Two facts arrived in my inbox this month that cannot both be true — or rather, cannot both mean what they seem to mean.
The first: independent bookstores are multiplying, book sales are at record highs, and the amount of time American adults spend reading has not meaningfully changed in decades, smartphone or no smartphone.
The second: the same college professors buying those books report that freshmen arrive able to sound out every word on the page and unable to say what the page argues.
If reading were dying, the bookstores would be closing. They are not. And if reading were fine, the professors would not be reaching for ChatGPT jokes to describe their students. They are.
This week I want to press on a sharper and more uncomfortable claim: that the crisis is not happening at the border between literate and illiterate. It is happening inside people who can already read.
The loud version of the story is the eulogy. Rose Horowitch's Atlantic cover, "The End of Reading Is Here," announces that America is going "post-literate" — still surrounded by words, steadily losing the capacity for the deep, sustained reading that once shaped everything from private thought to public life.
The numbers she marshals are genuinely grim. Fewer than half of U.S. adults read a book of any kind in 2022. The share who read for pleasure on a given day fell from 28 percent in 2004 to 16 percent in 2023. Gambling, last year, was reportedly more popular than reading.
But David Wallace-Wells offered a rejoinder that quietly dismantles the eulogy's premise. Adult reading time, he points out, has barely budged in decades — except among older people, who read less — and Americans were never especially avid readers to begin with.
Here is the fact that resolves the paradox and reframes the whole debate: last year, roughly 20 percent of adults accounted for more than 80 percent of all books read. The crisis, in other words, is not a decline in aggregate reading. It is distributional, and it is qualitative.
The crisis is not a decline in aggregate reading. It is distributional, and it is qualitative — happening inside people who can already read.
Which means "post-literate" is the wrong word for what's actually breaking. The panic misnames its own emergency.
Consider what the ground-level literacy data shows. According to the 2023 PIAAC survey, 21 percent of U.S. adults — more than one in five — read below a fifth-grade level and are classified as functionally illiterate. Fifty-four percent function below a sixth-grade level.
That is a real and severe problem, but it is a decoding problem, and it is not new. Notably, two-thirds of those functionally illiterate adults are native-born, not immigrants — which tells you this is a failure of instruction, not a story about who is arriving at the border.
The comprehension collapse is a different animal entirely, and it lives one floor up. It afflicts people who cleared the decoding bar years ago. Students at elite universities, John Ambrosio writes from his years teaching them, entered his classroom able to decode every word "but usually without the cognitive skills needed to comprehend, analyze, and synthesize complex text."
That is the load-bearing distinction the "post-literate" frame keeps blurring. Decoding — converting marks into words — and comprehension — converting words into arguments — are not the same skill, and being fluent at one guarantees nothing about the other.
In the work of Kara Federmeier and colleagues on the architecture of reading comprehension, meaning is, reportedly, not extracted from a page; it is constructed by the reader — inference, integration, the active building of a model of what the text is claiming.
This is the crucial point the declinists and the shruggers both miss. Comprehension is, reportedly, constructive and inferential, not automatic. It is built, and being built, it can reportedly be un-built — atrophying when it is never exercised on anything demanding.
And here is where the symptom everyone points at fits in. College students now turn to ChatGPT to "translate" dense texts they can technically read but cannot process.
I want to be precise about that, because it is easy to blame the tool. The AI is, per this argument, not the cause of the comprehension gap; it is a symptom of one that predates it. The students reaching for a translation are revealing that the constructive machinery — the part that turns a hard paragraph into a grasped idea through effort — was never built, or has gone slack.
What did build it, and what stopped? The attention data is the clearest tell. The average attention span on a screen reportedly fell from 2.5 minutes two decades ago to 47 seconds five years ago.
Comprehension of a complex argument requires holding many things in mind at once across minutes of sustained effort. Forty-seven seconds is not enough runway to construct anything. You cannot infer, integrate, and abstract in bursts the length of a video clip.
The neuroscience principle underneath all this is banal and unforgiving: you get good at what you practice. A brain that spends its reading time scrolling short-form text and video gets very good at scanning and very bad at sustaining.
Notice that this indicts the complacent position as much as the panicked one. "People are just reading in new forms" is true and beside the point — the new forms train a different and lesser skill. Form is not neutral to what the mind becomes.
So both loud responses fail. The eulogy is wrong on the facts, because reading isn't disappearing. The shrug is wrong on the stakes, because the specific capacity being lost — sustained, inferential, abstract comprehension — is the one that makes an argument legible rather than merely audible.
Which leaves a more demanding and more hopeful conclusion. If comprehension is a constructed skill and not a natural endowment, then it can be taught, and re-taught, deliberately.
Ambrosio, facing students who could decode but not think, did not lower the bar or abandon books. He reconceived his job: teaching students to distinguish personal opinion from evidence-based argument, to assess truth claims against credible sources and coherent reasoning — comprehension as an explicit curriculum, not an assumed prerequisite.
Robert Douglas-Fairhurst's Look Closer, distilling more than twenty years of teaching literature, is essentially a manual for the same project: the deliberate techniques of slowing down, taking note, and bringing a text to life. Deep reading treated as a practiced art with actual methods, not a mood you're either in or you aren't.
And Peter Schmidt's answer runs alongside it: build communal book culture, the reading group and the shared argument, precisely because young people who came of age in lockdown crave the human connection of asking questions and thinking together. Comprehension gets exercised best out loud, with others, on a book no one can finish in forty-seven seconds.
The correctives, in other words, already exist. What has to change first is the concept.
Stop treating literacy as a binary — can read, can't read — that a bookstore boom can reassure us about. And stop treating deep reading as either a dying folk practice to be mourned or an old habit painlessly replaced by new ones.
Treat it instead as what Federmeier's architecture and Douglas-Fairhurst's masterclass are reportedly said to describe: a specific, effortful, trainable cognitive skill with a syllabus.
The bookstores are full. The readers are decoding fluently. The question the post-literate panic keeps failing to ask is the only one that matters: once the words are sounded out, is anyone still building the argument they were meant to carry?