A bestseller, a critic, and a reply

Matthew Walker's Why We Sleep, published in 2017, was a bestseller and widely praised,² and you may well have met its claims second-hand without knowing where they came from.

This course has not read it. That is deliberate. What this lesson has instead is two documents about it, both free and both online: a long critique by a writer called Alexey Guzey, mainly of the book's first chapter, with appendices on others; and a post by Walker answering questions readers had raised about the book "in reviews, online forums and direct emails".³

The post is not a reply to the essay. It does not name Guzey, and Guzey notes near the top of his essay that it is "explicitly not a response to my essay" and that Walker "never responds to them directly".² But several of its sections take up the same sentences the essay does, and those are the ones this lesson uses. On three sentences, you can put the critic's account next to the author's own written account and see where they meet. On some of it the author concedes in writing. So you can watch specific claims change shape on the way from a research literature into a paragraph, and then watch some of them change back.

This lesson reaches no verdict about anybody. It is not a takedown or a defence. It is about three sentences.

Three documents, and how far this course got into each

Neither document's underlying studies were read. When Guzey cites a review, or Walker cites a study, this lesson reports that they did, and does not report what the study found as though this course had checked it.

What each says about its own purpose explains a lot of what follows.

Guzey on his scope: "I hope that going through one full chapter, rather than cherry-picking stuff from across the book, demonstrated the density of errors in the book."² He reports spending "more than 130 hours" on it. His thesis is that "the book consistently overstates the problem of lack of sleep, sometimes egregiously so."² And his reason for holding a popular book to a scholarly standard is that the book, which he quotes, calls itself "intended to serve as a scientifically accurate intervention".²

Walker on his aim: "The aim of the book, Why We Sleep, is to provide the general public access to a broad collection of sleep research." And on the corrections: "I very much appreciate being made aware of any errors in the book requiring revision. Furthermore, and as an essential part of what I see to be good scholarly conduct, I have made a full set of corrections to the book. These corrections have now been published in a second edition, which is now in print."³ This course has not seen the second edition and can't say what it changed.

The first sentence: who said it

The book's sentence, as Guzey quotes it from page 4: "[T]he World Health Organization (WHO) has now declared a sleep loss epidemic throughout industrialized nations."²

Guzey's verdict: "This is false. The WHO never declared a sleep loss epidemic throughout industrialized nations." He reports that the book's footnote for that sentence points to a National Geographic documentary, that he "watched the entire 88-minute long film twice", and that it "never mentions the WHO".²

Walker's post: "The Centers for Disease Control (CDC) has stated that, "Insufficient sleep is a public health epidemic."" And: "The book's misattribution of the CDC statement to the WHO will be corrected in the next edition."³

So the two documents agree on one thing: the declaration wasn't the WHO's. Walker gives the CDC, a United States federal agency, as the source instead. Guzey reports, in an appendix, that a CDC page with that title changed "epidemic" to "problem" "More than 2 years before Why We Sleep was published".² This course has not checked the CDC's pages, so whether the CDC's statement still stood when the book was written is a disagreement it reports and can't settle.

Walker's post adds one more sentence, and it is a different, much more general claim: "Both organizations have stated that inadequate and disrupted sleep is dangerous for human health."³ He supports the WHO half with a quotation from a researcher he describes as working at the WHO. This course has not checked it.

The second sentence: what it claims

The book's sentence, as Guzey quotes it from page 3: "Routinely sleeping less than six or seven hours a night demolishes your immune system, more than doubling your risk of cancer."²

Guzey's verdict opens: "This is false. Walker does not cite any studies that support this assertion anywhere in the book." On causes, he adds: "There do not appear to exist any experimental studies or studies that would reasonably be able to establish causality, that would support this claim." And he sets against it a 2018 review which, in his words, "involved 1,550,524 participants and 86,201 cancer cases, found that neither short nor long sleep duration was associated with increased cancer risk."² This course has not read that review.

Walker's post: "Short sleep is by no means associated with a higher risk of all forms of cancer. Several studies, however, have indicated that short sleep is associated with a doubling of risk for specific cancers". He lists four, and then: "A meta-analysis published after the book was written, which examined 9 main forms of cancer (spanning a total of 20 subtype cancers), did not find an association between an all-cancer category and sleep duration." And: "While there is ready evidence associating certain cancers with short sleep, this study indicates that there are numerous cancers that do not show such an association. This will be reflected in future editions of the book."³

He goes further on the verb than the book did: "It is not correct to suggest, based on epidemiological findings, that sleeping less than 6 or 7 hours causes cancer." And the very next sentence is his case against the critic's reading: "However, one similarly cannot state that sleeping less than 6 hours a night does not double your risk of cancer. The same evidence cannot establish causality in the negative: absence of evidence in observational epidemiology does not necessarily imply evidence of absence." He also lists limits of the meta-analysis, among them that its authors said its results "cannot be generalized to all cancer types".³

This is the most useful exchange in the lesson, and it is worth doing slowly.

Notice where the two documents meet, and where they don't. Both say the evidence is observational and can't show that short sleep causes cancer. Both report that pooled evidence did not find an association with cancer as a whole; whether they mean the same meta-analysis, this course can't tell. They still disagree about what follows. Guzey reads the sentence as false. Walker reads it as too broad, with a narrower version supported by the studies he lists, and argues that finding no association is not the same as showing there is none. Those are two readings of the same evidence, and this course, having read none of the studies, does not choose between them.

That is the second shape a claim takes: an association in a literature becomes a flat statement about causes in a paragraph, and a scope that was specific becomes general.

The third: the missing column

This one has no agreement in it, and it is the harder case.

A 2014 study of adolescent athletes reported injury rates by how many hours they slept, and Walker gives its size as 112 people.³ The book, in its sixth chapter, printed an adaptation of its graph. Guzey, crediting the observation to somebody else ("this problem was first noticed by Olli Haataja"), put it like this:

"The 5 hours of sleep column – which is associated with lower chance of injury than 6 hours of sleep – has simply disappeared."²

He adds that the 9-hour column, which the book kept, "is based on exactly 1 child being injured out of 6 children who reported sleeping for 9 hours".² The appendix he put it in is headed "what do you do when a part of the graph contradicts your argument? You cut it out, of course", and elsewhere he calls it "the clearest example of deliberate data manipulation".² That is a charge about intent, and it is his.

Walker's account: "As shown in the original study, as the athletes' sleep amount progressively decreased from 9 to 6 hours, injury risk increased. This was depicted in a graphic in the book. The study also reported a group of individuals getting 5 hours of sleep a night. This group's injury risk was also high, though lower than the adjacent 6-hour group. However, noted by the authors in communication, the 5 hour group was of a size small enough to challenge robust estimate of injury risk, and was not represented in the book's graphic." And: "Related, the authors averaged the data for statistical power, comparing those getting less than 8-hours with those getting more than this amount."³

And his remedy: "To counter any confusion, including power issues at either end of the sleep duration distribution, and to better represent the statistical analysis of the study authors, future editions of the book will include a new graphic. This figure will depict the pooled injury risk of those getting less than 8 hours of sleep, relative to those getting 8 hours or more." He adds that the revised book will cite a 2019 meta-analysis that "confirms this original result".³

This course has not read the study or the meta-analysis, so it can't say how small the 5-hour group was.

That is the third shape a claim takes: a small group that runs against a trend drops out, and the pattern that is left reads cleaner than the data did. Whether that was a judgement about noise or a loss of information, or something worse, depends on facts about the study you'd have to go and check, and on a question about intent that no document here can settle.

Why each side does what it does

The lesson promised both cases at their strongest. Stated in general, not about this book, they run like this.

Why an author writing for everybody compresses. A book for a general reader can't carry every confidence interval, every subgroup and every hedge, or nobody reads it. Its job, in Walker's words, is "to provide the general public access to a broad collection of sleep research", and compression is what makes a literature readable at all. Some errors will come with that. The author's side of the case is that they should be corrected when found, which the post says has been done in a second edition.

Why a critic thinks the compression changed the claims. A claim that is been compressed can become a different claim: a cause instead of an association, all cancer instead of some, a clean line instead of a messy one. Guzey answers the popular-audience defence head on, calling it "indefensible", on two grounds: the book's own claim to accuracy, and that "in 2019, Walker published an academic paper that cited Why We Sleep 4 times just on its first page", so errors in the book travel into the research literature.² And he argues the errors weren't random: the book, in his words, "consistently overstates the problem of lack of sleep".² In the three cases here, each change did run in the book's favour. That is this course's observation, from the two documents' own accounts, and it does not tell you why.[4]

And the critic concedes more than a reader might expect. Guzey has an appendix of "things I'm not saying in this essay", and it includes "sleep is not important", "sleeping well is not important" and "there are no people who naturally need 8 hours of sleep a night".² The dispute is not about whether sleep matters.

What this does and does not establish

It establishes that the author accepts two of the three sentences need changing: the WHO attribution is wrong, and the cancer sentence is broader than he now says the evidence supports. On the third, the author and the critic disagree about what the graph should have shown, and this course can't settle it.

It does not establish anything against the research in this course's other lessons. None of the course's own sources is what either document disputes, and in fact each side leans on one of them: Guzey cites lesson 3's meta-analysis in his own support, and Walker's post cites lesson 2's experiment.²³ Guzey does add a caution that reaches lesson 3, that "Self-reported data is notoriously unreliable", which is lesson 1's point about instruments. "The book was criticised" and "sleep science is unreliable" are different sentences.

And it does not establish anything about Walker as a scientist or Guzey as a critic, or about either man's motives. Guzey's essay does make a charge about motive, calling the graph "deliberate data manipulation".² The post does not answer it, and this course can't judge it. It has read part of one critique and part of one post, and neither the book nor the studies.

Four things people get wrong

"The book was debunked, so the science is wrong." The dispute is about specific sentences in one book. The studies this course is built on are not part of it, and both sides cite some of them.

"The critic proved the claims false." On the WHO sentence, the author agrees. On cancer, the critic reports a review this course has not read, and the author narrows rather than withdraws, arguing that no association is not proof of none. On the graph, the two disagree about a study most readers haven't opened. "Argued, with evidence, and partly conceded" is accurate. "Proved" is too strong for the second and third.

"The author responded and corrected it, so the critic's case is answered." The post does not name the essay, and the critic says it never takes up his points directly. On the graph the two still disagree, and the charge of intent is unanswered. The corrections are in a second edition this course has not seen. A post that concedes some points leaves the others where they were.

"An author correcting a book means it was dishonest." Correcting a book is what an author is supposed to do when errors are found. Whether any particular error was careless, motivated or honest is a question neither document can settle and this course does not try to.

Practice

Connections

Back. Lesson 3's line between an association and a cause is the one the cancer sentence crossed, and the author's post crossed back. Note-Taking lesson 3 followed a famous study that travelled without its conditions; Memory lesson 4 traced a famous study to its source. This lesson does the same to a book, with the unusual help of the author's own written account.

Forward. Lesson 8 asks you to read the next claim you meet about your own sleep. The three shapes here are what to look for.

Go deeper

• Guzey's essay. Free. This course read the sections on the three sentences above, his introduction, and the appendices on the graph, on the WHO sentence's origin, on scope, on the popular-audience defence and on what he is not saying. The page has been edited since 2019, so check the date on the version you read. He links his own thoughts on Walker's post, and a response from UC Berkeley; this course read neither.
• Walker's post. Free. This course read the introduction and the sections on the WHO, cancer and the injury graph. It addresses many more points than this lesson covers, and reading it after the essay is the exercise above.

Sources

1. Matthew Walker, Why We Sleep, 2017. Not read. Every quotation from the book in this lesson reaches it through source 2, which gives the page number each time, and the body says so where each appears.
2. Alexey Guzey, "Matthew Walker's 'Why We Sleep' Is Riddled with Scientific and Factual Errors", guzey.com/books/why-we-sleep, created 2019, modified
   2025. Read in part, in its current version: the introduction, the sections on the WHO sentence and the cancer sentence, the passage on self-reported data, and the appendices on the injury graph, on the WHO sentence's possible origin, on why only one chapter was checked, on the popular-audience defence, and on what the essay is not saying. Supports: every quotation from the book, every quotation from the essay, and the report of the 2018 review and the 2014 injury study, neither of which this course read.
3. Matthew Walker, "Why We Sleep: Responses to questions from readers", sleepdiplomat.wordpress.com, 19 December 2019, updated 18 February 2021. Read in part: the introduction and the sections on the CDC and WHO, on cancer, and on sleep and injury risk. Supports: every quotation from the post. The studies it lists were not read, and the second edition it mentions was not seen. The post names no critic and is not a reply to source 2, which the body says.
4. The course's own constructions, labelled where they appear. The three shapes a claim takes on the way into a paragraph are this course's way of organising the three cases, not a finding. The possible routes by which an attribution moves are this course's speculation and the predict says so. The strongest cases for leaving the column out and putting it back are this course's statements of each side, built from the two documents. The observation that in all three cases the error ran in the book's favour is this course's, from the two documents' own accounts. The magazine, the loneliness sentence and the liver habit in the quiz are invented.