Mentalization-Based Therapy

Abstract

Mentalization-based therapy (MBT) is a type of psychotherapy, developed by Peter Fonagy and Anthony Bateman for borderline personality disorder, that treats the capacity to understand behaviour in terms of underlying mental states as the target of treatment rather than a means to it. Its founding claim is that many symptoms of borderline pathology — affective instability, impulsivity, and turbulent relationships — reflect a fragile capacity for mentalizing that collapses under the arousal of close attachment relationships. When mentalizing fails, older prementalizing modes return, and the therapy works to rebuild it by keeping the patient's arousal within the range where mentalizing can operate. This article sets out the mentalizing construct, the non-mentalizing modes, the not-knowing stance, the measurement of reflective functioning, and the controlled-trial evidence, including the epistemic-trust reformulation that now frames how the treatment is understood.

Keywords: mentalization, reflective functioning, borderline personality disorder, epistemic trust

What Mentalization-Based Therapy Is

Mentalization-based therapy is the structured psychotherapy for borderline personality disorder developed by Peter Fonagy and Anthony Bateman at the turn of the century and first tested in a randomised trial of psychoanalytically oriented partial hospitalisation (#ref-bateman-1999). MeSH classifies it under psychotherapy, but its distinguishing feature is a single, sharply defined target: mentalizing, the capacity to interpret one's own and others' behaviour as arising from intentional mental states — thoughts, feelings, wishes, and beliefs. Where most therapies treat the patient's understanding of mind as the vehicle through which change is delivered, MBT treats that understanding as the thing to be repaired, on the argument that its instability is what generates the disorder's presentation.

The theory rests on a developmental account drawn from attachment research. An infant comes to know its own mind by first finding it reflected in the mind of a caregiver, whose responses to the child's distress are contingent on it and marked — displayed in a form the child can recognise as a representation of its state rather than the caregiver's own. Through this marked mirroring the child builds a second-order representation of its own affects and, with it, the capacity for affect regulation; where mirroring is absent, incongruent, or unmarked, the child internalises a distorted image Fonagy and colleagues term the alien self, an unassimilated presence experienced as not-me from within (#ref-fonagy-2002). Mentalizing is thus an achievement of secure attachment, and its fragility in borderline pathology is understood as the developmental signature of disrupted early relationships. The core relationship the theory posits — between attachment arousal and the capacity to mentalize — is set out in Figure 1.

Mentalizing: The Core Capacity

Mentalizing is not a single skill but a set of related capacities that vary along four dimensions, each a spectrum between two poles. The first opposes automatic mentalizing — fast, implicit, and effortless, the reflexive reading of a facial expression — to controlled mentalizing, the slow, deliberate, and effortful reasoning about mental states that reflection requires. The remaining three run between self and other, between the internal features of mind and its external and behavioural signs, and between the cognitive contents of mind (beliefs, intentions) and its affective ones (feelings). Balanced mentalizing means moving flexibly across all four; the imbalances characteristic of borderline pathology are a reliance on automatic over controlled processing, on external over internal cues, and a fusion of self with other under stress (#ref-luyten-2020). The demo below lets the reader move a profile across the four dimensions and see where it becomes unbalanced.

The four dimensions of mentalizing

Mentalizing is balanced when a person can move freely along each spectrum. Drag any slider, or load a profile. A pole held to an extreme (near either end) marks a loss of flexibility on that dimension.

The clinical importance of the dimensions is that they fail together and predictably. Under the low-arousal conditions of ordinary reflection, a person can hold their own perspective and another's in mind at once, weigh evidence about what someone intended, and revise a first impression. As attachment arousal rises — in intimacy, conflict, or the threat of abandonment — controlled mentalizing is progressively down-regulated in favour of the automatic, fast systems, until at the switch point of Figure 1 it drops out altogether (#ref-fonagy-2009). The signature of borderline personality disorder, on this account, is not a stable deficit but an unstable one: a switch point reached at a lower threshold of arousal, so that the very relationships that most demand mentalizing are the ones in which it is most readily lost.

When Mentalizing Fails: The Prementalizing Modes

When mentalizing collapses, experience does not become blank; it reverts to developmentally earlier ways of representing the mind that Fonagy and colleagues call the prementalizing modes, three forms of non-mentalizing that normally give way to full mentalizing in childhood but re-emerge when the capacity fails (#ref-bateman-2016). In psychic equivalence, the distinction between mind and world collapses so that whatever is thought or felt is taken to be literally, unquestionably real — a feared abandonment is not a fear but a fact, with the concreteness and terror that follow. In the teleological mode, mental states are acknowledged only when expressed in observable, physical outcomes: care is real only if it takes a tangible form, which is why reassurance in words fails where an action, or an act of self-harm that compels one, succeeds. In the pretend mode, thoughts and feelings are decoupled from reality altogether, producing a fluent but empty pseudo-mentalizing — long, plausible accounts of one's states that connect to nothing felt, the dissociated over-talking familiar from the consulting room. The three modes are contrasted in Table 1, and the demo below lets the reader run a single situation through each.

One event, four modes of mind

Choose a situation, then see how it is construed when mentalizing has collapsed into each prementalizing mode — and how the same event looks once mentalizing is restored.

The Therapeutic Stance

The clinical method follows from the theory: if mentalizing collapses at a lowered arousal threshold, the therapist's first task is to keep the patient within the band where it can operate, neither so under-aroused that nothing is engaged nor so over-aroused that the switch is thrown. This is achieved less by technique than by a consistent attitude MBT calls the not-knowing stance — an inquisitive, genuinely curious position in which the therapist treats mental states as opaque and provisional, models the effortful work of mentalizing by doing it aloud, and resists the pull to supply authoritative interpretations of what the patient really means. Interpretations of unconscious content, prized in the psychoanalysis from which MBT descends, are largely avoided: to a patient in psychic equivalence they land as facts and raise arousal, defeating the aim. The therapist instead marks and explores the moment mentalizing was lost, rewinds to just before it, and helps rebuild the sequence.

The framing that now organises this work is epistemic trust: the willingness to accept knowledge from another person as relevant and generalisable to oneself. Luyten, Campbell, Fonagy and colleagues argue that the developmental adversity underlying borderline pathology produces a protective epistemic vigilance — a mistrust that seals the person off from social learning, so that even good advice cannot get in and be used (#ref-luyten-2020). The notion is borrowed from Sperber and colleagues, who describe epistemic vigilance as the ordinarily adaptive filtering by which people guard against being accidentally or deliberately misinformed; in the mentalizing account it is not absent but stuck on, closing the channel through which knowledge from others would normally flow (#ref-sperber-2010). On this reading, a therapy works not chiefly through the specific content it delivers but by re-opening epistemic trust: the experience of being accurately mentalized by the therapist signals that this person is a safe source of knowledge, which reconnects the patient to learning from their wider social world. The demo below illustrates how the trust setting gates what social information a person can take up.

Epistemic trust and social learning

Epistemic trust is the willingness to take another person's knowledge as relevant to oneself. Move the dial: when trust is low, protective vigilance seals the person off, and even good information cannot get in. MBT works by raising this setting.

Setting: Guarded — 33% of available social knowledge is taken up.

Measuring Mentalizing

Mentalizing became a research variable once it could be scored. The operational form of the construct is reflective functioning, the observable capacity to reason about mental states, first measured by applying the Reflective Functioning Scale to transcripts of the Adult Attachment Interview and rating the sophistication with which a person represents their own and others' minds (#ref-fonagy-1997). Expert-rated interviews are costly, so the field developed a brief self-report instrument, the Reflective Functioning Questionnaire, whose items index the two ways reflective functioning typically fails: an excess of certainty about mental states, corresponding to psychic equivalence, and an excess of uncertainty, a claimed opacity of mind that shades into the pretend mode (#ref-fonagy-2016). The two poles matter clinically because they call for opposite corrections, and a single global score would hide them.

Reflective functioning gave the theory a testable middle term: if MBT works by restoring mentalizing, then improvement in reflective functioning should track improvement in symptoms and, ideally, mediate it. Reviews of the measurement literature find the construct coheres and predicts as expected across clinical and developmental samples, while cautioning that the self-report and interview measures capture overlapping but not identical things and that mediation evidence remains partial (#ref-katznelson-2014). The measurement programme is what separates MBT's mechanistic claim from an untestable one, and it is where the theory remains most exposed to disconfirmation.

The Evidence Base

MBT is among the better-tested psychotherapies for borderline personality disorder. The founding trial randomised patients to psychoanalytically oriented partial hospitalisation or general psychiatric care and found large advantages for the treated group in self-harm, suicide attempts, hospitalisation, and mood, sustained at the end of treatment (#ref-bateman-1999). A follow-up showed the gains were not only maintained but continued to widen over the eighteen months after treatment ended, a rebound-free pattern unusual in the field (#ref-bateman-2001). The most striking result came at eight years from the original entry: patients who had received the mentalization-based programme remained substantially better than the comparison group on suicidality, service use, and diagnostic status, though most no longer met full criteria in either arm (#ref-bateman-2008). A later trial moved the treatment out of the day hospital into a purely outpatient format and tested it against structured clinical management — a credible active control rather than treatment as usual — and still found MBT superior on the clinically important outcomes (#ref-bateman-2009).

Independent synthesis has been more measured. A systematic review of the MBT evidence base judged the results promising but noted that much of the strongest data came from the originators' own group and called for replication by independent teams (#ref-maldacastillo-2019). A narrative review reached a similar verdict while documenting the treatment's extension to adolescents, eating disorders, and other populations (#ref-volkert-2019). The Cochrane review of psychological therapies for borderline personality disorder placed MBT among the interventions with beneficial effects on the core outcomes while rating the overall certainty of the evidence as low to moderate, the standing limitation of a field with small trials and few independent replications (#ref-storebo-2020).

Worked Example

Trial and meta-analytic results for MBT are often reported as standardised mean differences, which become clinically legible once translated. Take an illustrative moderate effect of Cohen's d = 0.50, within the range reported for mentalization-based and other structured psychotherapies for borderline personality disorder against control conditions (#ref-storebo-2020).

The first re-expression is Cohen's U₃, the proportion of the treated group exceeding the median untreated outcome, given by the standard-normal cumulative probability Φ(d): Φ(0.50) = 0.691. About 69% of patients receiving MBT therefore end above the median control outcome, against the 50% baseline.

The second is the probability of superiority — the chance a randomly chosen treated patient outscores a randomly chosen control — computed as Φ(d / √2) = Φ(0.50 / 1.4142) = Φ(0.3536) = 0.638, close to a 64% chance. Expressed as a number needed to treat, 1 / (2 × 0.638 − 1) = 1 / 0.276 = 3.6, so between three and four patients must be treated for one to cross an improvement threshold who would not have on the control condition alone.

These figures describe a moderate but genuine effect, consistent with a treatment that helps a real proportion of a difficult-to-treat population beyond what a credible active control provides (#ref-bateman-2009). They are the quantitative face of the trial literature that moved MBT from a promising model to an evidence-supported treatment.

Discussion

MBT's contribution is twofold. It gave borderline personality disorder a single mechanistic account — an unstable mentalizing capacity that fails under attachment arousal — parsimonious enough to organise a treatment and specific enough to be tested, and it supplied that treatment with an operational measure in reflective functioning, so that the mechanism and not only the outcome could be examined (#ref-fonagy-1997). The developmental theory ties the clinical picture to a large body of attachment research, and the not-knowing stance translates the theory into a stance any clinician can adopt, which has helped the treatment disseminate and be delivered by generalists rather than only specialists (#ref-bateman-2016).

The criticisms are substantive. The strongest efficacy data still come disproportionately from the originating group, and independent replications, though accumulating, remain fewer than the confident theory would warrant (#ref-maldacastillo-2019). The central mechanistic claim — that MBT works by improving mentalizing — is supported by the right pattern of association but not yet by decisive mediation evidence, and the construct's breadth makes it hard to falsify cleanly (#ref-katznelson-2014). The epistemic-trust reformulation is theoretically elegant and reaches to explain why many different therapies work, but that same generality makes it difficult to test as a claim specific to MBT (#ref-luyten-2020). The defensible reading is that MBT is a well-evidenced treatment built on a partially confirmed theory: its clinical benefits are real and replicated, while the account of why it works remains a live and productive research question rather than a settled one.

Current Directions

The active research front has moved in two directions. The first is beyond borderline personality disorder and beyond adults: mentalization-based programmes have been adapted for adolescents, for children and their families, for eating disorders, antisocial personality disorder, and trauma, and the open question is how far a treatment built around one disorder's mentalizing profile transfers to populations whose imbalances differ (#ref-volkert-2019). The second is the consolidation of epistemic trust from a theoretical proposal into a measurable variable, with self-report instruments for trust, mistrust, and credulity now being validated so that the salutogenic mechanism the theory posits can be tested directly rather than assumed (#ref-luyten-2020).

Running through both is the field's central methodological task: independent replication. The Cochrane synthesis was explicit that the evidence for psychological therapies in borderline personality disorder, MBT included, is held back less by the size of the effects than by the small number of adequately powered trials conducted outside the treatments' originating centres (#ref-storebo-2020). The clearest sign of the model's maturity will be efficacy that holds when the people testing it are not the people who built it.

Common Misconceptions

Mentalizing is just empathy or emotional intelligence.

It is broader and more specific: mentalizing covers the self as well as others, cognition as well as affect, and automatic as well as deliberate processing, and its clinical point is its instability under attachment arousal rather than a person's average level of it (#ref-luyten-2020).

MBT is a form of long-term psychoanalysis.

It grew out of psychoanalytic thinking but deliberately avoids the interpretation of unconscious content, which raises arousal and defeats its aim; it is a structured, time-limited treatment organised around the not-knowing stance (#ref-bateman-2016).

Self-harm in borderline patients is manipulative attention-seeking.

In MBT's terms it is the teleological mode at work: when mentalizing has collapsed, only a physical act can make an internal state real or compel a caring response, which is a failure of mentalizing rather than a strategy (#ref-bateman-2008).

Glossary

Affect regulation.

The capacity to modulate the intensity and expression of one's emotional states; in MBT's developmental account it is built through the caregiver's marked mirroring of the infant's affects.

Alien self.

The distorted self-representation internalised when early mirroring is absent or incongruent, experienced as an unassimilated, not-me presence within the self.

Attachment.

The emotional bond between child and caregiver whose security, in this theory, is the developmental condition under which mentalizing is acquired and within whose arousal it is most readily lost.

Automatic mentalizing.

Fast, implicit, effortless reading of mental states; adaptive in familiar situations but error-prone when it displaces controlled reflection under stress.

Controlled mentalizing.

Slow, explicit, effortful reasoning about mental states; the deliberate reflection that is down-regulated and finally lost as attachment arousal rises.

Epistemic trust.

The willingness to accept knowledge from another as relevant and generalisable to oneself; its restoration is the mechanism by which MBT is now thought to reconnect patients to social learning.

Marked mirroring.

A caregiver's contingent reflection of an infant's affect, displayed in a recognisably marked form, through which the child builds a second-order representation of its own emotional states.

Mentalizing.

The capacity to interpret one's own and others' behaviour as arising from intentional mental states; the target that MBT is designed to repair.

Not-knowing stance.

The therapist's inquisitive, non-authoritative attitude in MBT, which treats mental states as opaque and provisional and models the effortful work of mentalizing rather than supplying interpretations.

Prementalizing modes.

The developmentally earlier ways of representing mental life — psychic equivalence, teleological mode, and pretend mode — that re-emerge when mentalizing collapses.

Pretend mode.

A prementalizing mode in which thoughts and feelings are decoupled from reality, producing fluent but empty pseudo-mentalizing that connects to nothing actually felt.

Psychic equivalence.

A prementalizing mode in which mind and world are fused, so that whatever is thought or felt is taken to be literally and unquestionably real.

Reflective Functioning Questionnaire.

A brief self-report measure of reflective functioning whose two scales index failures by excessive certainty and by excessive uncertainty about mental states.

Reflective functioning.

The operational, scoreable form of mentalizing: the observable capacity to reason about mental states, rated from attachment interviews or self-report.

Teleological mode.

A prementalizing mode in which mental states are acknowledged only when expressed in observable physical outcomes, so that care is believed real only if it takes a tangible form.

Key Researchers

Anthony Bateman (living). Consultant psychiatrist and psychotherapist who, with Peter Fonagy, developed mentalization-based treatment and led its first randomised trials for borderline personality disorder. Faculty page

Chloe Campbell (living). Deputy director of the UCL Psychoanalysis Unit and a principal architect, with Fonagy and Luyten, of the epistemic-trust account of how mentalization-based and other therapies work. ORCID

Peter Fonagy (b. 1952). British psychoanalyst who originated the mentalization framework and co-developed MBT, linking attachment theory, developmental research, and the treatment of personality disorder. ORCID

György Gergely (b. 1953). Developmental psychologist and co-author of the theory of marked mirroring and contingent affect regulation on which MBT's developmental account rests. ORCID

Elliot Jurist (living). Philosopher and psychologist, co-author of Affect Regulation, Mentalization, and the Development of the Self, who developed the allied construct of mentalized affectivity. ORCID

Patrick Luyten (living). Clinical psychologist whose work reformulated mentalizing as a set of dimensions and helped develop the epistemic-trust model of psychopathology and treatment. ORCID

Mary Target (living). Psychoanalyst and long-standing Fonagy collaborator on reflective function and its measurement from the Adult Attachment Interview. Faculty page

Svenja Taubner (b. 1973). Psychotherapy researcher at Heidelberg University who has advanced the efficacy and effectiveness evidence for mentalization-based treatment and its adaptation to adolescents. ORCID

Frequently Asked Questions

What is mentalization-based therapy? It is a structured psychotherapy developed by Peter Fonagy and Anthony Bateman, first for borderline personality disorder, that treats the capacity to understand behaviour in terms of mental states, called mentalizing, as the target of treatment and works to rebuild it (Bateman and Fonagy, 1999).

What does mentalizing mean? Mentalizing is the capacity to interpret one's own and others' actions as arising from intentional mental states such as thoughts, feelings, and intentions, and it varies along dimensions of self and other, automatic and controlled, internal and external, and cognitive and affective (Luyten et al., 2020).

Why does MBT focus on borderline personality disorder? The theory holds that borderline pathology reflects a mentalizing capacity that is not permanently deficient but unstable, collapsing at a lowered threshold of attachment arousal, so the disorder's turbulent relationships are where mentalizing is most readily lost (Fonagy and Luyten, 2009).

What are the prementalizing modes? They are three earlier ways of experiencing mental life that return when mentalizing fails: psychic equivalence, in which thoughts are treated as literal reality; the teleological mode, in which only physical outcomes count as real; and the pretend mode, in which thoughts float free of feeling (Bateman and Fonagy, 2016).

What is the not-knowing stance? It is the therapist's inquisitive, non-authoritative attitude in MBT, which treats mental states as opaque, models the effortful work of mentalizing, and avoids the interpretation of unconscious content that would raise the patient's arousal and defeat the aim (Bateman and Fonagy, 2016).

How is mentalizing measured? Through reflective functioning, first scored from Adult Attachment Interview transcripts and later by the self-report Reflective Functioning Questionnaire, whose scales capture failures of both excessive certainty and excessive uncertainty about mental states (Fonagy et al., 2016).

Does MBT work? Randomised trials show benefits for self-harm, suicidality, and service use in borderline personality disorder that persist at long follow-up, and the Cochrane review places it among the effective psychotherapies while rating the overall certainty of evidence as low to moderate (Storebø et al., 2020).

What is epistemic trust and why does it matter? Epistemic trust is the willingness to accept knowledge from another as relevant to oneself; the current theory proposes that MBT helps by restoring it, so that being accurately understood by the therapist reconnects the patient to learning from their wider social world (Luyten et al., 2020).

References