What books are for: a response to “Why books don’t work.”

Link post

...and a follow up to my recent post “Recommended Readings.”


Andy Matuschak’s 2019 essay “Why books don’t work” has been on my mind for a while now. I highly recommend you give it a read. The piece’s central claim is that books fail at their fundamental implicit task: conveying information to their reader. Let Matuschak speak for himself:

Picture some serious non-fiction tomes. The Selfish Gene; Thinking, Fast and Slow; Guns, Germs, and Steel; etc. Have you ever had a book like this—one you’d read—come up in conversation, only to discover that you’d absorbed what amounts to a few sentences? I’ll be honest: it happens to me regularly…
I know I’m not alone here. When I share this observation with others—even others, like myself, who take learning seriously—it seems that everyone has had a similar experience…

Why? Because

books have no carefully-considered cognitive model at their foundation, but the medium does have an implicit model. And like lectures, that model is transmissionism. Sequences of words in sequences of lines in sequences of pages, the form of a book suggests people absorb knowledge by reading sentences.

But transmissionism is basically false. Matuschak concludes that

as a medium, books are surprisingly bad at conveying knowledge, and readers mostly don’t realize it.

And the answer, he thinks, is not to change how we read but rather to replace books with a medium more conducive to understanding.

We agree so far

I basically buy the central thrust of Matuschak’s argument. A typical nonfiction book contains probably many thousands of individual statements and propositions, the vast majority of which the vast majority of us will never remember. And I agree that changes to the medium can help. For example, Spencer Greenberg’s excellent post “How You Can Gain Self Control Without ‘Self-Control’” includes built-in, interactive flashcards to help readers absorb and remember its content.

Books do work, sort of

Matuschak acknowledges that books aren’t useless*:*

I’m not suggesting that all those hours were wasted. Many readers enjoyed reading those books. That’s wonderful! Certainly most readers absorbed something, however ineffable: points of view, ways of thinking, norms, inspiration, and so on. Indeed, for many books (and in particular most fiction), these effects are the point.

But I do think that he gives short shrift to what seems to me books’ hidden function: making an idea or a set of ideas take up an appreciable amount of room on our cognitive real estate. And I don’t just mean for more abstract “points of view, ways of thinking, norms, inspiration, and so on” but also for plain old information, both uncontroversial statements of fact as well as more contentious claims or arguments

Against the ‘Light Switch’ model of knowledge

Matuschak’s essay implies something like a ‘light switch’ model of knowledge, and this probably works pretty well for, say, state capitals and other “trivia”-style facts. You know the tenth digit of pi, or you do not. The switch is on, or it is off. For so much else, though, a statement’s salience is more important than whether we technically do or do not “know” it.

An example

I recently finished The Precipice by Toby Ord. His headline claim is that humanity faces greater than a 15% risk of “existential catastrophe” in the next hundred years. Before downloading the audiobook, I already “knew” this statistic. In Matuschak’s model, the flip for this proposition was turned to “on” in my brain, in the same way that I “know” that Annapolis is the capital of Maryland.

Further, I have either already forgotten or will soon forget at least 95% of the book’s ancillary facts and arguments. Right now, I do remember Ord’s description of a few terrifyingly close nuclear calls during the cold war, that he thinks unaligned artificial intelligence constitutes about 10 percentage points of the 16.6% risk, and a few of his proposals to improve international coordination among other things. If I were to enumerate every single fact and argument I remember, though, I have no doubt it would be far fewer than 5% of all those in the book.

But most of the book’s impact comes not from flipping any of these “fact switches” to ‘on’ - it comes from making that “one in six chance of extinction” much more salient to me than it was before. How does it do this? I’m not a psychologist, but it seems likely that 99% of the substance of the book is basically a trick to get me to mull over a cluster of ideas for a while.

Before, I knew that a very smart person whom I admire and respect thinks that humanity is basically playing Russian Roulette. Now, I really know that a very smart person whom I admire and respect thinks that humanity is basically playing Russian Roulette.

This is important! Lots of people “know” that climate change is a big deal, but not enough people really know that climate change is a big deal. In more precise language, the salience of climate change is probably relatively low even for those who are seriously concerned about it.

To most of us, I’m guessing, COVID is a much more salient crisis. Perhaps this is appropriate, but I doubt it. What might increase the salience of climate change? Spending ~10 hours reading or listening to David Wallace Wells explain why (as the first line of An Uninhabitable Earth reads) “It is worse, much worse, than you think.”

Another one

The very first “serious” non-fiction book I remember reading on my own is A Hope in the Unseen, probably 12 or 13 years ago. Despite its 384 pages and likely tens of thousands of individual propositions, my entire memory of the book is little more than the single sentence a smart poor black kid struggles to get into college, but is finally admitted to Brown.

Under the ‘light switch’ model of knowledge, it would have been much more efficient to simply memorize that italicized sentence. In fact, I “know” more information in the book after skimming the Wikipedia page just now than I remembered from reading the whole thing.

But that’s not the point. Because the book probably took me ~8 hours of reading spread over three weeks or so, my brain was forced to process a single, coherent narrative for quite a bit of time. The sentence “a smart poor black kid struggles to get into college, but is finally admitted to Brown” and a maybe a few other vague associations with this sentence have taken up a real amount of my cognitive real estate for the last 12 years. Not much—I don’t walk around thinking about Cedrick, the protagonist, all day—but more than could have been achieved by reading the Wikipedia article no matter how recently.

Books are clandestine salience-building devices

My hypothesis is that—in agreement with Matuschak—the thousands upon thousands of bits of information conveyed in a typical nonfiction book don’t effectively transmit themselves to a typical reader, but they do force the reader to marinate in the book’s main few ideas for enough time for something important to happen.

Another lens

To put words in someone else’s mouth, I think Matuschak would say that, for the purpose of conveying information, it would be much more efficient to read a very short summary than to read an entire book. After all, we never remember more than the summary’s content, and generally remember much less.

I think this is incorrect.

After a few weeks, months, or years, a person who read the book won’t remember more than the short summary’s content, but the person who read the short summary won’t remember anything at all. Or, if the latter does remember something, the former will remember far more.

On one level, this is completely trivial. People remember more stuff if they learn more stuff, and a book has more stuff inside than a summary. Duh.

But the key point to remember is that (to a rough approximation), reading a book doesn’t convey any information not included in the summary. All the thousands of extra sentences and anecdotes and facts aren’t there to be remembered. They’re there to trick the reader into spending enough time mentally interacting with a few key ideas.

So, if you’re an author who wants to convey an actual set of, say, three or four main ideas, you could write a couple dozen articles or a couple hundred tweets and hope that a reader decides to read every single one. Or, you could write one book, filled with tens of thousands of quasi-redundant statements that will rapidly be forgotten, that you can reasonably expect a good proportion of initial readers to fully consume.


At a psychological or neuroscientific level, I don’t really know what’s going on, but I can speculate. We know that availability bias disposes us to treat the ease of recall of a memory as a proxy for its importance, and the quantity of recalled events as a proxy for actual frequency. That’s why it’s surprising that about as many Americans die from asthma in a single year than have been killed in terror attacks since 1970 (more than 75% on 9/​11).

Worth noting how similar the NYT and Guardian coverage is

Even for those in the intelligentsia who “know” that heart disease and vehicle accidents kill orders of magnitude more people than do terrorists, or that nearly two thirds of U.S. gun deaths are suicides and only 0.2% come from “mass shootings,” it is difficult to shed availability bias’s distorting influence. Mass shootings don’t intuitively seem like 0.2% of the problem. This isn’t to say that we should care about everything in exact proportion to its frequency or “objective” importance, but the two should probably be more closely linked than they are.

Back to books

Anyway, I think that books are basically mechanisms to leverage this availability heuristic. Even after reading that “mass shootings” (not sure how these are defined) constitute a tiny proportion of gun deaths, their disproportionate media coverage (perhaps for good reason—that’s not the point of this post) maintains their outsized share of our cognitive real estate.

The sheer length of time required to read a book is analogous to regularly watching violent crime reporting in the nightly news, albeit hopefully to a better end. You’ll rapidly forget 99% of the book’s statements just as you’ll rapidly forget 99% of the nuances of every news story. All those forgotten propositions are doing something important, though, for better or for worse: keeping you engaged with a few central ideas long enough for them to make an impact.