Twenty-five years. 3,617 adults. One finding that took two and a half decades of data to prove and about two paragraphs of a letter to argue: well-being doesn’t move in a straight line, and it doesn’t move at random either. It resolves into a small number of identifiable shapes, and where you start goes a long way toward determining which shape you get. That’s Depressive symptoms predict divergent trajectories of well-being in US adults, published August 12, 2026 in Nature Mental Health by Cassondra Lyman, Anthony Ong, and Jonathan Rottenberg. Seneca made the structural version of this argument in one of his letters to Lucilius, written in the last years of his life: nobody becomes good all at once. Progress happens day by day, and where you begin shapes — without fixing — where you end up.

The Quick Verdict

Question	What we found
Who ran the study?	Cassondra Lyman, Anthony Ong, and Jonathan Rottenberg
Where and when?	Nature Mental Health, published August 12, 2026
How big was it?	Americans’ Changing Lives cohort, N = 3,617, tracked for 25 years
What method?	Growth mixture modeling on repeated well-being measurements
What did they find?	Well-being resolves into four distinct 25-year trajectories, not one universal pattern
What predicted the trajectory?	Baseline depressive symptoms — independent of a person’s baseline well-being score
What did Seneca argue?	No one is good all at once; moral progress is gradual, and starting conditions shape the path without determining the outcome

The Quick Version

A 2026 Nature Mental Health study followed 3,617 adults for 25 years and found their well-being didn’t hold steady or drift randomly — it sorted into four measurable trajectories, and baseline depressive symptoms predicted which one a person was on. Seneca argued something structurally identical in a letter to Lucilius roughly two thousand years earlier: nobody becomes good in a single stroke. It’s built incrementally, and where you’re starting from matters, but it isn’t a verdict.

What the Study Actually Did

The Americans’ Changing Lives study is one of the older workhorses of American social science — a nationally representative panel that first surveyed 3,617 adults in 1986 and kept returning to them for decades. Lyman, Ong, and Rottenberg used that depth to ask a question most well-being research can’t touch: not “how is well-being right now,” but “what shape does well-being actually take across a person’s adult life.”

They ran growth mixture modeling on repeated well-being measures — life satisfaction, self-acceptance, sense of purpose, relationships, mastery over one’s circumstances — across the full 25-year span. Growth mixture modeling doesn’t assume everyone follows the same curve. It looks for distinct subgroups within the data, each with its own trajectory shape, and sorts people into them based on how their scores actually moved over time.

Four trajectories emerged. Most people — roughly 60% of the sample — showed persistent well-being, staying on a stable path across the full quarter-century. The rest split three ways: a decelerated decline, a gradual increase, and a marked, unsustained increase — a sharp rise that didn’t hold. The detail that matters most: baseline depressive symptoms predicted which of the non-persistent trajectories someone was likely to follow, and it did this independent of how high or low their well-being happened to be at the start. Two people could start at the same well-being level and diverge for decades based on a variable that isn’t well-being itself.

ScienceAlert covered the findings under the framing that how you feel today can echo for decades — not as fate, but as a variable that shapes trajectory selection. Vice’s write-up leaned into the same point: this isn’t a study saying your mental health is fixed. It’s a study saying your starting conditions are doing more structural work than most people assume.

Seneca Already Made This Argument

Seneca’s letter — number 75 in his Moral Letters to Lucilius, titled “On the Diseases of the Soul” in most translations — opens with Lucilius apparently complaining that Seneca’s previous letters were too polished, too much like a performance and not enough like plain talk between friends. Seneca’s answer is essentially: fine, here’s the plain version. And the plain version turns out to be an argument about how moral improvement actually works, stage by stage, instead of a claim that anyone simply arrives.

He sorts people who are making progress — a state he calls prokopē, a term that just means “advancing” — into three rough classes. The first class has escaped the worst diseases and passions of the mind but hasn’t been tested yet; they don’t fully trust their own footing. The second class has “laid aside both the greatest ills of the mind and its passions” but isn’t in secure possession of that freedom — they could still slip. The third class is beyond many vices, especially the large ones, but not beyond all of them: someone might have conquered greed and still lose their temper, or escaped lust and still be gnawed by ambition.

None of these three classes is “done.” That’s the whole point. Seneca’s own text puts it this way: “Though he who makes progress is still numbered with the fools, yet he is separated from them by a long interval.” And on the people furthest along, not yet aware of how far they’ve come: “They have already arrived at a point from which there is no slipping back, but they are not yet aware of the fact.” Progress, in Seneca’s account, isn’t a threshold you cross and then stand on. It’s closer to a direction you’ve been moving in long enough that reversal has become unlikely — which is a structural claim, not a moral pep talk.

That’s the same shape Lyman, Ong, and Rottenberg found in the data, translated into growth curves instead of Stoic taxonomy. Not one universal path toward wellness or away from it. Distinct trajectories, distinguishable early, shaped by where a person starts — and, crucially in both accounts, not fixed by it. The Americans in the gradual-increase and marked-increase trajectories didn’t start from persistent well-being. They started from something closer to Seneca’s first or second class: further out, still capable of the long interval closing.

What Is Prokopē?

Prokopē (Greek, adopted directly into Stoic vocabulary) means “progress” or “advancement” — the state of moving toward virtue without having fully arrived. Seneca used it to describe people who had shed major vices but not all of them, distinguishing progress from perfection. It’s the Stoic term for exactly the space between “still a fool” and “sage”: a real, measurable, but incomplete state of improvement.

What Are the Four Well-Being Trajectories in the Study?

1. Persistent well-being — roughly 60% of participants, holding a stable trajectory across the full 25 years.
2. Decelerated decline — well-being that worsens over time, but at a slowing rate rather than a constant one.
3. Gradual increase — a steady, incremental climb in well-being across the study period.
4. Marked, unsustained increase — a sharp rise in well-being that doesn’t hold, eventually giving back some of the gain.

Where This Doesn’t Hold Up

This isn’t a case for reading a 25-year cohort study as a Stoic parable with better statistics, and neither Seneca nor the researchers would sign off on that reading.

Start with the obvious category error: Seneca was describing moral progress toward virtue — a philosophical, ethical claim about character. Lyman, Ong, and Rottenberg were measuring psychological well-being and depressive symptoms — a clinical, empirical claim about mental health. The two aren’t the same thing, even where the shape of the argument rhymes. Conflating “became more virtuous” with “reported higher life satisfaction on a survey” flattens a distinction that matters.

The cohort itself has limits worth naming. Americans’ Changing Lives began surveying its participants in 1986 — a generation with a different economic and social starting point than the one Cicero’s happiness-curve research described struggling right now. Whether the same four trajectories would emerge from a cohort that came of age with smartphones, gig work, and a much harder housing market is an open question the study doesn’t answer.

Growth mixture modeling also assigns people to trajectory classes probabilistically, not with certainty. It’s a statistical best guess about which curve a person’s data most resembles, not a diagnosis stamped on an individual life. And baseline depressive symptoms predicting a trajectory is a correlation across thousands of people, not a forecast for any one of them.

If what’s actually happening is clinical depression rather than a rough patch, that’s a case for a therapist or psychiatrist — not a letter written to a friend in the first century CE, however good its structural instincts turned out to be.

A Practice Worth Trying

Neither Seneca nor the researchers built this to be a five-minute fix, but the shared shape of their arguments points somewhere specific.

1. Judge the trajectory, not the day. A single bad week looks the same on a decelerated-decline curve as it does on a gradual-increase one. The study only found the difference by tracking direction over years. Before concluding anything from one hard stretch, ask what the trend has actually looked like, not just what today felt like.
2. Borrow the nightly accounting Seneca described elsewhere. Seneca’s De Ira describes a nightly practice of asking what fault you resisted today, what habit you’re still working on, in what way you’re a little better than yesterday. It’s the same incrementalism as prokopē — small, repeated, not asking for a finished result.
3. Don’t wait to feel “beyond” something before working on it. Seneca’s third class of people still had vices left — anger, ambition, fear — even after conquering the biggest ones. Progress doesn’t require clearing the whole list first.
4. Treat a low starting point as a condition, not a conclusion. The study’s most interesting finding is that people with low well-being and elevated depressive symptoms at baseline weren’t locked into decline — some of them were the ones who saw gradual or marked increases. Small, consistent connection tends to move the needle more than any single milestone, which is a modest, doable place to start rather than a dramatic one.
5. Write it down instead of estimating it from memory. A simple daily journaling practice is close to the only way an individual can track their own trajectory the way growth mixture modeling tracks a cohort’s — you need repeated data points to see a curve, not just today’s mood.

The Data Caught Up to the Argument

Seneca wasn’t running a longitudinal cohort study. He was writing to a friend who thought his letters had gotten too polished, working out — in plain, unpolished prose, as requested — why nobody arrives at goodness in one motion, and why the people furthest along a hard road often don’t recognize how far they’ve come. He sorted that argument into three classes because that’s what the structure of moral improvement looked like to someone paying close attention without a dataset.

Lyman, Ong, and Rottenberg built the empirical version from a completely different direction — 3,617 people, 25 years, and a modeling technique that didn’t exist in any form Seneca could have imagined. What it found wasn’t one well-being curve. It was four, sorted by a variable — baseline depressive symptoms — that shapes the path without writing the ending. This site keeps running into the same pattern: a Stoic gets the structure right first, centuries ahead of the study that eventually measures it. This time the structure is almost embarrassingly literal. Seneca said progress is made day by day. The data just took twenty-five years of days to show it.

---

This post discusses a 2026 longitudinal cohort study on well-being and depressive symptoms, and a classical philosophical text. It isn’t a substitute for professional guidance on mental health. If persistent depressive symptoms — not just a difficult stretch — are the actual issue, a therapist or psychiatrist can help in ways a two-thousand-year-old letter can’t.