Encoding is the process by which information enters long-term memory. Craik and Lockhart's levels-of-processing framework proposes that deeper, more meaningful processing produces more durable memory traces than shallow processing such as analyzing only orthographic or phonological features. Effective encoding strategies include elaborative rehearsal, self-referential encoding, the generation effect, interleaving, and the use of imagery and mnemonics.
Compare retention after deep versus shallow encoding tasks: ask whether a word describes you personally versus whether it is printed in uppercase. The dramatic recall difference demonstrates that quality of encoding, not sheer repetition, determines memory strength.
From your working memory model, you know that information enters consciousness through short-term and working memory systems with severe capacity limits. Simply holding information there — cycling it through the phonological loop via maintenance rehearsal — keeps it active but does not reliably transfer it to long-term memory. Craik and Lockhart's levels of processing framework (1972) explains why: what determines whether information reaches long-term memory is not how long it is rehearsed but how *deeply* it is processed. Shallow processing analyzes only surface features (Is this word in capital letters? Does it rhyme with "cat?"). Deep processing analyzes meaning (Does this word fit in the sentence? Does it describe you?). The deeper the encoding, the richer and more distinctive the memory trace, and the more retrieval cues can later access it.
This matters practically because the most common study strategy — re-reading — is essentially maintenance rehearsal applied to text. It is effortful and time-consuming yet produces weak encoding because it does not require meaningful processing. Elaborative rehearsal is the alternative: connecting new information to things you already know, generating explanations, asking why, relating concepts to personal experience. When you connect the new idea to an existing schema in long-term memory (which you already know can store vastly more than working memory), you create multiple pathways back to the memory. Retrieval is essentially pattern matching, and more connections mean more patterns that can trigger successful recall.
The self-reference effect is a particularly powerful instance of deep processing: judging whether information applies to yourself ("Am I an organized person?") produces better retention than judging whether it applies to someone else or whether it is true in the abstract. This is because self-referential processing activates a rich, well-elaborated self-schema in long-term memory, producing an unusually well-integrated encoding. The generation effect operates differently: generating an answer rather than reading it — even if you generate the wrong answer first — produces significantly stronger memories, because generation requires deep retrieval-like processing during encoding. This is why practice testing outperforms re-reading even when testing itself produces no feedback.
Imagery and mnemonic techniques amplify encoding by adding visuospatial information to a verbal trace, effectively doubling the encoding routes. The method of loci (placing items to be remembered at locations on a familiar mental path) exploits the richness of spatial memory to encode otherwise arbitrary lists. Interleaving — mixing different types of problems or topics rather than blocking practice on one type at a time — feels harder and produces slower initial acquisition but substantially better long-term retention and transfer, because it forces the learner to retrieve and re-consolidate each schema rather than riding momentum within a single topic. The common thread across all effective encoding strategies is that they require *effortful engagement with meaning* — the brain prioritizes for long-term storage information that it has worked to process deeply, not information that has merely been in front of the eyes repeatedly.