Reading is not one activity

You have a three-hundred-page book, forty minutes on a train, and a meeting tomorrow where somebody will ask what you make of it. What do you do?

Almost everyone does the same thing. They open it at page one and read until the train arrives, and they get to about page forty. Then they sit in the meeting knowing the first eighth of a book and nothing about the rest of it, including whether the first eighth was the part that mattered.

Mortimer Adler watched people do this and named what it costs. Such readers, he wrote in 1972, "start a book on page one and plow steadily through it, without even reading the table of contents", and so "are thus faced with the task of achieving a superficial knowledge of the book at the same time that they are trying to understand it."² Two jobs at once, and the harder one gets no room.

That's one option out of several, and it is usually the worst available. The problem isn't reading too slowly. It is that no decision was made, because nobody said there was one to make.

Four hundred years ago somebody wrote this down

Francis Bacon first published his Essays in 1597 and kept enlarging them; the 1625 edition is the last and fullest he published. One of them, "Of Studies", is about a page long, and it contains a sentence that people quote constantly and almost always cut in half:

Some books are to be tasted, others to be swallowed, and some few to be chewed and digested; that is, some books are to be read only in parts; others to be read, but not curiously; and some few to be read wholly, and with diligence and attention.¹

The half that gets dropped is the half after the semicolon, and it is the useful half. Bacon isn't being poetic about books. He is telling you three different things to do with your eyes and your hands. Read only in parts. Read the whole thing but not closely. Read the whole thing closely.

Notice what he has done. He has taken one word, reading, and split it into three activities that share almost nothing except the alphabet.

The four levels

The fullest working out of Bacon's idea is Mortimer Adler's How to Read a Book, first published in 1940 and revised in 1972 with Charles Van Doren. It is still in print and in most public libraries; Open Library's record will find you a copy. This course leans on it hard.

Notice this before we start: Adler quotes that Bacon sentence too, and he stops at the semicolon, exactly as everyone else does. The best reader of the last century did the thing this lesson opened by warning you about, which is a fair measure of how easy it is.

Adler names four levels, and he is careful about the word:

There are four levels of reading. They are here called levels rather than kinds because kinds, strictly speaking, are distinct from one another, whereas it is characteristic of levels that higher ones include lower ones.²

That distinction matters more than it looks, so here is the picture.

Each level has its own question, and the question is the thing to remember rather than the name.

Elementary reading asks "What does the sentence say?" This is the level you finished in primary school, and its question is about the words themselves rather than about what they claim. You are doing it right now without noticing it, which is what having mastered a level feels like from the inside.

Inspectional reading asks "What is this about?" Adler ties it to time: it "is characterized by its special emphasis on time", and its aim "is to get the most out of a book within a given time", and he finishes the sentence by conceding that the time is "usually a relatively short time, and always (by definition) too short a time to get out of the book everything that can be gotten".² Keep that last clause. Inspection is not a cheaper way of doing what a full reading does, and Adler says so himself. Other questions at this level: what is its structure, what are its parts, what kind of book is it. Lesson 2 is entirely this level.

Analytical reading asks what is being said in detail, and how, and in the end whether it is true. Adler is insistent about the order: the level is "preeminently for the sake of understanding", and judgement comes last, after eight of his fifteen rules. His pair of definitions is the clearest thing anyone has written about the difference:

If inspectional reading is the best and most complete reading that's possible given a limited time, then analytical reading is the best and most complete reading that is possible given unlimited time.²

He also connects it straight back to Bacon, quoting the same sentence and adding: "Reading a book analytically is chewing and digesting it."

Syntopical reading asks your question rather than the book's. You read several books against one problem, and the aim is something none of them contains. In Adler's words, "the syntopical reader is able to construct an analysis of the subject that may not be in any of the books". It's the last lesson of this course.

What actually sets the effort

The four names are only useful if something decides which one you pick. This course's answer follows, and it is ours rather than any one author's.

The effort a text demands is set by what you need to be able to do when you put it down.

That's a synthesis of the three sources, and they do not all say it. Bacon attaches the treatment to the book: his three kinds are kinds of book, and elsewhere in the essay he ranks them by worth. Adler brings in purpose, noting that analytical reading "is hardly ever necessary if your goal in reading is simply information or entertainment", and Carver's whole model is built on reading goals. Put them together and the decision moves from the shelf to the reader.

Test it against what each level leaves you holding. To say what a sentence says, you need that sentence. To say what a book is about, you need its shape: the claim, the parts, the kind of thing it is. To say whether its argument holds, you need the argument itself, rebuilt well enough that you could run it forwards and find the step you doubt. To answer your own question across four books, you need all four arguments and a vocabulary you built yourself in which they can be compared.

Each of those is a bigger thing to have produced than the last. That is why the same twelve pages can honestly take four minutes or four hours, and why neither number is wrong.

A psychologist tried to measure this, and half of it did not survive

Adler was a philosopher working from experience, and nothing in his book is measured. So it is fair to ask what happens when somebody does measure it. The answer is instructive, and it doesn't all go the lesson's way.

In 1992 Ronald Carver proposed that readers shift between five gears, each with its own speed: reading for recall at about 140 words a minute, reading for recognition at 200, reading out of interest at 300, skimming at 450, and scanning at 650.³ It is a tidy model, and if it held it would be Adler's levels arrived at from the psychology side.

It mostly doesn't hold. Marc Brysbaert's 2019 review, which pooled 190 studies and 18,573 readers, went looking for the gears and gave its verdict as a section heading: there is no evidence for reading gears except for the distinction between reading and scanning.³ Specifically, he found no difference between reading for recognition and reading out of interest, and no difference in kind between reading for recall and either of them.

One article, four ways

Take a twelve-page article on a change to local planning rules. Here is what each level leaves you holding.

Scanned, in about ninety seconds: you find the date the rules take effect and close the tab. You now know one fact and nothing else, and if that was all you needed, you are finished and the other three options would have been waste. Scanning is the odd one out here: it is not one of Adler's four levels, and strictly it's not reading. It is the one distinction the measurements did support, which is why it earns a place beside them.

Inspected, in about ten minutes: you read the opening and closing sections, the subheadings, and the first line of each part. You can say it argues that the new rules will slow small developments, that it's built as three case studies followed by a recommendation, and that it is written by a body with an interest in the outcome. You can also say whether it is worth the next hour.

Read analytically, in about an hour: you can state its claim in a sentence, lay out the argument from the case studies to the recommendation, say that the second case study does most of the work, and point out that it never addresses whether the delays it measures were caused by the rules or by something else happening in the same period. That last sentence is the one you couldn't have produced at any lower level.

Read syntopically, across this and four other pieces over a week: you can say what the actual disagreement is between the people who study this, which turns out to be about how the delays are counted, and you can say which measurement would settle it.

When the level changes halfway down the page

The awkward case, and the common one, is a document that needs two different treatments in different places.

Take a tenancy agreement you are about to sign. Most of it is wording you have met before, and there is one clause about who pays for damage that could cost you a great deal of money.

Reading the whole thing analytically is the conscientious-looking choice, and it means spending your attention at a flat rate across twelve pages of boilerplate. Reading the whole thing inspectionally is faster and skims the one clause that could hurt you. Neither is right, and the reason is that the level was chosen for the document when it should have been chosen for the passage.

So: inspect the whole agreement first, which tells you what is in it and where. Then find the clause that carries the risk and chew it. Bacon's three treatments are not three kinds of book. They are three things you can do to any number of words, including to two different paragraphs on the same page.

You have inspected the agreement and found the damage clause. It says the tenant is responsible for "any damage beyond fair wear and tear".

What people get wrong

"Reading properly means reading every word." Reading every word is one level out of four, and for most of what crosses your desk it is not the one the job needs. The reader who insists on it is not being thorough, they are declining to choose, and they pay for it by abandoning books they could have used.

"Skimming is cheating." Skimming done badly is casual browsing. Skimming done well has a procedure, a time budget and an output, and Adler treats it as a level of reading in its own right. Lesson 2 teaches it as a discipline with six steps.

"A summary is as good as the book." Bacon answered this in 1625, and he was not gentle. "Some books also may be read by deputy, and extracts made of them by others; but that would be only in the less important arguments, and the meaner sort of books, else distilled books are like common distilled waters, flashy things."¹ He is describing the book-summary industry, the executive briefing, and asking a machine what a book says. Note that he is also ranking books by worth there, which is his judgement and not a claim this lesson makes; the part to take is what a distillation costs you, not which books deserve one. The last lesson comes back to it.

"Faster is better." Speed is an output of the level you chose, not a skill you apply on top. Lesson 4 takes the claim apart with the measurements, including the parts of it that survive.

"The level is a property of the book." It's a relation between the book, you and what you need, which is why the same tenancy agreement gets two levels in two paragraphs.

Practice

Where this goes next

You now have the decision that comes before reading: what am I trying to be able to do, and therefore how much of this do I have to build?

Lesson 2 takes the level with the best return, inspectional reading, and turns it into a procedure you can run on a real book inside an hour. Lesson 3 explains why some books defeat you even when you choose the right level, which turns out to be a fact about what you already know rather than about effort. Lesson 4 does the numbers on reading speed and settles what speed-reading training can and cannot give you.

The analytical rules arrive in lesson 5, once you can survey a book well enough to know where its argument lives.

Go deeper

• Bacon's "Of Studies", in the Essays at Project Gutenberg. One page, public domain, four hundred years old, and it contains this lesson plus several arguments the course does not make. Search the page for "Of Studies".
• Adler and Van Doren, How to Read a Book, chapter 2, "The Levels of Reading", five pages, and the source of the four levels. Chapter 1 is worth it too, for the distinction between reading for information and reading for understanding. It's in print and in most public libraries; Open Library will find you a copy.
• Brysbaert's meta-analysis of reading rate, if you want to know now where the 300-words-a-minute figure came from and why it is too high. Lesson 4 uses it properly.

Sources

1. Francis Bacon, "Of Studies", in Essays or Counsels, Civil and Moral (1625), read in full at Project Gutenberg. The tasted, swallowed and digested passage with Bacon's own gloss, and the passage on reading by deputy and "distilled books". Both are quoted here in full rather than in the half-sentence usually seen.
2. Mortimer J. Adler and Charles Van Doren, How to Read a Book, revised edition (1972), chapters 1 and 2. The four levels; the sentence on why they are levels and not kinds; readers who "plow steadily through" a book; the description of inspectional reading and its emphasis on time, quoted to the end of Adler's own sentence; the pair of definitions contrasting inspectional and analytical reading; analytical reading as "preeminently for the sake of understanding"; Adler's own quotation of Bacon; and syntopical reading as constructing an analysis "that may not be in any of the books". Adler's separate views about which books are worth reading are not taught here. The text consulted for this course was a scanned copy at the Internet Archive; because that upload carries no licence and the book is in copyright, this lesson points readers at libraries and Open Library instead of at the scan.
3. Ronald Carver's five reading gears (1992), and the test of them, both from Marc Brysbaert, "How many words do we read per minute? A review and meta-analysis of reading rate", Journal of Memory and Language 109 (2019), 104047. Carver's own papers were not read for this lesson, so the gears, their figures and the account of where his 300 words a minute came from are all Brysbaert's reporting. Brysbaert tests the model and largely rejects it: his general discussion is headed "There is no evidence for reading gears except for the distinction between reading and scanning", and he reports failing to find a distinction between reading for recognition and reading out of interest, finding no bimodal curve that would have saved the theory, and finding that reading for recall is slower not as a separate mode but through "an increased structuring and organization of text information" visible as "many regressions and rereadings". On the 300 figure he writes that Carver's estimate "was influenced by Buswell and S. E. Taylor... but was also established independently", from Carver's earlier work on compressed speech. The same paper supplies the corrected rate of about 238 words a minute for silent reading of non-fiction, from 190 studies and 18,573 participants, describing adult native English speakers reading for comprehension and staying on task.