Somewhere around the age of eighteen months, the child begins to develop what developmental psychologists call secondary intersubjectivity: the capacity to share attention with another person about an object or event in the world, to follow the direction of a gaze, to understand that the caregiver’s emotional response carries information about whether a situation is safe or dangerous. With secondary intersubjectivity, a third element enters the dyad: the world. And the caregiver’s face becomes, for the child, a referencing point for how to interpret that world. This is called social referencing. The child looks at a new toy and then looks at the caregiver’s face to find out what to feel about it. The caregiver smiles: approach. The caregiver shows alarm: retreat. The child does not yet have the language or the cognitive capacity to evaluate the caregiver’s response critically. The caregiver’s face is the evaluator. Whatever it says about the world is what the world is.
This is the developmental context in which the conclusion is reached. Not as a thought, not as a decision, not as the kind of explicit inference the adult mind produces. As a registration in the body of what the caregiver’s face consistently communicates about the child’s fullness. The child reaches for something and looks at the caregiver’s face. What does the face say about the reaching? The child becomes loud, excited, takes up space. What does the face say about the taking up of space? The child expresses anger or grief or need. What does the face say about the legitimacy of those states? The conclusion is not reached once. It is reached hundreds of times. Over time, it stabilizes into what Bowlby called the internal working model: a set of implicit predictions about what the self can safely be in relation to others.
The internal working model is not stored in the part of the brain that is accessible to reflection and revision. It is stored in the implicit memory systems, in the procedural memory and the emotional memory that operate below the threshold of conscious access. This is why the working model is so durable and so resistant to the kind of change that understanding alone can produce. You can know, consciously and with conviction, that your current relational environment is safe. You can know that the person in front of you is not going to withdraw their care. All of this can be known, and the body still makes its automatic calibrations, still finds the expression softening before it reaches the other person, still runs the monitoring program before the conscious mind has decided to monitor. The knowing is in the prefrontal cortex. The working model is in the amygdala and the body. The body acts faster than the prefrontal cortex can intervene.
The specific conclusion that produces the not-choosing loop is some version of this: the full version of me exceeds what this relationship can hold. It is almost never reached as an explicit belief. It arrives as a pattern, as the accumulated weight of evidence from the social referencing that the caregiver’s face provided. Evidence that enthusiasm was too much. Evidence that need was burdensome. Evidence that anger was dangerous. Evidence that grief required management. Evidence that the particular, unrepeatable aliveness of this child was, in its full form, more than the relationship could absorb without something changing in the relationship that the child could not afford to risk. The conclusion is adaptive. In the context in which it was reached, in the room where the caregiving adults had their own limits and their own loops, the conclusion was accurate. The full version was more than the room could hold. The calibration was the survival strategy. It worked.
The working model does not come with an expiration date. It does not contain any mechanism for automatically updating when the context changes. The child who concluded, at eighteen months or three years or seven years, that the full version of the self was not safe in relationship carries that conclusion forward into every subsequent relationship, not as a belief that can be examined and revised but as a prediction that organizes experience before it has been assessed. New relationships are perceived through the lens of the old prediction. The partner who is simply tired and distracted activates the same physiological response as the caregiver who was withdrawing. The colleague who gives critical feedback activates the same body-level alarm as the parent whose disappointment meant that the warmth would cool. The pattern-recognition system is working as it was designed to work. The problem is that the resemblance it is looking for is the resemblance of the original room, and that room no longer exists. The original caregivers are not in every room. But the body believes they are.
What can revise the working model is not understanding but experience. Specifically, accumulated experience of being more fully the self in a relational context and not having the predicted outcome occur. Not being met with withdrawal. Not finding the warmth cooling. Being, instead, met. The nervous system updates its predictions through evidence, not through insight. And the evidence it requires is not intellectual evidence but somatic evidence: the felt experience of being fully present and finding the world capable of holding it. This is the earned security the attachment literature describes. The working model can be revised. The conclusion reached before language can be reached again, differently, in different conditions. The self that was calibrated to the room can learn, in a different kind of room, that it was always allowed to be more.