Development in the Navigational Mind Framework
Working Papers ·
written 18 September 2026 · open
Download the original (.docx)
Dr. Toye Oyelese, MBBS, CCFP, FCFP
Clinical Associate Professor, Department of Family Practice
University of British Columbia
I have been aware of Erik Erikson's eight stages since medical school, and I have used them in practice for over forty years without ever being satisfied with them. Something about the model was obviously right and something about it was obviously wrong and for a long time I could not say which part was which. I want to be clear at the outset that I am not discarding Erikson. He is the foundation of what follows. What I am doing is reinterpreting the mechanism he proposed and completing the parts he left open, and it is worth saying that he left them open himself rather than claiming to have closed them.
This paper was worked out in an extended dialogue with an AI system running a governance scaffold I designed, over the course of a single long session. The research citations were retrieved during that session and several of my own formulations were corrected in the process, in some cases three or four times before I arrived at the word I actually meant. I mention this because how something arose matters when you are deciding how much weight to put on it. The observations are mine and more than forty years old. The tidiness is recent.
This paper is one part of a larger framework, and it will not make sense on its own, so I should say what the larger thing is before I use its words.
I call it the Navigational Mind Framework. It started as a handbook of practical tools I built for myself after arriving in Canada in 1987 with a medical degree that did not transfer, and it was fourteen years before I understood that the tools were all instances of one structure. The framework is the account I eventually gave of what they had in common. The tools came first. That order matters, because it means the framework is something recovered from practice rather than a theory that generated practice.
The starting premise is the only one I insist on. Uncertainty is not a problem to be solved. It is the medium. Everything is in constant change, and what I do not know will always exceed what I know, not as a failure of effort but by the structure of the situation. I call this the Null Hypothesis, and everything else follows from taking it as the condition rather than the obstacle.
From that premise the framework has four parts.
The Sphere of Ordered Experience is the theoretical framework that attempts to map and explain how experience is ordered, and I would suggest a review of the document although it is not a prerequisite to understand the information that I present here.
The Inner House: A very significant departure from Erikson where I make the argument that the stages are actually “Residents” and are all present from the start. They form a plural mind constructed like a house a structure made entirely from them. It is imperative that this document is read to fully appreciate how I propose that the house functions.
The Navigation Loop is the operation the inner house performs. Sense, interpret, Act, Reflect, Update, Orient. It is running in everybody all the time, usually without intention.
The Binary Outcome Framework BOF is not a theory of optimal choice. It is a discipline against confabulation under uncertainty.
That is the whole framework in four paragraphs, which is not enough to use it but should be enough to read this paper. What follows is only the developmental account: how the house comes to be arranged the way it is, and what that means for children. It is not an introduction to the framework, and it does not argue for it.
I have found over the years that most disagreement is actually about words, so I define my terms before I use them. I discovered this when I came to Canada in 1987 and found I was using English words outside their Canadian context and being misunderstood without knowing why.
Resident — a functional part of the mind with its own preference and its own question. I use the word "resident" rather than "part" or "drive" because a resident lives in the house permanently. You cannot evict any one of them. There are seven: Trust, Autonomy, Initiative, Industry, Identity, Intimacy, Generativity. The names are Erikson’s, and I have kept them deliberately.
The house — one person's whole internal arrangement, the residents and the rooms they occupy. I use it in this paper mostly to make one point, which is that at the beginning of life a single loop runs across two houses rather than one.
Rotational leadership — which resident is currently leading, and the fact that it changes. The others do not stop when one leads. They are all still producing, and one of them is in front.
Gating — I say a resident "gates" when every other resident's output has to pass through its question before it can proceed. I chose the word because a gate does not create traffic. It decides what gets through.
Stage — the stages and their timelines are Erikson's, and I agree with both. What I am giving is an account of what is happening during them: a stage is the period in which a particular resident is gating. That is where I depart from him — on the mechanism, not on the inventory or the timing.
Gating and rotational leadership are not the same mechanism. Rotational leadership changes moment to moment, by activation — whichever question is most urgent leads, and the others wait. Gating is stage-dependent, and under it the others do not wait their turn: their output passes through the gating resident's question. A gate does not create traffic. It decides what gets through.
The loop — the Navigation Loop, described above. One thing about it matters for this paper and I want to state it plainly, because it is not obvious. The six operations are not segments of time that happen one after another. They are all running all the time, and what I mean when I say a person is at one of them is that one of them is leading. So, the loop does not break by losing a piece. It breaks when the wrong operation leads for too long.
Reassurance — a resident's alarm answered by the other residents. I needed a word for this because "development" was being used to describe something that is not development.
Articulation — putting the internal state into words precise enough to navigate by. This is a tool in the framework and not the same thing as venting, which empties a feeling without locating it. Articulation requires language, which is why it arrives when it does and why it is unavailable to the residents that formed before language.
The Archivist — the narrating function. It produces the account of what the residents already did, and it is not a resident itself because it does not do anything. It explains.
Undesired Outcomes P(U) = P(Ur) + P(Uir)— from the Binary Outcome Framework. Where Ur is reversible and Uir is a catastrophic result that directly threatens and can end the ability to navigate uncertainty.
Navigability — the ability to navigate, in its ordinary sense: whether a person can still function in the medium of Uncertainty (change in every unit of change), pursuing their desired outcome P(D) while managing the undesired outcome P(Ur + Uir). The word is not a term of art. The criterion is the point — throughout this paper it replaces the question of whether a stage resolved favourably.
Erikson gave me the inventory. Which pressures exist, and roughly in what order they become the dominant concern. Trust in infancy, autonomy in toddlerhood, initiative in the preschool years, industry in the school years, identity in adolescence. I have never found a reason to argue with that list, and I have watched it hold in the consulting room for four decades. It is the most useful thing anyone has given me about children.
What I could never use was the mechanism. On Erikson's account each stage is a crisis with two poles, and the crisis resolves one way or the other, favourably or unfavourably, and a virtue is acquired, or a deficit is left behind. This is the part that made the model unusable for me clinically, because I kept meeting adults whose supposedly failed stages were running perfectly well under some conditions and failing under others, and I kept meeting children who were described as having resolved a stage badly and who looked to me like children under poor conditions rather than children with a permanent deficit.
It is worth noting that Erikson did not claim to have the mechanism. He built the model from observation and clinical experience rather than controlled research, he described his own work as more of a descriptive overview than an explanation of the processes driving psychosocial growth, and he left unspecified what exactly produces a favourable resolution. So, the part I am rejecting is the part he did not assert. It was added by everybody who came after him, including a great many people teaching it to medical students, and it hardened into something he would probably not recognise.
That is why I say I am completing him rather than replacing him.
My claim is that nothing is acquired at any stage. All seven residents are present and active from birth. What changes is which one is gating.
This sounds like a small adjustment, and it is not. On the acquisition model a child who did not get what the stage required has missed something and the window has closed. On my model there was nothing to acquire, so nothing can have been missed. What varies is how long the loop ran and under what conditions, and both of those remain adjustable at any age.
Look at a one-year-old and you can see all seven if you are willing to look for the operation rather than the sophistication. Autonomy is in the head turned away from the spoon. That is a boundary being asserted and it is not waiting for its stage. Initiative is in the reaching and in the dropping of things to see what happens. Industry is in the repetition, the same peg into the same hole forty times with no product and no audience. Identity is in mine and not-mine, which appears well before language. Intimacy is in wanting to be held by that person and not a substitute, which is arguably purer at eight months than at forty years. Generativity I am least certain about and I will come back to it, but I suspect it is in the offering, the half-eaten thing held out to be taken, which appears early and has no obvious function.
If they are all there, then Trust leading at 0–2 needs a different explanation, and the explanation is position rather than seniority. Every other resident's output route through Trust at that age. Initiative reaches, and whether the reach continues depends on whether it was safe. Autonomy refuses, and whether the refusal is survivable depends on the response. Industry repeats, and whether repetition is possible depends on someone maintaining the conditions under which repetition can happen. Trust is not first in a queue. It is the gate that all the traffic passes through.
The question is: is this safe?
The loop runs before language, and it is observable. In the still-face procedure, where a caregiver is asked to go unresponsive for two minutes while maintaining eye contact, infants first try to signal by smiling and orienting toward the caregiver, then escalate to crying and protest as those efforts fail, and when the bids fail altogether many of them avert their gaze or begin to self-soothe. That is a full cycle: register the situation, act on it, notice the action failed, change strategy. At five months. Whatever navigation is it is not waiting for cognition to arrive.
The same body of work tells us something I consider more important for parents than anything else in this paper. Mother-infant interaction is not characterised by high levels of coordination. It is characterised by phases of mismatch and repair. And successful repair is what gives the infant a sense of efficacy at having changed the interaction and is what builds the capacity to self-regulate.
I want to sit on that for a moment because it inverts the standard instruction. Parents are told to be consistent and responsive, which generates enormous guilt because it cannot be done, and which sets the target in the wrong place. The mismatches are not the failure. The mismatches are the normal condition of two people trying to coordinate. What matters is that they get repaired, and the repair is the thing that teaches.
One more feature of this stage which I think has been missed. The loop at 0–2 does not run inside one person. The infant senses and signals. The caregiver interprets and acts. The result comes back to the infant as sensation. It is one loop completing across two houses, and the caregiver is supplying the stages the infant cannot yet run. The caregiver is not building anything. They are keeping the loop turning until the child can turn it.
I should also be honest about a claim I would have made more strongly ten years ago. The link between secure attachment and exploration is well described, and the absence of a secure base does appear to prevent infants from exploring actively. But when two of the landmark longitudinal studies were subjected to registered analyses, the correlations between infant attachment and later socioemotional outcomes came out at r of 0.10 to 0.15 once demographics were controlled. So, I hold the mechanism and I no longer make strong claims about what it predicts twenty years later.
What is missing at this stage is not machinery. It is the handoff. Leadership is entirely reactive. Nobody is assigning.
The question is: is this mine?
Trust stops gating here and becomes something I did not have a word for until recently. It becomes the floor that Autonomy pushes from. The three-year-old who refuses help and insists on doing the shoe himself is doing it in a room with a particular person in it, and the refusal is addressed to someone who is expected to stay. A resident can be out of the “rotational leadership” and still be the reason the current leader can function at all.
I had assumed for years that Trust and Autonomy alternate at this age, and that the toddler switches between clinging and pushing. That is wrong and the tantrum research shows why. Potegal and his colleagues collected detailed parental narratives of tantrums from 335 children between eighteen and sixty months and found two independents but partially overlapping processes. Anger rises quickly, peaks at or near the beginning, and declines from there. Distress is spread evenly across the whole episode. Sadness runs continuously and anger spikes over the top of it.
So, they run at the same time. There is no switching. And the practical instruction falls straight out of it: get past the anger peaks and what remains is sadness, and a sad child reaches out for comfort. The handoff is a decay curve, not a switch.
It is also worth knowing the durations, because they change how the advice lands. The modal tantrum in that study lasted between thirty seconds and one minute, and seventy-five per cent were over within five minutes. That is a bounded event, not an endurance test.
This section is short, and it is the most important thing I learned in preparing this paper.
A child cannot hold two opposite feelings toward the same object until around the age of ten.
Harter and Buddin mapped five levels of understanding across ages four to twelve, running from no acknowledgement of simultaneous emotions at all, through same-valence emotions toward one target, through same-valence toward different targets, through opposite-valence toward different targets, and only at the fifth level, opposite-valence emotions directed toward the same target. Most four and five-year-olds reject the idea that a story character could feel happy and sad at once. Most seven and eight-year-olds accept it. Virtually all ten and eleven-year-olds accept it. Reissland found the same thing at a mean age of ten. Work comparing American and Chinese children found that recognition of mixed emotions remains difficult under the age of ten in both groups.
Every framework I have ever written assumed negotiation. Make a deal between the residents. Give this one what it needs now and that one what it needs later. I have taught it to patients, and I put it in Navvy Tales, a podcast series I made in an attempt to make this framework usable by eleven and twelve-year-olds. It cannot be done before about age ten, and I did not know that when I made it.
This is not a small correction. It means that for the whole of early and middle childhood, two residents in collision cannot be reconciled. The collision co-occurs and cannot be integrated. Whatever happens to it, it is not a deal.
The question is: is this mine to start?
Three residents are now eligible for “rotational leadership”, and the dominant pair is Autonomy with Initiative. At this age "I want to do it" and "I want to do it myself" arrive as one utterance. Trust is still the floor, further out. Initiative and Industry begin working together, though at this stage it is Initiative recruiting Industry rather than the two as partners.
What the third eligible leader adds is a route that did not exist at 2–4. When Initiative and Trust collide, the collision can be handed to Autonomy instead of oscillating. I call this "routing by displacement" and I want to be precise that it is not negotiation. A different resident takes over and the collision is left where it was.
Now, guilt. Erikson makes guilt the negative pole of this stage, a disposition that gets acquired if the crisis goes badly, and I never believed that. Here is what I think guilt is. The child starts something, it goes wrong, and two facts are now true: I began that well, and it broke. Holding two facts of opposite valence toward the same object is the fifth level and it is six years away. So, the two facts cannot be held together, and one of them has to go somewhere, and the only place available is the self.
Guilt on this account is an artefact of a missing capacity rather than an outcome of a badly resolved crisis. It should therefore soften as the capacity arrives, rather than persisting as a settled trait. I do not have evidence for that prediction, and I would like to see it tested.
The management instruction that follows is not encouragement. It is to lend the child the operation they will not have for another five years. Hold both facts out loud yourself. You started that well, and it broke. Both, in one breath, because the child cannot put them in one breath.
And it predicts a specific failure of the encouragement approach, which I have watched happen in families for years without understanding it. If you praise the attempt and say nothing about the breakage, the breakage is left resolved, and the child still has to put it somewhere. The only place available is themselves. I propose that praising the effort alone can produce the very self-attribution it was meant to prevent.
The question is: can I do this?
This is the stage I have thought about longest because it is the one my children's material is aimed at, and it contains a structural problem I only saw clearly this year.
Industry and Identity work together. Competence accrues and becomes a story about who you are, and that is the whole point of the stage, and it is what makes effort worth spending. Industry and Identity are also in conflict. To become competent you have to attempt, and attempting means failing visibly, and a visible failure threatens the story.
The same pair is both the alliance and the conflict. This matters because it rules out the obvious protective move. You cannot shield the child from the Identity channel to spare them the risk, because that is the channel competence travels down. Close it and nothing accrues.
So, the route through is not to protect the child from the story but to make sure the story is about the round rather than about the child. A tower fell. Not: I am someone whose towers fall. In the children's material I put this as a tablet that got too full and had to shut down, and that is what the image is for.
Which brings me to the distinction I now consider the single most useful thing in this paper.
You can teach hot without burning the child.
Zero probability of irreversible harm does not mean experience without consequences. It means experience without catastrophic consequences. If the break costs nothing the child will not attend to it and nothing is learned. The sting has to land. What must not happen is the burn. A tower falling is hot. Being the child who never gets the tower up is a burn. And the difference is not severity, it is recoverability: does the cost end, and does the next attempt start clean.
A score kept across a whole term does not hurt more than a falling tower because it is bigger. It hurts because it stops being information about the attempt and becomes evidence about the person.
The management instruction inverts the usual advice. Not: ensure success experiences. Ensure recoverable failure, repeatedly.
Generativity is quiet through this whole period, and I could not work out why for a long time. I think it is because Generativity requires having something to hand on, and a child's something is small. But it suggests a lever I have not tested: if the output of the child's effort is shown or given to someone, the story may land on the object rather than on the child.
The question is: does this fit who I am?
This is a different kind of gate from the four before it. Trust gates on safety, Autonomy on ownership, Initiative on permission, Industry on competence, and all four of those are conditions in the world. Identity gates on a representation, and the representation is itself under construction. Every other resident's output is now checked against a story that is being written while it is being used as a standard.
But the structural change at this age is not the new gate. It is that the capacity to hold opposite-valence facts about the same object has arrived. Every stage before this one had collisions that could only be displaced. From here they can be held. Which means that for the first time the child can do the thing that a parent has been doing on their behalf since they were four: hold "I failed at this" alongside "I am not this failure," without help.
This is why I call it the hinge and not simply the fifth stage. Before it, navigation runs but it is reactive, and it borrows operations from adults. After it, the same loop runs with the possibility of intention. Nothing further is acquired. What changes is that the “rotational leadership” can be handed deliberately.
A new conflict appears here that has no equivalent earlier, which is Identity against Identity across time. The story told at fourteen colliding with the story told at seventeen. And Intimacy stops being idle, because being genuinely known by one person requires that there be something to know.
I should flag a limitation. The research showing that the integrating capacity arrives around ten measured children's judgements about characters in stories. Whether a fifteen-year-old can hold two opposite facts about themselves, under real stakes, with something at risk, is a different question and I have not found work on it. My whole account of the hinge depends on it and I am aware that it is not established.
The question is: does this outlast me?
Generativity is a stage in my sense of the word, and it gates in adulthood. Everything else routes through whether it extends past the person producing it.
It has one property that no other resident has. The other six ask questions that can be answered inside one house. Is this safe. Is this mine. Can I do this. Generativity's question is answered after you, in someone else, and you never get to see the answer. It is the only investment whose return is unavailable to the investor.
I think this is also why Generativity looks absent in early childhood and is not. It is present from birth like the rest, and what changes in midlife is that it gates. Erikson read the stage off the point where it became conspicuous rather than the point where it started running.
And here is where it connects back to the beginning of this paper. The caregiver completing the infant's loop at 0–2 is Generativity gating in the adult while Trust gates in the child. One loop, two houses, two different residents “rotating leadership”. The mechanism by which development happens at all is a resident doing its job.
Erikson's eighth stage is integrity against despair, and I have left it out. I want to say why, because leaving out a stage looks like carelessness.
Integrity does not run anything. Every resident has an operation and a forward question that some action can answer. Integrity asks whether it was worth it, which points backwards at a finished set, and no act satisfies it. You cannot do anything about integrity.
It also fails a test I apply to all seven residents, which is that they are present from birth. An assessment of a completed life cannot be present at two years old, because there is nothing yet to assess.
And a terminal stage implies an endpoint, which my account of the structure does not permit. On the account of the Sphere given above, what looks static is a rate of change slow enough to be read as a thing. Stop the motion and there is nothing left over.
Integrity, in my framework, is not a resident. It is the narrating function producing an account of what the residents already did. It is real and people experience it, and I do not think it belongs in a developmental sequence.
There is a finding in the attachment literature that appears to contradict everything above, and I want to address it directly because it is the most serious objection to my account.
Over the preschool years the child's attachment behaviour is described as developing into an internal working model, a cognitive framework of mental representations of relationships. That looks like Trust developing, which would mean something is being acquired after all.
My position is that Trust stays pre-verbal, and what develops is the account of it. The internal working model is Identity and Articulation working on Trust's output. It is measured through language, through narrative and interview and doll-play, which is precisely why it appears at the age when language becomes adequate to it. The operation did not change. The report of the operation became possible.
What varies between people is not the strength of the signal but the strength of the reassurance, and there is a clean experimental result on this. Mikulincer, Dolev and Shaver had participants suppress thoughts about a relationship breakup and then perform a colour-naming task under either high or low cognitive load. Under low load, avoidant individuals showed no rebound of the suppressed thoughts and brought to mind positive things about themselves. Under high load, which was nothing more than holding a seven-digit number in mind, the suppression failed, and negative self-representations surfaced. The suppression is effortful. Tax the resources it needs and the underlying signal comes back unreassured, including the material the positive self-representations were covering. The same result has been produced by a different route, using mental fatigue instead of a concurrent task.
So, what looks like mistrust as a trait is, on my reading, Trust running with the reassurance off. The two accounts make different predictions. A trait predicts stable expression. Reassurance predicts load dependence. The load test favours reassurance.
I will not overstate it. Attachment classifications do change over twenty years, particularly after significant negative life events, and psychotherapy produces shifts toward security. So, something moves. Whether what moves is the underlying signal or the strength of the reassurance is not settled by anything I have found, and the field has been arguing about exactly this since at least 2002 under the names prototype and revisionist. I am aware that my position is not distinguishable from the alternative on the change data.
There is one clinical consequence that does not depend on winning that argument. If reassurance is effortful and collapses under load, then a child who is tired, hungry, unwell or overwhelmed has Trust running at full volume with nothing reassuring it. Which means the verbal tools are unavailable at exactly the moment they are most wanted, because the residents that perform them have gone offline. Attend to capacity first. Rest, food, quiet, reduced demand. Then talk.
I need to be careful with a word here. Erikson's stages are fixed. It would be easy to read my account as saying they are correctable instead, and that is not what I mean either, because correctable implies a right configuration to restore and I do not believe there is one.
What I mean is manageable.
There is no target state. Nothing has failed to be reached. The criterion is navigability: can this person pursue what they want while managing what they cannot resolve. Collisions that were never integrated are current load, not accumulated damage. Load gets managed. Deficits are permanent. They are different objects and confusing them is what makes the acquisition model so discouraging to families.
And the four instructions that came out of the stages turn out to be one instruction with four surfaces.
Supply the operation the person cannot yet run and let the cost end.
At 0–2 that is repair. At 2–4 it is timing. At 4–6 it is holding both facts aloud. At 6–12 it is recoverable failure. The same move each time, addressed to whichever resident is gating.
One more thing about the past, since a developmental account invites archaeology. How someone got here is always relevant and it is only important if it tells them how to get out. There is no chapter in my work about excavating origins and that is a commitment rather than an oversight. The question is where do I go from here.
It is not proven. It is a reinterpretation of an inventory I did not compile, using a vocabulary I built for other purposes, with research attached afterwards.
It does not generate Erikson's order. Nothing in my account explains why Trust gates before Autonomy before Initiative rather than some other sequence. Erikson at least offered a rationale in the child's changing physical and social situation. I have described the sequence without producing it, and I am borrowing the order from him. That is a real gap, and I do not want it papered over.
It is not a treatment protocol and nothing in it should be used to delay clinical assessment of a child who is unwell.
And it is coherent, which I have learned to distrust. A model built later, with an earlier model's anomalies visible during construction, will almost always be tidier than the one it revises. Coherence is not evidence.
Whether the integrating capacity that arrives around ten is actually deployed under personal affective load, or only available for judgements about other people. The hinge depends on it.
Whether the gating order can be derived from inside my framework or has to be borrowed from Erikson permanently.
Whether guilt softening as the mixed-emotion capacity arrives is something anyone has looked for.
Whether reassurance strength is teachable or maturational. If it is teachable, then the only route to Trust runs through deliberately assigning a resident to reassure it, and that would be a usable clinical technique. If it is maturational, it is not.
Whether routing by displacement at 4–6 leaves anything behind. I have said it is load rather than deficit and I believe that, but I have not tested it, and it is the place where my model would most resemble the one I am arguing against.
The still-face procedure and the mismatch-and-repair finding: Tronick and colleagues, 1978 onward; Gianino and Tronick, 1986; Mesman and colleagues' review and meta-analysis, 2009.
Registered analyses of infant attachment predictive validity: Minnesota Longitudinal Study of Risk and Adaptation and the NICHD Study of Early Child Care and Youth Development.
Tantrum structure and duration: Potegal, Kosorok and Davidson, 2003, on 335 children aged 18 to 60 months; Potegal and Davidson, 2003.
Mixed-emotion development: Harter and Buddin, 1987; Reissland, 1985; cross-cultural comparison of American and Chinese children, 2023.
Attachment and cognitive load: Mikulincer, Dolev and Shaver, Journal of Personality and Social Psychology, 2004; replication by cognitive depletion, Kohn and colleagues; neural correlates, Gillath and colleagues, 2005.
Erikson's own characterisation of his model as descriptive rather than explanatory is well documented in the secondary literature and in his own later writing.
A note on the spelling: the name is Erikson, with a k. I have spelled it Erickson for years and it is worth fixing before this goes anywhere.