How to Read a Book for Understanding: in a World which Publishes as Though Every Book is Purely Entertainment

Or “What Andy Matuschak missed in Why Books Don’t Work is the lost knowledge of how to read a book.”

Song by Alexis
  1. The Multiple Uses of the Book Medium
  2. Informative Titles
  3. The 20 Minute Speed Read
  4. Climbing the Ladder of Understanding
  5. On Tables of Contents Including Many Asides About the Abusive Tables we are now Nearly Always Subjected to
  6. The First Full Read and Types of Notetaking
  7. How the Uses of Books Should Inform the Writing of Books

The Multiple Uses of the Book Medium

Books have been on the defensive since the first batch of Kindles sold out in 2007. Since then audiobooks and podcasts have exploded in popularity, and the internet not only provides millions of archived and public domain books ripe for download, but also creates opportunities for literary experiments and experiences which could not have happened otherwise. Text adventures, web serials, blogs on every subject under the sun – costs couldn’t be lower and opportunities to write have never been cheaper. Nonetheless the old codex format of pages between two covers still has much to recommend it. Traditional books are not obsolete, but I do believe we have forgotten how to approach books in particular amidst the information proliferation. And since we have forgotten how to approach books, publishers have stopped publishing books that are approachable. In the past dozen years, codex technology has not only failed to advance, but the knowledge of how to read and write a book has backslid.

One of the interesting things about physical books is their versatility. The form of a book lends itself to many different readings and interactions. For example, sometimes I read a book to quote mine or find an author’s opinion on a certain topic, other times to introduce myself to a new field, other times to read deeper in a field I’m already familiar with. Each of these goals means I will interact with the book in a different way. I can skim, flip through, read forward linearly, or even backwards – from a conclusion back towards the premises. I can single out tables and diagrams and read those, or jump right to the bibliography for a list of more works to read, or flip to the end-notes to discover a citation for some dubious claim. The Table of Contents should offer an outline of the book in miniature and a short study of the contents, should prime me for the meat of the work coming later. And, of course, the thickness of the sections provides quick intuitive information about how much I am missing when I skip around. The physical interaction encourages active reading and the static pages of the book allow the user to choose a reading style which fits best with his/her purpose. Today, in fact, I even read an index to get a handle on what the core vocabulary I need to master is. If I get lost in a sea of terms, I can refer to the index again to help guide me to the light. Okay that’s a big laundry list of things, but I will revisit and explain more fully in a moment.

Now admittedly, a digital book is better for quote mining and is equivalent to a physical book in a variety of ways, and superior to it in a variety of others. One disadvantage of the digital book, is how much harder to remember where in a work a particular argument was laid out or curious diagram printed. But the lightweight portable nature of the digital might offset those costs. If you would never engage with the work or have it on hand when needed otherwise, more power to the medium! There are trade-offs both ways, and I am not trying to convince anyone that physical books are better in every circumstance. Instead I am trying to recover a sense of what the medium of the physical book has to offer in a world of other options so that readers (and even writers) can decide what medium aligns best with their goals.

The principal problem, as I see it, is not the internet or audiobooks or the unwashed masses not appreciating the aesthetics of books, but the problem is how to read a book for understanding in a world which publishes as though every book is for entertainment. It may not be obvious that I am indicating any real problem, but I think I can demonstrate the issue with a simple test and some comparisons. Pick up a book that will challenge you, that you want to learn something from, the type of book you would read to develop a deeper understanding. Tell me what can you learn about a given book in 20 minutes? And how would that book be formatted if it were designed to maximize the knowledge gained in 20 minutes of interaction?

Informative Titles

The title should be informative enough to let you know the subject matter. Honestly, The Elephant in the Brain: Hidden Motives in Everyday Life may not be the greatest title, but at least it tells you that this work concerns psychology or neurology and “something no one is thinking or speaking about” with practical applications for life – not bad for just a simple title. It would be a shame if it were misleading.

The 20 Minute Speed Read

The Table of Contents should then outline the structure of the argument of the book. Mortimer Adler provides an excellent synopsis of the table of contents in his highly recommended How to Read a Book:

Study the Table of Contents to obtain a general sense of the book’s structure; use it as you would a road map before taking a trip. It is astonishing how many people never even look at a book’s table of contents unless they wish to look something up in it. In fact, many authors spend a considerable amount of time in creating the table of contents, and it is sad to think their efforts are often wasted.

It used to be a common practice especially in expository works, but sometimes even in novels and poems, to write very full tables of contents, with the chapters or parts broken down into many subtitles indicative of the topics covered… Such summaries are no longer common although sometimes you do come across an analytical table of contents. One reason for the decline of the practice may be that people are less likely to read the table of contents than they once were. Also, publishers have come to feel that a less revealing table of contents is more seductive than a completely frank an open one. Readers, they feel, will be attracted to books with more or less mysterious chapter titles—they will want to read the book to find out what the chapters are about. Even so, a table of contents can be valuable, and you should read it carefully before going on to the rest of the book.

How to read a book

I have been following Adler’s advice faithfully for years, and it has helped me learn more and retain more from my reading, as well as help me quickly go back and benefit more fully from having my memory jogged.

Perhaps at this point we are at minute 2 – 4 of our 20 minute tour of the book-to-be-understood. Now we read the preface, where the subject, general scope, and purpose are laid out. Read this quickly or even skim it if it is especially long. I find reading the first and last sentences of paragraphs to be a fast way to find the paragraphs which are crucial to me.

We are at minute 12-15 now. Flip through the book and sample some paragraphs or even a few consecutive pages to get a flavor the work, its density, its style, the challenges, and sensibility you will have to develop to appreciate it.

In the last few minutes, go for the total spoiler and read the final pages. Adler recommends that if there is an Epilogue, go to the pages right before the Epilogue. Usually an author cannot stop himself from summarizing what he believes to be the big takeaways at the end of the work. In any case, it is good to see where you are going to end up at the end so that the unity of the work can become clearer.

With that we come to the end of our 20 minutes and we should know a lot about our book. We should now know clearly the topic, the scope, and the basic skeleton of the work (think “head, shoulders, knees and toes” not “clavicle, acromion, coracoid”), the flavor of the text, and where the author wishes to take us. To some people, this might be a foreign and unromantic way to read, but it is rather a very involved and dedicated way to read. Yes, it is superficial. That’s sort of the point. To achieve this superficial overview required effort and attention, not merely glazed eyes scrolling over the pages. And the ultimate goal is an intimate knowledge of the book. Sometimes, even this superficial reading of a book, disabuses the reader of the notion that the book in question is worthy of deep reading. Perhaps the book contains only one core insight and several hundred poorly told anecdotes (On Grand Strategy likely qualifies). Sometimes a superficial reading reveals a superficial book.

Climbing the Ladder of Understanding

I remember in high school, we read at least one Shakespeare play a year. I wanted to like them, because I liked being challenged and I like language. My teacher recommended I read a summary of every play before reading it. So I purchased Shakespeare A to Z and read the summary of every Shakespeare play before I read the text. Then before each scene I would reread that scene’s overview from Shakespeare A to Z. I noticed my comprehension went up when reading the actual text, and Shakespeare became more and more enjoyable, until one day I could comprehend large swathes of unseen Elizabethan writing without need of a summary. This is an example of climbing a ladder of challenge toward understanding.

On Tables of Contents Including Many Asides About the Abusive Tables we are now Nearly Always Subjected to

One impediment to developing a deeper understanding and keeping clear memories of a work are the abysmal tables of contents produced today. Like Adler, I have noticed a seriously sad state in TOCs (Tables of Contents).

For example, the TOC for Superforecasting: The Art and Science of Prediction goes as such*:*

When I compare Tetlock’s TOC to my copy of Aristotle’s Politics… call me a clock — I am ticked about this TOC. Tetlock’s TOC is not useless, but it is a far cry from Adler’s ideal of a useful road map. In Tetlock’s defense, I’m sure this was an editorial decision—all popular science books are written this way now. Furthermore, while there were subheadings to each of Superforecasting’s chapters, they were not included in the TOC. I imagine this was not Tetlock’s fault. I don’t know, but my guess is that clean, minimal TOCs of exactly one page are publishers’ choice right now.

With this Table of Contents, I can tell you Chapter 1 is about Tetlock’s position. Chapter 2 is about uncertainty… no, wait, it’s about experts and his previous book Expert Political Judgement. Chapter 3 is how a scoring system works. You get the idea, but the problem with this, is that each of these chapters actually contains far more than I can quickly recall from seeing the chapter title. The subsections of each chapter would help immensely.

Benjamin Jowett’s TOC for Aristotle’s Politics stretches an immense eleven pages. Here’s the table of contents of just a part of Book 5. (For those who don’t know, in most editions of ancient works ‘Book’ is used in a way we might use Chapter, and chapters are just a few pages.)


Chapters 5—12. Revolutions in particular States, and how revolutions may be avoided.

5. (a) In Democracies revolutions may arise from a persecution of the rich; or when a demagogue becomes a general, or when politicians compete for the favor of the mob.

6. (b) In Oligarchies the people may rebel against oppression; ambitious oligarchs may conspire, or appeal to the people, or set up a tyrant. Oligarchies are seldom destroyed except by the feuds of their own members; unless they employ a mercenary captain who may become a tyrant.

7. (c) In Aristocracies and Polities the injustice of the ruling class may lead to revolution, but less often in Polities. Aristocracies may also be ruined by an underprivileged class, or an ambitious man of talent. Aristocracies tend to become Oligarchies. Also they are liable to gradual dissolution; which is true of Polities as well.

8. The best precautions against sedition are these: to avoid illegality and frauds upon the unprivileged; to maintain good feeling between rulers and ruled; to watch destructive agencies; to alter property qualifications from time to time; to let no individual or class become too powerful; not to let magistracies to be a source of gain; to beware of class-oppression.

Okay. Now this might be a bit excessive, but it is both useful before reading the work, and as a reference while in the weeds to see where the current section is going. A quick bird’s eye review of the table of contents gives the reader a context for understanding, for example, Aristotle describes Hiero of Syracuse use of secret police in Chapter 11. The TOC for 5.11 tells us that “Tyranny may rely on the traditional expedients of demoralizing and dividing its subjects” and here Hiero is an example one such tyrant, who kept his adversaries from coordinating by keeping them in fear. Astute readers easily see then how this example fits into the larger work of Book V, and even the larger vision of Politics.

Besides terse chapter titles with no subheaders, another problem one runs across in TOCs are totally coy titles. As one friend told me, “Often, even after I’ve read a modern nonfiction book, I can’t recall what a given chapter is about from the table of contents because they all have titles like ‘The Mouse and the Octopus’ or ‘How to Play Cribbage in a Boiler Room’.” I didn’t ask, but I think he had Antifragile: Things That Gain From Disorder in mind:

I appreciate that Antifragile has all of its subheadings included in the TOC. Some of them are quite useful, and I remember some of the sections therein quite well. Why do I remember some of these sections quite well, but not others? I suspect that the common cause, besides the punchiness of the writing which sometimes sticks, is their descriptive quality. Some of these subheadings, however, I just have no idea about. What was “France is Messier Than You Think” about? I vaguely recall the phrase “protesting as a national sport”. (In fact, I only remember the protesting as national sport thing, because I went to look up the book or article he was referencing and couldn’t access it.) Despite a less than perfect score on the table of contents, Taleb has the redeeming quality of plainly stating the thesis of his book at the beginning and again at the end in two different formats–verbal and mathematical. That he does this clearly improves an otherwise droll book sevenfold. A clear thesis provides a framework to his soup of spiteful words, amusing descriptions, and insightful lessons.

Douglas Hofstadter, known for his tyrannical control over each aspect in the production of his books, provides a very pleasant six page Overview immediately after the two page Table of Contents in Gödel, Escher, Bach: an Eternal Golden Braid. For those who have read it they will know, GEB is not really a book with a thesis, it’s an experimental work. The Overview, I think, makes that clear.

Take my word for it, though, most authors are not Douglas Hofstadter and will not try to maximize the information available to the reader for the purpose of the learning journey. So what can we do given the state of TOCs? Because most publishers prefer mystique, and the general public is willing to endure nearly useless Tables of Contents, one has an opportunity to engage with the book deeply and create your own Table of Contents at the beginning of the book on the blank page and in all that white space publisher left for you. Maybe something like this for Chapter 3 of Superforecasting:

3. Keeping Score

Ballmer’s Forecast on the iPhone – imprecise predictions can’t be assessed

a. “A Holocaust…Will Occur”

Predictions about Chernenko’s successor – hindsight bias rife among experts

b. Judging Judgments

Imprecise phrases like “very likely” and “serious possibility” – Sherman Kent’s Solution to numericize language – it was never adopted – The wrong-side-of-maybe fallacy – what calibration means — overconfidence and underconfidence – Brier Scores

c. Meaning of The Math

Brier Score Meaning depends on the Difficulty of Predictions

d. Expert Political Judgement

EPJ Program to assess expert predictions 5- 10 years out

e. And The Results…

Ideologues did worse – hedgehog and fox distinction – prototypical hedgehog Larry Kudlow and his recession denial – Foxes are more boring than hedgehogs

f. Dragonfly Eye

Sir Francis Galton and The Wisdom of Crowds – why crowds work – foxes simulate a crowd – Richard Thaler’s Guess the Number Game – using different perspectives yields more accurate guesses – Seeing poker through the perspective of the opponent – the dichotomy is a simplification, a mere model

Obviously, you can’t make a new table of contents off of the 20 minute fling we discussed earlier. You need to have read the work at least once. But if you do choose to make your own Table of Contents after you have read the book, then you are probably well on your way to a deep understanding of its material. But we are getting ahead of ourselves. We should talk about the first full reading of a work.

The First Full Read and Types of Notetaking

Through your first reading of a difficult text it makes sense to keep a brisk pace. Your goal should be to read all the parts you can understand *at your current level of knowledge.* Even though the work may be in your native language officially, if it is the type of work which is challenging you, then you need to read it as though it is a foreign language. Look for the key repeated terms and don’t worry if you do not understand something. Don’t slow down to work out the math section which is above your level, or to look up that arcane word which has something to do with naval batteries; read everything that is in the 75% comprehensible range and above. Slow down a bit when you have to, but avoid getting bogged down. This first reading, though still superficial, should provoke you to ask all the questions you need answered to make total sense of the text next time around. At this point you can go back and build your personal table of contents and then either embark on a closer, analytical reading to make sense of the work, or do whatever it is you need to do with the text.

Perhaps the most common way for a reader to take control of a text is through note-taking. The theory of note-taking, however, is a swamp of preferences and methods. Ideally each person uses a method which fits their context. What type of notes a reader takes should depend upon the reader’s expertise in the field AND on their purpose. This is why interactive learning platforms are so hard to create. Learners have disparate purposes and come in to a topic with different holes in their knowledge and understanding. For these same reasons, it is nonsense to say that there is a right or wrong way of taking notes in the abstract. Furthermore, some scream sacrilege about writing in books, others feel it is essential to making the book their own possession. Some prefer typing for its speed, others love baroque note-taking systems, like the Cornell method. Despite the diversity of methods and the idiosyncrasies of users, it is worth surveying five purposes of note taking and methods for going about it, so that readers can choose the method which suits their purpose best.

  1. Structural notes outline the sequence of topics covered by the work. This can be done in the margin or in a notebook. One can make a “key word outline” or key phrases. Seeing the structure should facilitate understanding the purpose of the arguments and descriptions.
  2. Substantial notes summarize the key arguments, descriptions, and examples in order. For this, one would want to identify important sentences or sections. Rewrite them, highlight them, or indicate them with a vertical line in the margin. The examples, descriptions, and specific arguments put flesh on the airy concepts and add meat to the otherwise bony structure.
  3. Conceptual notes paraphrase several takeaway ideas from the work in your own words. These are probably not written in the work itself but in a separate document.
  4. Critical notes include your emotional and intellectual responses to the key sections, core arguments, and general ideas of the work. This really should be done last. Of course, our temptation as intelligent readers is to prejudge based on what we already know. Understanding the author on his own terms is an essential goal. I have no fleshed out strategy for balancing the competing need to be both a discerning reader and a lenient judge (at least at first). More ideas welcome.
  5. Dialectical notes cross-reference passages from the work with similar or contradictory passages from other works you have read and even can cross reference previous ideas from the book in question. These notes are crucial when overviewing a broad topic and seeking to understand the shape of a wider conversation and not merely one author’s voice in it.

Note taking, I think, for most people is an annoying exercise. It requires much attention and effort and crucially takes longer than reading. Paradoxically, patience with note-taking takes time to develop, especially because it takes a long time to bear fruit. A decent rule of thumb is that the more invested a person is in mastery, the more time will be spent note-taking. While the conscientious may go overboard for fear of missing something, most need only assess to what extent they are reading for enjoyment, and then what type of notes to take becomes clear.

Marking enjoyable sentences, difficult passages, crucial arguments, and genuine insights is something one can do even when reading mostly for pleasure. Fun notes offer a sense of completion and something to show for your time.

How the Uses of Books Should Inform the Writing of Books

Books lend themselves well to use by people of all levels of expertise, from professionals to novices to dilettantes. A professional who is clued in to the larger conversation can mine through a book quickly and discover the interesting and unique insights, a novice can read slowly and digest each element of the work making notes and outlines and summaries, the dilettante can sample and read superficially, reading for pleasure sometimes and at others for a deep understanding. No group is slowed down by interspersed flashcards or interactive elements, which may be useful for some, but for others superfluous. As a medium, standard books offer significant optionality to readers, a freedom to choose when to slow down, speed up, when to stop to take notes, when to skip a section. These decisions can be be made quickly, easily, and sometimes even subconsciously. While audiobooks whisk listeners onwards for hours, books progress only at the rate of your processing. Studies on eye movement reveal the advantage of a medium which does not assume the manner in which a reader will engage with the information. Assuming we are not vetting readers for expertise, book design should offer accessibility to both professional and novice readers though they read differently. Experts navigate across the page differently and chunk information more efficiently. Despite these differences, a difficult book still should offer a gateway into a subject for the novice (learnability) and seamless navigation for the expert (discoverability).

I have formulated a few ideas explaining what publishers already do and what they should do to improve the medium further. Some of these ideas transfer to long form online articles as well, and if I put together a website this year I am now on the hook to practice what I preach, otherwise you have permission to harangue me with strongly worded emails.

Authorship 101 says that a book needs a definite and discoverable structure. We’ve already talked about this with tables of contents. However, it is important to remember as well, that reading even a detailed skeleton of a work is not the work itself. If a work is all skeleton, then there will be too much room for abstract misinterpretation or the evaporation of the ideas into meaningless platitudes; there needs to be some meat, specific arguments and examples and anecdotes. However, the ratio needs to be right. Too much meat and we rightly call it fat.

Books require cues which remind readers of their location within the conceptual territory of the work. The chapter titles and or section titles restated at the top of the page, page numbers in the bottom, or even paragraph numbers at the beginning of paragraphs (which make citing nonfiction way more convenient across platforms. This is one of the pleasant things about ancient classical works for example, Republic 514a always refers to the allegory of the cave paragraph in Plato.) these all serve to contextualize the page. Footnotes give the reader assurances while reading, and can help readers generate further inquiries quickly. Guns, Germs, and Steel: The Fate of Human Societies lacks footnotes and has such lackluster citations in the endnotes that I found myself growing more and more suspicious of the narratives as I read.

The Landmark History Series does everything in its power to make old historical works accessible to novice readers of classical history as well as experts. Here is the explanation of the method from the series editor, consider this whole block as though it were bold and italicized. It’s that worthy of emphasis:

Text features in Landmark editions are designed to assist the reader including side notes which are found on the outside page margin at the beginning of the chapters into which the ancient text was divided long ago by Alexandrian scholars. Normally, the first two lines of the side note display the book and chapter number and the date (if known or applicable). The third line shows the location of where the action takes place (or in some cases, a topical title). Finally, there is a summary description of the contents of the chapter. Each chapter contains section numbers in square brackets, such as [2] to mark the divisions into which scholars have traditionally divided the text for ease of search, analysis, and discussion. Running heads are placed on the top of each page of the book which at a glance provide date and place and a brief summary of the action of the first complete chapter on the page. Footnotes not only refer place-names in the text to nearby maps, as mentioned above, but they may serve to connect certain points in the text to other relevant sections, or to the work of other ancient writers and poets. They also cite particular paragraphs in the Introduction or in one or more of the appendices where the reader will find discussion of the topics or events footnoted. On occasion, they provide background information that does not appear in any of the appendices. They may also point out and briefly describe some of the major scholarly controversies over interpretation, translation, or corruption of the text. A few explanatory footnotes are quite long and detailed, but they contain important information which could not be further condensed. Footnotes and map data are repeated throughout the work to assist those who will read only selections from it, or whose reading of the text is discontinuous.

The result of this editorial care is a historical series which is deeply informative. Twenty minutes with any book in this series always lends itself to progress in understanding. And this should be our goal when putting together written works – to make our medium serve as many readers as possible.

Other innovations can and should be developed, especially for presenting long essays online. Gwern’s long essays are probably the best I have seen formatting wise. Wikipedia is acceptable. The New Yorker’s website does a poor job providing a sense of place to the reader. Audiobooks are necessarily abysmal (frequently Audible does not even include the subsections or chapter titles in their navigation pane). But when it comes to the physical book the Landmark Series is the best I know.

Realist history in one beautiful volume

A beautiful volume inviting the reader to master it.