The main idea of levels of processing theory is that how well you remember something depends on how deeply you think about it, not where or how it gets stored in your brain. Proposed by Fergus Craik and Robert Lockhart in 1972, the theory argues that memory is a byproduct of the quality of mental processing that happens when you first encounter information. The deeper and more meaningfully you engage with material, the stronger and longer-lasting the memory trace.
The Core Idea: Processing Depth Drives Memory
Before levels of processing theory came along, the dominant view of memory was essentially a storage model: information moves through short-term and long-term memory bins, each with a set capacity. Craik and Lockhart challenged this by proposing that the degree to which an event is remembered is a function of how it’s processed, not a function of how the experience is stored. In other words, memory isn’t about putting things in the right box. It’s about what your brain does with the information in the moment you encounter it.
The theory describes a continuum from shallow to deep processing. At the shallow end, you’re barely engaging with information, using minimal attention and reasoning. At the deep end, you’re actively connecting new information to things you already know, analyzing meaning, and evaluating relevance. Shallow processing creates fragile, short-lived memories. Deep processing creates durable ones.
Three Levels: Structural, Phonemic, and Semantic
The theory identifies three broad levels at which you can process verbal information, each progressively deeper.
- Structural (visual) processing is the shallowest level. You notice what a word looks like: its font, whether it’s uppercase or lowercase, how many letters it has. You’re processing the surface appearance without engaging with meaning at all.
- Phonemic (sound) processing is the middle level. You think about how a word sounds, whether it rhymes with another word, or how you’d pronounce it. This engages more cognitive resources than visual processing but still doesn’t touch meaning.
- Semantic (meaning) processing is the deepest level. You think about what the word actually means, how it relates to other concepts, or how it fits into a sentence. This level produces the strongest, most lasting memories.
Classic experiments demonstrated this hierarchy by giving participants different tasks for the same list of words. When people were asked to judge the typeface of a word (structural), they remembered very little later. When asked whether a word rhymed with another word (phonemic), recall improved. When asked to categorize words by meaning or decide how well they fit in a sentence (semantic), recall was significantly better. The information was identical each time. Only the processing changed.
Why Semantic Processing Works So Well
Semantic processing creates stronger memories because it forces you to build connections. When you think about what something means, you’re linking it to your existing web of knowledge, giving your brain multiple pathways to retrieve that information later. This is sometimes called elaborative rehearsal: the process of actively relating new material to things you already understand.
This stands in sharp contrast to maintenance rehearsal, which is simply repeating information over and over. Rereading your notes ten times or silently repeating a phone number are examples of maintenance rehearsal. Research shows these two types of rehearsal produce qualitatively different memory experiences. Elaborative rehearsal leads to genuine remembering, the kind where you can mentally re-experience the learning event. Maintenance rehearsal, on the other hand, produces a weaker sense of familiarity, a feeling of “knowing” something without truly remembering how or when you learned it.
Brain imaging research supports this distinction. When people process information deeply, there’s greater activation in frontal and medial-temporal brain regions, areas associated with meaning-making and memory consolidation. Some sensory regions activated during the initial learning are also reactivated during later retrieval, suggesting that deep processing creates richer neural representations that are easier to reconstruct.
The Self-Reference Effect
One of the most striking extensions of levels of processing theory is the self-reference effect. When you process information in relation to yourself (“Does this word describe me?”), memory improves even beyond standard semantic processing. Studies consistently show that self-referential encoding produces significantly better recognition than semantic encoding, which in turn beats structural encoding. This pattern holds across all age groups.
Researchers believe self-referential processing doesn’t just add more semantic detail. It creates a qualitatively different kind of memory trace, one enriched by your self-concept and personal associations. Your sense of self is one of the most elaborate and well-connected knowledge structures in your brain, so tying new information to it gives that information an unusually rich network of retrieval cues.
Practical Applications for Learning
The theory translates directly into study strategies. Any technique that pushes you toward semantic processing will produce better retention than techniques that keep you at the surface level.
Rereading highlighted text is shallow. Asking yourself “What does this mean?” or “How does this connect to what I learned last week?” is deep. Writing a summary in your own words forces semantic processing because you have to understand the material well enough to rephrase it. Teaching a concept to someone else pushes you even deeper, since you need to organize the information logically and anticipate questions. Relating new material to your own life (leveraging the self-reference effect) adds another layer of depth.
Even something as simple as changing the questions you ask while studying can shift your processing level. Instead of “What does this term look like on the page?” try “Why does this matter?” or “How would I explain this to a friend?” The goal is always to move from passive exposure to active engagement with meaning.
Criticisms and Limitations
The biggest criticism of levels of processing theory is that the concept of “depth” is circular. Deep processing is defined as processing that leads to better memory, and better memory is then explained by deep processing. Without an independent way to measure depth before seeing the memory results, critics argue the theory isn’t truly testable in a rigorous scientific sense. Craik himself acknowledged this limitation in later writings.
Another important nuance came from research on transfer-appropriate processing, which showed that deep processing doesn’t always win. Memory is best when the type of processing at encoding matches the type of processing required at retrieval. If a test specifically asks about how words sounded (a phonemic retrieval task), then phonemic encoding can actually outperform semantic encoding for that particular test. The advantage of deep processing, then, may partly reflect the fact that most real-world memory tasks are meaning-based, so semantic encoding is usually the best match. One way to reconcile these views is that deep, semantic processing produces what researchers call “robust encoding,” memory traces that are useful across the widest variety of retrieval situations.
Despite these criticisms, the core insight remains influential: what you do with information when you first encounter it matters far more than how many times you encounter it.