Opinion Guest Essay

Aug. 9, 2026 (NYTimes.com)

By Daniel Duane

Mr. Duane is a writer and therapist who lives and works in San Francisco.See more of our coverage in your search results.Add The New York Times on Google 

1. The Therapy Crisis

I love being a therapist. I love being present for people over time, listening for mystery and pattern. I also know that psychotherapy done well can bring emotional freedom and even authenticity, that precious feeling of leading one’s own life. But therapy needs a therapist of its own.

The profession has never been more self-assured. The old stigma has faded. Demand is high, thanks to nightmarish politics, digital isolation and TV shows like “Shrinking” and “Couples Therapy.” But the work itself — the actual practice, in the room — faces relentless devaluation.

Companies like BetterHelp try to reinvent therapy as low-paid gig work: DoorDash for emotional support. A.I. start-ups raise many millions of dollars to create therapy-adjacent chatbots, as if whatever a human therapist does can be replicated by an algorithm utterly devoid of feeling, much less embodied experience.

It would be nice to blame opportunistic entrepreneurs, but we wouldn’t be having this conversation if therapy itself were not a mess — or, rather, a Tower of Babel with hundreds of approaches based on contradictory theory and research, and no agreement on what even constitutes a good outcome.

To be fair, those hundreds of therapies mostly fall under one of three main tents — psychodynamic, existential humanist and cognitive behavioral (better known as C.B.T.) — plus a fourth that we might as well call influencer-driven because its unifying element is TikTok. But you would not believe the range of thinking underlying each.

Psychodynamic therapy, for starters, grows out of Freudian psychoanalysis, which remains the intellectual taproot of the entire enterprise. But it also hosts a bewildering diversity of thought. The psychoanalytic community alone churns with dialogue among classical interest in unconscious conflict, the attachment-theory concern with how patterns of childhood relationships influence adult lives, and the so-called ontological drift identified by Dr. Thomas Ogden, an influential writer who has been my own primary teacher as I have turned myself into a psychoanalytic therapist.

In Dr. Ogden’s framing, therapy’s power lies less in whatever knowledge it brings and more in the dreaming of one’s un-dreamt dreams — ways of feeling more fully the half-digested experiences that lie walled-off somewhere inside ourselves.

At the opposite extreme, C.B.T., which gets by far the most research attention, grows out of early-20th-century studies of how positive and negative stimulus can modify behavior. Applied to emotional behavior like a fear response, or cognitive behavior like the persistent thought that one is worthless, this can be enormously helpful. And yet, the most influential innovation in the C.B.T. camp has been Acceptance and Commitment Therapy, based on the idea that trying to modify thoughts often backfires and that we should focus instead on whatever we value most.

To make matters more confusing, most states require therapists to take continuing-education courses. New products sell better than old ones, so new therapies get branded on roughly the same cadence as soda flavors, with their own acronymic names. The biggest player in this space, PESI.com, channels Fast Eddy’s Discount Therapy Training Emporium with online seminars in, for example, Internal Family Systems, (I.F.S., naturally) a $1,319.78 value on special for $399.99.

2. The Problem With Therapy Research

Psychotherapy research is both helpful and not. Bruce Wampold, a co-author of “The Great Psychotherapy Debate,” told me that the most durable finding is that talk therapy works (whew!) at least as well as antidepressants but often with longer-lasting benefit and fewer side effects. But research has also found that some therapists — “super shrinks,” in the literature — were much better than others for reasons that did not correlate with whether they had a Ph.D. in psychology or a Masters in social work, their preferred theory of therapy, or even how long they’d practiced.

Still more confounding, studies repeatedly find that it barely matters for patient outcomes which of those hundreds of approaches a therapist claims to use. What matters vastly more is that a particular kind of relationship exist between therapist and client. This makes intuitive sense but may sound like saying that you shouldn’t care if your doctor treats your sore knee with Adderall or rhinoplasty so long as you like her bedside manner.

These findings, plus hundreds of therapies, are also the door through which BetterHelp and A.I. chatbots are stepping — a cultural and scientific moment in which therapy is somehow everything and nothing, infinitely deep and shallower than a dry lake bed.

Now for a big ray of sunshine. Therapy, at long last, may have a way out of its embarrassing identity crisis. It comes from a heady source:the neuroscience of predictive processing, which is perhaps best characterized as the study of how our brains construct reality. Throughout the psychotherapy community, in private seminars and articles like one by Daniel Villiger in The Journal of Contemporary Psychotherapy, there is audible excitement that predictive processing might give therapy its first unified field theory, a Rosetta stone for explaining what makes talk therapy work and how to know if you’re getting the good stuff.

3. How Therapy Lost Its Way

The messy predicament of therapy dates back to the late 19th century, when the young Sigmund Freud became fascinated by physical ailments with seemingly psychological origins — like so-called hysterical blindness, in which sudden loss of sight could not be explained in terms of nerve or tissue damage. If Freud had his way, he would have studied these ailments by looking directly into patients’ brains.

That was not possible at the time. So, Freud observed what he could — patients’ speech and behavior. In so doing, he created the oddest of scientific beasts, psychoanalysis — a discipline dedicated to objective understanding of the human organism through methods not dissimilar to those of philosophy.

Because Freud came first and wrote so well, he gave the field its theoretical foundation, including core concepts like, to cite the most famous example, unconscious conflict between desire for a boy’s mother, murderous aggression toward his father and fearful inhibition against expressing either. Freud initially believed that bringing the unconscious to light would set us free. Over time, he noticed that even clients aware of their unconscious predilections kept manifesting the same lousy patterns in life.

Freud called this the repetition compulsion — or, alternately, the destiny neurosis. The entire subsequent history of psychotherapy can be viewed as one long effort to understand and disrupt this frustrating tendency. Why do we repeat harmful and hurtful emotional patterns even after we learn to recognize them? And how can we stop?

With no help from physiological brain science, that question could only be addressed through more hypotheses. By the 1930s, even as these hypotheses proliferated, therapists noticed that most approaches to therapy itself — regardless of which hypotheses those approaches were built upon — had similar effects. This came to be known as the Dodo Bird Verdict, borrowing from “Alice in Wonderland,” in which the Dodo Bird says, after a foot race, “Everybody has won and all must have prizes!”

Decades later, research emerged that supported the Dodo Bird Verdict by finding that different therapies get similar results on standardized measurement tools like the Beck Depression Inventory. This drove new research based on the premise that if all therapies are equally effective, that effectiveness must result from whatever all therapies have in common, not whatever sets them apart. This “common factors” research produced a stable list of key elements. These include a therapist who displays genuine empathy while holding the client in unconditional positive regard; a client with a sincere desire to change and confidence in the therapist; and, between the two, a spirit of collaboration.

The biggest objection to common-factors research has always been the Adderall-rhinoplasty problem. It defies common sense to believe that there is zero difference in impact between psychoanalytic exploration of childhood and C.B.T. work sheets for correcting, say, negative self-talk. One explanation has been that perhaps common-factors researchers, by measuring the benefit of therapy according to vaguest metrics, simply guaranteed a vague finding.

Still, the common factors might have gotten us moving in the same direction — if not for the collision with managed-care health insurance in the 1990s. Insurance benefit managers wanted scientific evidence in support of any treatment they reimbursed. So they looked for randomized controlled clinical trials based on what one might call the opposite of the common factors approach: as in, Yes, good research consistently finds that the specifics of different therapeutic modalities don’t matter, but let’s look for studies of specific modalities anyway.

In the way of such things, this insurance-driven quest for clinical trials supporting specific therapies encouraged researchers to conduct ever more clinical trials. This feedback loop led to an implicit Dodo 2.0 — namely, the widespread misconception that only C.B.T. and its variants are supported by scientific evidence. By the early 2000s, Dodo 2.0 was so pervasive that the term “evidence-based therapy” became synonymous with C.B.T.

This, too, might have gotten therapists on the same page — Oh well, C.B.T. wins! — if Dr. Wampold and a clinical psychologist named Jonathan Shedler hadn’t both started pointing out the remarkably poor fit between psychotherapy treatment and the gold-standard requirements of clinical trials. These include having a clear instruction manual for any treatment under study; all clinicians following this manual; all patients sharing a single diagnosis; all trials running for defined time periods that, according to Dr. Shedler, are typically between six and 12 weeks; and outcomes measured in terms of relief from clearly defined symptoms.

Sensible stuff, but no good therapist of any persuasion follows an instruction manual. So, as Dr. Shedler has observed, whatever happens in those trials bears little resemblance to what skilled therapists do with clients in the real world. Also, C.B.T. variants are, by and large, the only therapies designed around instruction manuals. So an overwhelming number of clinical trials have studied C.B.T.s and little else.

As for diagnoses, those all come from the Diagnostic and Statistical Manual of Mental Disorders, Fifth Edition (better known as the DSM-5,), which has limited use for research because it classifies nearly all disorders based upon observable symptoms without regard to cause — as if everyone who limped into an emergency room got diagnosed with Limping Disorder instead of, say, bone cancer, or playing pickleball while stoned.

Worse, according to Dr. Shedler, other research shows, with regard to clinical trials that run from six to 12 weeks, that starting therapy brings a mood bump, but that real psychological change doesn’t begin until months later.

Finally, a study published in Lancet Psychiatry confirmed what almost anybody could have told you — that people don’t seek therapy solely for relief from D.S.M.-5 symptoms. They seek therapy for self-awareness and self-compassion, new perspective on their situation, adaptive new ways of thinking and better relationships.

This means that clinical trials of therapy conducted according to instruction manuals, with patients sharing DSM-5 diagnoses, for short time periods, with outcomes measured in terms of symptom relief, are a magnificent way to produce evidence in favor of brief, instruction-manual C.B.T. without studying anything of great importance.

Dr. Shedler pointed out to me that some of the most widely publicized C.B.T. trials had high dropout rates and often delivered meaningful benefit to small percentages of participants. Some studies have also found that after C.B.T. trials stop, benefits fade.

By contrast, according to Dr. Shedler’s analysis of the research, the benefit of psychodynamic treatment — that is, traditional talk therapy — in clinical trials seems to keep growing after termination. Still more curious, according to other recent papers, the overall effectiveness of manual-driven C.B.T. has weakened in trials since the late 1970s, when the first C.B.T. instruction manual was published.

None of this means that C.B.T. therapy as practiced by good clinicians is less effective than psychodynamic therapy. What it does mean is that we live in a fun-house therapy world where the term “evidence-based practice” is still widely used to imply the superiority of treatments for which the evidence is poor.

4. How Neuroscience Can Help

Predictive processing is a vast field that draws on more than a century of philosophy, cognitive science and empirical research into neuroanatomy and electrophysiology. My own understanding of it comes largely from conversations with Mark Solms, a professor of neuropsychology at the University of Cape Town and author of “The Only Cure: Freud and the Neuroscience of Mental Healing.”

It helps to begin with our shared confidence that our eyes and ears merely show us what is going on in the world. This confidence, it turns out, is based on an illusion. The truth goes more like this: Our brains maintain an internal generative model of the world and use it, in every instant of life, to predict the imminent future as the future becomes the present. Importantly, though, we do not experience these predictions as predictions. We experience them as reality.

The slightly longer version of predictive processing is that our brains must keep us alive despite having zero direct access to the outside world — just inbound nerve signals from the sensory surfaces in our eyes, ears, noses, tongues and skin. But these signals are ambiguous and caused by things we cannot see, so if we used only those signals to make sense, we would be slow and clueless.

As a workaround, living creatures evolved a pair of related tools: memory, so that past experience can inform our understanding of the present and future; and that internal generative model, in which memories get reconstituted and recombined to create our sense of whatever’s happening.

Let’s say you hear a cat’s meow behind you. As you turn to look, neurons fire in your brain’s visual cortex, and those are the prediction signals. This pattern will be composed of subpatterns associated with past experiences and will correspond to whatever your brain considers most likely to appear — your open back door, maybe, and your own tabby cat on the porch. This anticipatory pattern will then head outbound toward your eyes.

Once you have turned around and your eyes capture light reflected from that direction, your retinas translate that light into a new signal pattern that now heads inbound toward your brain. Along the way, this newer inbound pattern encounters the outbound predictive one in a hierarchical fashion: first, deep fundamentals like the basic outlines of the room and the glare of sunlight, then ever-finer details like the color of the cat’s fur.

If the inbound sensory pattern and outbound predictive patterns turn out to be identical, the brain’s original prediction becomes what you see. If an inconsistency appears at any level — the fur on the cat appears to be black, not tabby — your brain must decide which is more precise, its own prediction or this new information. Most of the time, your brain makes the right choice — my porch, not my cat — and uses the error to update its internal model and make better predictions going forward.

But now consider the combat veteran who, after months of terrifying night operations, goes home to his family. Asleep one night in bed, the veteran is awakened by a soft noise. Even before he opens his eyes, the veteran’s brain starts modeling, based on memory, whatever might be happening.

As Dr. Solms pointed out, our brains do this in part by favoring memories strongly associated with survival. The veteran’s brain, in choosing between memories of his beloved’s footfalls and a midnight ambush, favors the latter. A corresponding nerve signal heads outbound toward the eyes.

The veteran’s eyes then open and capture reflected moonlight in a pattern not dissimilar to that of his wife in her pajamas. This light becomes a second nerve signal. When the two signal patterns fail to match, the brain makes the safer choice by sticking with its prediction: ambush.

The veteran leaps out of bed, frightening not just his wife, but also his children. Once the lights come on, he realizes the mistake. This happens again the next night and the next. He takes to sleeping in the garage.

Ideally, in the interest of the veteran’s family and its well-being, the veteran’s brain would use all those repetitive errors — identical mismatches between outbound prediction and inbound sensory evidence — to update his internal generative model and stop constructing reality in the same mistaken way. But, again, survival: If we have had an extremely threatening experience in the past, our brains have good reason to keep using a memory of it as a predictive tendency, readily assuming that this terrible experience is now happening again.

We can think of PTSD as an extreme version of a slippage that we all experience every day, especially in our social lives. Dr. Solms pointed out that the memories informing our most fundamental understanding of reality get laid down in infancy, when we are tiny and so dependent upon others that abandonment for even a very short time could mean death.

This means that these deepest memories are at once highly social and associated with survival, making them powerful go-tos for modeling reality. Those earliest memories also get stored in forms unavailable to conscious thought.

For the rest of our lives, in every exchange with everyone who matters to us, our brains construct reality — by which I mean we quite literally see other people and ourselves — based on childhood experiences that we cannot even recall. Worse, we respond to the realities that our brains construct largely by repeating behaviors that worked for similar situations in the vanished world of childhood but that no longer bear any resemblance to the behavior of a sane adult.

Fear and stress in the present moment, whether at work or in love, cause further trouble by forcing our reality-modeling systems to favor their own predictions rather than incongruous new information. The more threatened we feel — worrisome day at work, painful conversation with a parent or child — the more we perceive everything through the lens of our most frightened childhood selves.

5. A Unified Field Theory of Therapy

As for how this can help therapy out of its identity crisis, it would start with that bit about why people seek therapy in the first place: for deeper self-awareness and perspective and for adaptive new ways of thinking, as I mentioned earlier. Put in predictive processing terms people seek therapy because something about their internal generative model keeps constructing reality in ways that cause trouble. Therapy, then, looks like a way of getting your predictive model to update. On the question of how, exactly, therapy does that, predictive processing theory tells us that model updates depend upon two key factors: safety and precision.

Safety, in this context, means an environment of such emotional and physical security that a person can let her guard down, stop defaulting to old ways of seeing things and start recognizing her predictive errors. Precision means a steady diet of new emotional experiences tailored to elicit precisely the predictive errors that are causing problems — Even my therapist will abandon me — and, more important, to let her mind register the error and self-correct.

Any competent therapist will have noticed that this notion of safety correlates exactly with the findings of both common-factors research and everything the best theorists have been calling good therapy for more than a century. I am speaking here of the unique therapeutic setting and relationship, defined by its privacy, predictability, confidentiality and the therapist’s empathy for the client.

More than 50 years ago, Dr. Donald Winnicott, a leading figure in British psychoanalysis, called this a “holding environment” and described it as the sine qua non of effective therapy.

The same goes for that notion of precision. Amy Wallerstein Friedman, a therapist with whom I consult on clinical cases, described a kind of going-toward, a deliberate exploration of whatever is most difficult in order to update the client’s understanding of self and world, as what every psychoanalytic therapist works to master.

Victoria Beckner, a clinical psychologist in San Francisco, gave a remarkably similar description of C.B.T., in which she trained. “The whole game in C.B.T. is predicated on the idea that people come in with their world model and learned behavior, and that it’s all stuck somehow — you were rejected by mom so you think you’ll be rejected by everybody, and we need to get you to step out, take risks and discover that this is not true.”

Perhaps most striking, whenever therapists have developed integrative approaches — blending traditions — they have typically done so along these lines. I co-wrote a paper last year with colleagues at the San Francisco Psychotherapy Research Group showing this to be true of control mastery theory, which holds that people seek therapy because they have a maladaptive belief, such as It is never safe to share my unpleasant feelings. That belief is making them miserable. They get into therapy to test — and ideally disconfirm — that belief in a safe setting. The therapist’s job, then, is to create that safe setting, watch out for the test and react in ways that support the client’s goal of shedding the problematic belief and forcing a model update.

It may be too much to hope that I will never see an advertisement for online seminars in Predictive Processing Therapy on summer sale for $99.99. But I am optimistic that predictive processing will someday guide empirical research updating the absurdity of our current Dodo Bird Verdict, in which all therapies win and deserve prizes. I also believe that in the short term predictive processing will provide a way for therapists to speak with one voice about what makes therapy effective. And yes, the specific work that a C.B.T. therapist asks of a client will doubtless remain very different from that encouraged by someone like me, working in the psychoanalytic tradition.

Here, though, predictive processing theory promises to be even more helpful. Michael Garrett, a psychoanalyst, pointed out to me that people gravitate to different ways of thinking about the human psyche and suffering. We also resonate with different rituals for healing.

Some of us love the ornate and yet somehow penetrating language of psychoanalysis. Others much prefer the no-nonsense pragmatism of C.B.T. Every therapy is also unique for the simple reason that therapy done right is a fundamentally human endeavor undertaken by two unique human beings in a spontaneous, ever-evolving relationship over time.

Predictive processing theory allows us to say that, so long as a given therapist and client have a genuine human relationship, emotional safety and a reliable method for generating — and resolving — flawed ways of modeling the world and self, they are probably on the right track. And if we can get it to work for the guy who keeps hearing an ambush when his wife stirs in the night, we can certainly be of use to people who want to understand themselves better, break free of whatever problematic beliefs hold them back, and enjoy lives that feel more authentically their own.

Daniel Duane, a writer and therapist who lives and works in San Francisco, is the author of the forthcoming memoir “Yosemite Falls.”

The Times is committed to publishing a diversity of letters to the editor. We’d like to hear what you think about this or any of our articles. Here are some tips. And here’s our email: letters@nytimes.com.

Follow the New York Times Opinion section on Facebook, Instagram, TikTok, Bluesky, WhatsApp and Threads.

(Contributed by Michael Kelly, H.W.)