Research Note 003: What Happens Between the Modes?

Cognitive Occupancy, Deconstruction, and the Problem With Drawing Arrows

In the last research note, I gave myself a constraint that has become increasingly inconvenient: I can’t responsibly model transition weight until I’ve sufficiently characterized transition distance, and I can’t characterize transition distance until the modes on either side have defensible boundaries.

I described it a little more simply then: before I measure the distance between rooms, I need to draw the walls.

I still believe that, but since writing it, I’ve realized there was another assumption hiding inside the way I was drawing the map, because I was treating the movement between those rooms as though it were clean.

Research → Synthesis.

Synthesis → Decision.

Decision → Creation.

The arrows make the model easier to understand, but they also make it look like one cognitive state ends, another begins, and somewhere in between your brain politely completes the handoff before carrying on with the rest of your day. That is s far removed from what my experience looked like a few weeks ago and we need to talk about it.

I was dealing with something emotionally consuming while also trying to continue working, and I noticed that I couldn’t change tasks. Well, technically, I could. I knew how to open the next document, move to another project, sit in front of the work I was supposed to complete, but changing the thing in front of me didn’t mean my cognition had completely arrived at that new destination with me.

Some part of my cognitive capacity still seemed occupied by what came before it.

That was the moment I started using the phrase cognitive occupancy, and I’m using that term very intentionally as a provisional description because I don’t actually know yet what mechanism I’m describing.

Whether what I’m calling cognitive occupancy is residual activation from the previous mode, unresolved information still taking up cognitive resources, emotional load interfering with my ability to reorient or some combination of those things. If I’ve stumbled into something else entirely, I can’t decide which explanation I like best and build the theory around it.

What I can do is admit that the experience made me look at the arrows differently.

If a person moves from Research to Creation, when exactly has Research ended? Is it when they stop searching, when they open the document where they’re going to write, when they generate the first original sentence, or at some point that isn’t cleanly visible in their behavior at all?

Could the behaviors associated with Creation begin while some portion of the previous cognitive demand remains active, and if they can, does that mean two cognitive states are meaningfully overlapping, that one state hasn’t fully resolved, or that I’ve drawn the boundaries incorrectly in the first place?

I don’t know yet, which unfortunately means the arrow just became a research problem.

Feeling Something Isn’t the Same as Explaining It

You can feel ignored without someone intentionally ignoring you, feel behind without actually being behind, or feel like you can’t do something without yet understanding what’s making it difficult. The experience is still information, but the explanation is a hypothesis.

Annoying in therapy. Extremely useful in research.

That’s essentially the problem I have with cognitive occupancy right now.

The idea resonated quickly when I talked about it publicly, and that matters because other people recognizing an experience can tell me there may be something here worth investigating, but recognition can’t tell me what caused the experience or whether everyone recognizing it is actually describing the same thing.

That distinction is becoming increasingly important as CMT develops because this entire project started with lived experience. I noticed patterns in my own work, started giving myself language for describing them, shared those ideas publicly, and then discovered that other people recognized many of those same patterns in themselves.

Enough to generate a research question, but it isn’t enough to answer one.

If I experience enormous friction moving from one kind of thinking into another, I can’t work backward from that experience and decide the transition itself must have been cognitively expensive, and if several people tell me they experience the same thing, I still can’t assume we’re experiencing it for the same reason.

My last research note was largely about preserving the distinction between transition distance and transition weight, and I don’t want cognitive occupancy to become another variable I add to the model simply because it gives me an intuitively satisfying explanation for something I know I’ve experienced.

Before I can claim occupancy changes the weight of a transition, I need to establish that I can reliably identify the transition, which means I need to know what changed, where that change happened, and what evidence gives me any right to say that it did.

And before I can do that, I need a way to reconstruct the cognitive episode without assuming Cognitive Mode Theory is already correct.

That’s where Recursive Cognitive Deconstruction becomes much more important.

If the Task Isn’t the Task, What Exactly Am I Deconstructing?

The first question that eventually became CMT was:

What if the task isn’t actually the task?

A task like “write the article” is useful for telling me what I’m responsible for producing, but it tells me almost nothing about the cognition required to produce it because I may need to find information, understand what I found, determine which information matters, establish relationships between ideas, decide what I actually want to say, construct a representation of that argument, evaluate whether the representation communicates what I intended, and change it until it does.

From the outside, I spent the afternoon writing an article, and cognitively that description tells me almost nothing.

This creates a methodological problem because most of the evidence available to us is behavioral. We can observe someone reading, writing, searching, talking, editing or moving information around, but none of those behaviors automatically tells us what cognitive transformation is happening underneath them.

Reading is probably my favorite example because the behavior can remain almost perfectly stable while the cognition changes substantially.

I can read because I’m trying to discover information I don’t have, because I’m determining whether something is credible, because I’m comparing two competing explanations and I’m still reading the entire time, which means the behavior didn’t tell us the mode.

What changed was what I needed the information to become.

That distinction is forcing me to get much more precise about what Recursive Cognitive Deconstruction is actually supposed to recover, because if the observable behavior isn’t the thing I’m classifying, I need a defensible way to get underneath it.

Start With What Became True

RCD begins with a bounded piece of work and an observable outcome, because before I can ask what cognition occurred, I need to know what that cognition eventually produced.

Let’s return to the newsletter that has somehow become the unpaid research assistant of Cognitive Mode Theory.

I could define the work item as “write the newsletter,” but “write” immediately anchors the description to a behavior, so instead I need to ask what completion actually looked like.

The outcome might be:

An approved newsletter was delivered to subscribers.

Now I have something I can investigate because something is true at the end of this work episode that wasn’t true at the beginning, and instead of asking what I did to get there, RCD can begin asking:

What had to become true before this outcome was possible?

An approved newsletter couldn’t be delivered until something capable of being approved existed, that artifact couldn’t exist until its contents had been represented somehow, and those contents couldn’t be represented coherently until someone had determined what the newsletter actually needed to communicate.

That determination required relevant information to be available, relationships between that information to be understood, possibilities to be narrowed, and some commitment to what mattered enough to include, which means the further backward I go, the less useful it becomes to describe the process as “writing a newsletter.”

The important thing here is that I’m not trying to recreate a to-do list, and I’m not documenting a workflow just because one thing happened before another.

I’m trying to reconstruct dependencies between cognitive states.

The question isn’t merely whether B happened after A, but whether B could reasonably have become true without whatever transformation happened in A, because chronology alone doesn’t tell me which cognitive changes were actually necessary for the outcome.

I could check Instagram halfway through writing the newsletter or get up and make coffee, but none of those things automatically becomes part of the cognitive architecture of producing the newsletter just because it happened somewhere between the beginning and the end.

RCD is trying to recover the changes that made the outcome possible, not everything that happened while I was getting there.

The Language Is Also Lying to Me

Once I started doing this more deliberately, I ran into another problem, because some of the words we use to describe cognition are doing entirely too much work.

Take the word know.

Someone might tell me, “I needed to know the subject before I could make the decision,” which sounds reasonable until I ask what know actually means in that particular work episode.

Did they need to discover information they didn’t have, recall something they’d encountered before, interpret what the information meant or demonstrate that they could apply that understanding to the problem in front of them? Who's to say?

“Know” looks like one cognitive state in language while potentially containing several transformations underneath it, which means accepting it as a single state would allow the language to make a methodological decision for me before I’ve actually investigated what happened.

The same problem shows up when we say someone needed to learn, understand, research or analyze. Those words are incredibly useful conversationally but potentially disastrous if I start treating them as indivisible units of cognition simply because they sound cognitively meaningful.

And “research” is becoming particularly annoying for me here.

Someone can source information and synthesize information during what they would reasonably describe as “doing research,” even though those processes feel violently different in practice and may ultimately belong to completely different parts of the ontology.

If I accept “research” as the cognitive unit because that's what the person called the activity, I may have collapsed several transformations into one category before I’ve even begun investigating whether they belong together.

So RCD has to keep asking the annoying question:

What actually changed?

When the answer is still hiding several independently necessary transformations, I have to keep deconstructing it until I’ve reached a level of resolution where the state change is specific enough to investigate without pretending I’ve discovered some fundamental atom of human cognition.

That’s the recursive part.

RCD Can’t Use CMT to Find CMT

This may be the most important methodological boundary I’ve reached so far.

CMT currently proposes nine modes, which means it would be incredibly easy for me to look at a finished work episode already expecting to find Research → Synthesis → Decision → Creation, reconstruct the work around those categories, and then turn around and celebrate when CMT successfully identifies the exact modes I went looking for.

Except…girl, I put them there.

I cannot use the theory to manufacture the evidence and then use that evidence to validate the theory.

RCD can’t use CMT to find CMT. You see what I'm saying?

That means deconstruction has to happen before mode assignment, and during RCD I don’t want to ask where Research happened, whether a particular moment looks like Synthesis, or which of the nine proposed modes best describes what I’m seeing.

I want to know what became true, what had to be true immediately before it, whether that prerequisite was actually necessary, and what evidence supports my claim that the transformation occurred.

Then, reconstruction is stable enough to stand on its own should the ontology be allowed into the room. RCD’s job isn’t to prove Cognitive Mode Theory.

RCD’s job is to produce evidence that Cognitive Mode Theory has to explain.

The Evidence Has to Stay Attached

There’s another problem with reconstructing cognition retrospectively, which is that human beings are exceptionally good at making messy processes sound coherent once we already know how they ended.

That means every reconstructed state needs to retain some relationship to the evidence that justified including it, because otherwise I can very easily turn a complicated work episode into a beautiful sequence of transformations that makes perfect sense and may have absolutely nothing to do with what actually happened.

Sometimes that evidence may be direct because there’s a draft, search history, message, version change or another artifact showing that something changed, while other times the evidence may come from the person describing what they were trying to accomplish at that moment, and sometimes a state may need to remain an inference because the subsequent state appears to require it even though I don’t have direct evidence of the transformation itself.

Those aren't necessarily equivalent forms of evidence, and I don't want RCD pretending that they are.

If I know something because I can see it in an artifact, I should record that, and if I know it because the participant remembers it, I should record that too, while anything I believe because the dependency structure strongly suggests it happened needs to remain clearly identified as an inference rather than becoming fact somewhere along the way.

The uncertainty isn’t something I need to clean out of the dataset because the uncertainty is part of the dataset, and eventually the strength of that evidence is going to matter when I start assigning modes and reporting how confident I actually am in those classifications.

But that comes later.

First I need the reconstruction.

Which Brings Me Back to Cognitive Occupancy

This is where that emotional epiphany becomes relevant to me as a research question, because when I said some part of my cognition still felt occupied even though I had technically moved to another task, I was describing an experience that now needs to survive the same scrutiny as everything else I'm asking CMT to explain.

I mean, would it be me, if it didn't happen like this?

If cognitive occupancy is real in the way I’m currently imagining it, what evidence would distinguish it from emotional load, what would distinguish residual occupancy from an unresolved decision continuing to demand resources, and what would tell me that the previous cognitive mode remained active rather than the next mode simply failing to stabilize?

If two modes can overlap, what would that actually look like in the reconstruction, and if they can't, what does that tell me about the experience I'm currently calling occupancy?

Most importantly, what evidence would convince me that this explanation is wrong?

Those questions are less satisfying than drawing another box on the model, calling it Cognitive Occupancy, and telling everyone I figured it out, but they're also more useful because I don’t want CMT to become a collection of experiences that sound true.

I want to know which parts survive investigation.

I’m Still Drawing the Walls

Research Note 002 ended with me saying that I’m still drawing the map and, in some places, still figuring out whether I’ve drawn the walls in the right place, but I understand that statement differently now because drawing the walls can't mean deciding where I believe Research ends and Synthesis begins and then creating definitions that preserve the distinction.

It has to mean developing a process capable of showing me whether that distinction consistently appears in real work at all.

That's what RCD is becoming.

I start with a bounded work episode and define what became true, then I work backward through what had to become true before it, decompose states that are still hiding multiple transformations, preserve the evidence supporting each reconstruction, and keep the ontology out of the process long enough that the reconstruction has a genuine opportunity to disagree with it.

Only then do I attempt classification.

If the same structures begin appearing across different people, different kinds of work, and different observable behaviors, then I have something worth investigating further, but if Research and Synthesis consistently refuse to separate when I stop forcing the taxonomy onto the work, I need to know that too.

If cognitive occupancy turns out to describe something that's already better explained by another construct, I need to know that, and if the nine modes eventually become six, twelve, or something completely different because the evidence demands it, then the theory needs to move with the evidence rather than the other way around.

That's the point.

The question I'm trying to answer isn't whether I can look at someone's work and make it fit Cognitive Mode Theory anymore, because if that's all I wanted to do, I could probably build a very convincing framework and spend the rest of my time finding examples that make it look right.

I'm trying to build a method capable of showing me when Cognitive Mode Theory is wrong.

Because before I can measure how far someone traveled between cognitive modes, how expensive that movement was, whether cognitive occupancy affected their ability to arrive at the next one, or what they were carrying when they made the trip, I need to establish something much more basic:

What actually changed, and what evidence gives me the right to say that it did?

That's what RCD is for.

And once I have a defensible reconstruction of a cognitive state change, I inherit the next problem:

What the hell do I call it?