Deep vs. Shallow Processing: Why Meaning Often Sticks Better Than Repetition
By the end of this guide you will be able to tell a shallow study move from a deeper one, try three levels on one word list, and swap at least one shallow habit for a deeper one.
The idea comes from a 1972 paper by Craik and Lockhart. It is a framework, not a rulebook, and we say below what it cannot tell you.
The rest of this guide is locked
Play a round of five cards to open it and collect this guide’s stamp. Or read it now without playing, and skip the stamp.
The round uses cards from the Learning Science Glossary deck. Your stamps and progress stay in this browser.
The idea in plain words
Craik and Lockhart (1972) reviewed evidence for theories that describe memory as separate stores and pointed out difficulties with that approach. They outlined an alternative framework in terms of the depth, or levels, of processing. Put simply, the framework asks how deeply you work with information when you meet it.
We illustrate levels with three examples of our own. Noticing how a word looks is shallow processing. Noticing how it sounds sits in the middle. Thinking about what it means is deep processing. We read only the abstract and opening of the 1972 paper, so we do not attribute these examples, specific experimental results or a rule such as "deeper always works better" to it.
Try it yourself: one word list, three ways
Take these twelve words: lantern, harbor, pebble, ladder, whistle, blanket, orchard, compass, feather, anchor, thimble, copper. Split them into three groups of four, then study each group a different way.
- By letters: for lantern, harbor, pebble and ladder, count the letters and mark the vowels.
- By sound: for whistle, blanket, orchard and compass, count the syllables and think of a rhyme.
- By meaning: for feather, anchor, thimble and copper, write a sentence about your own day that uses the word.
Wait a day, then write every word you remember from a blank page and see which group did best. This is an informal test with one person and words of different difficulty, so it is a way to notice your own pattern, not evidence. If the meaning group comes out ahead, that fits the idea. If it does not, check the obvious things: were your sentences specific or generic, and were some words simply easier? Either result tells you which move to repeat.
Shallow moves and deeper moves
| Situation | Shallow move | Deeper move |
|---|---|---|
| A new definition | Reread it until it looks familiar | Close the book, write it in your own words, then check |
| A new vocabulary word | Copy it five times | Write a sentence about your own day that uses it |
| A textbook paragraph | Highlight the key terms | Turn the main claim into a question and answer it |
| A worked example | Read the solution again | Cover the solution, redo the steps, then compare |
The sources we cite touch some of these moves, but none tests the table as a whole. Dunlosky et al. (2013), a major review of ten learning techniques, rated rereading and highlighting as low utility. Elaborative interrogation and self-explanation received moderate ratings. In plain terms, those mean asking yourself why a stated fact is true and explaining to yourself how you reached a step.
What the studies add
Slamecka and Graf (1978) ran five experiments in which words that participants generated themselves were remembered better than the same words when simply read. The advantage showed up on recognition, free recall, cued recall and confidence-rating measures, and it persisted across variations in the experiments. Making your own sentence or answer is a form of generating. The work used word lists in a lab, and the authors said the effect was hard to explain with the theories of the time.
Notice what this does and does not show. Participants generated words under the experimenters' rules, which is close to, but not the same as, writing a sentence about your own day. None of the studies we cite tests the sentence-writing tip itself, so it is practical advice built on the generation effect.
Limits and critiques
- A framework is not a recipe. The 1972 paper is a conceptual paper, not an experiment report. It offers an idea to test.
- Deep is not the only lever. Karpicke and Blunt (2011) found that practicing retrieval of a science text produced greater gains in meaningful learning than elaborative studying with concept mapping, even on inference questions. The authors read this as support for retrieval-specific mechanisms.
- Ratings can change. The moderate ratings in Dunlosky et al. reflect evidence up to 2013 and the authors' judgment.
Our practical takeaway is to combine the two: work with meaning while you study, then test yourself without looking. Grouping by meaning is also one way to build chunks, covered in chunking, and why rereading feels like learning covers the shallow habit in detail. For picture-based methods, see mental imagery and memory. To rehearse these terms, try the learning science terms deck. The wider context is in how human memory works.
Common mistakes
- Assuming "deeper is always better". The 1972 paper is a framework. In Karpicke and Blunt (2011), retrieval practice produced greater gains than concept mapping, so pair deep study with self-testing.
- Calling a habit deep because it takes a long time. Time is not depth. Ask whether you are working with meaning, such as explaining or applying the idea in your own words.
- Copying notes word for word and calling it studying. Reword each idea in your own terms, then answer a question about it from memory.
- Skipping the check after a deep study move. Close the notes and test yourself later. Seeing meaning feels good but still needs a test.
Frequently asked questions
What is levels of processing?
It is a framework from Craik and Lockhart (1972) that describes memory in terms of how deeply information is processed. In our simple illustration, shallow processing attends to surface features such as how a word looks, and deep processing attends to meaning. It is a framework to guide research, not a study rule.
Is deep processing the same as understanding?
Not exactly. Deep processing is a way of describing how you work with the material. Understanding is the result you hope for. The sources we cite do not equate them.
Does making my own examples really help?
In Slamecka and Graf's experiments, words participants generated themselves were remembered better than words they read. Those were lab word lists, so treat your own examples as a promising habit, not a guarantee.
Is rereading shallow processing?
In this guide we treat rereading as a shallow move, because the words go by without you producing anything. Dunlosky et al. rated it low utility. That does not make it useless for a first pass, but the review suggests testing yourself instead of relying on it.
Sources
We opened each source below before citing it. See Methodology and Sources for how we check them.
- Craik, F. I. M., & Lockhart, R. S. (1972). Levels of processing: A framework for memory research. Journal of Verbal Learning and Verbal Behavior, 11(6), 671-684. Open the source
- Slamecka, N. J., & Graf, P. (1978). The generation effect: Delineation of a phenomenon. Journal of Experimental Psychology: Human Learning and Memory, 4(6), 592-604. Open the source
- Dunlosky, J., Rawson, K. A., Marsh, E. J., Nathan, M. J., & Willingham, D. T. (2013). Improving Students' Learning With Effective Learning Techniques: Promising Directions From Cognitive and Educational Psychology. Psychological Science in the Public Interest, 14(1), 4-58. Open the source
- Karpicke, J. D., & Blunt, J. R. (2011). Retrieval Practice Produces More Learning than Elaborative Studying with Concept Mapping. Science, 331(6018), 772-775. Open the source
This guide is general study help, not medical or professional advice. It was last reviewed on November 12, 2026. Spot a mistake? Tell us here.