Levels of Processing: Why How You Read Determines What You Remember

Quick Answer: The levels of processing framework, proposed by Craik and Lockhart in 1972, shows that memory strength depends on how deeply you process information during reading—not how many times you review it. Shallow processing (recognizing words, noting formatting) creates weak memories. Deep, semantic processing (understanding meaning, making connections, evaluating arguments) creates durable ones. For readers, annotating meaningfully, summarizing chapters in your own words, and asking "why?" as you read will always outperform passive re-reading. For a full retention system, see our guide on how to remember what you read.

In 1972, psychologists Fergus Craik and Robert Lockhart proposed an idea that changed how we understand memory. Their levels of processing framework argued that how deeply you process information at the time of learning determines how well you remember it later. The depth of processing matters more than how long you spend studying or how many times you repeat the material.

This has direct consequences for anyone who reads nonfiction. Two people can read the same book for the same amount of time and walk away with dramatically different levels of retention. The difference is not talent or intelligence. It is processing depth.

What the Levels of Processing Framework Says

Craik and Lockhart proposed that memory is not a separate system where information gets "stored." Instead, memory is a natural byproduct of how you process information. Process something shallowly, and the memory trace is weak and fades quickly. Process it deeply, and the memory trace is strong and durable.

They identified a continuum of processing depth.

Structural processing (shallow): You process the physical features of the information. For text, this means noticing the shapes of letters, the font, the layout on the page. Almost no meaning is extracted. This is what happens when your eyes move across a page but your mind is elsewhere.

Phonemic processing (intermediate): You process the sound of the words. You could read the sentence aloud, but you are focused on pronunciation and rhythm rather than meaning. This happens during passive reading where you hear the words in your head but are not engaging with the ideas.

Semantic processing (deep): You process the meaning. You think about what the words mean, how the ideas connect to each other, and how they relate to things you already know. This is where durable memories are formed.

The original experiments demonstrated this cleanly. When participants were asked shallow questions about words (does this word contain the letter E?), they remembered far fewer words than when asked semantic questions (does this word fit in the sentence "The girl sat on the ___?"). Same words, same study time, dramatically different recall. The depth of the question determined the depth of the memory.

Why This Matters for Readers

Most people read nonfiction at the phonemic level. They process the words, follow the sentences, and experience a sense of understanding as they go. But that subjective feeling of understanding is misleading. Robert Bjork calls this the "illusion of competence," and it is one of the biggest obstacles to effective learning from reading.

The illusion works like this: when you read a well-written explanation, the ideas flow smoothly and make sense in the moment. Your brain interprets this fluency as evidence of learning. But fluent processing and deep processing are not the same thing. You can follow an argument perfectly without encoding it in a way that you can retrieve later.

This explains a universal reading frustration. You finish a book, feel like you understood it thoroughly, and three weeks later cannot recall its main arguments. You did understand it in the moment. But your processing was too shallow to create durable memory traces. The information was never deeply encoded in the first place.

The Three Factors That Determine Processing Depth

Craik and Lockhart's framework has been refined by decades of subsequent research. Three factors consistently predict how deeply you process what you read.

1. Elaboration

Elaboration means connecting new information to things you already know. When you read that "spaced repetition leverages the spacing effect to optimize review timing," shallow processing registers this as a fact. Deep processing asks: how does this relate to the way I currently study? Why would spacing be better than cramming? What is the mechanism?

Research by Bradshaw and Anderson in 1982 demonstrated that sentences learned with elaborative context were recalled significantly better than sentences learned in isolation. The more connections you build between new information and existing knowledge, the more retrieval pathways your brain creates to access it later.

For readers, this means actively asking questions while you read. Elaborative interrogation, which involves asking "why is this true?" and "how does this work?", is one of the most effective techniques for deepening processing during reading.

2. Distinctiveness

Information that is distinctive, meaning it stands out as different from what you already know, is processed more deeply than information that blends in. This happens because your brain allocates more processing resources to surprising or unexpected input.

When reading, you can leverage distinctiveness by paying special attention to ideas that contradict your existing beliefs or that surprise you. These moments of cognitive friction are signals that deep processing is possible. Rather than smoothing past them, stop and engage. Why is this different from what you expected? What would have to be true for this to be correct?

The von Restorff effect, documented in the 1930s, shows that distinctive items in a list are remembered far better than common ones. In reading, the equivalent is that a finding which challenges your assumptions will be better remembered than one that confirms what you already believe, but only if you engage with the surprise rather than dismissing it.

3. Personal Relevance

Information that is relevant to your own life, goals, or identity is processed more deeply than abstract information with no personal connection. This is called the self-reference effect, and it is one of the most robust findings in memory research.

Rogers, Kuiper, and Kirker demonstrated in 1977 that asking "Does this word describe you?" produced better recall than even semantic processing tasks. When you connect information to yourself, you recruit additional brain networks involved in self-reflection, which creates richer and more accessible memory traces.

For readers, this means constantly asking: how does this apply to my situation? When have I experienced something like this? How would this change what I do tomorrow? These questions feel like interruptions to your reading flow, but that interruption is exactly the point. Smooth, uninterrupted reading is a symptom of shallow processing.

How to Read at a Deeper Level

Understanding the levels of processing framework is useful, but the practical question is: how do you actually shift your reading from shallow to deep? Here are concrete techniques, each tied to the research.

Ask Questions Before Reading

Before starting a chapter or section, scan the headings and ask yourself what you expect to learn. This priming effect creates a knowledge gap that your brain wants to close, which automatically increases processing depth when you encounter the relevant information.

Pre-reading strategies like SQ3R formalize this into a system: Survey the material, generate Questions, then Read with those questions in mind. The questions transform passive reading into an active search for answers, which pushes processing from the phonemic level to the semantic level.

Pause and Paraphrase

After each major section or chapter, close the book and explain what you just read in your own words. This is a form of active recall that forces semantic processing because you must reconstruct the meaning rather than just recognizing the words on the page.

If you can paraphrase the section confidently, you have processed it deeply. If you struggle, you have identified a gap where your processing was shallow. Either outcome is valuable.

Connect to What You Know

Deliberately look for connections between what you are reading and your existing knowledge. When an author describes a concept, ask: what does this remind me of? Have I seen a similar idea in another book? How does this relate to my work or personal experience?

These connections serve as elaborative encoding. Each connection creates an additional retrieval cue that makes the information more accessible later. Synthesizing ideas across books is one of the most powerful forms of deep processing because it forces you to understand each book's ideas well enough to compare them.

Argue with the Author

One of the deepest forms of processing is evaluating an argument's validity. When you read a claim, ask: is this supported by the evidence presented? Are there alternative explanations? What would the author's strongest critic say?

You do not need to actually disagree. The act of evaluating the argument forces you to understand it at a structural level, including its premises, logic, and conclusions. This structural understanding is far more durable than surface-level comprehension.

Write Marginal Notes

Marginalia is not just annotation. It is a physical manifestation of deep processing. When you write a note in the margin, you are forced to generate a response to what you read, which engages semantic processing. The best marginal notes are reactions, questions, and connections rather than underlines or highlights alone.

Research on the generation effect shows that information you generate yourself is remembered better than information you passively receive. A marginal note that says "contradicts Taleb's argument about fragility" creates a deeper memory trace than simply highlighting the passage.

Levels of Processing and Spaced Repetition

The levels of processing framework and spaced repetition address different aspects of the memory problem, and they work well together.

Levels of processing determines encoding quality, which is how well information is stored initially. Spaced repetition determines retrieval maintenance, which is how well you can access information over time. Deep processing creates strong initial memories. Spaced repetition keeps those memories accessible.

Without deep processing, spaced repetition reviews are less effective because you are trying to maintain a weak memory trace. Without spaced repetition, even deeply processed information fades over months as the forgetting curve takes its toll.

The optimal approach uses both. Read deeply using the techniques above, capture the most important ideas using a tool like Chapterly, and then review those ideas at spaced intervals. The deep processing ensures each review strengthens a solid memory trace rather than repeatedly re-encoding a shallow one.

Measuring Your Processing Depth

How do you know whether you are reading deeply or just going through the motions? Here are three self-tests.

The explanation test: Can you explain the key idea of the last section you read to someone unfamiliar with the topic? If you can, you processed it deeply. If you can only summarize the words without explaining the meaning, your processing was shallow.

The connection test: Can you identify at least one connection between what you just read and something you already know? If every concept feels isolated, you are not elaborating sufficiently.

The application test: Can you think of a specific situation where this idea would change your behavior or decision? If the information feels purely abstract with no practical implication, you have not processed it at the self-referential level.

These tests take thirty seconds but provide immediate feedback on your processing depth. If you fail any of them, re-read the section with deliberate attention to meaning, connections, and application.

The Cost of Deep Reading

There is an honest trade-off here. Deep processing takes more time and mental energy than shallow processing. You will read fewer pages per hour. You will finish books more slowly. Your brain will feel more tired after a reading session.

This is not a problem. It is the point. The additional effort is exactly what produces the deeper encoding. Desirable difficulties, a concept from Robert Bjork's research, shows that learning conditions that feel harder in the moment often produce better long-term retention. Easy, fluent reading feels productive but often is not.

The practical solution is not to read everything deeply. Reserve deep processing for material that matters. A novel you are reading for entertainment does not need semantic processing at every paragraph. A professional book whose ideas you want to use for years deserves the investment.

Choose what to read deeply, then commit to the effort for those selections.

Getting Started

The simplest change you can make right now is to add one practice to your next reading session: after every chapter, close the book and spend two minutes explaining what you just read, either in writing or out loud. That single habit shifts your processing from the phonemic level to the semantic level.

You will notice the difference immediately. Some chapters you can explain easily, confirming deep processing. Others you struggle with, revealing shallow processing you did not detect while reading. Both outcomes improve your reading because they give you accurate feedback about what you actually learned versus what you only think you learned.

Over time, this feedback loop trains you to read more deeply by default. You start asking questions while reading because you know you will need to explain the answers afterward. That anticipation changes how you process every sentence.

Tools like Chapterly support this practice by letting you capture and review the passages that represent your deepest processing moments. When you highlight a passage because it connected to something important in your life or challenged an assumption you held, that highlight is a marker of deep processing worth revisiting.

For a structured system that channels deep processing into a testable review format, see our guide on how to create a study guide from any book. And if you find yourself struggling with difficult material before it clicks, see productive failure in reading for why the struggle is the mechanism, not the obstacle.

Frequently Asked Questions

What is the levels of processing theory?

The levels of processing theory, proposed by Fergus Craik and Robert Lockhart in 1972, argues that memory strength is determined by the depth at which information is processed, not by how many times it is reviewed. Shallow processing—recognizing words as shapes or sounds—creates weak, short-lived memories. Deep, semantic processing—understanding meaning and connecting ideas to prior knowledge—creates strong, durable memories. The theory shifted memory research from studying storage to studying the quality of cognitive engagement during learning.

What are the three levels of processing?

The framework describes three main depths: structural (phonemic) processing, which focuses on the physical appearance of words; phonemic processing, which focuses on how words sound; and semantic processing, which focuses on what words mean and how they connect to each other and to prior knowledge. For reading retention, the critical distinction is between surface-level processing and deep semantic processing. Only semantic processing reliably produces long-term memories.

Why does deep reading produce better retention than re-reading?

Re-reading typically triggers shallow processing because familiar text is easy to recognize without being understood. Deep reading involves active engagement: asking questions, connecting ideas to prior knowledge, evaluating arguments, and generating your own examples. This active engagement triggers semantic processing, creating stronger memory traces. Spending five minutes actively engaging with a chapter after reading it produces better retention than thirty minutes of passive re-reading.

How can I read more deeply without slowing down too much?

Four strategies reliably deepen processing with minimal time cost: (1) read with a question in mind for each chapter so your brain is actively searching for the answer; (2) annotate by reaction rather than just underline—write a one-phrase response in the margin; (3) at every chapter break, pause and state the main idea aloud or in writing before reading on; (4) connect each new idea to something you already know by explicitly asking "where have I seen this before?" These habits add minutes per session but multiply retention over months.

What 2026 Has Changed for Levels-of-Processing Theory

Two updates worth noting since Craik and Lockhart's original 1972 paper. First, the cognitive-neuroscience work of the last decade has substantially clarified the neural signature of depth. ERP research, extended through the early 2020s, shows that semantic-level encoding produces a reliably different brain signal than phonemic encoding — the difference is detectable at the time of encoding, not only retrospectively at retrieval. The theory used to be inferred from behavioral memory data; in 2026 it is observable. The practical implication for readers is that "deep processing" is not a folk-psychological metaphor — it is a specific operation your brain either performs or doesn't, and the operation tracks closely with whether you can later recall vs only recognize the material.

Second, the rise of AI summarization tools has created a new category of false depth. Reading a fluent AI summary of a chapter feels semantically rich because the language is dense with meaning words, but the reader is processing the summary semantically, not the book. The downstream test — can you talk about the original argument, with its specific examples and qualifications, six months later? — exposes the gap. The 2026-honest version of levels-of-processing theory has to add a clause: depth is not just about whether you engaged with semantic content, but whether the semantic content you engaged with was the actual material you wanted to remember. AI summaries can be useful inputs to deep processing, but they are not substitutes for it.

The third update is empirical. Recent meta-analytic work tightened the picture: the depth-of-processing effect is robust on free-recall tasks (large effect across many studies) and weaker on recognition tasks (small-to-moderate effect). The asymmetry matches the recall-vs-recognition distinction: deep processing helps most when the test demands that you generate the material from no cue. For readers whose goal is being able to talk about a book, not merely identify it as familiar, depth is the variable that actually moves the outcome.

Chapterly is built around the deep-processing side of the theory: highlights become flashcards that demand semantic retrieval, and the AI tutor pulls you into elaboration rather than recognition. Try it free.