (1994). Emotion regulation: Influences of at-
Belsky, J., Rosenberger, K., & Crnic, C. (1995). The ori- tachment relationships. In N. A. Fox (Ed.), The devel-
gins of attachment security: “Classical” and contex- opment of attachment regulation. Monograph of the
tual determinants. In S. Goldberg, R. Muir, & J. Kerr Society for Research in Child Development (Serial
(Eds.), John Bowlby’s attachment theory: Historical, No. 240) (pp. 228–249).
clinical and social significance (pp. 153–184). Hills- Churchland, P. S. (1986). Neurophilosophy. Cambridge,
dale, NJ: Analytic Press. MA: MIT Press.
Bidell, T. R., & Fischer, K. W. (1994). Developmental Cicchetti, D., & Beeghly, M. (1987). Symbolic develop-
transitions in children’s early on-line planning. In M. ment in maltreated youngsters: An organizational per-
M. Haith, J. B. Benson, R. J. Roberts, & B. F. Pen- spective. In D. Cicchetti & M. Beeghly (Eds.), Atypi-
nington (Eds.), Development of future-oriented pro- cal symbolic development. New directions for child
cesses. Chicago: University of Chicago Press. development (Vol. 36, pp. 5–29). San Francisco:
Bion, W. R. (1962). Learning from experience. London: Jossey–Bass.
Heinemann. Cicchetti, D., & Beeghly, M. (1990). The self in transi-
Bleiberg, E. (1984). Narcissistic disorders in children. tion: Infancy and childhood. Chicago: University of
Bulletin of the Menninger Clinic, 48, 501–517. Chicago Press.
Bleiberg, E. (1994). Borderline disorders in children and Cicchetti, D., & Toth, S. L. (Eds.). (1994). Disorders and
adolescents: The concept, the diagnosis, and the con- dysfunctions of the self (Vol. 5). Rochester, NY: Uni-
troversies. Bulletin of the Menninger Clinic, 58, 169– versity of Rochester Press.
196. Cooley, C. H. (1964). Human Nature and the Social Or-
Block, J. H., Block, J., & Gjerde, P. F. (1986). The per- der (Rev. ed.). New York: Schocken Books (Original
sonality of children prior to divorce. Child Develop- work published 1902).
ment, 57, 827–840. Crittenden, P. A. (1992). Quality of attachment in the
Bolton, D., & Hill, J. (1996). Mind, meaning and mental preschool years. Development and Psychopathology,
disorder. Oxford: Oxford University Press. 4, 209–241.
