Research figure: John Bowlby · Written by Artificial Sapiens
John Bowlby's internal working models of attachment are experience-based mental representations through which a person anticipates how important caregivers will respond, what kind of care the person can expect, and how the self can seek protection and support. Bowlby proposed these models to explain how encounters with attachment figures become organized into expectations that guide attention, emotion, decisions, and behavior beyond the immediate encounter. In his account, people form related working models of attachment figures, the self, and the relationships between them; these models can persist and be revised through further experience. This mechanism is central to Bowlby's theory and remains an active research problem in developmental and relationship psychology. Bowlby, 1973; Bretherton, 1992; Sherman, Rice, and Cassidy, 2015.
An internal working model is a theoretical construct, inferred from patterns in cognition, narration, expectations, and behavior. The construct describes how relationships are mentally represented; it is not a psychiatric diagnosis, a single childhood memory, a fixed personality label, or an instrument that directly reveals a person's future. Contemporary studies test narrower, operational forms of this broad proposal, including caregiver-specific expectations, attachment-related information processing, narrative coherence, and secure-base scripts. Dykas and Cassidy, 2011; Waters, Zhu, and Roisman, 2025.
Why Bowlby Introduced Internal Working Models
Attachment theory began with a practical question: how does a child use experience with protective adults to regulate closeness, explore the environment, and cope with danger or separation? John Bowlby described attachment as a behavioral system oriented toward maintaining access to a familiar protective person. Proximity-seeking acts such as approaching, calling, clinging, or checking a caregiver's location are context-sensitive means of achieving protection. A child must also anticipate whether a particular caregiver is available and which action is likely to work. For Bowlby, the existence of such anticipation required some account of stored relational information. Bowlby, 1969; Bretherton, 1985.
The problem extends beyond physical proximity. A child may remain confident while exploring because previous interactions have made help predictable, or may remain vigilant because access to help has been inconsistent. Repeated interactions potentially change the child's predictions before any new request for care occurs. An account restricted to observable attachment actions cannot fully explain that continuity. The working-model concept supplies Bowlby's proposed link between actual relationships, learned expectations, and future regulation of behavior. It does not require that a child verbally articulate the model or consciously rehearse every expectation. Sherman, Rice, and Cassidy, 2015.
Intellectual Origins: Kenneth Craik, Psychoanalysis, and Control Systems
From the idea of a mental model to relational prediction
Bowlby adapted the working-model concept associated with psychologist Kenneth Craik's 1943 book The Nature of Explanation. Craik argued that organisms can use internal models of the environment to anticipate possible events and compare possible courses of action. Bowlby placed an analogous predictive capacity inside an attachment system: children learn something about who can protect them, how to find that person, and what will probably happen when they express a need. The historical origin of the general model concept belongs to Craik; Bowlby's contribution was to incorporate it into a developmental theory of affectional bonds. Bretherton, 1992; Sherman, Rice, and Cassidy, 2015.
The word working indicates a model that can operate in real time. It supports forecasts and choices, rather than serving only as an archive of what happened. If a child expects an adult to respond promptly, the child may initiate contact and return to play after receiving reassurance. If previous bids have repeatedly failed, the child may anticipate nonresponse and regulate contact differently. These are theoretical illustrations of prediction guided by experience, not deterministic descriptions of any particular child.
Psychoanalytic continuity and a change in explanatory language
Bowlby trained in psychoanalysis and inherited longstanding questions about early relationships and internalized objects. His account retained the importance of the child's relationships with real people while using ethology, developmental observation, cognitive representation, and feedback-oriented control systems to specify how attachment could operate. The working-model idea translated parts of an older discussion of internal representations into hypotheses about expectations, selection of action, and updating. Bretherton, 1992.
This intellectual synthesis matters for attribution. Bowlby proposed a general representational mechanism. Mary Ainsworth's observational studies gave attachment researchers influential procedures and distinctions for studying caregiver sensitivity, secure-base behavior, and infant attachment patterns. Later researchers introduced discourse-based measures, scripts, self-report scales, and information-processing paradigms. The resulting research program developed across generations and cannot be assigned to Bowlby alone. Main, Kaplan, and Cassidy, 1985; Waters and Waters, 2006.
What an Internal Working Model Represents
Expectations about an attachment figure
A central element is a forecast about a particular person's availability, accessibility, and responsiveness. Availability concerns whether the person can be reached when protection is needed. Responsiveness concerns the likely quality and timing of that person's answer. A model can distinguish one caregiver from another: reliable support from a grandparent and inconsistent support from a parent need not become the same expectation. The child's social world contains multiple attachment relationships, and models may be partly specific to them. Sherman, Rice, and Cassidy, 2015.
Repeated experiences may support conditional predictions: what happens when the adult is busy, frightened, emotionally available, or physically absent? Because the model incorporates contexts and goals, the relevant question is more detailed than whether an individual is simply good or bad. The extent of contextual differentiation in actual mental representation is an empirical question, with cognitive research still refining its answer. Waters, Zhu, and Roisman, 2025.
Expectations about the self
Bowlby's formulation also linked models of others to a model of the self. Children learn whether their expressions of need appear to elicit attention and whether they can effectively reach sources of safety. Repeated treatment may shape a sense of being worthy of concern and capable of requesting it. This is why the working-model concept extends beyond beliefs about a caregiver: it concerns what a person expects to happen in an encounter between the self and an important other. Bowlby, 1973; Bretherton, 1985.
Self-representation is multidimensional. Confidence about being comforted can coexist with self-doubt in an unrelated domain. Expectations of relational support are also distinguishable from global self-esteem, temperament, and current mood. The theory predicts connections among these constructs in some circumstances, but their measurements and causal explanations must be evaluated separately.
Relationships as patterns of interaction
An attachment working model includes sequences, not only isolated judgments. A child notices distress, signals a need, encounters a response, and eventually experiences some form of resolution or continued distress. With repeated opportunities, that sequence can become a generalized expectation about how support usually unfolds. Contemporary secure-base script research attempts to specify such event structure more precisely. Waters and Waters, 2006.
A person may hold expectations that differ across parents, friends, partners, and changing circumstances. A useful model of attachment must therefore accommodate both recurring organization and flexibility, as well as potentially incompatible experiences within the same relationship. Treating a working model as a single immutable score removes much of the complexity Bowlby's term was intended to capture.
How Models Develop During Infancy
Bowlby situated the emergence of rudimentary models in the child's increasing ability to discriminate familiar people, remember experiences, anticipate events, and coordinate attachment behavior. Early learning need not resemble adult autobiographical reasoning. Infants can use regularities in voice, face, touch, proximity, timing, and emotional response long before explaining them verbally. The 2015 review by Sherman, Rice, and Cassidy examined infant memory, social cognition, learning, and other capacities relevant to this proposition; it also emphasized how much remains to be learned about the precise assembly of caregiver-specific representations. Sherman, Rice, and Cassidy, 2015.
A developmental account begins with many concrete episodes. The adult picks up the child, talks, returns after an absence, notices a change in expression, or responds to crying. Episodes differ and may contain mixed signals. Across time, a child can detect which cues are predictive in a given relationship. The hypothesis that patterns consolidate into expectations is broadly compatible with developmental cognitive capacities, while the formation and structure of a specific internal working model cannot be observed directly as a single event.
Early experience influences opportunities for learning, but it is only one part of developmental history. Caregiving environments, family resources, illness, safety, cultural practices, changes in caregivers, and later relationships may all affect what experiences become available. Children form attachments in varied family constellations. The historical language of a unique mother-child dyad should be read in the light of evidence on maternal and paternal sensitivity and the actual presence of multiple caregivers. Madigan et al., 2024.
From Experience to Expectation: The Proposed Mechanism
Encoding and integration
In representational terms, the first step is encoding information from relational encounters: what happened, what cues preceded it, how distress changed, and which action preceded a response. Generalization then integrates partially similar experiences into predictions useful in later situations. The existence of memory and learning capacities is established; the precise architecture by which attachment events become a particular durable working model remains a topic of current theory and investigation. Sherman, Rice, and Cassidy, 2015; Waters, Zhu, and Roisman, 2025.
Prediction and behavioral planning
A working model becomes relevant when it affects what a person expects next. When discomfort rises, the individual may forecast whether another person will answer, whether expressing vulnerability is safe, and which behavior will increase access to help. Those forecasts can influence whether the person requests support, postpones a request, intensifies signaling, or shifts attention elsewhere. These pathways are theoretical mechanisms that researchers investigate through observed behavior, reported expectations, experimental tasks, and narrative assessments. Dykas and Cassidy, 2011.
Feedback and revision
An expectation can be confirmed or contradicted. Effective feedback requires noticing that an encounter offers information relevant to the model. Bowlby's use of working models leaves room for correction when relationships change. If a formerly unavailable caregiver becomes consistently responsive, or a dependable person becomes inaccessible, updated experience can support revised predictions. Existing patterns may influence which new evidence is noticed and interpreted, helping explain both continuity and change. Contemporary work examines the relative contribution of these processes rather than treating either rigidity or unlimited malleability as proven. Fraley, 2002; Waters, Zhu, and Roisman, 2025.
Attachment Activation, Exploration, and the Secure Base
The attachment behavioral system is especially relevant when a person perceives threat, separation, fatigue, or uncertainty. A working model may then affect which figure comes to mind, how likely support seems, and how quickly distress diminishes after a response. Under relatively safe conditions, confidence in support can accompany exploration, learning, and independent activity. These are functions of the broader attachment system; working models are Bowlby's proposed representational mechanism connecting prior caregiving experience to current anticipation. Bowlby, 1969; Bretherton, 1985.
Ainsworth's work is essential at this point. Her field observations and laboratory research refined the secure-base concept as an observable organization of exploration and reunion behavior and investigated the relationship between caregiving and attachment security. Ainsworth's Strange Situation procedure classifies behavioral patterns in a structured infant-caregiver context. A working model, by contrast, is an inferred psychological representation invoked to explain how patterns of experience can influence responses. An observed classification should not be casually equated with a complete map of the child's inner expectations. Bretherton, 1992.
Why the Model Is Described as Internal and Working
Internal refers to psychological representation: information about relationships is carried forward between encounters and can be used even when an attachment figure is physically absent. Working emphasizes active use, including selection, interpretation, anticipation, and adjustment. A working model should be able to support a person in evaluating possibilities, such as whether asking for reassurance now is likely to be helpful. Bowlby, 1973.
The terminology does not establish that the model is always conscious, symbolically verbal, or localized in a single neural region. An adult may articulate an expectation in a sentence; an infant may demonstrate anticipatory behavior without language. Neither observation alone identifies the exact representation producing the response. Detailed claims about neural implementation need their own neuroscientific evidence, beyond the historical attachment framework.
Internal Working Models and Attachment-Related Information Processing
Bowlby's explanatory proposal extends to the way relational cues are processed. A person who anticipates dependable support may interpret ambiguous contact differently from someone expecting dismissal or uncertainty. Attention to signs of availability, memory for support or rejection, expectations about conflict, and interpretations of interpersonal behavior are candidate processes through which representations influence later interactions. Dykas and Cassidy's integrative review assessed attachment-related social information processing across childhood, adolescence, and adulthood and organized evidence around variation in processing associated with attachment histories and security. Dykas and Cassidy, 2011.
This field faces important design challenges. A correlation between an attachment measure and interpretation of a social event can arise from several pathways: the representation may affect processing, a current relationship may affect both, or a third variable may shape each. Experimental tasks can test immediate attention or interpretation under specified conditions, while longitudinal studies provide evidence about temporal ordering. A claim that one particular memory bias causes an adult attachment pattern requires a design stronger than a cross-sectional association.
Defensive Exclusion and Incongruent Information
Bowlby linked working models to the selective accessibility of attachment-relevant information and to defensive exclusion: the theoretical idea that painful or threatening relational information may be restricted from processing or awareness in particular circumstances. This aspect of his thinking drew on clinical questions about loss, conflict, and the possibility that people maintain representations that are partly incompatible. Later attachment research explored idealization, gaps in narrative coherence, and other features of discourse, but the operational constructs of particular instruments were developed after Bowlby's initial theoretical statements. Bowlby, 1980; Main, Kaplan, and Cassidy, 1985.
Defensive exclusion is a theoretical interpretation. Narrative inconsistency can have many explanations, including ordinary limits of memory, cultural conversational styles, emotional arousal, and the conditions of assessment. It should not be used to infer trauma or a psychiatric disorder from a brief interaction. The scientific task is to formulate testable links between specific processing operations and documented histories or outcomes.
From Bowlby's Theory to Mary Main's Representational Research
One major later development was a methodological move from observing infant behavior toward studying attachment at the representational level. Main, Kaplan, and Cassidy's 1985 work examined how early relationships might be organized in language, memory, and discourse across development. Related research used the Adult Attachment Interview to evaluate states of mind regarding attachment, emphasizing the organization and coherence of discussion rather than counting favorable childhood statements. These developments are associated with Main and colleagues and should not be retroactively credited to Bowlby as his own assessment technique. Main, Kaplan, and Cassidy, 1985.
An adult interview classification concerns a person's discourse about attachment experiences under standardized coding rules. An infant Strange Situation classification concerns observed behavior with a particular caregiver in a particular procedure. The two measures address related but different phenomena; they are not interchangeable measures of a single hidden essence. Early research on intergenerational associations found meaningful concordance, while subsequent large syntheses tested how strongly such relations generalize and how much of them caregiving sensitivity explains. van IJzendoorn, 1995; Verhage et al., 2016.
The Secure-Base Script: A More Specific Modern Proposal
A major criticism of internal working models is their breadth. If a model can mean almost any memory, belief, emotion, strategy, or relationship schema, the theory becomes difficult to test decisively. Harriet Waters and Everett Waters proposed one more specific component: a secure-base script, an organized expectation about the sequence in which a problem emerges, a person seeks or receives support, an attachment figure responds effectively, and the person can resume constructive engagement. Waters and Waters, 2006.
Researchers investigate secure-base script knowledge through narrative methods that ask participants to construct stories from specified prompts and then code how clearly the stories express the relevant support sequence. This operationalization has the advantage of stating what representational content the investigators seek. It also introduces methodological questions about language ability, storytelling conventions, context, coder agreement, and the extent to which a narrative reflects expectations carried into real relationships. These questions are central to assessing construct validity. Waters and Roisman, 2019; Waters, Zhu, and Roisman, 2025.
Research on script knowledge has examined antecedents in caregiving, links with other attachment measures, longitudinal continuity, and associations with relationships and parenting. Waters, Zhu, and Roisman's 2025 review describes meaningful progress while emphasizing that the full psychological mechanisms and the boundaries of the construct still require investigation. Evidence about one secure-base script component therefore advances, but does not exhaust, the larger Bowlby working-model proposal. Waters, Zhu, and Roisman, 2025.
Different Ways Researchers Assess Attachment Representation
Narrative interviews and discourse
Interview procedures can study how a person organizes discussion of childhood or current relationships. They may reveal coherence, contradictions, perspective taking, and ways of discussing distress. Their interpretation depends on validated coding systems, standardized administration, and appropriate populations. The Adult Attachment Interview is a specialized research and clinical-assessment instrument; a spontaneous conversation or an online questionnaire is not an equivalent substitute. Main, Kaplan, and Cassidy, 1985.
Child stories and play-based methods
Story-completion and doll-play tasks invite children to respond to hypothetical separation, fear, or reunion scenes. Trained coders evaluate features defined by the method. These techniques extend inquiry into representations beyond the infant laboratory, but their validity must be judged for the child's age, language, cultural context, and the exact measurement protocol. A story is an elicited performance affected by task demands; it is not a literal transcription of unconscious content. Sherman, Rice, and Cassidy, 2015.
Secure-base script narratives
Script tasks focus on a particular event structure and allow investigators to formulate clearer hypotheses about expectations and predictions. Their narrower target is scientifically useful, yet script knowledge and all aspects of attachment representation cannot be treated as synonymous. Waters and Waters, 2006; Waters, Zhu, and Roisman, 2025.
Self-report and social-cognitive experiments
Adult attachment questionnaires often measure dimensions such as anxiety and avoidance in close relationships. Those measures are relevant to later attachment research, but they do not directly read out Bowlby's original caregiver-specific working models. Social-cognitive tasks can investigate attention, recall, interpretation, or response to support cues. The strongest conclusions arise when different methods converge on well-defined predictions while their differences are retained. For the established thematic treatment of adult orientation measures, see Adult Attachment Theory: How Anxiety and Avoidance Shape Relationships. Dykas and Cassidy, 2011.
What Does the Evidence Actually Support?
The evidence is strongest for the existence of organized attachment relationships, measurable differences in caregiving and security, and reproducible associations among defined representational measures and some later relational outcomes. The broad assertion that a single internal working model entirely mediates every pathway from infancy to adult relationships has a different evidentiary status. It remains a theoretical proposal with only partial direct mechanistic testing. Sherman, Rice, and Cassidy, 2015; Waters, Zhu, and Roisman, 2025.
Madigan and colleagues' 2024 meta-analysis assembled 174 samples, 230 effect sizes, and 22,914 caregiver-child dyads and found a pooled association between sensitivity and attachment security of r = .25, with a 95% confidence interval from .22 to .28. Maternal and paternal sensitivity were both positively associated with security. This is evidence for a meaningful average relation between care and attachment, not a direct numerical estimate of the formation of an internal working model and not a rule that fixes the outcome for any child. Heterogeneity and study methodology matter. Madigan et al., 2024.
Research on attachment-related representations adds another layer. Verhage and colleagues' 2016 synthesis covered 95 samples and 4,819 participants in analyses of intergenerational transmission and found positive, heterogeneous associations, including a secure-autonomous association of approximately r = .31. The authors identified a remaining transmission gap: established caregiving-sensitivity measures explained only part of the connection between parental representations and children's attachment. Such findings support further study of mechanisms while making simplistic inheritance narratives untenable. Verhage et al., 2016.
Stability, Developmental Continuity, and Change
Bowlby's framework anticipates continuity because prior experience influences present expectations, which in turn can affect how new relationships are approached. If a person expects rejection, the anticipation may change what is noticed, requested, or interpreted in a new encounter. Yet new experiences can also modify the prediction. Continuity and change are both theoretically relevant. Fraley's 2002 meta-analysis and modeling study considered prototype and revisionist explanations of attachment stability and found moderate continuity during the first 19 years, with the investigated data better fitting a prototype-like dynamic. These are population-level longitudinal patterns and model comparisons, not a prediction for an individual life. Fraley, 2002.
The phrase early experiences shape later relationships is therefore best understood probabilistically. Some continuity can emerge because environments remain similar, relationships recur, or individuals carry expectations into new contexts. Some change can occur because of supportive relationships, transitions, stress, development, or altered circumstances. Longitudinal studies must separate the influence of a person's earlier relationship experience from continuing features of the environment. Neither a single infant observation nor an adult questionnaire can reconstruct that full developmental history.
Multiple Caregivers and Multiple Working Models
Bowlby's historical writing frequently emphasized the child's principal attachment figure, often described as the mother. Contemporary empirical work considers fathers, grandparents, adoptive parents, other stable caregivers, and the wider ecology of care. Different relationships can supply different patterns of availability and different opportunities to learn. A child may hold relatively secure expectations in one relationship and uncertainty in another. These possibilities follow from the caregiver-specific character of representations and are compatible with a developmental literature examining multiple relationships. Sherman, Rice, and Cassidy, 2015; Madigan et al., 2024.
The scientific question concerns the quality, continuity, and circumstances of actual care. A historical monotropy proposition should be distinguished from a claim that only one biological mother can provide an attachment relationship. An individual caregiver's responsiveness may help a child establish a particular expectation, while broader security and adaptation reflect a network of experiences. The size and organization of such effects should be measured rather than presumed.
Internal Working Models Beyond Childhood
Bowlby conceived attachment as relevant across the life course. Later researchers investigated close friendships, romantic partners, parenting, and adult responses to threat or separation. Working-model language offers a way to ask how people anticipate support and interpret their own entitlement to care. The adult attachment literature, however, includes distinct dimensional self-report traditions and interview-based traditions with different measurement assumptions. They can inform one another without collapsing into a single test or classification. Dykas and Cassidy, 2011; Fraley, 2002.
A person might approach one partner with confidence and still fear disapproval in another relationship, or feel secure during ordinary circumstances yet expect abandonment under severe stress. Working models may differ in specificity, accessibility, and responsiveness to context. These examples illustrate the framework's analytical range; individual claims require valid evidence about the person and their circumstances. Broader questions about adult romantic-relationship patterns belong to the dedicated relationship research, not a redefinition of Bowlby's historical contribution.
Clinical Relevance and the Limits of Clinical Inference
Working-model concepts can help clinicians explore a patient's expectations of support, vulnerability, separation, and interpersonal safety. A therapist may attend to how earlier experiences and current encounters interact and how opportunities for more dependable relationships might support revised expectations. Bowlby's clinical writing contributed to this language of relational experience. A theoretical formulation, however, is not by itself proof that a specific treatment works or that changing a particular representation causes recovery. Clinical efficacy must be established for the actual intervention, population, outcome, and study design. Bowlby, 1977; Dykas and Cassidy, 2011.
Neither an insecure attachment classification nor a low secure-base script score is a psychiatric diagnosis. Attachment-related expectations can coexist with depression, anxiety, trauma-related problems, or ordinary interpersonal distress, but they cannot alone diagnose any of them. Separation anxiety disorder has its own diagnostic criteria and clinical evaluation; it must not be confused with an infant's normative protest at separation or with Bowlby's historical explanation of attachment behavior. Psychological assessment requires clinical context and instruments appropriate to the question.
Methodological Challenges and Scientific Criticism
Defining the construct narrowly enough to test
The internal working model has functioned as a broad explanatory umbrella. Its strength is the integration of cognition, affect, and relational history. Its corresponding scientific challenge is specificity: researchers must state which information is represented, in what format, at which age, through which process, and with what measurable outcome. Waters and Waters made this criticism explicit when advancing secure-base scripts, and the 2025 review continues to examine where the evidence supports the narrower construct and where mechanistic explanation remains incomplete. Waters and Waters, 2006; Waters, Zhu, and Roisman, 2025.
Direct observation and circular explanation
A model cannot be confirmed merely because a behavior is called secure and then explained by an assumed secure model. Stronger studies define representation and outcome through partly independent measures, use blinded coding when possible, compare rival explanations, and test whether representations predict outcomes beyond present caregiving, stress, temperament, and other relevant factors. Establishing a statistical pathway requires temporal ordering and attention to confounding. The intergenerational transmission gap shows why a theoretically plausible mediator still needs direct research. Verhage et al., 2016.
Culture, language, and measurement
What support looks like, how it is requested, and how relationships are narrated vary by family and social context. Narrative coding and self-report methods can be sensitive to language, schooling, cultural expectations, and task familiarity. Researchers must demonstrate reliability and comparable interpretation across groups rather than assuming that every measure transports unchanged. Theoretical attention to relational availability can be retained while hypotheses about its expression are evaluated within actual contexts. Waters, Zhu, and Roisman, 2025.
Causal direction and alternative mechanisms
Caregiver behavior and children's responses affect one another. Changes in family circumstances may influence both caregiving and a child's security; child characteristics may affect how responses are elicited or perceived. Repeated measures, experimental interventions, and appropriate statistical models provide stronger evidence than simple associations, though each design has limitations. An observational link between sensitivity and security is important, but it does not directly identify an internal representational mechanism. Madigan et al., 2024.
What Bowlby Proposed and What Later Research Added
Bowlby's contribution was the developmental idea that representations of attachment figures and self guide predictions about protection and support; that such representations are shaped by interaction; and that they can organize later adaptation while remaining open to revision. His major exposition appears across Attachment and Loss, especially the 1973 volume Separation: Anxiety and Anger, and is illuminated by his wider account of the attachment behavioral system. Bowlby, 1969; Bowlby, 1973.
Ainsworth contributed foundational field and observational evidence and methodological refinements of attachment security. Main, Kaplan, Cassidy, and later researchers developed representational approaches and interviews. Waters and collaborators advanced the narrower secure-base script formulation and its measurement. Contemporary meta-analyses study effect sizes and heterogeneity that Bowlby could not have estimated from the available evidence of his period. Preserving these attributions makes the intellectual history and the modern evidence understandable on their own terms. Bretherton, 1992; Main, Kaplan, and Cassidy, 1985; Waters, Zhu, and Roisman, 2025.
The Place of Internal Working Models in Contemporary Attachment Science
Bowlby's concept remains valuable because it identifies a mechanism worth explaining: experiences in emotionally consequential relationships can affect how individuals anticipate future support. Current cognitive-developmental research increasingly asks what precisely is learned, how it is represented, when it is retrieved, and when it changes. The secure-base script is one promising operational form, and information-processing studies add testable predictions about attention, memory, and social interpretation. The evidence supports a sustained research program while requiring restraint about direct access to another person's mental representations or deterministic forecasting from infancy. Dykas and Cassidy, 2011; Sherman, Rice, and Cassidy, 2015; Waters, Zhu, and Roisman, 2025.
The enduring scientific value of the working-model proposition lies in its specificity as a question: by what psychological processes do experiences of care become expectations that guide future seeking of care? That question connects Bowlby's historical theoretical work with present research on developmental learning, social cognition, autobiographical and semantic memory, narrative, interpersonal prediction, and the revision of expectations through experience.
Frequently Asked Questions
What is an internal working model in John Bowlby's attachment theory?
It is a developing representation of how attachment figures are likely to behave, how the self can obtain their care, and what to expect in the relationship. Its proposed function is to guide predictions and action during attachment-relevant situations. The model is inferred from evidence; it is not directly visible as an object or reducible to a single questionnaire score. Bowlby, 1973.
Did Bowlby invent the working-model concept?
He adapted the general concept associated with Kenneth Craik's earlier work on mental models and made it a central mechanism within attachment theory. The historical credit for the general idea and Bowlby's developmental application are distinct. Bretherton, 1992.
Are internal working models the same as attachment styles?
They are related concepts that operate at different levels of explanation. A working model is a proposed representational mechanism, whereas an attachment classification or self-report orientation is an operational description derived from a defined measure. The measures can be relevant to the theory without being equivalent to a complete map of representation. Main, Kaplan, and Cassidy, 1985.
Can an internal working model change in adulthood?
Bowlby's account permits updating as new relationship information becomes available. Longitudinal evidence indicates both continuity and change in attachment-related organization; a particular person's future cannot be inferred from a single early measurement. Fraley, 2002.
Does the theory require one exclusive mother-child relationship?
Working models can concern different attachment figures and relationships. Modern evidence identifies meaningful associations between both maternal and paternal sensitivity and child attachment security, supporting a more differentiated view of caregiving than exclusive-mother interpretations of historical theory. Madigan et al., 2024.
How are internal working models studied empirically?
Researchers use standardized interviews, narrative and secure-base script tasks, developmentally appropriate child measures, longitudinal observations, and social-information-processing experiments. Each investigates a specified aspect and requires its own reliability and validity evidence. Waters and Waters, 2006; Dykas and Cassidy, 2011.
Related Articles
John Bowlby — A comprehensive scientific history of Bowlby's life, attachment theory, major publications, and legacy.
Attachment in Infancy and Early Childhood: Security and Development — Observational foundations, caregiving sensitivity, and the evidence concerning early attachment security.
Adult Attachment Theory: How Anxiety and Avoidance Shape Relationships — Contemporary approaches to adult attachment, relational expectations, and dimensional measurement.
John Bowlby's Attachment Theory: Origins, Principles, and Legacy — The original attachment theory and its conceptual development.
John Bowlby's Defensive Exclusion and Information Processing — Defensive exclusion and attachment-related information processing.
John Bowlby's Attachment Across the Lifespan and Later Research — How attachment-related representations informed later lifespan research.
References
Bowlby, J. (1969). Attachment and loss. Vol. 1: Attachment. Basic Books. Publisher-edition bibliographic record.
Bowlby, J. (1973). Attachment and loss. Vol. 2: Separation: Anxiety and anger. Basic Books. Bibliographic record.
Bowlby, J. (1977). The making and breaking of affectional bonds: I. Aetiology and psychopathology in the light of attachment theory. British Journal of Psychiatry, 130, 201–210. https://doi.org/10.1192/bjp.130.3.201.
Bowlby, J. (1980). Attachment and loss. Vol. 3: Loss: Sadness and depression. Basic Books. WorldCat bibliographic record.
Bretherton, I. (1985). Attachment theory: Retrospect and prospect. Monographs of the Society for Research in Child Development, 50(1–2), 3–35. https://doi.org/10.2307/3333824.
Bretherton, I. (1992). The origins of attachment theory: John Bowlby and Mary Ainsworth. Developmental Psychology, 28(5), 759–775. https://doi.org/10.1037/0012-1649.28.5.759.
Dykas, M. J., and Cassidy, J. (2011). Attachment and the processing of social information across the life span: Theory and evidence. Psychological Bulletin, 137(1), 19–46. https://doi.org/10.1037/a0021367.
Fraley, R. C. (2002). Attachment stability from infancy to adulthood: Meta-analysis and dynamic modeling of developmental mechanisms. Personality and Social Psychology Review, 6(2), 123–151. https://doi.org/10.1207/S15327957PSPR0602_03.
Madigan, S., Deneault, A.-A., Duschinsky, R., Bakermans-Kranenburg, M. J., Schuengel, C., van IJzendoorn, M. H., Ly, A., Fearon, R. M. P., Eirich, R., and Verhage, M. L. (2024). Maternal and paternal sensitivity: Key determinants of child attachment security examined through meta-analysis. Psychological Bulletin, 150(7), 839–872. https://doi.org/10.1037/bul0000433.
Main, M., Kaplan, N., and Cassidy, J. (1985). Security in infancy, childhood, and adulthood: A move to the level of representation. Monographs of the Society for Research in Child Development, 50(1–2), 66–104. https://doi.org/10.2307/3333827.
Sherman, L. J., Rice, K., and Cassidy, J. (2015). Infant capacities related to building internal working models of attachment figures: A theoretical and empirical review. Developmental Review, 37, 109–141. https://doi.org/10.1016/j.dr.2015.06.001.
van IJzendoorn, M. H. (1995). Adult attachment representations, parental responsiveness, and infant attachment: A meta-analysis on the predictive validity of the Adult Attachment Interview. Psychological Bulletin, 117(3), 387–403. PubMed.
Verhage, M. L., Schuengel, C., Madigan, S., Fearon, R. M. P., Oosterman, M., Cassibba, R., Bakermans-Kranenburg, M. J., and van IJzendoorn, M. H. (2016). Narrowing the transmission gap: A synthesis of three decades of research on intergenerational transmission of attachment. Psychological Bulletin, 142(4), 337–366. https://doi.org/10.1037/bul0000038.
Waters, H. S., and Waters, E. (2006). The attachment working models concept: Among other things, we build script-like representations of secure base experiences. Attachment and Human Development, 8(3), 185–197. https://doi.org/10.1080/14616730600856016.
Waters, T. E. A., and Roisman, G. I. (2019). The secure base script concept: An overview. Current Opinion in Psychology, 25, 162–166. https://doi.org/10.1016/j.copsyc.2018.08.002.
Waters, T. E. A., Zhu, V. L., and Roisman, G. I. (2025). A secure base script perspective on attachment: Progress, promise, and prospects. Advances in Child Development and Behavior, 69, 39–67. https://doi.org/10.1016/bs.acdb.2025.07.002.