← Thoughts

In the last fifteen years, English picked up a vocabulary it didn’t have before.

Toxic, which used to mean chemicals, started meaning relationships. Attachment, which used to mean legal documents, started meaning Tinder bios. Depression, which once meant a worldwide economic catastrophe, became a clinical category that has been steadily climbing for forty years. Newer phrases arrived to do newer jobs: executive dysfunction, intrusive thoughts, dysregulation. Older clinical terms quietly retreated.

This is a piece about that vocabulary. About what it looks like when you measure it, in three different ways, against three different sources. Google Books for the long view. Google Trends for the recent acceleration. CDC and SAMHSA for what was happening to the people who were saying these words.

The argument is short. The vocabulary is real. So is the distress underneath it. Both have been climbing, and neither alone explains the other.

Below: seven scenes. Three case studies, a wider grid, a methodological caveat, a data overlay, and a coda. Ten minutes of reading.

Essays · Vocabulary

Toxic.

What this word meant in 1990, and what it means now.

Words don’t expire. They get reassigned. A word can spend a century pointing at one thing in the world, peak, drift, and quietly take a job pointing at something else. Sometimes something nearly unrelated. The shape of that drift is hard to feel from inside a single decade. It’s easier to see on a graph.

Take toxic. For most of the twentieth century the word lived in chemistry textbooks and EPA reports. By the late 2010s it had moved into the language of breakups, group chats, and HR memos. The reassignment happened in a single human lifetime.

Scroll on.

Original / New
Toxic — relative use in published books SCENE 1 / 3

Each curve is normalized to its own peak. In absolute book frequency, the original meaning still leads — toxic waste and toxic chemicals together outnumber the relational phrases by roughly 22-to-1 across the full record. But the trajectories tell the story. Source: Google Books Ngrams, English corpus.

Scene 01 — 1990

For most of the twentieth century, when a book said toxic it meant poisonous. Industrial. Chemical. The word peaks in print in 1990, at the high-water mark of an era of public anxiety about contamination: Love Canal, the Superfund Act, the daily television vocabulary of leaking drums and red tape.

The blue curve is toxic waste and toxic chemicals, summed.

Scene 02 — 2010s

Now watch the second curve appear. Same word, same corpus, new referent: toxic relationship, toxic people, toxic family. Through the 1900s it sits near zero. It does not begin to climb until the late 1990s. By 2010 it is steep. By 2019 it has nearly caught up to its own peak.

The rust curve is toxic relationship, toxic people, toxic person, toxic family, summed.

Scene 03 — Both

Side by side, the two curves trade places. The chemical meaning is in retreat from its 1990 high. The relational meaning is on a vertical climb. By 2015, the relational meaning is rising as steeply as the chemical meaning is falling. The trajectories have crossed even though the volumes haven’t.

The word didn’t lose its old job. It just took a second one.

Coda

The same word, two unrelated referents. The word didn’t expand. It changed jobs.

Toxic is the cleanest case in the data. One word, one century-long job, full reassignment in fifteen years. Attachment is messier. The original meaning is still alive in the corpus, retreating slowly since the 1980s, while a new meaning born in psychology climbs into the space it leaves.

Essays · Vocabulary · No. 02

Attachment.

What this word meant in 1955. What it means now.

Some words don’t get replaced. They get demoted. A word can stay in the language while its job description shifts under it; what it pointed to in your grandparents’ decade is not what it points to in yours.

Take attachment. For most of the twentieth century the word lived in legal documents and emotional shorthand. “Writ of attachment.” “A strong attachment to one’s country.” By the 2010s it had become a category on dating-app bios and a four-question Instagram poll.

Scroll on.

Original / New
Attachment — relative use in published books SCENE 1 / 3

Each curve is normalized to its own peak. In absolute book frequency, the original meaning still leads, but the trajectories tell the story. The asymmetry by 2019 is roughly 2-to-1 — gentler than toxic’s 22-to-1, and the changeover is still in motion. Source: Google Books Ngrams, English corpus.

Scene 01 — The plateau

For most of the twentieth century, when a book said attachment it meant something legal, financial, or sentimental. The frequency plateaus at high volume from the 1920s through the 1980s, supported by court decisions, biographies, and ordinary descriptions of human warmth. Then, around 1985, it begins a slow and steady decline.

The blue curve is attachment of, writ of attachment, by attachment, warm attachment, personal attachment, and strong attachment, summed.

Scene 02 — The clinic

Now watch the second curve. Same word, same corpus, naming something almost entirely different. Bowlby first wrote about attachment in mother–infant pairs in 1958. Ainsworth’s Strange Situation experiment ran in 1970. Hazan and Shaver moved the framework to adult romance in 1987. For thirty years it lived in journals. Then it didn’t.

The rust curve is attachment style, attachment theory, anxious / avoidant / secure attachment, and attachment styles, summed.

Scene 03 — Both

Both meanings sit on the same word. The legal usage is in slow retreat. The relational usage is on a steady climb. By 2019 the rust curve has reached its highest point in the record, with most of that growth concentrated in the last fifteen years.

The reassignment isn’t finished. It’s in motion.

Coda

The clinic, in 1958, was a small room. Today the clinic is the whole conversation.

The clinical meaning of attachment took thirty years to leak from psychology journals into common speech. Depression is older and runs the other direction. The word did not start in the clinic and migrate outward. An external event picked it up — for a single decade, large enough that the word meant only that. Fifty years later, the clinic found the word and brought it back inside.

Essays · Vocabulary · No. 03

Depression.

What this word meant in 1934. What it means now.

Some words don’t drift. They get borrowed by an emergency, then returned. Then, decades later, an entirely different field shows up to claim them.

Take depression. For a single decade in the middle of the twentieth century, the word meant a worldwide economic catastrophe. After the war ended, economists handed it back to ordinary speech. By the 1980s, a different profession, psychiatry, had picked it up and was carrying it forward.

Scroll on.

Original / New
Depression — relative use in published books SCENE 1 / 3

Each curve is normalized to its own peak. In absolute book frequency, the 1934 economic spike is roughly an order of magnitude larger than any year of the clinical curve — but the two curves operate at completely different timescales, and one is event-driven while the other is structural. Source: Google Books Ngrams, English corpus.

Scene 01 — The 1934 spike

The 1934 spike is one of the sharpest single-decade vocabulary events in the English corpus. The Great Depression turned the word into the dominant frame for a global economic catastrophe. Books, newspapers, and government documents adopted it inside a couple of years. Then the war came, and afterwards economists quietly migrated to a softer word: recession. The economic meaning of depression has been in slow decline ever since, kept alive mostly by historians and reference works.

The blue curve is great depression, the depression, economic depression, and depression era, summed.

Scene 02 — The clinic arrives

Now watch the second curve. It barely registers until 1980, when the DSM-III formally defined Major Depressive Disorder as a discrete diagnosis. In 1987, Prozac arrived. Through the 1990s and 2000s, direct-to-consumer pharmaceutical advertising made the clinical vocabulary mainstream. The curve is still climbing in 2019.

The rust curve is clinical depression, severe depression, depression treatment, depression symptoms, and postpartum depression, summed.

Scene 03 — Taking turns

Two different forces, on the same word, in non-overlapping eras. The economic meaning was event-driven: one global crisis that owned the language for a decade. The clinical meaning is structural: a slow buildout of diagnosis, treatment, and language, compounding for forty years and still going.

Not a handover. A relay.

Coda

Some words get reassigned. Some get demoted. Others get borrowed by one century and conscripted by the next.

Three words. Three mechanisms. Toxic, attachment, depression — each with its own dates, institutions, and driving culture. But three is a small number, and the same things are happening to many more. The grid below pulls back to a wider slice, all at once.

Essays · Vocabulary · No. 04

Everywhere.

Three case studies, three mechanisms. The same thing is happening across the language.

Three words don’t make a pattern. Three thousand might.

The case studies were specific by necessity. Step back from individual cases and the patterns repeat. Below: the same kinds of shapes, in the same time windows, on words we haven’t discussed. Each panel is unannotated. The grouping is the argument.

Look down the rows.

Born in the clinic.

Row 1 · 5 panels

Words that didn’t exist in books before about 1990. Now they do.


Old words, new jobs.

Row 2 · 4 panels · two curves each

Words that already meant something else. Each panel shows the original meaning declining (blue) and the new meaning rising (rust).


Words that left the conversation.

Row 3 · 3 panels

Words that peaked earlier in the twentieth century and have been declining ever since.


The small charts above are not the whole vocabulary. Just twelve panels of it. But they’re representative. Pick another five clinical-origin words and they will mostly look like the top row. Pick another five faded words and they will look like the bottom row. The middle row is the strange one — the row this piece is about.

Something is happening across the language. Not in three words. In hundreds.

Coda

The case study is a useful trick. It can also hide how much else is happening.

What the case studies and the grid share is a single source: Google Books. Books are a fine corpus for the long sweep of English, a poor one for this year’s vocabulary. They are slow. They are written, edited, indexed, and shipped years after the conversations they try to capture; they cannot show what is happening in the language right now. The lag shows up clearest on three words.

Essays · Vocabulary · No. 05

What books haven’t seen.

Three words that exploded online while the books we read were still catching up.

There is a problem with reading books to understand what people are saying.

Books are slow. They are written, edited, fact-checked, designed, printed, distributed, and indexed. By the time a word appears in a single book, it has usually been alive in the language for years. To register in a corpus the size of Google Books Ngrams, it has been alive for decades.

What we have been showing in this piece is the lagged signal. The actual phenomenon is bigger and faster than the books say.

Red flag.

In books, the dating sense of red flag barely shifts the overall frequency, which is dominated by gun policy and historical usage. In search interest, the term roughly quadruples over the same window — and continues to climb after the books data ends.

Books, 2010–2019Search interest, 2010–2025

Gaslighting.

Books picked up gaslighting around 2014 and showed steady growth through 2019. Search peaks in 2022, the year Merriam-Webster named it word of the year, far above anything in the books window. That peak happened entirely in the years books couldn’t see.

Books, 2010–2019Search interest, 2010–2025

Neurodivergent.

Books and search disagree about when this word arrived. Books show a steady climb from around 2014. Search shows almost nothing until 2020, then an explosion. The two data sources tell different stories about the same term, separated by roughly six years.

Books, 2010–2019Search interest, 2010–2025

Coda

Everything in the case studies is real, but undercounted. Whatever is happening, it is bigger than the books say it is.

That undercounted vocabulary landed against a measurable rise in adolescent depression, in suicide rates among adolescent girls, and in the share of high school students reporting persistent sadness. One chart, then, putting the rising vocabulary against the rising data, and asking what it means that they have been climbing together.

Essays · Vocabulary · No. 06

Real, and rising.

The vocabulary is rising. The data is rising. The two trajectories share a curve.

The piece has been about words. Here, finally, are the things the words are pointing at.

Below: the rising therapy-speak vocabulary on one curve, and three independent measures of adolescent female distress on three others. The Centers for Disease Control’s suicide rate for ages 15 to 24. The National Survey on Drug Use and Health’s rate of past-year major depressive episodes among adolescent girls. The CDC’s Youth Risk Behavior Survey, asking high school students whether they have felt persistent sadness in the past year. Three agencies, three methodologies, three sources of error, all pointing the same direction.

Read the rust curve against the blue ones.

Vocabulary & reality — four curves, each normalized to its own peak 2005 – 2023
Vocabulary (books) NSDUH — female MDE % YRBS — female persistent sadness % CDC — female 15–24 suicide rate

Each curve normalized to its own peak. Vocabulary aggregate is the sum of five clinical-origin phrases (PTSD, executive dysfunction, intrusive thoughts, panic attack, dysregulation) and four recolonized-meaning bigram groups (toxic relationship, attachment style, clinical depression, shadow work and related). Distress measures are sex-specific (female adolescents) where available because the rises are most pronounced in that group. Books data ends 2019; CDC suicide rate through 2022; NSDUH through 2022; YRBS through 2023. Sources: Google Books Ngrams; CDC NCHS Data Brief 509 (2024); SAMHSA NSDUH (2004–2022); CDC YRBS Trends Report (2009–2023).

Coda

The words did not arrive in an empty room. They arrived into a real and rising distress. Once it had words, the distress became more nameable. More shareable. More felt.

One more page.

Essays · Vocabulary · Coda

Both.

What the seven scenes add up to.

You have read about the word toxic, peaking in 1990 when it meant chemicals and recolonized after 2010 when it meant relationships. About attachment, where the legal usage retreated as the clinical took over. About depression, borrowed by an economic crisis in the 1930s and reclaimed by psychiatry fifty years later. About a grid of other words doing the same things. About the gap between what books saw and what people were searching for. And about the rising mental-health curves climbing alongside the rising vocabulary, each making the other visible.

Coda

Here is the piece, in one sentence. There is a rising vocabulary. There is a rising distress. They have been building each other.