Deep Reading: What It Is, Examples, and How to Do It

What Is Deep Reading? Meaning and Definition

Deep reading is slow, attentive reading aimed at understanding a text rather than just getting through it. A deep reader draws inferences, connects the text to what they already know, tests the author's argument, and reflects on what it changes for them. Skimming tells you what a text says; deep reading tells you what it means, and whether it's right.

The literary critic Sven Birkerts coined the term in this sense. In The Gutenberg Elegies: The Fate of Reading in an Electronic Age (1994), he wrote, "the term I coin for this is deep reading: the slow and meditative possession of a book," and he worried that electronic media would push it aside. The cognitive neuroscientist Maryanne Wolf, who has directed UCLA's Center for Dyslexia, Diverse Learners, and Social Justice, gave the idea its scientific shape in Proust and the Squid (2007), in a 2009 article with Mirit Barzillai, and in Reader, Come Home: The Reading Brain in a Digital World (2018).

Wolf and Barzillai's definition is the one researchers cite most often: "By deep reading, we mean the array of sophisticated processes that propel comprehension and that include inferential and deductive reasoning, analogical skills, critical analysis, reflection, and insight." Every item on that list is something the mind does with the words once it has them, and none of it is about speed or vocabulary.

The Characteristics of Deep Reading

Put Birkerts and Wolf together and deep reading has six recognizable features:

• Inference: You read between the lines, working out what the author implies but doesn't say.
• Background knowledge and analogy: You connect the text to what you already know, and notice when it resembles something from another field.
• Critical analysis: You weigh the evidence, spot the assumptions, and ask what's missing.
• Perspective-taking: You temporarily inhabit someone else's point of view, which is why Wolf counts empathy among the deep reading processes.
• Reflection and insight: You form a view of your own, which may differ from the author's.
• Time: You give these processes room to run. They're the slower part of reading, so they're the first thing cut when you hurry.

What Deep Reading Is Not

Deep reading isn't speed reading. It isn't skimming for key points or scanning for the one fact you need. Those are legitimate reading modes with their own uses, but they skip most of the list above. It isn't quite the same as slow reading either, though the two overlap: slowing down gives the deeper processes time, but pace alone doesn't make reading deep. Deep reading needs time, attention, and a willingness to sit with difficulty.

The practice is older than the name. The 1972 edition of How to Read a Book, by Mortimer Adler and Charles Van Doren, calls it "analytical reading," the third of four levels: the point where you stop asking only what the author said and start asking whether the author is right, and why.

---

Deep Reading Examples: What It Looks Like in Practice

The clearest way to see deep reading is to compare two reads of the same text. The difference shows up in what you can do after you close the page:

A Worked Example

Take one famous sentence from Francis Bacon's essay "Of Studies" (first published 1597, revised 1625): "Some books are to be tasted, others to be swallowed, and some few to be chewed and digested." Each deep reading move pulls something different out of it:

1. Literal meaning: Books deserve different amounts of attention.
2. Inference: Bacon is describing a way of reading rather than ranking books by quality, and his next clause confirms it: some books "are to be read only in parts."
3. Connection: Tasting, swallowing, and chewing map neatly onto skimming, ordinary reading, and Adler's analytical reading, more than three centuries before Adler.
4. Critical analysis: Who decides which books to chew? You can't know a book deserves digesting until you've tasted it, so triage has to come first.
5. Reflection: Which of the last five things you read deserved chewing, and did any of them get it?

That's deep reading in miniature: five passes of attention over one sentence, each pulling out something the previous pass missed.

How to Tell If You Read Something Deeply

You probably did if you can do these things without looking back at the text:

• Explain the main argument in two or three sentences of your own.
• Name the strongest piece of evidence and the weakest.
• State one objection, and say how the author would answer it.
• Connect it to something else you've read or lived through.
• Say what you now believe, or will do, differently.

If you can't, you've just found the part to reread.

---

The Deep Reading Circuit: What Happens in Your Brain

The deep reading circuit is the network of brain regions a skilled reader draws on to understand a text rather than merely recognize its words. Every reader has to build it, and it keeps adapting to how they read.

"We were never born to read." Wolf opens the first chapter of Proust and the Squid with that sentence, and the neuroscience follows from it. Writing is only a little over 5,000 years old, far too recent for evolution to have produced a dedicated reading organ. Instead, every child who learns to read builds a new circuit by linking regions that evolved for other jobs: recognizing objects, processing spoken language, storing memories, and planning.

That circuit is plastic. As Wolf put it in a 2018 essay for The Guardian, the reading circuit "is not given to human beings through a genetic blueprint like vision or language; it needs an environment to develop," and it adapts to whatever that environment demands. In Reader, Come Home, she describes a skilled reader who slows down doing far more than recognizing words: evaluating, connecting the text to background knowledge, and responding to its emotional weight.

Hurry, and the circuit does less, because the slower deep reading processes are the first to be cut. Over time, Wolf warns, a reader who mostly skims can lose what she calls "cognitive patience," the willingness to stay with a difficult text.

One brain-imaging study suggests that sustained reading leaves traces. Gregory Berns and colleagues at Emory University gave 21 undergraduates resting-state fMRI scans on 19 consecutive days: five days before, nine days during which they read a ninth of Robert Harris's novel Pompeii each evening, and five days after (Brain Connectivity, 2013). Connectivity rose in the left temporal cortex, an area linked to language, the morning after reading sessions, and changes in sensorimotor regions persisted for several days after participants finished the book. Not every change lasted, and the sample was small, but nine evenings with a novel were followed by measurable changes in the resting brain.

Is the Medium Part of the Problem?

Partly, and less than you might think. The most-cited meta-analysis, by Pablo Delgado, Cristina Vargas, Rakefet Ackerman, and Ladislao Salmerón (Educational Research Review, 2018), pooled 54 studies published between 2000 and 2017, covering 171,055 participants. Paper beat screens for comprehension by a small margin (g = −0.21). The gap was larger for informational text and under time pressure, it was absent in studies that used only narrative texts, and it grew over the years rather than shrinking as people got used to screens.

Virginia Clinton's 2019 meta-analysis found the same small paper advantage across 33 experiments, and in the studies that measured it, people reading on screens judged their own understanding less accurately, so the loss is easy to miss. In 2026 the same researcher, now publishing as Clinton-Lisell, and Christine Litzinger added a network meta-analysis showing that paper's advantage was clearest when the text had to be scrolled.

One leading explanation is the shallowing hypothesis: years of quick, fragmented screen use train a habit of skimming, and the screen cues that habit. That's the hopeful part, because a habit can be overridden. Wolf's own prescription isn't to abandon screens. "We need to cultivate a new kind of brain," she wrote in the same Guardian essay: a "bi-literate" reading brain "capable of the deepest forms of thought in either digital or traditional mediums." For the full evidence, see our breakdown of reading on screen vs paper.

---

Why Deep Reading Is Declining

Nicholas Carr's 2010 book The Shallows: What the Internet Is Doing to Our Brains argued that the web was training us out of sustained, linear thought. Some critics called it alarmism. The research since then hasn't proved his strongest claims about rewired brains, but it does document the behavior he described.

Attention is fragmented. In a 2016 study, Gloria Mark of UC Irvine and her colleagues tracked 40 information workers at one company and found they spent an average of 47 seconds on one computer screen before switching (the median was 40). The often-quoted claim that it takes "23 minutes and 15 seconds" to refocus after an interruption doesn't appear in the study it's usually credited to; it traces back to a 2008 magazine article. What the 2016 study does show is less dramatic and more relevant: at work, long, unbroken stretches on one screen are rare. It measured screen-switching in one office, not anyone's capacity to concentrate.

Fewer people read for pleasure. A 2025 study in iScience by Jessica Bone and colleagues analyzed American Time Use Survey diaries from 236,270 people between 2003 and 2023. The share of Americans who read for pleasure on a given day fell from a peak of 28% in 2004 to 16% in 2023, a relative decline of about 3% a year. People who still read spent longer at it, though, up from 1 hour 23 minutes a day to 1 hour 37 minutes. What shrank was the number of readers, not their reading time.

Students are struggling with complex text. In the 2024 National Assessment of Educational Progress, 30% of US eighth-graders scored at or above NAEP Proficient in reading, and the share scoring below NAEP Basic was the largest in the assessment's history. Twelfth-graders posted the lowest average reading score since the assessment began in 1992. Internationally, the OECD average reading score in PISA 2022 fell 10 points from 2018, a drop the OECD called twice the previous record. Screens aren't the only suspect (the pandemic is an obvious one), and nobody has proved causation, but the trend runs the same way.

Summaries remove the incentive. When you can get the gist of any book in 30 seconds, 10 hours with the full text starts to feel optional. AI summaries, podcast recaps, and video explainers offer a low-friction path to surface knowledge. They're useful. They also skip the work that produces understanding.

Wolf doesn't blame the technology itself so much as the habit it encourages. In the same Guardian essay, she described what skimming costs: "When the reading brain skims like this, it reduces time allocated to deep reading processes. In other words, we don't have time to grasp complexity, to understand another's feelings, to perceive beauty, and to create thoughts of the reader's own."

---

The AI Paradox: Summarization vs. Understanding

AI can compress a 300-page book into a five-minute read, pull out the key arguments, and write discussion questions. Those are real capabilities, and they create a convincing illusion: the feeling that you understand something you've only been told about. That's the paradox.

Understanding isn't information. You can know every fact about swimming and still sink. Reading a summary of Thinking, Fast and Slow gives you the names of the biases. Working through Kahneman's examples, arguing with him in the margins, and testing his claims against your own decisions is how you come to understand them.

Shiri Melumad and Jin Ho Yun measured the gap directly in seven experiments with 10,462 participants, published in PNAS Nexus in October 2025. People learned about everyday topics, such as planting a vegetable garden, either from an AI-generated synthesis or from ordinary web search results, and then wrote advice for someone else. Those who learned from the AI summary wrote sparser, less original advice, and other people shown that advice were less likely to adopt it.

The core facts were held constant. What the summary removed, the authors argue, was the work of finding and connecting sources yourself, and that's where the learning happens. Adding links to the sources didn't fix it: only 26% of participants clicked them.

So treat AI summaries as a triage tool. Let AI scan 20 articles and pick the three worth reading carefully, or preview a lecture with YouTube Summary before deciding to watch the whole thing. Once you've read something deeply, quiz yourself on it with Glasp's AI chat. Let AI tell you where to look; deep reading is what you do once you get there.

As summaries get cheaper, the readers who stand out will be the ones who can do what a summary can't: form original interpretations, make unexpected connections across fields, take an author's perspective seriously, and test ideas against their own experience. All of those are products of deep reading.

---

Close Reading Strategies That Work

Close reading is deep reading with structure. In American schools it's the first of the Common Core's anchor standards for reading: "Read closely to determine what the text says explicitly and to make logical inferences from it; cite specific textual evidence when writing or speaking to support conclusions drawn from the text." Three frameworks make that practical.

SQ3R (Survey, Question, Read, Recite, Review)

Francis P. Robinson introduced SQ3R in his 1946 book Effective Study. The idea is that each step forces a different kind of processing.

1. Survey: Skim the headings, bold terms, figures, and conclusion. Build a mental map before you go in.
2. Question: Turn each heading into a question. "The Causes of Inflation" becomes "What causes inflation?" Now you're reading for answers.
3. Read: Read the section carefully, looking for those answers. Highlight key passages and note your reactions.
4. Recite: Look away and summarize the section in your own words. This is retrieval practice, which Karpicke and Blunt (Science, 2011) found produced more learning than elaborative studying with concept maps.
5. Review: When you're done, go back through your highlights and notes, connect ideas across sections, and state the overall argument.

Adler's Four Levels of Reading

Mortimer Adler first published How to Read a Book in 1940. The 1972 revision, written with Charles Van Doren, organizes reading into four levels, each containing the ones before it:

1. Elementary reading: Understanding what the sentences say.
2. Inspectional reading: Systematic skimming to learn a book's structure and scope in limited time.
3. Analytical reading: Full engagement, which means identifying the argument, judging the evidence, and reaching your own verdict. This is deep reading.
4. Syntopical reading: Reading several books on one subject and building an analysis none of them contains.

Moving from inspectional to analytical reading is a decision: slow down, ask questions, and wrestle with the text.

Notice and Note Signposts

For fiction, Kylene Beers and Robert E. Probst's Notice and Note: Strategies for Close Reading (2012) teaches readers six "signposts" that flag moments worth slowing down for, each paired with a question:

• Contrasts and Contradictions: A character acts in a way you didn't expect. Ask why.
• Aha Moment: A character suddenly realizes something. Ask how it might change things.
• Tough Questions: A character asks themselves a hard question. Ask what the question makes you wonder about.
• Words of the Wiser: An older, wiser character gives advice. Ask what the life lesson is.
• Again and Again: An image, phrase, or event keeps recurring. Ask why the author repeats it.
• Memory Moment: The story pauses for a memory. Ask why it matters now.

The same habit carries over to nonfiction. When an argument contradicts itself or a speech keeps returning to one phrase, that's your cue to stop and ask why.

Comparing the Strategies

SQ3R and Adler's levels both treat deep reading as more than one pass, with a different purpose each time, while Notice and Note tells you where to stop on the first. If you're just starting, SQ3R is the easiest to follow because its steps are concrete and sequential.

---

Deep Reading vs. Skimming: A Comparison

Both modes have a place. Trouble starts when skimming becomes the default for every text, including the ones that deserve more. The two compare like this:

The speed figures come from Marc Brysbaert's 2019 review in the Journal of Memory and Language: 238 wpm is the average from his meta-analysis of 190 studies, and about 100 wpm is his separate estimate for reading a text you'll have to recall. On the skimming side, a 2016 review led by Keith Rayner in Psychological Science in the Public Interest concluded that faster reading comes with lower comprehension, and that no technique escapes the trade-off.

Skim freely, as long as you can still switch to deep reading when a text deserves it.

---

How Annotation Powers Deep Reading

Annotation is how reading turns into thinking. When you highlight a passage and write down why it matters, you're doing three things at once: selecting (deciding it's important), encoding (putting the idea in your own words), and organizing (connecting it to a larger framework).

The research on highlighting, the most common kind of annotation, comes with a catch. A 2022 meta-analysis by Ponce, Mayer, and Méndez in Educational Psychology Review pooled 85 effect sizes from 36 articles. Highlighting you do yourself improved memory for the material (d ≈ 0.36) but didn't reliably improve comprehension (d ≈ 0.20). Mason, Ronconi, and colleagues point the same way for comprehension: in their 2024 study of digital reading, highlighting by itself didn't help, but students whose highlights captured more of the ideas experts rated as important scored higher on literal comprehension and transfer.

So what matters is what you choose to mark, and what you do next. Highlighting a sentence because it "sounds good" is passive. Highlighting it and writing "this contradicts what Smith claims in chapter 2" is active, because the note forces you to test the idea against something else you know.

Effective annotation habits for deep reading:

• Questioning: Write questions in the margin. "Why does the author assume X?" "Where's the evidence?"
• Connecting: Link ideas to other texts or to your own experience. "This is Kahneman's framing effect again."
• Disagreeing: Push back on weak claims. "This ignores Y."
• Summarizing: After each section, write one sentence in your own words.
• Color-coding: Use multiple highlight colors by role: one for key claims, one for evidence, one for ideas to revisit, one for disagreements.

Highlights plus notes turn a private mental exercise into a visible record of your thinking. Months later, they show you not just what the author said but how you engaged with it. For more on what the research does and doesn't support, see the science of highlighting.

---

Deep Reading on Screens: Web Pages, PDFs, E-Books, and Video

A persistent myth says deep reading only works with printed books. It doesn't, but reading deeply on a screen takes deliberate setup, because the defaults are built for skimming: links pull you away, notifications interrupt, and infinite scroll keeps you moving.

Web pages. The biggest enemy is the open tab. Reading with 15 other tabs in view splits your attention before you start. Close what you don't need and annotate as you go with Glasp's web highlighter. Highlighting and taking notes on the page pull you out of scroll mode.

PDFs and academic papers. PDFs behave more like paper, with fixed layouts, page numbers, and fewer distractions. Use single-page view instead of continuous scroll, since the 2026 network meta-analysis found screens fell furthest behind paper when readers had to scroll. A research paper's predictable structure (abstract, methods, results, discussion) also suits a multi-pass strategy like SQ3R.

E-books. Most e-reader apps support highlights and notes, so use them. Check yourself at chapter breaks by summarizing what you just read without looking. People reading on screens tend to judge their understanding less accurately than paper readers, most likely by overestimating it (Clinton, 2019), and that quick self-test is the cheapest correction there is.

Video. Lectures, interviews, and documentaries carry complex ideas too, and the default way of watching (press play and let it wash over you) is passive by design. A transcript turns a video into text you can read at your own pace. YouTube Summary by Glasp gives you a timestamped transcript, so you can read ahead to see the argument's structure, highlight key statements as you watch, note your reactions in the moment, and connect them to highlights from related articles and books.

A 90-minute conversation runs to thousands of words. Without a transcript you keep fragments. With an annotated one, you keep the structure.

A workflow for deep reading on any screen:

1. Triage: Skim, or use AI, to decide which texts are worth reading deeply.
2. Set up: Close tabs, silence notifications, switch to a paginated view where you can, and set a 30- to 60-minute timer.
3. First pass: Read the whole thing for its shape, highlighting only what stops you.
4. Second pass: Reread the parts you marked and annotate them: key claims, weak evidence, and connections.
5. Synthesis: Write a short summary and one question the text leaves open.
6. Archive: Keep your highlights and notes somewhere searchable. Glasp does this automatically, so your reading builds into a library you can search later.

For a complete system for keeping what you read, see how to remember what you read.

---

Building a Daily Deep Reading Practice

Deep reading is a skill. It responds to practice and fades with neglect. If you've spent years in skim mode, getting your patience back takes time, but Wolf points to "an old rule in neuroscience that does not alter with age: use it or lose it." The rule cuts both ways, so practice can rebuild what disuse wore down.

Start with 20 to 30 minutes a day. That's enough to engage seriously without turning it into a chore.

A long-running health study adds a footnote. In a 2016 study in Social Science & Medicine, Avni Bavishi, Martin Slade, and Becca Levy followed 3,635 adults over 50 for up to 12 years. Book readers had a 20% lower risk of dying during that period than people who read no books, after adjusting for factors including education, wealth, and self-rated health. People who read books for 3.5 hours or more a week, about 30 minutes a day, had a 23% lower risk. The survival advantage was significantly larger for books than for newspapers and magazines, but the study is observational, so it can't show that reading caused the difference.

Choose material that rewards depth. Not everything does. A listicle doesn't need deep reading. A chapter of a serious book does. Match your reading mode to the text and your goal.

Make distraction effortful. Put your phone in another room. Close email. On a screen, keep one window open. The aim is to make distraction harder than reading, the opposite of the default.

Use a timer. Set it and read until it goes off. Handing the clock to a timer saves you from constantly deciding whether you've read "enough."

Annotate from the start. Don't save annotation for a second pass that may never happen. Even one highlight per page and one note per section keeps you in active mode.

End each session with two minutes of reflection. Write down the main ideas, what surprised you, what you disagree with, and what you want to learn next. Recalling it this way does more for memory than rereading. For a system that makes the most of these notes, see how to take smart notes.

Track your progress. A reading log or a growing library of highlights makes the habit visible. If you read on Kindle, you can bring your Kindle highlights into Glasp so the books you read deeply sit next to the articles.

Mix genres. Alternate dense nonfiction with a novel you can't put down. Both exercise the same deep reading processes, and fiction is where perspective-taking gets its workout.

Read with other people. Explaining a text to someone else forces you to articulate and defend your interpretation, which deepens it. Share your highlights publicly on Glasp and you'll find readers working through the same texts.

---

Frequently Asked Questions

What is deep reading in simple terms?

Deep reading is reading slowly and attentively enough to understand a text, not just recognize its words. A deep reader makes inferences, connects the text to what they know, questions the argument, and reflects on what it means. Sven Birkerts coined the term in 1994, and the most-cited definition, from Maryanne Wolf and Mirit Barzillai (2009), describes it as the "sophisticated processes that propel comprehension," including inference, analogy, critical analysis, reflection, and insight.

What is an example of deep reading?

One example is a news investigation. A deep reader finishes it able to say who the sources were, what the documents showed, and what the reporters couldn't confirm; a skimmer remembers the headline. Other examples include checking a study's sample size and limitations instead of stopping at the abstract, and working out why a character in a novel acts against their own interests.

What is the difference between deep reading and close reading?

Deep reading is the broad term for immersive, reflective engagement with any text. Close reading is a narrower, structured practice from literary criticism, now a core goal of school standards: careful, often repeated reading of a short passage with attention to its language, structure, and evidence. All close reading is deep reading, but most deep reading doesn't follow a formal close reading protocol.

Is deep reading the same as slow reading?

Not quite, though they overlap. Slow reading is a practice of choosing fewer texts and reading them at a pace your thinking can keep up with. Deep reading names what your mind does with the text, such as inference, connection, critique, and reflection. Slowing down gives those processes time to run, but you can read slowly and still drift, so pace alone doesn't make reading deep. Our guide to slow reading covers the practice and the speed numbers.

Can you read deeply on a screen?

Yes, with some effort. Meta-analyses find a small comprehension advantage for paper, larger for informational text and under time pressure (Delgado et al., 2018), and clearest when the text has to be scrolled (Clinton-Lisell and Litzinger, 2026). Taking the time pressure off and reading in a paginated view instead of scrolling target those conditions directly, and a quick self-test at each break guards against the misplaced confidence screens seem to encourage (Clinton, 2019).

Is deep reading compatible with AI tools?

Yes, if AI handles the triage and you handle the reading. AI is good at picking which texts deserve your attention, drafting pre-reading questions, and quizzing you afterward. It's a poor substitute for the reading itself: in Melumad and Yun's 2025 experiments, people who learned from an AI synthesis ended up with shallower knowledge than people who learned through ordinary web search.

How long does it take to build a deep reading habit?

Expect about two months, with wide variation. In a well-known study by Phillippa Lally and colleagues (2010), a new daily behavior took a median of 66 days to become nearly as automatic as it would get, with estimates ranging from 18 to 254 days (the longest are projections from a 12-week study). The behaviors were simple eating, drinking, and exercise habits, not reading, so treat the number as a rough guide. Short daily sessions beat occasional long ones, because consistency is what builds the habit.

Does deep reading improve empathy and critical thinking?

Fiction reading has a modest link to empathy, though the famous result needs a caveat. A 2013 study by David Kidd and Emanuele Castano in Science reported that reading a short piece of literary fiction improved performance on theory-of-mind tests right afterward. The result hasn't held up well in replications: a three-lab study of 792 people (Panero et al., 2016) found no literary-fiction advantage, and a large 2018 replication project couldn't reproduce the first experiment. A 2018 meta-analysis by David Dodell-Feder and Diana Tamir, covering 14 studies, found a small positive effect of fiction reading on social cognition (g = 0.15). For critical thinking, the case rests on the practice itself: evaluating claims, spotting assumptions, and weighing evidence are what deep reading consists of.

---

Conclusion: Read Deeply, Think Clearly

The ability to stay with a hard text until you understand it is getting rarer, and that makes it worth more.

The evidence points the same way from several directions. Reading is a learned circuit that adapts to how you use it. Hurry and scrolling cost comprehension. Fewer Americans read for pleasure. And learning from an AI synthesis left people with shallower knowledge than working through the sources themselves.

AI has made surface knowledge almost free. That's a gift, because it saves your limited reading time for the texts that reward depth. Use AI to scout the territory, then pick your ground and read it deeply.

The tools exist for any medium. Glasp's web highlighter lets you annotate articles as you read. YouTube Summary gives you transcripts to mark up. Color-coded highlights and notes build a searchable archive of your thinking.

Tools only support the practice. The practice is 20 to 30 minutes a day with a text that deserves it, a highlighter in hand, and two minutes at the end to write down what you think. Keep that up, and what you understand will be your own, which is the one thing a summary can't hand you.