Section I

How emotions are made. That is not a question. There are people out there that think they have the answer to this. The most prominent is Lisa Feldman Barrett and what is called the theory of constructed emotion. Constructionist theory is the most visible and institutionally powerful theory of emotion right now and Barrett is one of the most cited psychologists alive. She literally wrote the book How Emotions Are Made. The biggest problem with constructionists is not that their work is inherently incorrect (they get plenty right), it’s that they don’t engage with literature that doesn’t support what they already believe. Given the constructionist prominence and influence on how most people currently understand the neuroscience of emotion, I think it deserves a bit of scrutiny.

In order to tell this story, first we need to embark on a tangential but hopefully breezy history. The original sin of both psychiatry and psychology is that their empirical traditions and subsequent foundations are rooted in top-down processes and top-down observations. Can you blame them? It’s not like I can seriously fault Freud for not having an MRI scanner to help him develop his theories. The giants of the 20^th century used the tools they had available. The tools they had available were mostly behavioral observation and narrative data from subjects that could be collected in mass quantities and analyzed. Ye olde psychiatrists and psychologists of yesteryear did their best and their observations were often indeed canny. They built our current nosology and mental health system.

Stop and think about that for a second: the current nosology is built on folk psychological observations, and as neuroscience developed, as the tools improved, we started grafting newly discovered neural correlates onto a folk system. This is madness! If you were designing this from scratch, knowing what we know now, you’d build it bottom-up, starting from how the brain actually works and building categories from valid processes, not starting from behavior and hunting for correlates afterward. This bottom-up approach has never been done.

Back to our researchers of yesteryear: as they leaned heavily on subjectively reported information and behavioral observations to explain how the brain worked, they collected their lexical information from patients or research subjects and this became psychometric data. With this new label it now sounds super important and objective even though it is possibly just a college student who filled out a questionnaire. Plenty of people have built their careers in the psychometric wing of psychology, and while I don’t mean to be insulting, what I’m about to say will obviously come across as insulting: the psychometric wing is low hanging fruit. This is low effort research to engage in. It is also very productive. I’m using productive here to mean it’s pretty straightforward to build a research career and publish rather than perish.

Cognitive neuroscience is a field that also conceptually suffers from the same top-down problem. Cognitive neuroscience is rooted in studying the cortex. The cortex is literally the top part of the brain. It is the most accessible part of the brain so it is the easiest part to study. It’s more forgiving in that a metaphorical stray nick in the cortex will hopefully not have the same catastrophic consequences compared to a stray nick in the midbrain or brainstem. The cortex is also much easier to image and much easier to get an fMRI as the vessels in the cortex are large compared to the smaller vessels in the lower parts of the brain. Cognitive neuroscience also came of age at the same time as the computing revolution. I was not alive, but this must've been a very exciting time to be able to have actual computing power to crunch reams of data, and at the same time also have new techniques to study the living computers inside our own heads.

Cognitive neuroscience also sort of has a similar productivity problem like psychometric psychology. As a career move it’s much more straightforward to engage in research that has an obvious path towards results and publication. So people do it. Then those same people that are successful and productive build their careers and they train the next group of researchers in their labs, and voilà institutional bias.

There is nothing wrong with studying the cortex, observing behavior, or collecting lexical data. Lots of fine research has come out of it. However, I think there is definitely something wrong with only studying these things and trying to apply what has been learned but then ignoring any information that doesn’t line up with what has already been found.

My favorite story illustrating something like this is Panksepp corresponding with larger-than-life radical behaviorist BF Skinner. Panksepp was asking him to help him build bridges and create a more universal cross discipline marriage between neuroscience and psychology. Skinner responded saying I don’t really care how the brain works I just study the behavior and it doesn’t really matter how the brain works anyway because the behavior is the behavior regardless. There you go, a lot of people doing research don’t really care about the big picture they care about their little silo. I think this is also why I had such an immediate gravitation towards Panksepp. He had big questions about emotions and he took the hard road. He did not get sucked into yet another productive career. He could see the only way to get the deeper answers would be very hard and difficult. He saw the research that would be needed to be done in humans to answer his questions about emotions was not possible in his day, and still is really difficult now. So, he said fuck it, let’s do it with animals and then build carefully towards humans in any way that we can over time.

The cortico-centric bias is real. As a reminder, usually when you encounter the term affective neuroscience in the wild you will be reading the same tired cortico-centric concepts using the term affective to apply cognitive neuroscience to emotions. Barrett’s lab is called Interdisciplinary Affective Science Lab and the URL is affective-science.org. I guess nobody really gets to own words, but when I hear people talk about affective neuroscience, they are rarely talking about primary process affective systems.

The good news though is that cognitive neuroscience, if you pare it down to its own realm, has done a pretty good job. The actual cortical processes part of the constructionist work is most certainly correct. Interestingly, the hill they want to die on is insisting that any and all subcortical systems are merely survival circuits disconnected from emotion and are not relevant for feelings. Further, they hold the view that there are no specific affective systems at all, rather just one general positive or negative valence arousal system.

I know what any reader might be thinking: who is this lunatic ranting about the most cited psychology researcher alive today, what a tool. Yes, that is exactly what is happening. Me, a nobody, just a bright-eyed-bushytailed-eager-beaver early career psychiatrist trying to look at the big picture. I have no research career. I have no stakes to protect. I just see patients every day and have been blown away with how disconnected from emotion psychiatry is at the nosological level, so it only makes sense for me to go out and read emotion research and follow that where it leads. So, let’s look at the bigger picture and take a dive into what constructionism is and the evidence behind it. Then let’s stack it up against affective neuroscience and see if we can’t arrange a marriage of affective neuroscience and cognitive neuroscience in order to integrate basic affective systems with emotional constructionism. Who doesn’t want a complete understanding of emotions? Constructionists, I guess.

Section II

Barrett’s claim is that subcortical systems provide nothing except valence and arousal. Nonspecific valence and arousal generated largely by interoceptive and allostatic circuitry: brainstem and hypothalamus. This is literally Russell’s circumplex model from 1980. His paper reports four separate studies to create and synthesize this model.

Study 1: 36 undergraduates sorted 28 emotion words into eight preset categories, then arranged those categories in a circle putting opposites across from each other. Statistics and analysis were used to turn these results into PSYCHOMETRIC DATA.

Study 2: 34 different undergraduates, same 28 words. More sorting. More analysis. More PSYCHOMETRIC DATA generated.

Study 3: No total number of subjects reported; minimum 27 subjects per word. They reused previously published rating data plus new ratings with 51 new subjects scoring the same 28 words on pleasure-displeasure and arousal semantic differential scales, plotted on two orthogonal axes. More rating. More analysis. More PSYCHOMETRIC DATA generated.

Study 4: the big one, 343 UBC students between 1976 and 1979, split between psych-course credit and paid campus recruits, rating 518 affect-descriptive adjectives on how accurately each one described how they currently felt, on a 1-to-8 scale, with more statistical analysis to yield yet more PSYCHOMETRIC DATA.

What we get in all four studies: all undergraduates, all word-sorting and self-rating tasks with multidimensional scaling and factor analysis to create an underlying structure. No anatomy, no physiology, no neurochemicals, no lesions, no knockout genes, no chop chop of the midbrain, no electrical stimulation. Just undergraduate students and statistics. I do not consider this serious evidence. I think it’s bloody sad that this paper is a foundation for how we understand emotions today. Because, that’s it, there is nothing else to it… this is Russell’s Circumplex Model of Affect.

In 1998 Barrett teamed up with Russell and published Independence and Bipolarity in the Structure of Current Affect. This paper is essentially the empirical foundation under everything that Barrett later builds as “core affect” in How Emotions Are Made. This paper tries to tackle an age-old psychometric argument (psychometric argument; an oxymoron?). Is pleasant affect the flipside of unpleasant affect on a single continuum, or are pleasantness and unpleasantness independent of each other?

They ran three studies, undergrads at Boston College and UBC, 129, 225, and 316 subjects, using adjective checklists, agree/disagree statements, semantic differential ratings, and Watson’s PANAS scales, then ran confirmatory factor analysis with correlated-error modeling. Well, considering that it was a team up, you wouldn’t expect it to disprove Russell’s work. I’m not saying the fix was in, and we will see why this was destined to be, but surprise surprise, their PSYCHOMETRIC DATA did not disappoint the authors. They found that the two-dimensional circle with valence on one axis and arousal on the other really is indeed all there is to core affect.

Osgood’s semantic differential work from the 1950s means something inconvenient for this whole architecture. Factor-analyze ratings of almost any concept on bipolar adjective scales and you get the same three dominant dimensions every time: evaluation, potency, and activity. Not for emotion words specifically. For everything! Osgood ran this across dozens of cultures and found the same three-factor structure holding up cross-culturally. Later work pointed it at logos, colors, brand names, products, services, and typefaces. Any concept a researcher wants; same evaluation-and-activity backbone, every time.

Evaluation looks a great deal like valence. Activity looks a great deal like arousal.

Adjective-rating-plus-factor-analysis, as a method, spits out a good-bad axis and an active-passive axis no matter what you point it at. Wake up psychologists! That is a property of how people use rating scales and how factor analysis finds variance. It is not a property of the thing being rated. Rate video games, historical figures, or ice cream, and you get a similar shape. This literally tells you nothing about what is actually being rated!

It is also worth noting that even researchers who accept this method’s authority on its own terms using the same tools can’t replicate the work to get Russell and Barrett’s specific two-dimensional circle to survive its own internal test. In 2021 Polish researchers with a sample size of 863 participants ages 16 to 83 (finally not just undergraduate students) found no support for the model.

The low-arousal pole, sleepiness, deactivation, calm, whatever you want to call it, had the worst fit and the lowest explained variance every time. The researchers found that if you ipsatize the data, a statistical normalization technique that rescales each person’s responses relative to their own average, the fit improves and starts looking like a proper circle. They say this outright: “this can be treated as an analytical artefact rather than as insight into the data.”

Their hypothesis for why the model fails is a measurement problem. Arousal survey items seem to smuggle in valence. “I’m filled with energy” reads positive, “I’m feeling stirred up” reads negative. Arousal may not be cleanly separable from valence in the first place. They also posit that arousal might not be a true bipolar dimension like valence. Arousal might be unipolar, activated versus not-activated rather than a real opposite-pole continuum.

I’m not invested in how psychometric psychologists argue with each other over how psychometrics explains the function of the brain. I just included it to demonstrate that people seriously argue about this stuff. Their methods say nothing other than how humans use language to rate and compare everything. Why is subjective report from college students that is run through factor analysis supposed to support a claim this large about how the brain works? How is this what passes as neuroscience?

That is the treatment that core affect gets in Barrett’s model. Subcortical structures, all work together to just send up undifferentiated valence (this is good vs. this is bad) and arousal (high intensity vs. low intensity) to the cortex. This is what she calls core affect. She explicitly frames core affect as elementary, undirected feeling, not yet about anything or aimed at anything, prior to being about a specific object or having a specific quality of any kind. She dies on the hill that core affect is only two continuous dimensions, prior to any conceptual work, and everything qualitatively specific gets added later by categorization.

She goes on to claim that core affect is categorized through a situated conceptualization process she locates heavily in paralimbic cortex: orbitofrontal cortex, anterior cingulate cortex, and anterior insula. She argues that this process is largely using stored concept knowledge rather than pure neocortical elaboration. This part of her work, that cortical processing and prediction take affect and shapes our experiences, is fine. I have no qualms with it (neither did Panksepp).

Section III

So, why is she so motivated to say subcortical circuits are detached from emotion?

She is fairly explicit about this and I can find three obvious motivations. The first is stated outright, and it’s political. She opens the book by walking through what she thinks is wrong with the classical and essentialist view of emotion and how this is bad for the world. She uses the TSA’s SPOT program (screen passengers by observation, $900 million, based on the premise that innermost feelings leak out through fixed facial and bodily signatures) as a direct casualty of essentialism. She uses the finding that women presenting with chest pain are more likely to get diagnosed with anxiety and sent home while men with identical symptoms get treated for heart disease, which she attributes partly to “classical view beliefs that they can detect emotions like anxiety, and that women are inherently more emotional than men.” Fortunately for a critic like me, she defines essentialism:

"The belief in essences is called essentialism. It presupposes that certain categories, sadness and fear, dogs and cats, African and European Americans, men and women, good and evil, each have a true reality or nature."

She explicitly draws the line from emotion-essentialism to the same cognitive move that underwrites race and sex essentialism, writing that social categories “must not be mistaken for physical, biological categories with deep dividing lines in nature.” For Barrett, saying “emotions have fixed, locatable essences in the brain” and “races or sexes have fixed, locatable essences” are instances of the same error, essentialism, and she’s explicitly hunting the general error, not just the emotion-specific case. A subcortical circuit for RAGE existing is, in her framework, structurally the same kind of claim as a gene for criminality. This seems to be a big part of why she’s not neutral about it.

Fuck, who is going to tell her that biological sex, no matter how many variations, no matter how humans define it, comes down to which gamete an organism produces? I have no interest in anyone's culture war here. The point is only her own inconsistency using her own example. She files sex alongside race and emotion as a social category we're warned not to mistake for a biological one with "deep dividing lines in nature." Does an organism produce a small and mobile gamete or a large and immobile gamete? It does not matter how many chromosomes, how many chromosomal abnormalities, or what hormonal disturbances exist, this is a near universal bimodal fact rooted in the essence of animals. Spoiler: this is true for affect too. Her test for essentialist error just flagged one of the cleanest biological boundaries in animals. She also denies an essence to subcortical circuits with fifty years of lesion and stimulation data behind them. If "deep dividing lines in nature" are real enough to build a biological category out of when the topic is sex, they're real enough to build one out of when the topic is affect, a circuit Panksepp mapped with a scalpel and an electrode.

She says “The neuroscientist Jaak Panksepp, for example, writes that an emotion’s essence is a circuit in the subcortical regions of your brain,” right alongside Pinker’s “genes” and Ekman’s “programs.” I died a little here. She files Panksepp under essentialism. This is an LMAO moment. The biggest problem here is that Pinker and Ekman are just more of the standard 20^th century psychologists with theories not rooted in neuroscience, but rooted in observation. Pinker is a psycholinguist and Ekman studied facial expressions. Wow, what a shock, psychologists using the traditions of psychology a.k.a. the only tools they had available: observations and lexical data, a fancy term for writing down whatever somebody tells you. Top-down psychologists…and she lumps in Panksepp with those guys! Panksepp the only bottom-up pioneer out there! The guy going chop chop on brain tissue of multiple mammals for fifty plus years! How can you just lump him in with them and call him an essentialist! It really is a bold move as a way to dismiss the strongest evidence, just lump Panksepp categorically with other weaker evidence and dismiss his work. We will get there eventually. We will look closer at the neuroscience program Panksepp built and why it is absurd to lump him in with psychologists relying on self-report of college students and the observations of facial expressions.

Her actual contribution to the field isn’t “emotions are constructed,” it’s a general claim that perception, memory, concepts, the self, and emotion are all instances of the same predictive mechanism, an interoceptive network running allostatic simulations and a control network arbitrating between competing predictions. That’s her unifying move. If subcortical circuits for FEAR, RAGE, LUST, and so on turn out to be real, discrete, and doing serious causal work independent of concept-driven prediction, then emotion possibly stops being an instance of the general predictive-brain story and becomes a special case with its own dedicated machinery, exactly the modular picture she’s spent her career arguing against. Conceding Panksepp’s circuits isn’t a local concession about one phenomenon. It threatens the unification that is the actual novel claim of her work.

Hold your horses. We will get to this. There is no reason the arousal from a specific subcortical circuit can't be fed up into the very cortical conceptualization and prediction networks she describes. Other than not offending Barrett, the circuit and the construction story fit together fine.

There is also a practical promise she’s making to the reader that only works if concepts are doing real causal work. Chapter 9 is literally called “Mastering Your Emotions,” and the whole pitch is that because emotion is constructed from concepts and prediction rather than triggered by a fixed circuit, you can retrain it: change your concepts, change your interoceptive predictions, change what you feel. That is a compelling promise. It also offers clinical utility. I disagree with it completely. Humans just cannot control their thoughts and feelings. As a thought experiment to demonstrate this, I often ask patients to tell me the last time they planned to have a thought or the last time they planned to feel a certain way. Humans do not plan their thoughts. Even a child looking forward to Santa isn’t planning to feel excited. The anticipation is there and the excitement just shows up on its own. Humans do not plan their feelings. Humans do not control internal experiences very well. I do think you can get better at this over time, the same way a basketball player gets better at sinking three-pointers, through practice. When you practice mindfulness, contact the present moment, and build psychological flexibility, the same processes many present-focused therapies already train, you can get better at observing the internal experiences you have no control over, accepting the difficult ones, not getting stuck in them, and engaging in values-directed committed action. That leads to a better life. People who are doing those things over time will feel better, but that’s a very different process of building a better life and later happening to feel better versus engaging in a process to control how you feel when feelings can’t really be reliably controlled.

Her self-help promise gets a lot weaker if Panksepp is right that discrete circuits generate specific affective states and concepts are just a late gloss on top. You can try and relabel your rage, but you can’t out-concept a functioning RAGE circuit. So, there’s real motivated reasoning here, not cynical, but structural and personal. Her science, her unifying theory, and her practical promise to readers all sit on the same load-bearing wall, and the wall is “concepts are causally central, not epiphenomenal.” Subcortical circuits with their own discrete, pre-conceptual organization are the one finding that, if taken at full strength, cracks all this at once.

Panksepp started in psychology, then because he realized he wanted real answers, made the jump into bench neuroscience, a move Barrett never made. Beyond lumping Panksepp in with other observational psychologists, this is how she deals with the subcortical systems:

“It is doubtful, however, that these proposed emotion circuits exist in any animal brain. You do have survival circuits for behaviors like the famous 'four F's' (fighting, fleeing, feeding, and mating); they're controlled by body-budgeting regions in your interoceptive network, and they cause bodily changes that you experience as affect, but they are not dedicated to emotion. For emotion, you also need emotion concepts for categorization.”

Barrett’s actual move with Panksepp’s work is a mix of definitional and empirical. She doesn't grant Panksepp's circuits exist at all, she disputes that directly, and only concedes a generic, non-discrete kind of survival circuitry that isn't even what Panksepp is describing in the first place. She just declares the circuits as “not dedicated to emotion” by stipulation. Boring. That’s a huge cop-out. A serious scientist with a theory under real pressure from data does one of two things: makes a sharper prediction the rival theory doesn’t make, or updates. Redefining the disputed term so the data no longer counts against you is neither. That’s not a scientist following evidence to an uncomfortable place. That’s a psychometric psychologist protecting her premise everything else depends on.

To Be Continued…

This will be my first article that will be split up. I know I made a lot of promises about where this is going, what’s to come, and I asked yall to hold your horses. Let’s take stock of where Part One actually leaves us before we get to Part Two. We looked at the tools twentieth century psychiatry and psychology had to work with, and why cortical, observational, easy-to-study evidence ended up doing most of the talking. We looked at how weak the actual origins of Barrett’s treatment of affect are, adjective checklists and college students standing in for anatomy and physiology. And we looked at why she’s motivated to keep that definition as thin as it is, political, structural, and personal, all three at once.

Next time we will pick this up by looking at the lines of evidence Barrett uses to support emotional construction. The story started with some of constructionism’s weakest spots; the constructionist camp does have some stronger stuff waiting in the wings. We will stack that all up side by side with the lines of evidence that support subcortical primacy of affect. Eventually this will conclude with a shotgun wedding of sorts. Hopefully after all this you will agree with me in that any serious person who follows the evidence will have no problem watching these two camps say their vows. I guess in this analogy the emotional constructionists are the groom at the barrel of a shotgun, as they don’t seem to want the wedding. Pankseppians fit the jilted bride well enough, left standing alone at the altar for decades. Well, you can build your empire and spend your career only looking at a certain silo of evidence, but this marriage will happen eventually because no matter how hard you try, you can’t outrun the science.

If you have a patient who murders their children you will get scrutiny. Scrutiny is not inherently good or bad. Some people avoid scrutiny as best they can. Some people enjoy scrutiny and clearly seek it out. Some people even pretend to hate scrutiny but crave it anyway. In psychiatry land it is pretty much impossible right now to avoid discussion about Lindsay Clancy at the proverbial water cooler. Like anything that gets national attention, nuance and complexity get lost. People split into camps, take sides, and form strong opinions. Everyone who provided Lindsay Clancy care is getting an enormous amount of scrutiny at the national level. There are important things to learn from the broad strokes of this case that don’t require specific details, violating the Goldwater rule, or weighing in on an outcome that is in the hands of the legal system. There are takeaways if one wants to be better prepared for scrutiny.

I don’t need to pontificate on a diagnosis just to say that based on the outcome we are all aware of, everyone involved in her care was likely wrong in one way or another. Being wrong does not mean being negligent. In general, physicians are allowed to be wrong. We are just not allowed to be negligent. An important part of not being negligent is that your assessment and plan match each other. If you diagnose the patient with depression and the plan is to start 5 mg of Haldol, you already have some problems. This might seem obvious, but psychiatry has lots of games in regard to diagnosis. As I hit on in Shutter Island Syndrome, there are many forces and incentives that push and pull towards preferred diagnosis. In some cases, specific diagnoses are needed to access certain services or levels of care regardless of their veracity. So, in this regard as a consequence of how the current nosological system is structured, psychiatry deals with a fair amount of gamified diagnosing. Be mindful of that and always be cognizant of whether the diagnosis and plan match.

I grew up as a clinician in the era of private equity, corporatism, and physicians becoming employees. I think I have a pretty good view of how incentive structures and systemic problems interact with a very reliable but not very valid system. Very short appointment times do not leave time to formulate patients properly or address complex patients. Corporatization, private equity, for-profit care, mid-levels, all the things that have changed what it means to be a physician over the last several decades, are driving some part of this. There are serious systemic problems in healthcare, but that won’t save you, so the next important thing to protect oneself from appearing negligent is documenting well. Demonstrate that you considered the differential diagnosis and what your thinking process was for how you arrived at the conclusion that you did. Below are three examples.

Example 1

Assessment: Discussed MDD versus bipolar versus personality with the patient. Patient has failed multiple antidepressants, but no history of psychosis. No family history of bipolar/psychosis. Reassuring mental status exam today and by history no concern for mania/psychosis now. Considered Lithium vs starting Amitriptyline. Discussed the risks and benefits of both options with the patient. Patient prefers trial of Amitriptyline. Discussed precautions should symptoms worsen including for emergent mania/psychosis. She has no suicidal or homicidal ideations or plans. She voices understanding of how to access 988 crisis line or emergency services if needed.

Diagnosis: unspecified mood disorder.

Plan: start amitriptyline 25 mg by mouth at bedtime.

Example 2

Assessment: Patient has failed multiple antidepressants. Still reporting symptoms. Discussed the risks and benefits of starting a TCA that can help with sleep.

Diagnosis: MDD

Plan: start amitriptyline 25 mg by mouth at bedtime.

Example 3

Assessment: MDD

Plan: start amitriptyline 25 mg by mouth at bedtime.

Subscribe now

If you are a lawyer and you have a client who just murdered her children and you are reviewing the documentation, which example do you think will best support a defense that your client received poor care? If you are a lawyer involved in a malpractice case and you have a shot at suing, which example makes you think you have a case? If you are taking the stand in a murder trial and you are being cross-examined, which example allows you to be confident in the care you provided based only on your documentation?

When you demonstrate that you considered the risks, benefits, and alternatives, this forces you to consider a differential diagnosis and put some of the differential diagnosis on paper.

To any trainees out there, major depressive disorder is not always a valid diagnosis. There are valid cases of depression, but as it is currently in the DSM, it is an extraordinarily heterogeneous diagnosis, to the point that it is almost meaningless. It is so overused, including in cases of personality, temperament, and life stressors, that I think it communicates next to nothing about the underlying condition. Do not be scared to use an unspecified diagnosis. If you are not confident that the person has an episodic mood disorder, and if you are not confident that they are currently in an episode of that episodic mood disorder, then do not diagnose major depressive disorder. Many patients use the language of depression to describe all points of time in their life, that does not inherently mean they were in an episode of depression. Yes, as psychiatrists we often have our work cut out for us. This is sometimes really difficult to do in a 20-minute appointment.

This will be fun, let’s go on a digression about treatment resistant depression (TRD). There are some patients who have an episodic mood disorder who do not respond to medications and who might actually be conceptualized as having TRD, but I am very wary of using this diagnosis. If someone has a major depressive disorder diagnosis and they are not responding to multiple treatments for depression, I reconsider the diagnosis before I just casually say yep, they have treatment resistant depression. I am reformulating the patient. I am going back and ensuring I did not miss childhood adversity and trauma. I am going back and trying to get collateral information and reevaluate their level of functioning and adult milestones. I am seriously reconsidering the mental status exam and considering bipolar and bipolar spectrum problems and other conditions. Think about it, with TRD we are saying to the patient the medications I am prescribing are not helping you, you are resistant, here are more very similar but different medications. Sometimes that is indeed the right answer and that is what is going on, but not always, and you should always demonstrate that you considered alternatives.

Am I missing bipolar/psychosis?

Am I missing personality factors complicating treatment?

Am I missing drug use?

Am I missing trauma?

Am I missing that the patient is lying and just enjoys psychiatric care (exceedingly rare)?

Am I missing nothing and the patient needs further pharmacologic management for depression?

I would document that I considered other things before prescribing someone their fifth antidepressant in such a short period of time. Again, this is sometimes really difficult to do in a 20-minute appointment.

My big project and the reason I started writing at all, is that I believe affective neuroscience provides a mechanism for personality pathology and that this greatly affects how we should consider personality and therefore how we formulate patients. I always formulate the patient’s personality prior to adding any diagnosis on top of it. So, for me personally, I have a very high threshold to consider someone TRD. I’m much more worried about why a patient is still struggling or not improving with treatment. I’m worried that I as the psychiatrist am missing something.

If you do not formulate patients properly, if you engage in low effort prescribing, or good effort prescribing but you do not document very well, and there is a very bad outcome, you are going to have a bad time under scrutiny. That does not mean you were actually wrong, and it does not mean you were actually negligent, but you will face scrutiny. Therefore, if you think a patient has a personality problem and you use depression as the diagnosis and have a crazy outcome leading to scrutiny, you will have a more difficult time explaining yourself. It’s pretty easy to jot down a line about personality impacting care and suggest the patient engage in psychotherapy. Many patients do not want to go and will never go to therapy. You can write that down as well. Our job as psychiatrists is to get patients the correct diagnosis and the best treatment plan for their problem. If you prescribe based on what the patient wants or doesn’t want when it is not really what is needed for their problem that’s usually not great.

You can also check the PMP. You can see if someone is getting multiple controlled medications from multiple different physicians or nurse practitioners. You are allowed to ask the patient about this! You are allowed to discuss your uncertainty. I can literally ask the patient about what is going on and why they have seen so many different people for their problem. This is often an indication the patient is in distress or does not tolerate distress well. It can be a red flag for substance abuse. It can be a red flag for doctor shopping. It can be a red flag for cluster b traits and splitting or disagreeing frequently and having an all or nothing response to conflict.

Physicians usually want to relieve patients’ distress. We want nothing more than to help patients! Sometimes relieving distress is not the right thing to do. Sometimes when there seems to be pressure to relieve distress that right thing to do is slow down or even do nothing! If a fentanyl addict comes to me and begs for morphine, I do not prescribe that even though it would provide relief. Once more, this is a lot to handle in a 20-minute appointment. The incentives all push to just prescribing something and moving on to the next patient.

Hypothetically if someone has been diagnosed with depression and insomnia, has been given multiple medications, is not getting better, and you are just the last person in a long line of prescribers to add a different antidepressant or sleep medication, and that patient goes and kills their children. You will get a lot of negative attention. Again, that does not mean you did not do your job. No one knows what is happening in the room. You could be the most kind, compassionate person, have the strongest therapeutic alliance, and be working harder than any other psychiatrist on the planet, but if that happens, you get the scrutiny.

I personally feel no obligation to prescribe to someone if I am uncertain about the problem. When I have uncertainty I will tell patients straight up, I am uncertain about your problem, the things you are telling me do not add up, here are the issues I am having, here is what I see on your mental status exam, here is what I see about your life versus what you are telling me, I am confused about these discrepancies. Why are you bouncing around seeing different nurse practitioners? What are you looking for exactly? Or better question, what am I missing that explains these discrepancies? Again, if your patient kills their children and you prescribed them their fifth antidepressant when antidepressants were not working, you probably want documentation demonstrating you considered many things and why you chose what you did.

I had a new patient recently whose chief complaint was that he needed to get his depression medications right. He also said medications had never really worked for him and listed fifteen prior antidepressant trials across several decades. He was adamant none of them had helped. Digging into why he wanted back on one now, he said he’d just always taken something for his depression and anxiety but had been off antidepressants for several months because a nurse practitioner told him they lower the seizure threshold, and he’d been having too many seizures, “adult absence seizures” he called it, and had been told to stop going to the hospital for them. He said he recently started having more violent shaking seizures. His chart also was concerning for a different functional neurologic disorder diagnosis in addition to his non-epileptic seizures. The neurologist’s own note said the patient reported a past history of a condition, had no current evidence or symptoms for the condition, and needed no treatment for the condition.

So here is a patient with a difficult life history, a completely normal mental status exam, red flags for somatic symptoms and functional neurologic disorders, whose nurse practitioner was uncomfortable prescribing to him, demanding I get his medications right, while telling me medications have never once helped him. I asked him point blank. If they don’t work, why do you want one? Why now? Do you see the contradiction in saying medications don’t work while asking for one?

Patients cannot fix a problem they do not know they have. He said no one had ever discussed functional neurologic disorder with him (I doubt that is true). I explained what it is, how it’s treated, and that medications weren’t part of today’s plan if he wanted to establish care with me. I asked if he’d ever engaged in psychotherapy. He said no, and that it sounded like a waste of time.

I’m not a fan of rating scales but new patients fill them out as part of the intake paperwork. He endorsed every symptom of depression on a PHQ-9. I could have taken that PHQ-9, called his problem depression, written a script for a random SSRI, and been done on time. My assessment and plan would have matched. I’d have had the data to prove it. I just don’t think that would help anyone. His problem is he has functional seizures and is having life stressors piling up as consequences from his functional seizures: big EMS and ED medical bills, exhaustion of friends and family. It is a vicious cycle where the added stress is causing more functional seizures.

I explained to the patient why I did not think his problem was depression and why it would not be fixed by an antidepressant. All of this to a patient who disagreed with me the whole time, disagreed about the characterization of his functional neurological problems, who has no reason to trust a plan he didn’t want, and who will probably just go find someone else willing to write the script he wants. I also had to spend an extra forty minutes documenting. Laying out on paper why despite the patient using the language of depression, I did not think that was the problem. I have a life. I have things I’d rather do with forty minutes than spend time documenting. I did it anyway.

This speaks to the next thing: the path of least resistance. Our current system basically forces people to take the path of least resistance. If I spent that much time with every single patient, the system would not work, it is not possible. This forces us not to deal with complex problems. This forces us to neatly slot patients into categories and move on. We have all forgotten that these categories are not inherently valid. When psychiatrists become prescribers only, when the only thing we can offer patients is medication, that is all we do. We do not consider that we do not have to offer medication. We do not consider that we can stop and reformulate the patient. These really are systems problems. These are deep, deep system and structural problems. Document accordingly, here's what that looks like.

If the problem is personality, label that, and discuss the discussion you had with the patient in your note. Talk about how you are working to build insight and work on that problem. Do not prescribe medication that you do not believe is going to help. If the problem is personality and you think medication will help, describe why you are using it off label. This patient gets very, very angry and has difficulty controlling her anger, we discussed mechanisms of action of serotonin reuptake medications, including emotional blunting. We reviewed the potential benefits of a trial of this medication to see if it is helpful with this patient’s anger and irritability in her daily life while we continue to pursue psychotherapy. If the problem is personality, do not write major depressive disorder and engage in gaslighting prescribing for depression.

If the problem is bipolar disorder or psychosis and you recommend lithium and olanzapine and the patient refuses, document that, and do not prescribe them an antidepressant.

If the problem is uncertain, document that the problem is uncertain. Document why you are uncertain. Document that you discussed with the patient why it is uncertain. Document the things you need to get done to resolve that uncertainty so you can treat the problem. Surely that will look better than diagnosing depression and prescribing an antidepressant when you are uncertain of the problem.

I think these are practical takeaways on documenting, following the standard of care, not being negligent, and making yourself a hard target for litigation. Again, if a lawyer reads your notes and just hangs his head knowing he has no shot at you: this guy could be wrong as hell, could be the biggest tool on the planet, but fuck, he was not negligent, look at all the shit he considered.

If you engage in low effort, do nothing prescribing, if you do not formulate your patients properly, if you let the weight of a shitty system push you down the path of least resistance, if you assume that someone who says they’re depressed has treatment resistant depression, then you will reap what you sow.

As with all patient stories I write about, details have been altered or composited to protect confidentiality.

Subscribe now

When I was a resident, like anyone new learning psychiatry, I still had so many questions. Way too many things felt off with how we use major depressive disorder to talk about patients with depression, so when I watched a psychofarm video talking about understanding depression with affective neuroscience, I was captivated. I’m excited to get to expand on and share this story. I hope you find it as compelling as I did.

Oh, of course, I do mean Pankseppian affective neuroscience specifically, you know, the decades of wet lab bench work mapping the primary-process affective systems of the mammalian brain using lesion studies, deep brain stimulation, pharmacological manipulation, and direct neurochemical measurement across multiple species. Much of what is called affective neuroscience today is really just cognitive neuroscience, built around language, self-report, and fMRI studies, the same tired corticocentric constructionist narratives being applied to emotions while using the term affective. An invective for another time.

John Bowlby was born in 1907 and he was writing about attachment while arguments were still being had on how neurons even communicated with each other. He posited that children come into the world with a biological drive to form an emotional bond with a caregiver, that this bond is a survival mechanism, and that a child forms one primary attachment more important than any other relationship. This is of course the mother. He described a critical window in the first two and a half years of life during which failure to form that bond, or continual disruption of it, could produce lasting cognitive, social, and emotional damage. Out of this attachment, he argued, a child builds the internal working model that will guide every relationship that follows. Voila, attachment theory!

Attachment theory holds up remarkably well with affective neuroscience. All mammals are born with subcortical primary process systems, and one of them is PANIC/GRIEF, the system responsible for separation distress. This is especially important for mammals with very dependent infants. Mammals needed a way to motivate an infant to reestablish contact with its caregiver if separated, so evolution built a system that makes separation feel bad. A rat pup separated from the nest produces distress vocalizations. These vocalizations can be measured and they have a specific frequency. These distress vocalizations can be reproduced by either electrical stimulation of specific areas in periaqueductal grey (PAG) or environmental manipulation. Either way, it just feels bad, and that dysphoric feeling motivates behavior to get attention to hopefully get back to the primary caregiver.

Witness the origin point of emotional pain and dysphoria in humans.

Bowlby also went on to describe a protest-despair sequence in infants separated from their caregiver. First there is protest: the infant screams, clings, tries to prevent the parent from leaving. If the attachment figure doesn’t return, the child moves into despair. The protest stops. The child looks calm, still upset, but withdrawn, often refusing comfort. A third stage follows, which Bowlby called detachment and others called denial, where the child stops protesting the separation altogether and can act indifferent to caregivers.

Separation distress activates PANIC/GRIEF, producing distress calls aimed at restoring contact. However, continued distress signaling stops being adaptive if the attachment isn’t restored within some window. Think about it, drawing attention to yourself while helpless and unprotected is, a good way to get eaten and not pass on your sweet sweet genetic code. So, if PANIC/GRIEF activation runs too long without resolution, it terminates into a conserved shutdown, collapse of the SEEKING system. That collapse produces loss of motivation and anhedonia. This shutdown is adaptive. Curl up. Conserve energy. Stop signaling. Wait for rescue. Avoid danger. This bears out in the animal research as a coherent strategy for a helpless infant, and the same circuitry appears capable of activating in adult humans under analogous conditions of loss and sustained PANIC/GRIEF arousal.

Humans complicate the picture because we do not only attach to caregivers, we can attach to abstract and symbolic objects. Pokémon cards, a job, an identity, a role we’ve built a life around, all of these can function as attachment objects in the affective sense, even without a literal caregiver anywhere in the picture.

If you're still reading, you're probably thinking, what the hell is the SEEKING system? Fair question. Panksepp often referred to SEEKING as a goad without a goal. SEEKING is the original primary process system, generating the feeling of enthusiasm that lies underneath curiosity, planning, foraging, ambition, and the simple willingness to get up and find out what's out there, before any object of that urge has been specified. Without SEEKING, nothing else has a reason to happen. The evolutionary logic here is essential, animals needed a way to be motivated to engage with their environment, before there was any specific thing worth engaging with.

The relevant circuitry has a name and an experimental history. So, bear with me while we digress into the anatomy and physiology of SEEKING (for anyone who wants even more technical details here is a review paper on the SEEKING system)

I tend to agree with Panksepp that SEEKING is the oldest system. SEEKING’s antiquity isn’t about where the Ventral Tegmental Area (VTA) sits in the midbrain, rather think phylogenetic distribution. How far back across the tree of life do we find something homologous to appetitive approach behavior driven by dopamine and dopamine-like signaling? Pretty darn far back, as it turns out, the distribution is enormous. Dopamine-driven approach, or octopamine driven approach, dopamine’s close chemical cousin and the dominant signal in insects, is everywhere! Organisms using dopamine driven approach include nematodes, fruit flies, honeybees, and crayfish, all creatures with nervous systems far simpler than any mammals. RAGE and FEAR have reasonably deep phylogenetic roots too, but PANIC/GRIEF is considered a comparative latecomer, since separation distress as a system only makes evolutionary sense once you have sustained parental care to be separated from, largely a mammal and bird story. SEEKING doesn’t have any dependency. Something needs to move an organism toward what it needs before any more specific system, say attachment, aggression, or threat response has anything to work with. That’s a functional argument for priority. SEEKING is close to a precondition for the other systems to matter at all.

This is also the system Bowlby was watching collapse in his children, even though he had no name for it and no anatomy to point to. What Bowlby called protest is PANIC/GRIEF doing its job creating dysphoria. Despair is what happens once SEEKING has gone quiet underneath it.

The medial forebrain bundle runs through the lateral hypothalamus, connecting lower brainstem and midbrain structures up to the medial frontal cortex. Olds and Milner identified this pathway in 1954 after observing that rats would compulsively self-stimulate it to the point of physical exhaustion, which is how it earned the name the brain reward system. Bilateral damage to this tract leaves animals unable to care for themselves, in a state that looks, behaviorally, like profound depression, and these animals when lesioned commonly die without intensive nursing support. Direct stimulation of the tract reverses it. That gives the SEEKING-collapse model of depression a concrete anatomical substrate, not just a behavioral analogy borrowed from watching infants.

Clinically, in my opinion, anhedonia, the loss of the capacity to feel pleasure, is one of the most reliable markers I track for symptoms specific for depression. I think it’s also the symptom that really tracks with a mental status exam congruent with severe depression. Anyone who engages in any kind of clinical work is familiar with patients who carry an MDD diagnosis whose mental status exams are not technically impressive for depression. Once someone has lost their motivation and drive and that ‘can do’ enthusiasm that gives us the feeling of energy to go out and take what we need, we stop consuming rewards. Of course nothing will feel pleasurable, because we are no longer even seeking pleasure.

The dopaminergic core of the medial forebrain bundle originates in the ventral tegmental area and splits into two main projections, one running to the nucleus accumbens and one running forward into the prefrontal cortex, the circuit we still call mesolimbic and mesocortical dopamine. Again, it is often described as a reward center or pleasure circuit (it isn’t). Ventral tegmental dopamine neurons fire hardest in anticipation of a reward, not during its consumption. Once the reward arrives, firing drops off. This is the whole reason the reward system label was wrong from the start. The system generates wanting, not liking, an appetitive pull toward the world rather than the satisfaction of having reached it. You can dissociate the two pharmacologically. Block dopamine and an animal stops working for a reward it still clearly enjoys once it’s placed in front of it. Block opioid signaling in a small hedonic hotspot in the accumbens shell instead, and the animal keeps working for the reward but stops visibly enjoying it when it arrives. Two different chemistries, two different jobs, and calling this the reward system flattened a distinction that patients live out every day, the difference between still being pulled toward something and still getting anything out of it once you’re there.

That distinction is why anhedonia and SEEKING collapse are not quite the same event, even though they travel together. Loss of wanting comes first, mechanistically. Loss of enjoyment is often what’s left over once wanting has already gone quiet, because there’s nothing left pulling the patient toward the things that used to reward them. A patient who says nothing sounds good anymore is often, underneath that complaint, a patient who no longer feels pulled toward anything to find out.

The lateral hypothalamus sits in the middle of this circuit both anatomically and functionally, which is part of why Olds and Milner found what they found where they found it. More than just a relay, it receives direct input from the medial hypothalamic nuclei that track hunger, thirst, and temperature, and it translates those homeostatic imbalances into general appetitive arousal. This is also where the PANIC/GRIEF connection stops being a metaphor and starts being wiring, as primary process systems are densely interconnected. The same general region that feeds homeostatic drive into SEEKING sits adjacent to, and interacts with, the PAG, the structure most implicated in separation distress and the aversive end of primary affect.

Direct activation of the SEEKING system’s dopaminergic pathway, whether by amphetamine microinjected into the nucleus accumbens, a terminal field of the medial forebrain bundle, or by electrical stimulation along the bundle itself, reliably produces 50-kHz positive-affect vocalizations in rats, which is part of why those calls are trusted as a readout of SEEKING system activity in the first place. Repeated activation of that adjacent aversive PAG territory produces the mirror-image result: a measurable and durable drop in those same positive-affect vocalizations, a suppression that persists for weeks after the aversive stimulation has stopped. Turn SEEKING on directly and the calls reappear. Stimulate PANIC/GRIEF territory and the vocalizations that would normally appear on their own stop for weeks on tests that have nothing to do with the original stimulation. That is about as close as bench neuroscience gets to demonstrating, rather than just implying, that sustained separation distress leaves a lasting thumbprint on the SEEKING system specifically, not just on mood in some vaguer sense.

The clearest confirmation of this dual-system architecture in humans, not rats, comes from Volker Coenen and Thomas Schlaepfer, a neurosurgeon and a psychiatrist who ran a proof of concept deep brain stimulation trial for treatment-resistant depression. They reported the first human data testing the PANIC/GRIEF-SEEKING model directly, stimulation of the medial forebrain bundle itself in seven patients with treatment-resistant depression. What they found is exactly what the two-system model predicts and exactly what the wanting-versus-liking distinction predicts. Stimulating the bundle did not produce euphoria or anything resembling mania. It produced exploratory behavior! Patients visually scanned the room. They reported sudden interest in travel or activities they hadn't wanted to do in years. That is SEEKING coming back online in its purest form, appetitive pull with no consummatory payoff yet attached.

One might wonder just why in the world DBS for the MFB has never been pursued further. The seven-patient case series wasn't actually the end of that story. Coenen and Schlaepfer's own group in Bonn followed it with a larger, sham-controlled trial, sixteen patients this time, randomized to real or sham stimulation for the first two months before everyone moved to open-label treatment. After moving to open label every patient met response criterion. Half were in remission at one year. A separate team at UT Houston set out specifically to replicate the finding at an independent site, and did, in a smaller cohort that added a placebo and insertion-effect control the original Bonn design never had. The Houston group published longer term follow up data in 2022. Four of five patients who achieved the 5-year time point analysis continued to be responders.

The broader field of DBS for depression had already failed hard once before any of this happened. BROADEN, a large multisite sham-controlled trial targeting a different structure, the subgenual cingulate, was halted by the FDA in 2013 after failing a futility analysis partway through, meaning the interim data made it clear the trial was very unlikely to hit its primary endpoint even if it ran to completion.

Abbott, which acquired St. Jude Medical, the device maker behind BROADEN, in 2017, picked the field back up in 2022 with an FDA breakthrough device designation. That work is now running at Mount Sinai as the TRANSCEND study, roughly a hundred patients, randomized to immediate versus delayed stimulation, with a full year before anyone gets unblinded. The target is the subgenual cingulate again, the same structure BROADEN already failed on, not the medial forebrain bundle with two positive cohorts and an independent replication behind it. Mayberg's group had published long-term follow-up on the original BROADEN patients showing many did well on sustained stimulation, which is probably why Abbott built a full year into the blind this time instead of the shorter window BROADEN used.

That's its own cautionary tale, DBS at the subgenual cingulate may work on a slower clock than the standard antidepressant trial is built to measure, and industry has apparently taken that lesson seriously. What it hasn't done is point that appetite at the target with the stronger early data. None of this means the MFB findings are wrong. It means psychiatry had the money and the appetite for one more DBS depression trial, and in 2022 spent both again on the structure that already failed once.

The reason isn’t hard to find. There are very few players with the money, the device platform, and the existing FDA relationship to run trials like this. Abbott chose re-running the target that already had a decade of surgical protocol, device programming, and regulatory history, rather than opening a fresh IDE process for a target that would mean starting almost all of that from zero. Once a company has that much sunk cost in one target, switching targets is a much bigger ask than continuing to refine the one it already knows.

There's a field-level logjam behind that corporate one, too. Subgenual cingulate, MFB, ventral capsule/ventral striatum, nucleus accumbens, and habenula all have small positive pilot data behind them at this point, and it’s not like anybody can just run a head-to-head trial to say which one is actually best. A funder backing one more expensive, invasive, sham-controlled trial has to place a bet on a single target with no comparative data to justify the choice, which pushes toward the option with the most institutional history, meaning subgenual cingulate, almost by default. There's also at least one reported hemorrhage in the small existing MFB literature, the kind of signal that makes both a company and a regulator want more safety data before committing to something larger.

The patients with MFB DBS who lit up with exploratory behavior are real data. Getting from that data to a trial that could change practice still requires exactly the kind of larger, expensive, sham-controlled trial nobody has pointed at this particular target.

Another connective tissue between the two systems, the reason grief and lost drive show up in the same patient instead of as two unrelated complaints, is very likely opioid. Attachment and proximity to an attachment figure keep endogenous mu-opioid tone comfortably high. Separation drops it. That drop is registered by the nervous system in the same currency as physical pain, it is a shared receptor system doing double duty. Sustained PANIC/GRIEF arousal, sustained low opioid tone, appears to place a brake on the dopaminergic SEEKING circuitry running through that same lateral hypothalamic territory.

There is a rich history of trying to treat depression with opioids, and the results have been mostly mixed to negative. Opium was a standard treatment for melancholia through the 19th century. A German case series of 97 patients from 1889 reported that 31 of 39 melancholic patients recovered on opium, worse results showed up in mania and in melancholia with hallucinations, which is itself a strange and mostly forgotten datapoint. It’s a neat signal that a subset of melancholic patients recovered with opium and what were probably the bipolar/psychotic patients did worse (one does have to wonder if the term melancholia meant the same thing in 1889 Germany).

Open-label and small pilot trials of buprenorphine have shown striking effects, in one small inpatient trial average HAM-D scores fell drastically. Sadly, if you were rooting for buprenorphine, but comforting because it follows the expected pattern of open label to RCT trials, the best-controlled evidence runs the other way. The largest multisite RCT to date, built specifically around opioid receptor target engagement with PET and fMRI, found no significant clinical benefit over placebo. That pattern, real early signal that a rigorous trial can’t reproduce, points at one of the weakest areas in all of medicine: translational work from animal models to humans. This kind of translational work is not easy, so that weakness isn’t surprising. Human brains and human behavior carry a complexity no animal model captures, and I wouldn’t expect a single ligand acting on a single receptor to produce the behavioral effects we are hoping for. This may be part of why affective neuroscience hasn’t been taken up more seriously in psychiatry. The animal data is clean, and while I think much of the anatomical and physiological homology holds up just fine, the human clinical translation work is rarely simple.

There’s also a narrower point. None of these trials enrolled patients on anything resembling PANIC/GRIEF-driven SEEKING collapse. They enrolled on standard treatment-resistance criteria, nonresponse to an SSRI or SNRI, a MADRS or HAM-D cutoff. Opioids would only be expected to work for this specific mechanism, sustained separation distress suppressing SEEKING, not for the broad, mechanistically mixed population that meets criteria for MDD. A trial that enrolls on symptom severity alone is asking a mechanistically narrow intervention to move the average of a mechanistically heterogeneous group. That mismatch, not the pharmacology, may be a real part of why the trials keep coming back negative.

Speaking of someone who loved a good narcotic, Freud wrote Mourning and Melancholia in 1917. He described ordinary mourning as a person consciously knowing what has been lost in working through the loss over time by gradually withdrawing emotional investment from the lost object until the ego is again free to attach to something new. He wrote that mourning is painful but ultimately self-limiting. He described melancholy as looking similar on the surface but that the melancholic does not consciously know what they’ve lost and that person is incapable of letting go of the lost object as it has been internalized. He commented on the harsh self-criticism melancholic patients level at themselves as being misplaced: the shadow of the object fell upon the ego. This is a foundational text for object relations thinking in general. While I ultimately have mixed feelings about psychoanalysis, I think anyone who engages seriously with it can’t help but concede many of the observations made were indeed canny.

Panksepp’s work identifying these subcortical circuits gives a neuroanatomical and physiological address that overlaps so plainly with what early psychiatrists and psychologists observed and wrote about prior to having any understanding of neuroscience. A lot of our constructs are based on these old ideas and observations, so it has just been mind blowing to me that a medical discipline like psychiatry would be so resistant to engaging fully with affective neuroscience as a discipline. The SEEKING arousal that has been driving me is the sense of enthusiasm that I know psychiatry cannot outrun the neuroscience forever. Eventually it all has to come back to the brain, and this elegant overlap of early attachment work and bench neuroscience is just a sliver of what Panksepp’s program accomplished.

I don’t think anyone who is intellectually engaged with psychiatric nosology and who also engages in clinical work can deny what a ridiculous and unwieldy category major depressive disorder has become. I think this is borne out very clearly by the number of patients who get given a diagnosis of major depressive disorder when the mental status exam does not support anything pathological. We are all trying to do our best for patients, but clearly as a category and for how we conduct research, it is just frankly too broad to take seriously. Every study I read about depression I have to wonder what does this even mean? What kind of patients are these?

We already carve out seasonal depression, depression with anxious distress, atypical depression, and several more. I think prolonged PANIC/GRIEF activation leading to SEEKING collapse deserves the same kind of recognition, a cluster with a mental status exam that would match: psychomotor retardation and a melancholic affect that is actually reflective of the anhedonia and loss of motivation. I’m not calling for a new box in the DSM. I’m not a fan of categorical diagnosis at all, and I’m not the first psychiatrist to struggle with this and have big ideas about nosology that went nowhere. RDoC and HiTOP exist precisely because the categorical model keeps failing to carve nature at its joints. When I say PANIC/GRIEF-SEEKING collapse, I mean it the way RDoC means a domain, a mechanism-based axis a patient sits somewhere along, not a checklist a patient either meets or doesn’t. The carveout I want isn’t a new line in a manual. It’s clinicians recognizing this specific mechanism when they see it.

We all know that patients with depression are not all the same. Part of the resistance to using the mental status exam this way, I think, is fear of invalidating the patient’s experience or fear of withholding the best treatment. But the mental status exam is telling you something, specifically the highest level of functioning a patient can currently sustain. Someone who can still laugh, still smile, still remain functional and engaged with the world while feeling sad or depressed is a different patient than the one whose mental status exam shows psychomotor retardation and a melancholic minimally reactive affect. I suspect the deeper resistance might be a fear of creating a ‘real’ depression versus ‘fake’ depression divide in the public eye, so there’s no real appetite to change much about MDD. All patients deserve engagement and treatment for whatever they present with. That should not stop us from following the neuroscience.

Getting fired, watching a marriage end, losing someone important, all of these can trigger PANIC/GRIEF, and prolonged activation can terminate the SEEKING system. None of these erases how complicated real cases get. Rumination and hopelessness carry their own psychological weight. Someone with chronic low-grade inflammation from carrying extra weight and smoking is bringing a physiological variable into the mix that has nothing to do with attachment. There are likely thousands of variables, both biological and psychological, that influence depression. I’m not pretending this explanation is clean. But I do think this model, with a history running from attachment theory through bench neuroscience, and human trials explains a real type of depression, even if it doesn’t yet have a recognized name in the DSM.

Something different this week in that there are not a ton of tie-ins to affective neuroscience. Fair warning, this one gets into the weeds on receptor pharmacology. I rolled my eyes on the day the approval was announced, but what pushed me to write this was an article in Psychiatric News, the APA’s own newspaper, titled Auvelity expands treatment options for agitation in Alzheimer’s. My favorite line: “this will help clinicians, families, and caregivers who are struggling to take care of individuals with agitation”. Frankly, reading that I felt a bit of RAGE arousal.

Subscribe now

One of the hats that I wear is as the medical director of a small rural inpatient geriatric psychiatric hospital. I treat severely demented patients with significant neuropsychiatric symptoms, agitation included, as a matter of daily work. This is a brutally difficult population to prescribe for. Discerning effect from side effect is mostly observational, built on close coordination with the techs and nurses who are with these patients around the clock, because the patients themselves often cannot tell you what is happening to them. At baseline this population has extreme difficulty communicating their needs. Exposing a demented patient to more medication rather than less should be the last resort, not the first ask. So, let’s dive into this combo pill Auvelity and look much closer at dextromethorphan specifically.

Auvelity is a combination of bupropion and dextromethorphan. It was approved by the FDA for major depressive disorder in August 2022. Not a single patient 65 or older was included in the trials or pharmacokinetic studies that supported that approval. In April 2026 it was approved for agitation in Alzheimer’s dementia, an indication that by definition targets a population the drug was never studied in to begin with. Fortunately, I am not the first person to notice how weak the latter approval is.

The ADVANCE-1 trial (n=308) was a five week RCT with change from baseline on the Cohen-Mansfield Agitation Inventory (CMAI) as the primary endpoint. The result was statistically significant. Drug arm dropped 14.9 points, placebo dropped 11.6, a treatment difference of 3.3 points (95% CI -5.8 to -0.8). A three point difference on a scale that runs to over two hundred points is not clinically meaningful by any reasonable standard. The more telling finding is ADVANCE-2, the larger five week parallel group replication trial (n=408). It missed its primary CMAI endpoint entirely, 13.8 versus 12.6. This was the trial that was supposed to confirm ADVANCE-1. It failed.

A recent and good piece by Kevin Kennedy at Polypharmacy goes further into both studies and expands on an additional flawed withdrawal design trial. He is also appropriately critical of the FDA and their lack of transparency with this approval.

In summary, one five week trial hit a three point CMAI difference. The larger replication trial did not. Two of the four authors are Axsome Therapeutics employees.

This approval tracks closely with the brexpiprazole approval for the same indication. The CMAI is low hanging fruit for drug companies. It is such a massive scale with so many points that squeezing out statistical significance is not hard. I do not think there is anyone who genuinely believes brexpiprazole produces a different result than risperidone or any other antipsychotic would in the same population vs. placebo. I do not think there is anyone who believes Auvelity produces a more meaningful response than sertraline or any other SRI would in a similar study vs. placebo. An FDA approval is only that: someone ran a study and applied. The incentive to run that study only exists when a company has something patented to sell at a markup, without competition.

But really to understand the Auvelity story we have to look very closely at dextromethorphan (DXM). I enlisted in the Army after high school. Enlisted soldiers are young and prone to degenerate debauchery. Soldiers are also regularly drug tested, which makes any population more likely to reach for substances that do not show up on a standard panel. I am not saying DXM abuse was rampant, but it was a known barracks drug, precisely because it was available and carried no consequence for a Monday morning screen.

Erowid is a harm reduction forum for using drugs recreationally. Figures for oral DXM lay out what it actually takes to reach dissociative NMDA territory recreationally. A light dose starts at 100 to 200 mg. Common recreational use runs 200 to 400 mg. Strong effects start at 300 mg and heavy dissociation runs 600 to 1500 mg. Auvelity’s total daily dose, 90 mg, sits below even the light threshold. That comparison is misleading. It describes DXM taken alone, and Auvelity is never DXM alone. Bupropion’s CYP2D6 inhibition changes the picture. We will get to that.

Now let’s look at Ki. Ki is the binding coefficient, the strength with which a ligand holds a given receptor. Lower Ki means stronger binding. DXM’s tightest affinity by a wide margin is SERT, 23 to 40 nM. The NMDA receptor comes in a distant sixteenth on the binding table, at 2,120 to 8,945 nM, roughly two orders of magnitude weaker than the SERT number. Fifteen other targets sit ahead of it, all binding DXM more readily than NMDAR does. Binding is simultaneous and probabilistic, governed by mass action at each receptor independently. At any given moment, some fraction of DXM molecules are bound at SERT, some fraction at NMDAR, some fraction free, all at once, in proportions set by the concentration relative to each receptor’s own Ki. The table below, from the dextromethorphan Wikipedia article, lists Ki values drawn from rat data. Species differences in receptor pharmacology exist, but ionotropic glutamate receptors like NMDAR are among the more conserved targets across mammals, so these numbers are considered unlikely to shift dramatically between rat and human.

Before getting into how Auvelity’s dextromethorphan concentration was calculated, it helps to know what that number is actually for. Every receptor’s Ki is measured in nanomolar, a unit of concentration. To know how much of a given receptor gets occupied by a drug, you need the drug’s own concentration in that same unit, specifically the free, unbound concentration reaching the brain, not the total dose swallowed and not the total amount floating around in the blood. That free concentration is the figure that gets compared directly against Ki to predict occupancy. Getting there from a lab report takes a few steps, and none of them are complicated.

Auvelity's own FDA submission reports a steady state Cmax, the peak blood concentration reached only after several days of consistent twice-daily dosing, not after a single dose. That figure comes out to roughly 70 to 122 ng/mL of dextromethorphan, drawn from three separate human trials. To convert that into nanomolar, divide the concentration by the drug’s molecular weight, 271.4 grams per mole for dextromethorphan. That division tells you how many molecules of drug are actually packed into that same volume of blood, which is what nanomolar measures. Run the math and 70 ng/mL comes out to roughly 260 nM, while 122 ng/mL comes out to roughly 450 nM. But that is the total amount of drug in the blood, not the amount actually free to reach a receptor. Dextromethorphan runs 60 to 70 percent bound to plasma proteins, locked up and unavailable, according to the drug’s own label. That leaves 30 to 40 percent, the unbound fraction, free to act. Apply that fraction to the total concentration range and the free, active concentration lands somewhere around 75 to 180 nM. That is the number the occupancy math below is built on.

Ki predicts occupancy, the fraction of a given receptor type that is bound at a given free drug concentration. The relationship is simple: occupancy equals concentration divided by concentration plus Ki. A receptor with a Ki far below the drug’s concentration gets nearly saturated. A receptor with a Ki far above it barely gets touched. Plug in Auvelity’s estimated free plasma concentration, 75 to 180 nM, and that is exactly what happens here. Against SERT, Ki around 30 to 40 nM, that concentration produces roughly 65 to 85 percent occupancy. SERT is close to saturated. Against NMDAR, using the most potent Ki reported that I could find, 962 nM from a 1989 rat cortex radioligand binding study, occupancy comes out to roughly 7 to 16 percent. Using the range from the above table it drops to roughly 1 to 8 percent. Even the most generous reading of a genuinely inconsistent literature caps NMDAR engagement at a modest fraction of receptors, nowhere close to the near total occupancy DXM achieves at SERT on the same dose of drug.

IC50 is a different kind of number than Ki. Where Ki comes from a binding assay, radioactively tagged drug competing for a receptor in a test tube, IC50 comes from a functional assay, an actual electrical measurement of what a receptor does when the drug is present. To get it, researchers take isolated neurons, hold them at a fixed voltage, and apply a sustained, continuous dose of NMDA to keep the channel open. Then they add dextromethorphan at various concentrations and measure how much the resulting current shrinks. The concentration where the current drops by half is the IC50. For dextromethorphan at the NMDA receptor, that figure has been reported anywhere from roughly 550 to 6,100 nM depending on the study, a more than tenfold spread across labs measuring the same drug at the same receptor. The most favorable of these studies, a 1994 patch-clamp experiment in cultured mouse neocortical neurons, found that 100 nM of dextromethorphan reduced NMDA receptor current by about 20 percent.

The more important detail in all of these studies is the experimental setup itself. Holding a channel open continuously with sustained agonist is close to a best-case scenario for a channel blocker to find its way in. It looks nothing like how NMDA receptors actually behave in a living brain, where channels open in brief bursts tied to real synaptic firing, not held open on demand. Dextromethorphan is also a fast-off blocker, meaning it moves in and out of the channel quickly rather than getting trapped inside the way a drug like MK-801 does, or ketamine does. That kinetic profile is exactly why dextromethorphan is tolerable at all, but it also means it has far less opportunity to build up any meaningful block during the brief, intermittent windows a channel is actually open during normal neuronal activity. The IC50 and the 20 percent figure both measure the drug under conditions optimized to maximize its access to the receptor. Real neuronal firing is close to the opposite condition, which means even the most favorable numbers on this receptor are likely a substantial overestimate of what dextromethorphan is doing at NMDA in an actual human brain, not an underestimate.

Stack all these numbers together and the picture narrows even further than either number suggests on its own. Put plainly: at best, under the most generous reading of the data, the free drug concentration Auvelity actually produces yields something like a 20 percent reduction in current in at most 16 percent of NMDA receptors. Given the current enthusiasm for ketamine and its NMDA effects, the Auvelity marketing leans hard on implying a novel, deliberate NMDA-based mechanism of action. What the data actually supports is something far less confident: a real but small and poorly characterized effect, capped from two independent directions, never once measured in a living human brain, sitting on top of a serotonin reuptake effect that is neither novel nor in dispute. The marketing speaks with a certainty the science has not earned.

That ceiling isn’t even the full picture, either. DXM has a metabolite, dextrorphan (DXO), produced when CYP2D6 does its normal job of breaking dextromethorphan down. DXO binds NMDA considerably more tightly than DXM itself does, 486 to 906 nM versus DXM’s 2,120 to 8,945 nM. Ordinarily, when DXM is metabolized quickly, DXO is the dominant compound in the blood, and DXO would be doing more of whatever NMDA work happens. But Auvelity is built to prevent exactly that. Bupropion’s whole job in this pill is blocking CYP2D6, which suppresses the conversion to DXO and shifts the ratio back toward DXM, the weaker of the two binders. So the drug ends up keeping more of the weaker NMDA compound in circulation and less of the stronger one, even as it elevates total DXM exposure by blocking its main clearance route. What that nets out to for overall NMDA engagement isn’t something the ratio alone settles. Whatever the actual intent behind that tradeoff, the practical effect is keeping DXM levels elevated and consistent regardless of a patient’s CYP2D6 status, the same design logic Nuedexta uses with quinidine.

So to wrap this up clinically, the proposal is to take geriatric patients, whom the original studies excluded, who are already usually on multiple medications, and add an expensive combination medication that includes an inhibitor, bupropion, which itself can be activating, can worsen insomnia, and can lower seizure threshold, all for a clinically meaningless shift on a bloated rating scale even after the larger replication trial completely failed. And I’m supposed to be excited about this opportunity to help patients and families?

To be direct about it: the patient here is not complaining of agitation. Families and caregivers notice it, and it is a symptom that contributes to caregiver burden. Addressing that with an expensive combination medication, because a company angled its way to an FDA approval, is insulting. It is equally insulting to watch what amounts to drug company talking points get repeated by supposed thought leaders inside the APA’s own publication.

The reps are out in force. Their target is anyone without a strong grip on the underlying literature. There is no secret sauce here. Whatever fraction of NMDA receptors this drug is actually engaging, the clinical trial data gives no reason to think it matters. A genuinely novel, clinically meaningful mechanism of action should show up in outcomes. ADVANCE-1 barely cleared statistical significance, and ADVANCE-2 failed outright. Whether a three point difference on the CMAI is clinically meaningful is, in my opinion, worth debating in the first place. Either way, if the NMDA story were doing real work, that is not what the data would look like. Whatever modest effect shows up in these trials is most parsimoniously explained by SERT binding. If you treat dementia patients with agitation and your first reach is Auvelity, you are not demonstrating clinical sophistication. You are demonstrating that the marketing worked.

Sertraline is a better default for an undifferentiated dementia patient with agitation and difficulty expressing their needs. No meaningful drug-drug interactions, once daily morning dosing, a single medication rather than an unnecessary additional one layered onto a frail patient. Or, if a low dose antipsychotic is warranted, risperidone does the job at a fraction of the cost. Either option is cheap and off patent, which is precisely why neither one gets a Psychiatric News writeup. It sure seems like the physicians who work at the FDA are pretty far removed from treating patients and have no real grasp of how to evaluate whether something is clinically meaningful. Sadly, now I have to wonder about the clinicians at the APA who are so ready to give lip service to an expensive marketing ploy.

Subscribe now

Now that I’ve started writing openly about affective neuroscience, people have started asking me directly whether I actually formulate patients this way. Yes. I do. And I still practice evidence-based medicine and follow the standard of care while I do it.

A resident posted on Reddit recently, asking how to tolerate patients with borderline personality disorder (BPD). They described being hurt more easily by borderline behavior than most of their peers, having grown up with a mother and a sister who both have BPD, and having a much lower frustration tolerance for it as a result. They described the behavior as deeply triggering for them. They said this wasn’t normal for them at all, that they’re usually calm and collected. Despite reading books, training in dialectical behavioral therapy (DBT), and studying psychoanalytic concepts like projective identification, they still found their heart racing, felt an internal loss of control, got angry, and stayed irritated for days after certain patient encounters.

I responded with some thoughts from affective neuroscience, and they landed well. That response is worth expanding and organizing properly, so let’s go deeper into how I actually formulate personality using affective neuroscience. The single most important move is treating personality and personality functioning as the baseline against which any additional pathology gets considered.

Affective neuroscience is what gives mechanism to personality pathology. Look carefully at what the animal research established and at the converging evidence in humans, and you’re left concluding that we carry homologous brainstem and midbrain structures no different in kind from any other mammal. Affective arousal is a primary process, one that even Kernberg acknowledged as the driving force of development. Primary process means the circuit is coded innately, present from its developmental origin, and requires no learning to function. It’s capable of producing its characteristic output on its own. The valenced arousals from these systems exist to drive and motivate behavior. Secondary processes, learning, classical and operant conditioning, the whole apparatus behaviorism spent the twentieth century mapping, exist to give these primary affects something to attach to, to form emotional habits around. Then you get a cortex layered on top, capable of tertiary processes: abstraction, prediction, association, reappraisal. How well these cortical skills come online varies enormously from person to person, shaped by life and development, and everyone ends up with a different profile of strengths and weaknesses in regard to cortical capacity.

Most of mainstream personality research, meanwhile, is still built on trait factor analysis, and has rewarded us with the five-factor model. An enormous amount of effort has gone into the five-factor model over the decades, and the trait folks have refined reams of data and are confident in their findings. None of that changes the fact that it remains fairly useless for saying anything meaningful about personality pathology.

What matters most for pathology is that subcortical systems have real primacy. Affective arousal has the ability to overwhelm cognition. Emotional regulation literature is fairly consistent on this: the stronger the activation of subcortical systems, the stronger the felt state, and the weaker the activation of cortical systems. The inverse holds too, the weaker the subcortical activation, the weaker the felt state, and the stronger the cortical activation. Broadly speaking, this is what affective dysregulation is. It’s when arousal is strong enough that the tertiary process cortical systems normally used to weigh and select behavioral output lose their footing, the evolutionarily conserved subcortical circuits take over, and what’s left is a much narrower channel of behavioral output.

Personality functioning, in the sense that matters for what we broadly label psychopathology, can be thought of as what happens at that junction between subcortical affect and cortical regulation. It’s a mismatch between subcortical affect and cortical regulation, a person in a situation that demands one kind of behavior while their affect pulls them toward another, without the capacity to modulate the gap. If person A reacts to a stimulus with a 1 out of 10 annoyance and person B reacts to the same stimulus with a 10 out of 10 rage, person B just looks crazy. This is what makes someone look unhinged next to someone else who can hold the same affect and still choose their behavior. People with BPD are, in a real sense, walking around with PANIC/GRIEF and RAGE set on a hair trigger. That makes life very hard for them.

Building on this, one way to think about personality problems: they’re the observable, aberrant behaviors due to affective dysregulation that bring a patient to clinical attention in the first place, produced by that same mismatch between subcortical affect and cortical regulation. Once you’ve seen enough of these patients, it gets hard to keep defending many of our discrete categorical diagnoses. Instead of saying a patient has bulimia, or functional neurologic disorder, I think it’s more accurate to say this patient has a general affective-regulatory problem that shaped how their personality developed and how they make choices, and this happens to be the behavioral channel it’s coming out through right now.

Look at the operationalized criteria for BPD in the DSM: difficult-to-control anger, affective dysregulation, transient paranoia and dissociation, all of which I’d argue are mechanistically secondary to intense affect arousal and overwhelmed cortical regulation and resolve once the affect arousal dies down. The cutting, extreme suicidal gestures, and erratic behaviors are also usually carried out under that same affective arousal state. The affective-regulatory failure is what accounts for the high rates of emergency room visits and inpatient admissions. This is what actually accounts for the psychopathology part.

So every patient, to me, is a walking affective-regulatory system. The question is always whether the patient is currently having trouble with their behavioral output, and if so, whether that trouble represents a normal, fully functioning affective-regulatory system running into a mismatch with its environment, or whether there’s an additional, valid pathology intruding on top of the patient’s baseline functioning. Getting very angry very fast, and acting on that anger, is not inherently a pathological process. Ten thousand years ago, being the first person to act on your anger was probably advantageous. In the present environment, the same response is considered wrong and tends to carry consequences most people can’t tolerate.

This is what I think should be considered personality pathology, and I think a lot of what’s currently scattered across the DSM should be folded back into it. There is no cutting disorder. Cutting gets folded into a personality construct. Bulimia, meanwhile, gets its own separate category, and I have never encountered a patient with bulimia who didn’t also exude Cluster B personality traits. The idea that a normally functioning person can just catch or develop bulimia is pretty ridiculous.

The most extreme forms of personality pathology are rooted in affect dysregulation itself: Some patients cut. Some drink. Some use meth or opioids. Some purge their food. Some have pseudoseizures. Some just enjoy using healthcare, the somatic-factitious presentations. A great deal of what gets diagnosed as discrete categorical disorders is personality pathology wearing a different costume depending on which behavioral channel it happens to come out through. After all, there are many ways to engage in experiential avoidance.

Rather than condemning a patient with the idea that their personality itself is disordered, using a diagnostic system built on traits that are themselves downstream constructs, imagine if the criteria were built around affect and the pathological actions taken under it instead. Sure, there are walking caricatures out there who’ll report fears of abandonment and chronic feelings of emptiness, and there’s absolutely something real captured in the borderline construct itself. But a dimensional account of personality, grounded in neuroscience, seems far more in line with what psychiatry is actually trying to do. I can’t imagine that if someone started today, from scratch, with all the data and knowledge currently available, they’d build the system we actually have. That system is obviously a top-down product of its history. If you started today, from scratch, with everything we now know, it’s hard to imagine building anything except from the bottom up, starting with the actual affective systems the brain runs on.

This is why I think the affective neuroscience viewpoint is a better way to capture personality pathology than how psychiatry traditionally considers things through the DSM. It’s a better way to view extreme presentations, grounded in how the brain actually functions, and it lends itself to valid future research, for psychotherapy and for pharmacology, for any condition that’s ultimately derived from intense affect. Imagine developing medications for a valid existing brain system rather than trying to develop them for an invalid DSM construct like MDD (see A closer look at RAGE).

I get that this is something different. This framing is genuinely foreign to how psychiatrists and psychologists are trained to think about patients and personality. This is why it’s important to be precise about how to consider additional pathology in the context of personality.

So, after accounting for a person’s baseline personality functioning, the question becomes whether there’s an additional, valid pathology intruding on top of it. Even when a patient’s baseline personality functioning is itself producing pathological behavior, is there something else on top of that, a mood episode, bipolar disorder, OCD, schizophrenia? Schizophrenia is a useful contrast case, because there you have obvious pathology across multiple dopaminergic pathways, too much dopamine in the mesolimbic tract, too little in the mesocortical tract, disrupted functional network connectivity, aberrant synaptic pruning, glutamatergic signaling abnormalities, and so on. I’d conceptualize a patient with schizophrenia as having valid pathology intruding on an otherwise (hopefully) normally functioning affective-regulatory system.

Everyone has a personality, though, and in many cases there’s no additional pathology at all. The patient simply has an affective-regulatory system whose current calibration is producing dysfunction in their environment. Psychiatry’s traditions largely predate any real capability in neuroscience, which is part of why personality hasn’t historically been considered this way. From an affective neuroscience standpoint, a patient with BPD isn’t choosing volatility as a character trait. Their affective systems fire at an intensity and speed that outpaces whatever cortical capacity they managed to form and grow during development.

A brief callback to the toddler analogy I used previously. A toddler hasn’t yet built the cortical machinery to modulate a strong affect, so when RAGE or PANIC/GRIEF fires hard, there’s no regulatory system online to intervene. The behavioral repertoire collapses to a narrow set of primitive outputs. Reasoning with a toddler mid-tantrum doesn’t work, not because the child is being willfully unreasonable, but because the cortical systems that process reasoning are offline due to age. The connections simply don’t exist yet. In an adult with clinically apparent BPD, the same thing happens for a different reason: the subcortical arousal is running the show because it’s too strong relative to whatever cortical regulation is available, whether that regulation is itself underdeveloped or is simply outmatched by the intensity of the affect. The tantrum ends when the affect discharges, not when an argument lands.

This is really the crux of severe personality presentations: everything is fine for them until it’s not! They lack the capacity to handle environmental stimuli that trigger negative-valence affective arousal. They aren’t able to use the cortex to hold affect as a signal passing through consciousness, information to be weighed, rather than as an imperative that demands immediate action. So they make choices in the moment that they regret twenty minutes later, once the affect has passed and they no longer feel the same way. Society expects a certain baseline capacity to tolerate and modulate affect, as part of what it means to function as a regular adult. These patients just can’t keep up.

The affect-versus-emotion distinction matters more than psychiatry currently treats it. Affect is the raw subcortical valence and arousal state, present from birth in infants and animals, a highly conserved primary process with a neuroanatomical address. Emotion, more precisely, is that affect after it’s been shaped by cortical appraisal, memory, and narrative, layered with secondary-process learning and tertiary-process cortical abstraction. Take shame. It looks different in Japan than in England, shaped by culture and learning. Shame is a cortically represented emotional construct. Not an affect.

The DSM largely collapses this distinction. It treats reported feeling-states as the basic unit, which is part of why personality disorder categories end up sounding like character judgments instead of descriptions of a subcortical-cortical regulatory deficit. Reformulating personality disorders as a subcortical-cortical interface problem, rather than a personality problem, turns the clinical stance from managing a difficult person into treating an underdeveloped regulatory system. Proper psychotherapy is neuromodulation. The willful, mindful practice of psychological skills, and mastery of those skills, is accomplished through neurons in the cortex literally making new connections.

There’s a clinical paradox that insight-based therapies keep running into, and it comes down to how affect and cognition are wired relative to each other. Skills learned in DBT or therapy get encoded under a regulated state. But retrieval is state-dependent: access to that stored material depends heavily on the state you’re in when you try to reach it. When a strong affective arousal hits, subcortical systems take primacy, the whole neurophysiological context shifts, and skills built during a calm, cortically online state become far harder to reach, not because they’ve been forgotten, but because the retrieval conditions have changed. Executive function, working memory, deliberate skill recall, all of it goes offline under intense affective arousal, while more primitive systems take control of behavior. What you’re watching is a real-time shift in which brain systems are running the organism, not a loss of the underlying knowledge.

That’s also why insight and outcome correlate so weakly in severe presentations. This isn’t to say concepts like object relations or splitting have no value, there’s real clinical utility in them, but they describe the pattern rather than the mechanism. Someone can build a sophisticated model of their own patterns, understand their history, describe their own defenses accurately, and none of it changes what happens once affect takes over, because insight lives in a system that gets sidelined right when it matters most. The direct implication for treatment: skills need to be practiced under some degree of affective activation, not only in the calm state where they were first taught, if they’re going to survive the transition into a dysregulated moment.

Imagine going through life as a slave to strong affective arousal, bouncing from one to the next. How difficult is it to be self-motivated and self-directed when any arousal strong enough can pull you off track at any moment? It’s tragic, and it can feel more like spiritual suffering than psychiatric suffering, watching someone be this directionless, this aimless in their own life, this unable to plan their actions around their own stated values, just reacting to whatever affect comes up until it finally dies down and they can use their own brain again.

This is also why the condition is so treatable. Go to therapy. Identify the problem early. Grow the connections needed to modulate strong affect. Recognize that this isn’t a character deficit, and it isn’t a trait problem.

Which brings this back to the resident who asked the original question. Their mother and sister aren’t just difficult people. They’re attachment figures whose dysregulation this person was exposed to repeatedly during their own development, which is likely part of why their circuitry answers with more charge than a colleague’s does to the same behavior in a stranger. They’re showing a milder version of the same architecture they’re describing in their mother and sister: real cortical insight, built from books and DBT training and a working knowledge of projective identification, that holds up fine right up until the affect actually arrives, at which point it can’t keep up.

Picture the mother and sister living with subcortical systems that fire this hard as a constant condition, not an occasional lapse. They’re very likely aware of the harm, and have probably tried to do better. But when the affect is strong enough, all of that gets overridden, and what’s left looks a lot closer to a toddler mid-tantrum: no access to the memory of past damage, no access to whatever skills were once built to avoid repeating it.

It tends to be easier to hold onto empathy when you frame it this way rather than as a personality problem.

I don’t think I’ll be able to get Shutter Island syndrome into DSM-VI, maybe the next next DSM.

Subscribe now

A patient tried to assault me exactly once in my career. I was an intern. I learned enough from that experience that it has never happened again (knock on wood).

She had been transferred to our inpatient unit from an ER more than three hours away. That distance should have told me everything before I walked in the room. She had to pass by every psychiatric hospital in two different metro areas with around a million people apiece in order to make it to my hospital. Every hospital had already turned her down. They knew her. They wanted nothing to do with her.

She was well past middle age, and her chart said schizophrenia. Her mental status exam said otherwise. She spoke fluently, with no thought blocking, no poverty of speech, nothing resembling a negative symptom. Nothing indicating cognitive impairment. Her affect was reactive. Her vocabulary was large and precise. The only things anyone could point to as psychotic were the subjective auditory and visual hallucinations she described with something close to relish, and a life history full of milestones that patients with schizophrenia rarely accumulate.

I was young enough to think this was good news, and young enough to say so out loud. I told her she didn’t have schizophrenia. I told her this was probably borderline personality disorder, that it was treatable, that psychotherapy could help her build a different life once she left the hospital. I emphasized that she didn’t need to take multiple antipsychotics anymore, and a good thing too as they didn’t seem to be working very well for her hallucinations. I meant it as reassurance. I had just told her she didn’t have a chronic, progressive, degenerative brain disease. She came across the table at me with her hands curled like talons.

I have thought about that moment for years, and here is what I take from it now. If a patient has been turned away by every hospital between here and three hours in every direction, the pretest probability for personality pathology should be quite high. Far more salient, personality pathology is easiest to address when the patient is still young. As long as one does not have dementia, the capacity to change is never gone; the cortex keeps growing connections whether you are twenty or sixty, but it grows slower and harder with age, and every year spent inside a wrong diagnosis is a year not spent building what is needed and locking in what is maladaptive. I see children who started cutting at twelve, who spent six years in and out of hospitals for it, and who were told the whole time that the problem was depression. Personality was absent from the discussion and documentation. Then they turn eighteen and something strange happens. They arrive at every encounter fluent in the language of depression, demanding a medication change after every crisis, having fully absorbed a diagnosis that was never quite theirs to begin with and having remained woefully ignorant of anything to do with personality. This kind of obstacle also makes the necessary treatment harder no matter the patient’s age.

All of that is to say, the sooner personality gets addressed, the better, and that I think is best demonstrated through the story I actually want to tell. This story happened later, when I was a PGY-2 on the consult service. It was a slow day. We got called for schizophrenia, a patient disruptive on the medical floor. I don’t remember what she was admitted for medically. The nurse said she was singing and wouldn’t keep her clothes on.

The day was slow, so I ended up spending two hours with her. Nothing in her mental status exam pointed toward schizophrenia. Several times during the conversation she dropped into a baby voice, sing-song cadence, simple childlike sentence structure, requests for a teddy bear or a hug. Then she would surface and talk about her actual life in completely normal adult speech, well organized, expressive, no negative symptoms, no cognitive impairment, no abnormal movements. No barriers or impediments to communication or expression of ideas. A normal mental status exam, which really doesn’t fit a longstanding schizophrenia diagnosis.

Her history was the kind of history you find at the root of this presentation almost every time. Severe childhood trauma, sexual and physical, starting young. Self-harm not long after. She was living in a group home for patients with schizophrenia, on disability, on a medication list that included multiple antipsychotics, antidepressants, and a mood stabilizer.

What struck me most was how ordinary she sounded whenever we talked about her actual life. What she liked. What she had lost. What she still wanted. As the two hours came to a close, I told her what I had learned. I said she didn’t really have schizophrenia. I said she spoke too well, expressed herself too clearly, was too sharp for that diagnosis. I asked her why she did this.

She looked at me and said if she didn’t have schizophrenia, who was she. Just a fucking loser who failed at life. She said with schizophrenia her family cares about her and visits her and expects nothing of her. She lifted her scrubs over my objection. Cuts ran from under her chin down her neck, across her chest, down her abdomen. Hundreds of them, all horizontal, old scarring layered under new. She said her family thought she was a fucking loser and that without the diagnosis of schizophrenia she was nothing, a stupid idiot who meant nothing to anyone. Then she pulled her scrubs off entirely, dropped back into the baby voice, and walked out of the room singing, louder and louder as the nurses converged on her.

Anyone who has seen Shutter Island knows the ending. Leonardo DiCaprio’s character is offered a choice: change, or be lobotomized. He chooses the lobotomy, but not before giving his psychiatrist a look that says he knows exactly what he is doing, and prefers the fantasy to the truth. That was this patient. She had made her choice long before I met her.

Given her behavior and the weight of the chart, I couldn't imagine how to write a note explaining why schizophrenia was not accurate and why the antipsychotics weren't touching the actual pathology. A note like that would have taken hours to untangle years of prior documentation, and it would have accomplished nothing. Notes get skimmed, not read, and nobody except the social security administration would benefit from relitigating a diagnosis that had already calcified into her disability determination, her group home placement, and her sense of who she was. By the time I met her, she was already locked into a low-functioning, disabled life built on a diagnosis she never had. I made a small, meaningless medication adjustment and moved on, because there was no way to explain to a floor nurse, however skilled and however dedicated, the nuance of why a mutilated, singing, naked patient does not really have schizophrenia. I took the path of least resistance, which is what most of us take in that moment. It is also, I think, what kept me in inpatient psychiatry: the hope of catching the severe pathology early, in someone still young enough to change, before what I now call Shutter Island syndrome sets in.

Shutter Island syndrome is severe Cluster B pathology that has stopped negotiating with reality and started defending its own fantasy at any cost. It’s the horror show when a patient rides their chosen externalizing pathology to the moon, consequences be damned. It doesn’t have to be under the guise of schizophrenia although that is common in Shutter Island syndrome. Anyone who has worked an inpatient unit long enough has discharged a patient who did not want to leave, and watched that patient return to the ED within the day, glass or nails swallowed, telling the story as though the hospital had wronged them by believing they were well enough to discharge. Try explaining to a surgeon prepping to remove glass that the patient chose this. The extreme behavior is, in its own terms, the least emotionally painful path available to them, a way of securing a role that beats every other role currently on offer in their life, however strange that calculus looks from the outside. There is sometimes more in it than relief. There can be a grim satisfaction in it too, the feeling of having won against an antagonist, even at the cost of a stomach full of glass. That satisfaction is not something to romanticize. It’s something to take seriously as a driver, because a behavior reinforced by a felt sense of victory is much harder to extinguish than one driven by pain alone. Primary process RAGE does not feel good in itself. Bathing in the blood of your enemies, metaphorically or otherwise, sometimes does. That gap, between the raw affect and the emotions and meaning humans are capable of building on top of it, deserves its own piece.

I am not actually arguing for Shutter Island syndrome as a diagnosis. I am arguing that stories like these expose what happens when personality pathology has nowhere legitimate to live. It really is a travesty that in 2026 our diagnostic model for personality disorders is still categorical, still built on a checklist logic that traces back to a psychoanalytic era that ended decades ago. There is an alternative model tucked into the appendix, a step in the right direction, but it ignores the affective foundations of personality. The system we are trained to use shapes how we think, and the DSM currently trains every medical student, nurse practitioner, therapist, and trainee to conceptualize personality pathology through a framework that has already failed them.

There are more patients walking around with Shutter Island syndrome than any of us are comfortable admitting. That discomfort is not incidental. It is the point. Somewhere, collectively, as a field, we failed to treat this before it calcified. Everyone took the path of least resistance. The patient wants schizophrenia, the family wants schizophrenia, the insurance company pays for schizophrenia, for-profit psych hospitals want beds full of people with schizophrenia, the forces and incentives are vast. An accurate diagnosis that is dissatisfying for everyone and is stigmatized and not reimbursed falls by the wayside. Who among us has not caved in and just given the people what they want? That failure feels bad, and it should. It is its own quiet source of shame, one the field has gotten very good at labeling as schizophrenia.

As with all patient stories I write about, details have been altered or composited to protect confidentiality.

It looks like an ADHD story, but is it?

Lieslehto et al. published a registry study in 2023 examining pharmacotherapy and suicide risk in borderline personality disorder (BPD). It comes out of Sweden, hooray, another European country whose universal healthcare system makes this kind of study possible. 22,601 patients, ages 16 to 65, with registered treatment contact for BPD from 2006 to 2021.

They used a within-individual design. Every patient served as their own control. Because the study was observational, they omitted the first one or two months of medication exposure from the analysis, an attempt to control for protopathic bias, the tendency for a drug started during a crisis to look falsely dangerous simply because the crisis came first.

The primary exposure was class-level medication treatment: antidepressants, mood stabilizers, antipsychotics, benzodiazepines, and ADHD medication. The primary outcome was attempted or completed suicide. Over the follow-up period they captured 8,513 hospitalizations for attempted suicide and 316 completed suicides.

The headline finding: ADHD medications, including lisdexamphetamine, methylphenidate, atomoxetine, and dexamfetamine, were associated with a reduced risk of attempted or completed suicide. Benzodiazepines were associated with an increased risk, consistently, across every cut of the data. 56% of the entire cohort received a benzodiazepine at some point. Lithium trended toward protective relative to other mood stabilizers, though it did not reach statistical significance.

Subscribe now

So what do we make of this?

The obvious inference is that because 17.2% of these patients carried a comorbid ADHD diagnosis, the patients on stimulants were the patients with ADHD, and treating ADHD reduces suicide and suicide attempts in ADHD patients. That is a reasonable inference. It is also not what the study shows. The paper never ties the protective effect to the diagnosed subgroup specifically. It reports an aggregate hazard ratio for anyone exposed to ADHD medication versus anyone not exposed, without stratifying the outcome by whether the responders were the diagnosed-ADHD patients or the broader BPD population.

Look at what else moved in the data.

Benzodiazepines were associated with the highest suicide risk of any class, roughly a fourfold increase in completed suicide. Mechanistically, benzodiazepines potentiate GABA-A broadly across cortical-limbic circuits. That degrades top-down prefrontal control over subcortical affect-driven arousal, in a population whose top-down control is already the core deficit. Stimulants moved in the opposite direction, and stimulants have the opposite mechanism at the relevant dose range. They enhance prefrontal catecholamine signaling and strengthen exactly the gating that benzodiazepines dissolve.

That is a clean, symmetric, dose-dependent story about affect regulation architecture. It does not require postulating that most of these patients had a distinct neurodevelopmental attention disorder sitting underneath their BPD. It only requires that BPD itself involves impaired top-down control over intense, bottom-up affect. I do not think this is a controversial claim.

This looks like an impulsivity story. I do not think it is an ADHD-impulsivity story. The benzodiazepine-stimulant symmetry points to the impulsivity being generated by a dimensional affective-regulatory deficit, the inability to gate intense, rapidly shifting affect before it becomes action, rather than by a discrete neurodevelopmental trait that a subset of these patients happens to carry alongside their BPD.

None of this rules out valid comorbid ADHD driving some of the effect. It is likely a mix. But the symmetry is the tell. An affect-regulation mechanism predicts this exact shape of result even in patients who never had ADHD at all.

If you are still prescribing benzodiazepines to patients with BPD, this study is telling you something about your practice. The study also signals that there are real scenarios where treating BPD with a stimulant might be reasonable, not because the patient has ADHD, but because the medication improves processing speed and attention and strengthens the capacity to gate affect before it becomes impulsive behavior. The goal is helping the patient build a better life and stay connected to what matters to them. In terms of prescribing, the value here comes from the mechanism, not the diagnostic label.

There is a second problem underneath this one, worth naming on its own.

If this were an American cohort, I would trust the ADHD inference even less. ADHD has become a preferred and sought-after diagnosis in the United States, and it has been having its own cultural moment for several years now, right alongside autism. More and more of what people mean when they talk about ADHD today has less and less to do with attention and more and more to do with concepts like rejection sensitivity and emotional dysregulation. Those symptoms are too general to easily attribute to ADHD specifically, and they get harder to attribute to ADHD specifically with every year the category expands to absorb them.

Some of the deficits that define ADHD are impaired fronto-striatal and fronto-cerebellar circuitry, underpinned by impaired catecholaminergic signaling. Essentially, cortical regulatory circuitry has not matured on the expected timeline. Emotional dysregulation is not a specific symptom of that particular immaturity. Emotional dysregulation is what happens any time top-down cortical regulation cannot keep pace with the subcortical affect underneath it, whatever the reason the regulation is insufficient. Call it ADHD. Call it BPD. Call it a toddler who has not built the circuitry yet, or a woman in her sixties whose circuitry is coming apart. The dysregulation looks the same because the same architecture is failing.

Because this is Sweden, the pretest probability that a patient on an ADHD medication actually has ADHD is higher than it would be in an American sample. I will grant that much. But a higher pretest probability for the diagnosis does not settle the question of mechanism, and it does not touch the deeper problem, which is that emotional dysregulation keeps getting handed to whichever diagnosis is fashionable at the moment. First it was anxiety and depression. Then DID, more than once. Then bipolar. Then CPTSD. Now it is ADHD and autism. No one stops to ask whether emotional dysregulation belongs to a diagnosis at all.

Impaired concentration is now a symptom patients diagnose in themselves before they walk into the room. Social media has taught people the vocabulary of ADHD. At worst, they walk in with a checklist already primed to recite and a label already chosen. An adult who presents with emotional dysregulation, rejection sensitivity, and concentration complaints deserves careful evaluation before any of it gets filed as a complex neurodevelopmental disorder, especially one with a specific developmental trajectory that onsets in childhood and waxes and wanes with cortical growth and environmental demands. Concentration in adults is affected by a great deal that has nothing to do with ADHD. Depression, anxiety, trauma, stress, cannabis, alcohol, poor sleep, overwork, a sedentary life. All of that needs to be worked through first.

Concentration complaints in adults have to get shoehorned somewhere. A mood disorder, an anxiety disorder, a substance use disorder, or a childhood neurodevelopmental disorder that is genuinely difficult to establish retrospectively, since a patient’s current functioning biases how they report their own childhood symptoms, especially when the patient is specifically seeking a particular diagnosis or treatment. Adults who receive an ADHD diagnosis often describe it in identity terms. This explains my whole life. This changed my life. The medication itself is experienced as positively reinforcing, and patients describe it that way. None of that makes the diagnosis inherently wrong. It does mean the pattern deserves more scrutiny than it currently gets.

Worldwide ADHD prevalence held stable at around 5% for a long time. The last CDC survey put ADHD diagnosis among adolescents at close to 15%. Going from 5% to 15% in one generation is a real jump, and it is the base rate already entering adulthood. Conflating diagnostic inflation in children with diagnostic inflation in adults is a mistake, but both rates are climbing together, in adolescents and in adults, and a simultaneous rise across both populations is its own problem, whatever turns out to be driving the increases.

It would be tempting to argue that adults presenting with concentration complaints need a diagnostic category of their own, one that does not require tying concentration complaints to a neurodevelopmental disorder. I do not think that is a workable path. Minting a new and likely non-valid category just adds another box for the same phenomenon to get sorted into incorrectly. The nosology already lets patients satisfy criteria for almost any diagnosis if they are motivated to, and a new category does nothing to fix that. What has actually been missing this whole time is a mechanism, not a category.

Here is what does not get discussed enough. Patients with early developmental adversity and trauma carry the largest burden of affective dysregulation, intense emotional experience that produces disturbing behavioral output. It is difficult to concentrate while having affective arousals of that kind on a daily basis. The literature is fairly clear that when affect runs high, cortical function narrows and behavioral repertoire shrinks. Concentration impairment can be downstream of affective dysregulation itself, a routine, ordinary cause that psychiatrists rarely name directly. It gets filed under attention instead, when it should be understood as downstream of arousal.

This study is worth taking seriously for the danger of benzodiazepines in borderline personality disorder, and for the fact that there is something real in the symmetry between ADHD medication and benzodiazepines in this population.

Subscribe now

AFFECTIVE NEUROSCIENCE PRIMER: Sixty years of Affective Neuroscience investigations show there are evolutionarily conserved, subcortically generated affective states shared across mammals that are primary and not derived from cognition, that are consciously felt before they are understood, and that precede and partially determine the higher-order conscious experience that tries to explain them. Jaak Panksepp founded this field. Following his convention, Affects are written in capitals to distinguish them from folk-psychological uses of the same terms. The positive valence systems (SEEKING, LUST, CARE, PLAY) drive mammals toward what they need. The negative valence systems (FEAR, RAGE, PANIC/GRIEF) protect mammals from what can destroy them. Psychiatry has mostly ignored the Affective Neuroscience literature due to its foundations in animal models rather than human ones (studying subcortical systems in humans is a heavy lift). The appropriate response is not to dismiss the animal literature but to build upon it carefully in human contexts, using the converging lines of evidence that are available. I have tried to make Affective Neuroscience visible within psychiatry and compiled the converging lines of evidence in humans HERE.

Subscribe now

We need to talk about panic disorder.

We stopped talking about it a long time ago. We named it, filed it under the category of anxiety, and moved on. The category is where the field went wrong.

Two of the systems that Panksepp’s work identified keep getting mistaken for each other. FEAR mobilizes escape and avoidance in the presence of predatory or environmental threat. FEAR arousal leads to fight, flight, or freeze, and FEAR exists to get an animal away from the thing that can destroy it. PANIC is the separation distress signal, the powerful aversive state that arises when social bonds are ruptured and that motivates the urgent reestablishment of contact with attachment figures, a system without which no socially dependent mammal would survive infancy. PANIC arousal gives us the cry of the infant who cannot find its mother. PANIC is why we feel pain during and after a fight or breakup that puts distance between our closest attachments. Panksepp eventually renamed it PANIC/GRIEF, because clinicians kept hearing the word panic and thinking anxiety. Psychiatry took the word panic, attached it to the wrong system, and built a diagnosis on the confusion.

As neuroscience demonstrates, they are separate machinery.

Separate an infant mammal from its mother and it cries. There are specific distress vocalizations that can be measured and replicated by both environmental manipulation and electrical stimulation. Guinea pig pups make them. Chicks make them. Puppies make them. Rat pups make them. Panksepp built his models on the guinea pig and the chick because the response is so reliable. Distress vocalizations are one of the outputs of the PANIC/GRIEF system, and most importantly they are definitely not FEAR. The animal is not fleeing a predator. It is protesting a broken bond. What calms a frightened animal and what calms a lonely one are not the same things.

Panksepp mapped the PANIC/GRIEF circuit by finding the places where a small current produces the separation distress vocalizations. It runs through the periaqueductal gray, the dorsomedial thalamus, the ventral septum, and up into the anterior cingulate. The fear circuit lives somewhere else. FEAR runs through the amygdala. Two systems, two addresses. Stimulate one and the animal organizes to escape danger. Stimulate the other and it calls for what it has lost.

The chemistry is distinct as well. In animal models the PANIC/GRIEF response of separation distress is quieted by opioids. Panksepp and Herman showed that morphine, far below any dose that would blunt pain or bring sleep, will silence the isolation call in a puppy or a guinea pig. Block the opioid system with naloxone and the crying gets worse. Now take an animal whose isolation call has been silenced by morphine and show it a predator. The fear response is not changed in any way. Opioids modulate separation distress and leave FEAR untouched.

Clinicians noticed the same split long before we had the mechanism. Donald Klein saw that panic and ordinary anxiety came apart under treatment. Imipramine decreased panic attacks but did nothing for everyday apprehension. Klein also noted that childhood separation anxiety was a frequent precursor of adult panic disorder. It is a common observation that the patients whose problems are most deeply rooted in affective dysregulation are the patients with the most invalidating, adverse, and sometimes traumatic childhoods, and are also the patients most vulnerable to opioid abuse, who also tend to have the most volatile adult relationships. Almost as if PANIC/GRIEF is firing fast and strong in these patients, going through life with their separation distress circuitry operating on a hair-trigger.

I say the child who cries when the mother leaves the room and the adult who has panic attacks are running the same circuit, PANIC/GRIEF. The pharmacology and the epidemiology point at the same place, and it is not the FEAR system. It is not anxiety.

This is what subcortical primacy of affect means. The primary direction of influence is bottom-up. Affect is generated below and the cortex receives it, modulates it, often works to dampen it, in service of executive function. The toddler, the borderline patient, and the older patient with mild cognitive impairment all showed us the same thing from the other side: affect intact, regulation failing. Psychiatry was built the other way around, from the top down, out of what behavior could be observed. Subcortical primacy of affect is a staked-out and falsifiable claim, and I believe it will bear out, because it is what the bench neuroscience points to.

A panic attack is not a FEAR response. That is the first thing to see. If FEAR were driving it, the attack would carry the logic of fear. There would be a threat, and the body would organize around escaping it: fight, flight, or freeze. A panic attack has none of that structure. Just look at the laundry list of symptoms in the DSM-5 for a panic attack: sweating, trembling or shaking, unsteady, dizziness, lightheaded or faint, derealization or depersonalization, elevated heart rate, palpitations, nausea, abdominal distress, tingling, numbness, paresthesias, shortness of breath, fear of losing control or “going crazy”, fear of dying, choking feelings, chest pain, and chills or heat sensations. That is not an adaptive response to a threat. There is no predator in the room but if there were, having a panic attack would surely not be helpful in dealing with it. Nothing says escape from danger quite like depersonalization. There is nothing to fight and nowhere to run. The panic response is not adaptive to any threat the person is facing, because it was never about threat. Any of our ancestors who faced danger and responded with PANIC/GRIEF activation did not pass on their genes.

PANIC/GRIEF is about separation distress. Due to the complexity of the human cortex and the myriad of developmental and behavioral choices we display compared to other animals, humans alone have the ability to process reams of information and make symbolic associations giving us the capacity to form strong attachments to things other than our mother such as friends, jobs, favorite toys or possessions, celebrities, leaders, and even extended family relations and in-laws. Said simply, humans have a fancy cortex.

The story we tell about panic is that it comes from nowhere. The attack arrives unbidden, untriggered, disconnected from anything in the person’s life. A condition where there are recurrent unexpected panic attacks, in the absence of triggers. This is the part I want to argue with.

Panic attacks do not come from nowhere. The person having them is in separation distress and does not know it. Get close enough to the life and the triggers are not hidden. They are obvious. The job under threat, which is never only a job, because work is where a life is bound to other people. Work is also often an obligation to dependents. The friendship that is fracturing. The family situation that will not resolve. These are the things that trigger separation distress in adult humans who have grown up and no longer cry when mommy leaves the room. The panic attack is not disconnected from the life. It is the life, moving through a system built to register exactly this kind of loss. The affect arrives without its object, and so the patient cannot name it. That is what separation distress does. It is also what grief is. The same system, acute in the attack and chronic in the mourning.

None of this means FEAR plays no part. It arrives late. The attack begins, the heart pounds, the chest tightens, the body floods, and the person reaches for an explanation. I am dying. I am having a heart attack. Now FEAR has something to work with. The catastrophic reading of the body becomes the threat, and FEAR activates, downstream of the panic that started it. Primary PANIC/GRIEF sometimes with secondary FEAR. Distinct in kind, and inside the attack, separated only by degrees.

Let’s run this down with some clinical cases.

Case one: a young woman came to me after her primary care provider started her on BuSpar. She told me the medication had made her sick. She had vomited every day for a week. She was having panic attacks every day. She was certain the BuSpar had increased their frequency and made everything worse even though she only took one dose of the medication. The reaction was alarming enough that her PCP expedited a psychiatric referral. I was the first psychiatrist she ever saw.

She swore there were no stressors in her life. Nothing she could name. Nothing. By all accounts she was well adjusted, never had mental health care before, college graduate and working as an accountant. Happy at home with her partner.

BuSpar does not do this. A single dose does not cause a week of vomiting. It does not drive up the frequency of panic attacks. That is not its side effect profile. I joke with students that if a patient reports side effects from BuSpar, they probably have a personality disorder. It is tongue in cheek. It is not true. But like every joke, there is a little truth buried in it. The patients who report reactions to usually inert drugs are the patients whose separation distress circuitry is running hot. That is the same population Klein described. What the joke gets wrong is that it hears a hot circuit and calls it character. Sometimes it is character. Sometimes it is a week. This woman had no history of anything, and that week her PANIC/GRIEF was firing hard enough to turn her stomach and her cortex handed the whole thing to the pill. The reaction was real. It just did not come from the pill.

I did not tell her she was wrong. I was reminded of Francis Steven’s Affective Neuroscience and Psychotherapy, and I told her something else instead. We often do not know why we feel what we feel. The cortex likes to make associations. It reaches for connections and hands us explanations for our own internal states. Sometimes it is right. Sometimes it is wrong. We rarely know for certain what an internal experience means or why it came to us. We assign it meaning anyway. I told her it is not common for BuSpar to cause a week of vomiting from a single dose, and not common for it to increase the frequency of panic attacks. I asked her to sit with not knowing why she felt what she felt and be open to the idea that all of us should be less certain about why we feel anything.

Case two: a young man came in on a police hold. He had been having a panic attack in his car. He pulled over and called the police himself and told them he had thoughts of self-harm and was going to kill himself, because he felt so bad. He had never seen a psychiatrist. He had never taken a psychiatric medication.

He had a history of self-harm by cutting in high school. He had once crashed his car on purpose. That attempt never reached clinical attention, because he reported it as an accident. He told the officers he had fallen asleep at the wheel. He told me the truth. He had been driving home from work, overwhelmed, and he crashed the car deliberately. When he realized he was alive and unhurt, he understood it was a mistake, and he regretted it.

He too denied that anything was wrong in his life. Unsure why driving home from work he had a panic attack so distressing it led to a police hold and involuntary hospitalization.

Eventually it came together. He had fallen for a scam at work. He had been taking money from the cash register to send to a man who told him he was a Secret Service agent investigating counterfeit bills.

When I asked him about this, he had a panic attack.

According to the DSM, this is an untriggered event.

The woman came back about a month later. She was doing great.

She had quit her job and found a better one. She told me she had been under enormous stress at the old job and had not put it together at the time. She had been due for a promotion, because the person above her was leaving. That person was leaving because the position was intolerable. The supervisor she would have reported to was mean, degrading, abusive. On paper the promotion was good news. In reality it was more work for barely more pay and daily degradation from the person directly above her. She thought about it. She found another job before it came to that. She had not had a panic attack since. She was not on medication.

Consider the other path. I could have told her she was having panic attacks. The diagnosis would have organized her attention around the attacks, because that is what a diagnosis is for. I could have told her we needed to keep this from becoming panic disorder. I could have handed her an SSRI and told her, correctly, that the evidence shows she would have fewer attacks while she took it. Most patients get better this way. This is the standard of care and it is the standard of care for good reason.

The medication would have done something. It probably would have helped to some degree. But I doubt it would have helped as much as the thing that actually helped, which was planting a seed. She discovered for herself that she does not really know why she feels what she feels. She learned to pause before handing a feeling to a medication or to a convenient cause. She sat with the not knowing. In that space she found what was actually happening to her, and she changed it.

The panic did not come from nowhere. It came from a job that degraded her every day and threatened to bind her to more of the same. She could not see it until she stopped explaining it.

Both patients were out of touch with what was happening in their lives. Not stupid. Not evasive. Out of touch in the precise sense that matters here. The affect had arrived without its object.

We need to talk about panic disorder because the category tells clinicians and patients that it has to do with anxiety (it doesn’t) and panic tells the patient the enemy is the attack (it is not). The panic attack is the messenger. Silence the messenger and you have quieted your ancestral inheritance: a PANIC/GRIEF system that is trying, in the only language it has, to tell someone their most important attachments, human or symbolic, are under threat. Ask what the affect is about, and sometimes the person walks out without a diagnosis and without a prescription, having found the thing they could not see. That is not a failure of psychiatry. That is the point.

Subscribe now

AFFECTIVE NEUROSCIENCE PRIMER: Sixty years of Affective Neuroscience investigations show there are evolutionarily conserved, subcortically generated affective states shared across mammals that are primary and not derived from cognition, that are consciously felt before they are understood, and that precede and partially determine the higher-order conscious experience that tries to explain them. Jaak Panksepp founded this field. Following his convention, Affects are written in capitals to distinguish them from folk-psychological uses of the same terms. The positive valence systems (SEEKING, LUST, CARE, PLAY) drive mammals toward what they need. The negative valence systems (FEAR, RAGE, PANIC/GRIEF) protect mammals from what can destroy them. Psychiatry has mostly ignored the Affective Neuroscience literature due to its foundations in animal models rather than human ones (studying subcortical systems in humans is a heavy lift). The appropriate response is not to dismiss the animal literature but to build upon it carefully in human contexts, using the converging lines of evidence that are available. I have tried to make Affective Neuroscience visible within psychiatry and compiled the converging lines of evidence in humans HERE.

Subscribe now

There are really only three conditions in the DSM that address anger. One of the criteria for borderline personality disorder (BPD) is Inappropriate, intense anger or difficulty controlling anger. Then there is intermittent explosive disorder (IED), and for the kiddos there is disruptive mood dysregulation disorder (DMDD). None of these diagnoses do a good job capturing the problems of anger. Many patients struggle with anger and difficult to control anger without necessarily becoming explosively violent and destructive in the way IED is meant to capture. Many patients struggle with anger and difficult to control anger without meeting a single other criterion for BPD. DMDD was conjured out of thin air to try and lower the rates of bipolar diagnosis in children, I’m not sure what it’s really supposed to capture. Psychiatry has intentionally remained more or less agnostic about mechanism. When it comes to the emotional foundations of personality and the affective dysregulation that drives many presentations, psychiatry is behind the times and has not kept up with the neuroscience. This is one area we really no longer have to be agnostic about mechanism and as a field we really can do much better.

To my delight, I happened to have two such patient encounters in one day this week that fit this bill. Both chief complaints were boiled down to anger. So, let’s take a closer look at what Panksepp described as RAGE, and then let’s look at some comical clinical trials of psychiatry ineptly trying to study Aprepitant for depression.

Going back to Bard in 1928, the original decortication study in cats led to the term ‘sham rage’ due to the decorticate animals’ exaggerated and intense rage responses to trivial and neutral stimuli compared to their cortically intact counterparts. The obvious implication of this is one does not need a cortex to experience the negative valence that comes from the affective arousal of the subcortical RAGE system.

Panksepp’s work documented that substance P microinjections into the periaqueductal gray (PAG) and hypothalamic attack areas facilitated rage and aggressive behavior in animal models. Neurokinin 1 (NK1) antagonists in those same circuits attenuated the response. This fit with the broader finding that substance P is released during social defeat and threat situations, and that the NK1 system is part of the affective coloring of defensive aggression rather than just the motor output of it.

There is also literature showing NK1 knockout mice display reduced aggression, which was interpreted as supporting the substance P / NK1 system as a modulator of the RAGE substrate specifically, not just pain or general arousal.

Even in fruit flies, tachykinin is responsible for aggression. Tachykinin-expressing aSP-g neurons promote wing threat, lunging, and tussling toward other males in Drosophila. The lunge is the primary quantifiable aggression measure in fly models, and tachykinin signaling is directly upstream of it. Tachykinin from a single male-specific neuronal type recruits two separate downstream groups of neurons, mapping out the microcircuit architecture through which tachykinin drives aggressive behavior. Critically, activation of tachykinin-expressing neurons could partially restore aggression even when food, a normally required resource for fly aggression, was eliminated from the arena, meaning tachykinin doesn’t just facilitate contextually triggered aggression, it can drive it intrinsically.

The evolutionary conservation angle here is powerful for the argument. The tachykinin system driving aggressive behavior is present from insects to mammals. In flies it drives lunging. In rodents substance P in the PAG and hypothalamic attack areas drives rage. In humans the NK1 system is distributed through the same homologous structures. That kind of phylogenetic depth is exactly what Panksepp meant by primary process affect. RAGE is not a human construct. It is ancient, conserved circuitry, and tachykinin is part of its machinery across the entire animal kingdom. The DSM has no category for it.

They say science moves forward one funeral at a time, I take that to mean paradigm changes are hard due to the weight of tradition. Nothing illustrates this more than the sad but hilarious attempt to use the NK1 antagonist aprepitant as an antidepressant for major depressive disorder.

In the late 90s initial studies showed aprepitant’s antidepressant activity equivalent to paroxetine, with a 4.3-point greater reduction in Hamilton depression (HAM-D) scores after 6 weeks and improved insomnia compared to paroxetine. This generated enormous excitement when Merck made a public announcement calling it a breakthrough. Subsequent phase III trials failed to confirm any efficacy. Further study looked at combination antidepressant and aprepitant as augmentation. Surprise surprise, this also failed. There was a retrospective analysis published in 2017 titled NK1 receptor antagonists for depression: Why a validated concept was abandoned. This piece essentially argues that the dose was not high enough. At no point in the retrospective was anger or RAGE mentioned one time.

None of the researchers stopped to think about the mechanism they were studying and what that means for major depressive disorder. To design a study and to get funding, studies have to use a DSM diagnosis. The investigators’ hands were tied because at that time there was no way to conceptually design a study looking at RAGE or anger in humans because the DSM does not really have any good descriptions of this category. This has improved somewhat since the adoption of RDoC in terms of alternative ways to fund studies outside of the DSM, but RDoC has not looked at or adopted Affective Neuroscience in any meaningful way despite the obvious implications Affective Neuroscience has on the emotional foundations of personality.

It is in some ways glorious to sit back and sift through decades of research on aprepitant for MDD: The scientists were starting from a faulty premise; they ignored decades of well-done bench work and animal literature to try and shoehorn a complex neurotransmitter system that every animal model suggests is responsible for subcortical RAGE valence into outcomes for depression. Of course, any study is going to fail on any outcome when looking at depression. The study would have to be designed to look at its effect on RAGE in humans. That can’t really be done using a HAM-D as the outcome. No study has been conducted using the Affective Neuroscience Personality Scales (ANPS) in patients who complain of difficult to control anger and otherwise don’t obviously meet criteria for borderline personality disorder or intermittent explosive disorder to see whether there is an appreciable effect of aprepitant in a human on anything other than a HAM-D.

This is what psychiatry has lost and continues to lose by choosing to ignore Affective Neuroscience and to intentionally remain agnostic about mechanism in areas where we no longer need to be. Almost no psychiatrists are aware of Panksepp’s work or its implications, and the field is at a crossroads as many far wiser than me have recognized that researching non-valid DSM conditions leads to decades of garbage in garbage out research that is not moving the field forward. Paradigm changes are hard, and this project straddles many diverse fields and silos including neuroscience, psychiatry, psychology, nosology, psychopathology, and personality. Alternatives such as HiTOP and RDoC are taking shape by recognizing the importance of dimensional conceptualizations rather than categorical ones, but Affective Neuroscience remains functionally invisible within psychiatry.

One of the things I found most jarring transitioning from student to trainee to psychiatrist is that there is almost nothing in the DSM about emotions. Until I discovered Affective Neuroscience as a trainee, I was guilty of shoehorning patients with a chief complaint of anger into a diagnosis of depression or anxiety, and I still see countless patients suffer the same fate. There was always a gnawing sense of unease about this practice, and Panksepp’s work is what finally explained it. As I dove deeper into this literature, psychiatry began to make sense in a way it never had. I began to see in my own patients the seven primary process affective systems and their respective subcortical substrates driving so much of daily clinical life, and the consequences of jamming patients into the jagged categories the DSM has carved were no longer invisible.

This Substack is a continuation of that project, an effort to make Affective Neuroscience visible within psychiatry in the hope that as a field we can move toward a more valid system of classifying psychopathology that leads to better research and better outcomes for patients.

Until then, patients with difficult to control anger will keep being told they are depressed, anxious, or borderline, and no one will even think to ask why.

References:

Keller M, Montgomery S, Ball W, Morrison M, Snavely D, Liu G, Hargreaves R, Hietala J, Lines C, Beebe K, Reines S. Lack of Efficacy of the Substance P (Neurokinin1 Receptor) Antagonist Aprepitant in the Treatment of Major Depressive Disorder. Biol Psychiatry. 2006;59(3):216-223. doi:10.1016/j.biopsych.2005.07.013

Ball WA, Snavely DB, Hargreaves RJ, Szegedi A, Lines C, Reines SA. Addition of an NK1 receptor antagonist to an SSRI did not enhance the antidepressant effects of SSRI monotherapy: results from a randomized clinical trial in patients with major depressive disorder. Hum Psychopharmacol. 2014 Nov;29(6):568-77. doi: 10.1002/hup.2444. Epub 2014 Oct 20. PMID: 25330122.

De Felipe C, Herrero JF, O’Brien JA, Palmer JA, Doyle CA, Smith AJ, Laird JM, Belmonte C, Cervero F, Hunt SP. Altered nociception, analgesia and aggression in mice lacking the receptor for substance P. Nature. 1998;392:394-397. doi:10.1038/32904

Rupniak, N.M.J., & Kramer, M.S. (2017). NK1 receptor antagonists for depression: Why a validated concept was abandoned. Journal of Affective Disorders, 223, 121-125. https://doi.org/10.1016/j.jad.2017.07.042

Asahina K, Watanabe K, Duistermars BJ, Hoopfer E, González CR, Eyjólfsdóttir EA, Perona P, Anderson DJ. Tachykinin-expressing neurons control male-specific aggressive arousal in Drosophila. Cell. 2014 Jan 16;156(1-2):221-35. doi: 10.1016/j.cell.2013.11.045. PMID: 24439378; PMCID: PMC3978814.

Panksepp, J. (1998). Affective Neuroscience: The Foundations of Human and Animal Emotions. Oxford University Press.

Pop quiz, hotshot. What do the toddler, the borderline, and the forgetful have in common? It’s not a trick question. It just looks like one from inside the DSM.

Mammals have a separation distress signal, the powerful aversive state that arises when social bonds are ruptured and that motivates the urgent reestablishment of contact with attachment figures, a system without which no socially dependent mammal would survive infancy. Jaak Panksepp named this system GRIEF/PANIC and compellingly describes its evolutionary origins as the instinct for generating negative valence when faced with separation distress. Whether it’s a rat separated from its mother or a small child who just lost her favorite toy, it just feels bad.

Panksepp spent decades mapping the subcortical systems that generate primary process affect in mammals. As I have argued before, many converging lines of evidence support the extension of his work to humans. The systems responsible for generating the affective valence that motivates and drives behavior are subcortical and present and intact from birth. The cortical systems that regulate affect are grown. When cortical capacity to regulate affect is not sufficient, affective dysregulation is observed.

Psychiatry has been naming that dysregulation for over a century. The naming project is phenomenological, and it is inconsistent. Bulimia gets its own category. There is no cutting disorder, so self-harm gets folded into a personality construct. Some presentations get no category at all. The inconsistency is not a problem of diagnostic refinement. It is what you get when you build a nosology on the wrong foundation, when you start with the behavior rather than the mechanism, with the surface rather than the substrate.

Using GRIEF/PANIC, three cases illustrate the affective-regulatory principle. A toddler, a borderline patient, and a woman in her mid-sixties with mild cognitive impairment.

The toddler does not have a disorder. She has a cortex that is not finished yet.

Watch her when her mother leaves the room. The affect is immediate, total, and disproportionate to the actual threat. She screams. She cannot be reasoned with. She cannot be told that her mother will be back in two minutes, because the neuronal connections needed that would receive that information and use it to modulate the panic are not yet online. The subcortical systems that generate negative valence for separation distress are fully operational. The cortical systems that regulate it are not. This is not pathology. This is development. The dysregulation is exactly what you would predict from the substrate. The toddler is forced to live in the moment reacting to every emotion as it comes. When angry the toddler is the embodiment of RAGE. When scared she is the embodiment of FEAR. When you attempt to tickle her, she is the embodiment of PLAY.

The borderline patient is a version of the same problem, arrived at differently. Subcortical affect generation is intact. PANIC/GRIEF, RAGE: these fire at full intensity. What did not develop adequately were the cortical connections that would allow her to feel those states without being commanded by them. The borderline lacks the capacity to observe affective experience without being governed by it, to hold primary-process signals as events passing through consciousness rather than as imperatives demanding immediate behavioral response. This is not a character flaw and it is not a choice. It is what happens when the conditions required for that development, safety, consistency, attuned relationship, were not reliably present when the developing cortex needed them most. She did not fail to grow. She grew what she could, with what she had. What she grew is not sufficient. She cannot escape being brought to clinical attention. When that aversive separation distress signal is activated, she too appears like the toddler unable to modulate her behavior when affect is aroused. Cortical capacity is overwhelmed until the affect dies down. Every fight with her partner and every breakup leads to dramatic and potentially life-changing consequences.

The third case is the most telling.

She is in her mid-sixties. No past psychiatric history. Lives alone, manages her own affairs, meets all her IADLs/ADLs without assistance. She comes to her first appointment with her family in tow, they are worried. They describe new onset anxiety, depression, and panic. They also describe something more specific: when anyone visits, and then tries to leave, she transforms. She cries. She screams. She hits the wall. She runs to the car and puts herself on the hood. She will not let anyone go.

This is all new. It started recently. It is getting worse.

By the time the family reaches my office, she has seen two nurse practitioners and a psychiatrist. She has been hospitalized. She has been on multiple medications. Nothing is working, by which the family means: she still does this, so the medications must be wrong, or the dose must be wrong, or the diagnosis must be wrong.

The diagnosis is wrong. But not in the way they mean.

Despite seeing three previous providers and spending weeks in a psychiatric hospital, I was the first person to perform any kind of cognitive testing. Her Montreal Cognitive Assessment score is twenty-three out of thirty. This is, roughly, the average score one would expect from someone with mild cognitive impairment. The cortex is beginning to go. Not dramatically, not yet. She is still functional in most domains. But the capacity that allowed her to feel separation distress and not be overwhelmed by it, the capacity she had been building and exercising her entire life, that capacity is eroding.

The subcortical PANIC/GRIEF system does not know this. It is doing exactly what it has always done. Someone she loves is leaving. The alarm fires. What is no longer there, or no longer there reliably, is the cortical regulatory apparatus that would receive that signal, contextualize it, and modulate the response. She felt this her whole life. She no longer can feel it without being consumed by it.

No medication treats that. Not because the medications are inadequate, but because depression, anxiety, and panic are not what she has. Those are labels for the output. They describe what dysregulated affect looks like from the outside. They do not explain it, and they do not point toward anything that would help.

What helped was this: explaining the mechanism to the family. The cortex that had regulated her emotional life for six decades is losing capacity. The feelings she is having are real. The grief when people leave is real. What is impaired is the ability to hold that grief without it taking over. The medications were not going to fix that, but adjusting expectations about what medications could do, and reframing the goal from symptom elimination to management of reactivity, changed what the family was trying to accomplish.

And then something simpler. The family learned to say, before leaving: we are going now. We will be back. They learned to say it clearly, and to say it again if needed, and to understand that they were not managing a behavioral problem. They were doing, from the outside, what her cortex could no longer do from the inside.

This is what affect before diagnosis means in practice. Not as a theoretical preference. As a clinical fact.

The toddler cannot regulate because the cortex is not yet built. The borderline patient cannot regulate because the cortex was not built well enough, under conditions that made adequate construction impossible. The woman on the hood of the car cannot regulate because the cortex she spent a lifetime building is coming apart.

Three different etiologies. One mechanism. Subcortical affect generation intact. Cortical regulatory capacity impaired.

If you start with the diagnosis, you get depression, anxiety, and panic, and a medication list that does not work, and a family that keeps being told the problem should be responding to treatment when it was never going to respond to that treatment, because that was never the problem.

If you start with the affect, you see what is actually there. And you know what to do.

She is nineteen years old. She works the night shift as a CNA, is enrolled in nursing school, and has full custody of her two-year-old son. She lives with her father, the man who adopted her, in a house outside of town, in a county where it’s not rare to find rattlesnakes in the yard.

She came to psychiatry because she is struggling. Not collapsing. Struggling.

Her mood shifts fast and without warning. She gets angry in a way she describes as a switch. Sudden, total, and strangely satisfying when it lands on someone else. She feels sad often, but not every day, and when she feels sad, she feels like the feeling will last forever. She doesn’t sleep well because she ruminates. She worries about things that aren’t likely to happen. She has passive thoughts that her family would be better off without her, not suicidal ideation in the active sense, but a recurring background conviction that her absence would be a net positive for the people she loves. She says she has tried to ignore all of it, to bottle it up. She says this hasn’t worked. She feels exhausted.

She graduated high school at sixteen, pregnant. She kept her son. She went back to school. She is employed in a physically demanding job that requires clinical competence and emotional regulation under pressure. By any reasonable measure, she is functioning. Yet she is right that something is wrong.

This is exactly the kind of patient who reveals the limits of the DSM and exactly the kind of patient who does better when we treat those limits seriously.

What the DSM does with her

The DSM is a descriptive document. This is not a criticism, it is a statement of design intent. It describes clusters of symptoms that reliably co-occur, names those clusters, and provides operational criteria for identifying them. What it does not do, and has never claimed to do, is explain why the clusters exist. Etiology is largely outside its scope. Mechanism is not its business.

How does one apply a DSM diagnosis to this patient?

She is still working. Still studying. Still parenting. She deals with flat tires and difficult shifts and the ordinary catastrophes of a life that does not stop because she is struggling.

She describes feeling sad and depressed, but not every day, and she retains the capacity for pleasure and joy, real pleasure, real joy, when she is with her father and her child. If you hand her a checklist and ask directly, she will endorse depression, anhedonia, SI, low energy, poor sleep, and poor concentration. She ruminates when she tries to sleep. But she is busy all day. She does not stop. She has no choice. And so when she lays her head down at night and it is dark and still, that is the moment, the first moment she has had, when the anxieties surface and the distractions are gone. Of course she cannot sleep. Of course she ruminates. Hand her that same checklist for generalized anxiety disorder and she will endorse every item because when she feels anxious, she feels like she will be anxious forever.

Her mental status exam tells a different story. She is bright and reactive. She laughs. She smiles. She is slightly tearful at appropriate moments, discussing the hardest periods of her life with a total stranger. When pressed, she is not anxious all day, she is busy all day.

She witnessed a friend die in a car accident as a teenager. She carries a history of early adverse experiences that, depending on the clinical judgment applied to the criteria, could support a PTSD diagnosis. Her moods shift. Some days she has more energy than others. Her sleep is poor. Why not bipolar II?

There is one thing she knows for sure: she has always been this way. She says she has felt this way her whole life.

Defining a discrete episode is difficult, because there is no discrete episode. There is no period in her history where she was functioning well and then, following an identifiable stressor or traumatic event, became symptomatic. She did not grow up and then encounter adversity. She grew up inside it. The trauma and the stress and the dysregulation are not events that happened to a previously well person. They are the water she learned to swim in. Her symptoms did not have an onset. They have a history.

This matters diagnostically more than it might appear. Every DSM condition that might plausibly apply to her. Depression, anxiety, PTSD, Bipolar disorder, carries an implicit assumption of punctuation. Something changed. There was a before and an after. The criteria are written for episodes, for discrete periods where the clinical picture was meaningfully different from whatever preceded it. She cannot give you that. Not because her history is unclear, but because the history is unbroken. She has felt this way as far back as she can remember, which is about as far back as memory goes.

The problem with giving her any DSM diagnosis is that every diagnosis implies a disease, and every disease implies a treatment, and the implicit contract is: this is what is wrong with you, and this is what we will do about it.

I do not believe that contract can be honored here.

I have no hope that telling her she has major depressive disorder and treating her accordingly will remit her symptoms. I have no hope that if I tell her she has PTSD, which I could justify, given what she has survived, and she follows a trauma-focused treatment plan to the letter, her symptoms will resolve. I have no hope that an anxiety diagnosis and a treatment plan for anxiety will do better. I have no hope that a bipolar diagnosis, with mood stabilizers or antipsychotics calibrated to that target, will move the needle in any meaningful way.

This is not nihilism. It is precision. The treatments I just described are not wrong in the abstract. They are wrong for what she actually has. And what she actually has is not a disease that arrived and can be made to leave. It is an architecture. It is the shape of a nervous system that was built under particular conditions and has been running ever since.

Naming the wrong thing does not help. It just gives the wrong thing a Latin title and a billing code, however any of these could be written on a chart today and defended at peer review.

What the affective-regulatory framework does with her

The affective-regulatory account begins not with her symptoms but with her system.

Jaak Panksepp’s work and the substantial neuroscientific literature that followed it established that the primary emotional systems are subcortically generated, evolutionarily conserved, and largely beyond voluntary control. The RAGE system, the FEAR system, the PANIC/GRIEF system: these are not constructs. They are identifiable neural circuits with documented anatomical substrates, characteristic neurochemistry, and behavioral outputs that are consistent across mammalian species. They generate affective experience in a bottom-up fashion, sending activation upward from the midbrain to the cortex.

The cortex, for its part, is supposed to receive that activation and do something with it, contextualize it, modulate it, insert a delay between impulse and response. This is not metaphor. The orbitofrontal cortex, the anterior cingulate, the prefrontal regions more broadly: these structures provide top-down inhibitory regulation of subcortical emotional circuits. They are the brakes.

Here is what is critical about those brakes: they are not innate. They are grown. Cortical regulatory capacity develops postnatally, shaped by experience, environment, and attachment relationships. The window from birth to approximately age five is the period of fastest growth. What happens during that window whether the child is consistently soothed or chronically unsoothed, whether the environment is predictable or chaotic, whether caregivers model regulation or dysregulation directly and durably affects the regulatory architecture that emerges.

This patient’s early environment was, by her account, characterized by maternal neglect, emotional unpredictability, and intermittent physical aggression. Her primary regulating caregiver was herself dysregulated significantly, with the same phenotypic features now presenting in her daughter. Her father, who appears to have been a stabilizing presence, was largely absent for work. She was parentified by the time her younger sister arrived. She was, functionally, raising herself.

Simultaneously, she carries documented genetic loading for the very traits she is presenting with. She describes her mother, maternal grandmother, and maternal aunt sharing the same phenotype that includes affective intensity, mood instability, and anxiety. She says she thinks they all have bipolar disorder, yet to her knowledge none have ever taken medication or been hospitalized. Temperament is heritable. The subcortical emotional systems that generate high-intensity emotional output are, at least in part, a matter of baseline calibration and that calibration runs in this family.

What we have, then, is not a mysterious disorder or any categorical DSM diagnosis. We have a young woman with a constitutionally reactive affective system, shaped during its most plastic period by an environment that provided insufficient regulatory scaffolding, who is now at nineteen, sleep-deprived from shift work, newly bereaved of her grandfather and her dog in the same week, single-parenting a toddler in a city she recently moved back to running that system under load.

She is not sick in the way a broken bone is sick. She is struggling in the way that any system will struggle when it was built under difficult conditions and is now being asked to perform under difficult conditions.

This distinction is not semantic. Here is where the two frameworks diverge most consequentially, and where the implications for the patient experience of treatment are sharpest.

A DSM diagnosis implies a target. It implies that something is wrong, that the wrong thing has now been named, and that treatment is the process of correcting it. Remission. Response. Recovery. The language of psychiatry’s outcome literature is the language of completion of moving from a diseased state to a well state, of checking a goal off a list.

This is clinically useful for some conditions. Major depressive episodes remit. Acute psychosis resolves. Specific phobias respond to exposure with a success rate that would be the envy of most of medicine.

But it maps poorly onto something like this patient’s presentation and onto a great deal of what psychiatry actually sees. Affective-regulatory dysfunction is not a disease that is cured. It is a developmental gap that is filled, gradually, through practice, skill acquisition, and neuroplasticity. The cortical regulatory capacity she did not get to build in optimal conditions can still be built. That is not wishful thinking, it is the neuroscience of experience-dependent plasticity but it is built the way any complex skill is built: imperfectly, incrementally, and without a finish line.

This matters for what we tell patients. When a clinician implicitly or explicitly frames treatment as the correction of a disorder, the patient’s expectation becomes: I will take this medication, I will attend this therapy, and I will get better. Better meaning back to baseline. Better meaning the problem goes away.

This patient said, directly, that she wants to feel normal. That is an understandable thing to want. It is also, as a treatment goal, a setup for failure because it frames the endpoint as a feeling state, and feeling states are not reliably controlled, even by excellent treatment.

The affective-regulatory frame offers something more honest and, in the long run, more useful: you are not trying to feel a certain way. You are trying to build something. The cortical capacity to observe your emotional experience without being driven by it, to insert a pause between stimulus and response, to act from values when feelings are pulling hard in another direction this is what is being grown. It does not have an endpoint. It is more like the practice of medicine than the treatment of a broken leg. You keep getting better at it. You never arrive.

This reframe is not merely philosophical. It changes what patients do with hard weeks. When a patient with a DSM diagnosis has a bad month, the natural interpretation is that treatment isn’t working that the disorder is reasserting itself, that the goal of correction is slipping away. When a patient who understands her own regulatory system has a bad month especially one that includes shift work disruption, acute grief, and poor sleep she has the conceptual equipment to understand why. The system is under load. That doesn’t mean the system isn’t changing. It means right now the conditions are hard.

The journey framing, to use a word that risks sounding like a wellness influencer, is not idealism. It is accuracy. And accuracy, in psychiatry, is a clinical tool.

What she carries out of the room

At nineteen, she is having her first contact with psychiatry. Whatever framework she absorbs in that encounter whatever story she is handed about what is wrong with her and what can be done about it, she will carry for years, possibly decades. It will shape how she explains herself to future clinicians, how she interprets hard periods, what she expects from treatment, and how she understands her own agency in her own recovery.

If she walks out with a DSM diagnosis she carries a label and a conceptual frame that centers disorder over development, pathology over plasticity. She may encounter clinicians who see the diagnosis before they see her.

If she walks out of that office understanding that her brain is the sum of the connections her neurons have made, that she feels things intensely, that she is emotionally reactive, not because something is wrong with her but because of how she developed and where she came from, she has something more useful than a diagnosis. She has a map.

Her brain was shaped by genetics and environment in ways that were not her fault. The cortex she needed to build had suboptimal conditions during its most critical window. The regulatory capacity that develops in early childhood, the capacity to observe an emotion without being commanded by it, to pause between impulse and action, to feel the full force of something and still choose, that capacity is grown through relationship and repetition and safety, and she did not have enough of any of those things when it mattered most.

But the cortex does not stop growing. That is not a motivational claim. That is neuroscience. The connections she needs can still be built. They are built the way any complex skill is built, through practice, through repetition, through the slow accumulation of experiences that teach the nervous system something different than what it learned before.

And here is what she should also understand: she has already been doing this. Under tremendous stress and difficulty, with no map and no one handing her the framework, she has already grown the connections she needed to function. She kept her son. She went back to school. She is still showing up. That is not nothing. That is evidence. Her brain is already capable of building what it needs. We are just going to do it more deliberately now.

She does not have a disease. She has a history, a biology, and a nervous system that is still, at nineteen, remarkably plastic.

That is not a consolation prize. That is the best possible news.

She has a two-year-old watching her. She has a lot of growing left to do.

The question is whether psychiatry meets her there, or hands her a category and sends her on her way.

As with all patient stories I write about, details have been altered or composited to protect confidentiality

.