We named the outputs and called them the things. Depression. Shame. Anger. Guilt. These words describe what the surface looks like — the compressed downstream product of a system that has been running something much more specific. When we use the same word for mechanistically different states, we make it impossible to research them accurately, treat them effectively, or even experience them clearly.
There is a problem built into the way we talk about inner experience — in clinics, in research, in daily life. It is not a problem of not having enough words. We have plenty of words. The problem is where the words are pointed.
Every label we use for a mental state — whether a clinical diagnosis or an emotion word — is drawn from the most downstream, most compressed, most observable layer of what the nervous system is doing. We named the outputs. We built entire fields of inquiry around those names. And in doing so, we created a map that describes the surface with great specificity while leaving the generating mechanism almost entirely invisible.
This has consequences. Not minor ones.
Consider depression. Not as a lived experience — that is real, and devastating, and not in question. But as a category. What does the word actually pick out?
The DSM description of depression is a cluster of observable outputs: low mood, reduced interest, fatigue, changes in sleep and appetite, difficulty concentrating, feelings of worthlessness, diminished motivation. These are described as symptoms of a single condition. The assumption underneath this — the one that never gets stated — is that these outputs are being generated by the same underlying mechanism in everyone who meets the criteria.
They aren't.
The same surface presentation — the flatness, the withdrawal, the inability to move — can be generated by at least three mechanistically distinct system states. They look identical from the outside. They require different interventions. They respond to different things. The word "depression" does not distinguish between them.
These three presentations are aggregated under one word, studied as a single condition, and treated with the same first-line interventions. When those interventions work for some people and not others, we call it "treatment-resistant depression" — as if the problem is in the patient's resistance rather than in the category's imprecision.
The category isn't wrong, exactly. The outputs really are similar. But the map was drawn at the wrong layer.
Now extend this to the emotion words we use every day — not just clinical diagnoses but the ordinary vocabulary of inner life. Shame. Anger. Guilt. Anxiety. Grief. These words feel precise because they feel familiar. Everyone knows what shame feels like. The problem is that "what it feels like" and "what the system is doing" are not the same thing — and the word lives entirely in the first category.
Take shame. The word points at a cluster of outputs: heat in the face, a desire to disappear, a collapse in the chest, a sense of being fundamentally exposed or wrong. This is real. Every person who has felt shame recognizes it. But the mechanism generating that cluster can be operating at completely different points in the system.
In each case, the word names the output — what the surface looks and feels like. But the generating mechanism can be operating at completely different points in the system: at the body level, at the prior layer, at the concept layer, in the tank, in the burners. The same word. Different signal. Different layer. Different leverage point.
When we mistake the label for the thing, we create three compounding problems — in research, in treatment, and in experience itself.
If you run a study on "depression" and your sample contains people in all three mechanistic states — depletion, futility prior, stuck loop — you will get noisy data. Your intervention will work well for some, partially for others, and not at all for the rest. You will average across these groups and conclude that the intervention "works," with high variability. The variability is not statistical noise. It is mechanistic signal that the category is collapsing.
This is what depression research has been doing for decades. It is what shame research does. It is what anxiety research does. The constructs are real as experiences. They are not coherent as mechanisms. And you cannot build a reliable science of mechanism on a categorization scheme that ignores mechanism.
If a clinician sees "depression" and prescribes the standard first-line response — medication, behavioral activation, CBT — they are making a bet that the mechanism in front of them is the one the intervention targets. Sometimes the bet pays off. When it doesn't, the dominant clinical frame tends to locate the failure in the patient: treatment-resistant, not ready, not compliant, not doing the work.
But a person whose flatness is generated by an empty tank does not need behavioral activation. They need resource. A person running a stuck loop does not primarily need a futility prior addressed. They need the loop interrupted. Applying the wrong lever to the right surface symptom is not a neutral act — it can reinforce the problem. It can confirm, at the premise level, that trying doesn't work.
This one is the quietest damage, and possibly the most pervasive.
When you have a word for what you feel, the word becomes the container. You organize your experience around it. You explain yourself to others with it. You look for treatments that address it. And in doing this, the word can actually prevent you from getting closer to what's happening — because the word is at the far downstream end of the chain, and what's actually generating the state is somewhere upstream, running silently.
The person who calls themselves "an anxious person" has taken a downstream output, turned it into an identity, and in doing so may have foreclosed the inquiry into what the system is actually tracking, and why, and at what layer.
The label doesn't just describe — it also organizes. It shapes what questions you ask, what help you seek, what you believe is possible for you. A label drawn from the output layer will organize your self-understanding around the output layer. It will keep you working at the level of the word — managing it, reducing it, coping with it — rather than moving toward whatever signal the system is actually carrying.
This isn't an argument that emotion words are wrong, or that we should stop using them. They are fast, shared, and human — and they carry real meaning. The point is to understand what layer they're operating at and what they can and can't tell you.
A word like "depression" or "shame" or "anxiety" can be a starting point — a flag that says something is happening here. What it can't do is tell you what's happening at the generating layer. For that, you need different questions: Where in the system is this originating? Which burner is this? Is this a tank problem or a prior problem? Is this a signal that something needs attention, or a loop that needs interrupting? Is this downstream of a premise the system accepted as fact before there were words?
Those questions don't fit easily on a diagnostic form. They aren't measurable the way a symptom checklist is measurable. But they are the questions that actually point toward the mechanism — and mechanism is where leverage lives.
The label is the compression artifact. The signal is still running. The inquiry is: what is actually upstream of this word?