Subjects and lessons to explore.

Law

Education · Educational Psychology

Memory and Learning

13 min read 5 practice questions
10 key termsSourcesReport an error
Want it in plain words first? Jump to Eli explains — the same idea, no jargon.
On this page 9 sections
  1. In 30 seconds
  2. Why this matters
  3. The college version
  4. Eli explains
  5. Worked example
  6. Key takeaway
  7. Quick check
  8. Study tools
  9. Sources & references

In 30 seconds

Memory is not one store, and forgetting is not one process. Encoding, storage, and retrieval fail in different ways and need different repairs. Long-term memory splits into memory for events, memory for facts, and memory for skills. Most study time goes into rereading and highlighting, which the evidence rates poorly, while the two techniques that survive every test - spacing practice out and retrieving instead of reviewing - feel harder and get used least.

Why this matters

Teacher preparation hands you two accounts of memory that look alike and are not: a folk psychology built from metaphors of recording and storage, and a research literature that has spent a century measuring what actually survives. The gap between them is where bad study advice lives. Students who reread until a page feels familiar are measuring fluency, not learning, and they keep doing it because it works on Friday's quiz and fails on June's final. Knowing how retrieval, spacing, and interference behave lets you build a review schedule instead of improvising one, explain to a student why a method failed, and recognize the memory myths that still circulate in staff rooms and professional development slides.

The college version

Three stages, three different failures

Memory research separates the initial learning of information, its persistence over time, and the act of getting it back. The separation earns its keep because all three failures look identical from outside - a student who cannot answer - while each calls for a different response. Material never attended to was never encoded. Material encoded but never consolidated may genuinely be gone. Material still present but unreachable is a retrieval problem, and retrieval problems are the ones a teacher can usually solve, because retrieval runs on cues. The National Academies' How People Learn II puts the point in a form worth keeping: retrieval is triggered and guided by the retrieval cues available in the learner's environment, so what a student can produce depends on what the question and the format offer her. That gives you a ten-second diagnostic. Offer a hint. If the answer arrives, the memory was there all along and only unreachable - the trace existed, the route did not. Psychologists mark this as the difference between what is available in storage and what is accessible right now, and it runs underneath everything else in this lesson.

Working memory has parts, and long-term memory has kinds

How much fits in working memory at once, and how to design materials around that limit, is the subject of the cognitivism lesson in this subject; this section is about structure rather than size. Baddeley and Hitch's multicomponent model divides temporary storage by code instead of treating it as one box. A phonological loop holds verbal and acoustic material - the inner voice repeating a phone number. A visuospatial sketchpad holds visual and spatial material. A central executive directs attention among them and controls what gets manipulated. In 2000 Baddeley added a fourth component, the episodic buffer, a limited store that binds material from the subsystems and from long-term memory into integrated episodes. The framework replaced a single short-term store because working memory does not merely hold information, it operates on it. Long-term memory is likewise not one thing. Squire and Dede split it into declarative memory, knowledge available to conscious recollection and dependent on the hippocampus and surrounding medial temporal lobe, and nondeclarative memory, an umbrella covering skills and habits in the striatum, priming, emotional associations in the amygdala, and classical conditioning in the cerebellum. Declarative memory subdivides again into for specific events located in time and place, and for facts detached from the occasion of learning them. The clinching evidence is patient H.M., who after medial temporal lobe surgery could not form new declarative memories yet still acquired motor skills across days of practice, improving at a task he had no recollection of ever performing. Memory, in Squire's phrase, is not a single entity. The classroom consequence follows directly: a student can execute a titration flawlessly and be unable to say what it demonstrates, and if your objective names a skill, a definitional test will not measure it.

Depth of processing, and the problem it never solved

Craik and Lockhart's 1972 levels-of-processing framework proposed that retention is a byproduct of how material is analyzed rather than of how long it is held. Analyzing surface form - what a word looks like, what it rhymes with - leaves a weaker trace than analyzing meaning. The practical instruction that follows is the framework's most durable contribution and is still what the open memory literature tells students: at study, think about the meaning of the material, form associations, and build vivid mental images. Elaborate, do not repeat. The theoretical core is shakier, and Craik said so himself. His 2002 retrospective takes up the standing criticisms directly: whether memory can be treated as pure processing, how depth could be measured, how encoding and retrieval conditions interact, and whether levels are real in any strong sense. The sharpest objection is circularity - if depth is inferred from how well material was remembered, and remembering is then explained by depth, the framework has redescribed the result rather than explained it. Treat as a reliable heuristic for designing study tasks, not a dial with units on it.

The forgetting curve, told honestly

In 1879 and 1880 Hermann Ebbinghaus ran 163 double tests on himself, learning eight series of thirteen nonsense syllables (six in some sessions) until he could recite them twice without error and then relearning them after about a third of an hour, an hour, nine hours, one day, two days, six days, or thirty-one days. He measured retention as : the proportion of the original learning work he no longer needed on relearning. The result is the curve everyone has seen - a steep early drop flattening into a long tail. Roughly 58 percent of the work was saved after twenty minutes, 44 percent after an hour, 34 percent after a day, a quarter after six days, and about a fifth after a month. Now the caveats, the first of which is Ebbinghaus's own: he called the numbers special, individual, and uncertain, and judged the 33.7 percent one-day value one or two points too high. Second, the sample was one person, himself. Third, the material was deliberately meaningless - he chose nonsense syllables because ordinary words and poetry evoked variable meanings and added measurement error, which is to say he engineered out the feature that most defines classroom material. Murre and Dros tested the curve again in 2015 with the same method: one subject, about seventy hours of learning and relearning across the same seven intervals. Their curve closely resembled Ebbinghaus's at nearly every point, diverging sharply only at thirty-one days, where savings fell to 0.041 against Ebbinghaus's 0.211, and they concluded the curve is not perfectly smooth, most likely jumping upward around the twenty-four-hour point. So the shape is real and has been reproduced across more than a century; the numbers describe one person relearning meaningless syllables. Cite the shape, not the percentages.

Why we forget: decay, interference, retrieval failure

Three explanations compete, and a fourth possibility - that the material was never encoded - has to be ruled out first. Decay theory says traces fade with the passage of time. It is the oldest account and the weakest, because disuse does not reliably produce forgetting; contemporary versions talk about consolidation being disrupted rather than about time doing the damage. Interference says memories compete. is older learning blocking newer: the first two conjugation patterns make the third harder to acquire. is newer blocking older: the third pattern makes the first two harder to retrieve. Retrieval failure says the memory is intact but the cue inadequate, which is what a tip-of-the-tongue state is. All three contribute, and the honest position is that they are not tidy rivals. What matters for teaching is that they are distinguishable in practice. Give a cue: if the answer comes back, you were looking at retrieval failure, and the fix is retrieval practice under varied cues rather than more exposure. Look at what was learned nearby: if the errors are systematic confusions between similar items rather than blanks, you are looking at interference, and the fix is to separate confusable material in time and practice the discrimination directly.

Encoding specificity, and what context really buys you

The states that a cue is effective to the degree that it overlaps the memory trace of the original experience. It is the most usable idea in the retrieval literature, and it is routinely oversold. The famous demonstration is Godden and Baddeley's 1975 study of divers who learned word lists on land or underwater and recalled them in the same or the other environment; watch what happened to it. Smith and Vela's 2001 meta-analysis found environmental context effects reliable overall but heavily moderated - they shrink when learners use non-contextual cues during study (overshadowing), when such cues are available at test (outshining), and when learners simply imagine the original setting at test. Then in 2021 Jaap Murre reran the diving experiment with sixteen divers and failed to replicate it: the crucial interaction between learning and recall environment was not significant, F = 1.999, p = 0.163. A 2022 addendum reanalyzed the data as a repeated-measures ANOVA with an order covariate, which did make the interaction significant at p = 0.017, but in a pattern that did not match the original, and Murre restated that the replication was unsuccessful. So separate two claims that get bundled together. The principle - cues work in proportion to their overlap with what was encoded - is well supported and directly actionable. The strong environmental version, that a student must be tested where she studied, is not something to build practice on. Match the cues that carry meaning: practice retrieving in the format, with the prompt wording, and against the kind of question the assessment will use.

What actually works, in order

Dunlosky, Rawson, Marsh, Nathan and Willingham reviewed ten study techniques in 2013, judging each against four dimensions: learning conditions, student characteristics, materials, and criterion tasks. Two earned high utility - practice testing and distributed practice - because their benefits held across ages, abilities, materials, and outcome measures, including in real educational settings. Three earned moderate utility: elaborative interrogation, self-explanation, and interleaved practice. Five earned low utility: summarization, highlighting and underlining, the keyword mnemonic, imagery use for text learning, and rereading. Read the middle tier correctly - moderate means thin evidence, not demonstrated failure. The is the anchor result. Roediger and Karpicke had students read prose passages and then either take immediate free-recall tests with no feedback or restudy the passage the same number of times. Five minutes later, restudying won. At two days and one week, the tested students retained substantially more, while the restudying students were the more confident group. That reversal is the finding to carry around: immediate fluency and durable retention can point in opposite directions. Spacing is the other high-utility technique, and Cepeda's meta-analysis of 839 assessments across 317 experiments in 184 articles established the design rule - the gap between study sessions that maximises final retention grows as the delay before the test grows. needs the most care, because a later meta-analysis sharpened it. Brunmair and Richter pooled 59 studies and 238 effect sizes across 158 samples and found a moderate overall benefit, Hedges' g = 0.42, with the material doing most of the work: 0.67 for paintings and other visual categories, 0.34 for mathematics, non-significant for expository texts, and minus 0.39 for word learning, where blocking beat interleaving. So interleave problem types and categories that students confuse; do not interleave a vocabulary list. Elaborative interrogation - asking why a stated fact is true - and self-explanation - saying how a new idea relates to what you know - work by forcing generation rather than recognition. The two techniques students report using most, rereading and highlighting, do not consistently improve performance, and the recommendation is substitution rather than addition: practice testing instead of rereading, not after it.

Two myths, and one precise correction

The claim that we use ten percent of our brain is false, and teachers believe it. Dekker and colleagues surveyed 242 teachers interested enough in the neuroscience of learning to take part - 137 in the United Kingdom, 105 in the Netherlands - and 48 percent and 46 percent respectively endorsed it. The refutation is straightforward. The brain is about two percent of body weight and consumes roughly twenty percent of the body's calories, and that consumption does not drop when a task is simple or when a person is asleep. Recovery after the loss of an entire hemisphere in early childhood reflects rewiring, not spare capacity sitting idle. The idea appears to descend from William James's 1907 remark that people use only a small part of their possible resources - a claim about effort, not anatomy. The second myth is subtler, because the correction gets overstated in turn. How People Learn II rejects the file-cabinet picture outright: memory systems give people the ability to produce knowledge without storing copies of it, and memories are reconstructed anew each time they are recalled, taking into account current knowledge, expectations, and context. Loftus's thirty-year program on the misinformation effect shows the consequence - memory for an event is impaired by misleading information encountered afterwards, warnings help only in limited circumstances, and people can be led to believe they experienced complex events that never happened. Be precise about what that licenses. Reconstruction does not mean memory is generally unreliable; the same systems support accurate recall constantly. It means retrieval is an active process that can incorporate material from outside the original episode, so confidence is not a reliable index of accuracy. The classroom corollary is uncomfortable and useful: a confidently recalled misconception feels exactly like a confidently recalled fact, which is an argument for retrieval practice with feedback rather than retrieval practice alone.

Eli, the EliExplains learning guide

Eli explains

The same idea, in plain words

Explain it like I’m 10

Your brain does not save a copy of a lesson the way a phone saves a photo. It keeps a web of connections, and remembering means rebuilding the thing out of that web. That is why three completely different things can go wrong. Sometimes you never built it, because you were not really paying attention. Sometimes you built it and it came apart. And sometimes it is still in there but you cannot find the way to it, which is why one small hint can bring back a name you were sure you had lost for good. Studying works best when you practice the rebuilding instead of practicing the looking.

Picture it like this

Learning something is like wearing a path across a field of tall grass. Every time you walk it yourself, the path gets clearer and easier to find again. Rereading your notes is like standing at the edge of the field looking at a photo of the path. It feels familiar, and it does almost nothing to the grass. Testing yourself is walking it. Spacing your practice out is coming back before the grass has grown all the way in, so each trip does real work.

Where the picture stops working

The path picture gets two things wrong. Real memories are not fixed routes. Each time you walk one you can change it a little, so a memory can come back altered after you hear someone else's version of the story - a path cannot pick up a bend from a rumor. And a field has one kind of ground, while memory has several systems that behave differently. Remembering your first day of school, knowing that a triangle has three sides, and being able to ride a bike run on different machinery, and damage that wipes out one can leave the others working fine.

Worked example

A student has a cumulative biology final in six weeks and plans to reread the textbook chapters twice, highlighting as she goes. Rebuild the plan from the utility ratings. Rereading and highlighting are both low utility, so they get substituted rather than supplemented. Weeks one to five: after each chapter she closes the book and free-recalls everything she can onto blank paper, then checks against the text and corrects what she got wrong - practice testing with feedback, so retrieval strengthens the right answer rather than her guess. She revisits each chapter on a widening schedule instead of once, because Cepeda's synthesis shows the optimal gap grows with the retention interval, and her interval is six weeks. Cell respiration and photosynthesis, which she keeps confusing, get interleaved in mixed practice sets and compared directly, since interference concentrates on similar material. Vocabulary drilling stays blocked - Brunmair and Richter found interleaving hurt word learning. For each process she asks why each step must happen in that order, which is elaborative interrogation. Final week: full practice exams in the format the real test will use, matching cues to the assessment rather than to her notes.

Key takeaway

Treat memory as three separable processes running over several distinct systems, and study strategy follows from the diagnosis: retrieval failures need cued practice, interference needs separation and discrimination, and the methods that feel hardest - self-testing and spaced review - are the two the evidence rates highest.

Quick check

3 questions here, of 5 in this lesson’s practice set. Answers stay hidden until you check.

Question 1 of 3foundational

In Baddeley and Hitch's multicomponent working memory model, which component was added in 2000, and what does it do?

Choose an answer, then check it.
Question 2 of 3intermediate

A professional development slide states that students forget about two thirds of a lesson within twenty-four hours, citing the Ebbinghaus forgetting curve. What is the strongest objection?

Choose an answer, then check it.
Question 3 of 3intermediate

A student preparing for a cumulative exam in six weeks currently rereads each chapter twice and highlights as she goes. Which revision is best supported by the utility ratings and the underlying studies?

Choose an answer, then check it.
Practice all 5

Keep learning

Ready to build on this? Continue to the next lesson.

Practice this lesson
Study tools & related lessonsYou’ll learn to · Common mistakes · Easily confused · Key vocabulary · Related

You’ll learn to

  • Distinguish encoding, storage, and retrieval failures, and use cue-supported recall to tell an inaccessible memory from an absent one.
  • Describe the components of Baddeley and Hitch's working memory model and the declarative/nondeclarative division of long-term memory, and explain what each division predicts about assessment.
  • Explain the Ebbinghaus forgetting curve together with the methodological limits of the single-subject nonsense-syllable studies that produced it, including the 2015 replication.
  • Analyze a case of forgetting in terms of decay, interference, and retrieval failure, and identify which explanation the available evidence supports.
  • Apply the Dunlosky et al. (2013) utility ratings to evaluate or redesign a student's study plan, including where interleaving should not be used.
  • Evaluate the claims that we use ten percent of our brain and that memory records experience, and state precisely what reconstructive memory does and does not imply.

Common mistakes

  • Quoting the Ebbinghaus forgetting curve as the rate at which students forget lessons - for instance, telling teachers that pupils lose about 70 percent of a lesson within a day.

    The curve's shape has replicated, but its percentages come from single-subject studies in which one man relearned deliberately meaningless syllables using the savings method. Ebbinghaus himself called the values special, individual, and uncertain. Use the shape as a reason to schedule early review; do not quote the numbers as classroom forecasts.

  • Judging a study method by how fluent it feels, so that rereading and highlighting keep their place because the material feels familiar afterwards.

    Roediger and Karpicke found restudying beat testing after five minutes and lost badly after two days and a week, while the restudying students were the more confident group. Fluency in the moment and retention at the delay can point in opposite directions, so the felt sense of knowing is not evidence.

  • Interleaving everything once you learn that interleaving works, including vocabulary and word lists.

    Brunmair and Richter's meta-analysis found a moderate average benefit but a reversal for word materials, where blocking outperformed interleaving. Interleave problem types and categories learners confuse; keep straightforward word learning blocked.

  • Reading reconstructive memory as proof that memory is generally unreliable, or, in the other direction, treating a confident recollection as a recording of what happened.

    Memory produces knowledge rather than replaying stored copies, and post-event information can alter what is recalled - so confidence is not an index of accuracy. That does not make everyday recall untrustworthy. The teaching consequence is narrower and more useful: retrieval practice needs feedback, because unchecked retrieval strengthens whatever the student produced, including the misconception.

  • Repeating that we only use ten percent of our brain, or implying students have untapped memory capacity waiting to be unlocked.

    Roughly half the teachers in Dekker's UK and Dutch samples endorsed this, and it is false. The brain is about two percent of body weight and uses about twenty percent of the body's calories, activity continues throughout the brain during sleep, and the idea traces to a remark by William James about unused effort, not unused tissue.

Easily confused

Episodic memory vs. Semantic memory

Both are declarative and consciously accessible, but episodic memory reinstates a particular event in time and place, while semantic memory holds facts stripped of the occasion on which they were learned. A student can lose the memory of the lesson and keep the content.

Declarative memory vs. Procedural memory

Declarative memory is expressed by saying or recognizing; procedural memory is expressed by doing. They depend on different brain systems, which is why H.M. could improve at a motor skill across days while having no recollection of practicing it.

Decay vs. Retrieval failure

Decay claims the trace is gone; retrieval failure claims the trace is intact but unreached. The test is a cue: if a hint restores the answer, nothing had decayed, and the fix is retrieval practice rather than reteaching.

Proactive interference vs. Retroactive interference

Both involve competition between similar memories; the direction differs. Proactive means older material obstructs newer, retroactive means newer material obstructs older. Which one you have determines whether the fix is spacing the new material out or re-establishing the old.

Blocked practice vs. Interleaved practice

Blocking finishes one kind of problem before starting the next; interleaving mixes them. Interleaving helps most when categories are confusable and the material is complex, and it can hurt for simple word learning, so the schedule should follow the material rather than a slogan.

The encoding specificity principle vs. The claim that students should be tested where they studied

The principle - cues help in proportion to their overlap with encoding - is well supported. The environmental version is fragile: meta-analytic context effects shrink under overshadowing, outshining, and mental reinstatement, and the famous diving study failed to replicate in 2021. Match the format and the prompts, not the furniture.

Key vocabulary

encoding specificity principle
The rule that a hint or prompt helps in proportion to how much it overlaps with what was actually registered at the time of learning.
savings
Ebbinghaus's measure of retention: the proportion of the original learning effort no longer needed when the same material is learned a second time.
episodic memory
Conscious recollection of particular events from one's own life, located in a specific time and place.
semantic memory
Conscious knowledge of facts, concepts, and word meanings, held independently of when or where it was acquired.
procedural memory
A form of nondeclarative memory expressed in performance rather than recollection, covering motor and cognitive skills and habits.
proactive interference
Disruption in which material learned earlier competes with and blocks newer, related material.
retroactive interference
Disruption in which recently learned material competes with and blocks the recall of older, related material.
testing effect
The finding that attempting to recall studied material improves later retention more than restudying it for the same amount of time.
interleaving
A practice schedule that mixes problems or categories together rather than completing all of one kind before moving to the next.
levels of processing
Craik and Lockhart's framework holding that retention depends on whether material was analyzed for meaning or only for surface features.

Sources & references

  1. Memory (Encoding, Storage, Retrieval) (Noba Textbook Series: Psychology) — Kathleen B. McDermott and Henry L. Roediger III / Noba Project, DEF Publishers
  2. Forgetting and Amnesia — Nicole Dudukovic and Brice Kuhl / Noba Project, DEF Publishers
  3. Working Memory From the Psychological and Neurosciences Perspectives: A Review — Wen Jia Chai, Aini Ismafairus Abd Hamid, Jafri Malin Abdullah; Frontiers in Psychology 9:401
  4. From short-term store to multicomponent working memory: The role of the modal model (Memory and Cognition, 47(4), 575-588) — Alan D. Baddeley, Graham J. Hitch, Richard J. Allen; author accepted manuscript, White Rose Research Online
  5. Conscious and Unconscious Memory Systems — Larry R. Squire and Adam J. O. Dede; Cold Spring Harbor Perspectives in Biology, via PubMed Central PMC4355270
  6. Levels of processing: past, present. and future? — Fergus I. M. Craik; Memory 10(5-6), 305-318, 2002; PubMed record 12396643
  7. Memory: A Contribution to Experimental Psychology, Chapter VII: Retention and Obliviscence as a Function of the Time — Hermann Ebbinghaus (1885), translated by Henry A. Ruger and Clara E. Bussenius (1913); Classics in the History of Psychology, ed. Christopher D. Green, York University
  8. Replication and Analysis of Ebbinghaus' Forgetting Curve — Jaap M. J. Murre and Joeri Dros; PLoS ONE 10(7):e0120644, via PubMed Central PMC4492928
  9. Environmental context-dependent memory: a review and meta-analysis — Steven M. Smith and Edward Vela; Psychonomic Bulletin & Review 8(2), 203-220, 2001; PubMed record 11495110
  10. The Godden and Baddeley (1975) experiment on context-dependent memory on land and underwater: a replication — Jaap M. J. Murre; Royal Society Open Science 8(11):200724, via PubMed Central PMC8568063
  11. Addendum to: Murre (2021). The Godden and Baddeley (1975) experiment on context-dependent memory on land and underwater: a replication — Jaap M. J. Murre; Royal Society Open Science 9(1):211924, via PubMed Central PMC8728159
  12. Improving Students' Learning With Effective Learning Techniques: Promising Directions From Cognitive and Educational Psychology — John Dunlosky, Katherine A. Rawson, Elizabeth J. Marsh, Mitchell J. Nathan, Daniel T. Willingham; Psychological Science in the Public Interest 14(1), 4-58, 2013; PubMed record 26173288
  13. Strengthening the Student Toolbox: Study Strategies to Boost Learning — John Dunlosky; American Educator (American Federation of Teachers), Fall 2013
  14. Test-enhanced learning: taking memory tests improves long-term retention (Psychological Science, 17, 249-255) — Henry L. Roediger III, Jeffrey D. Karpicke; PubMed record 16507066
  15. Distributed practice in verbal recall tasks: A review and quantitative synthesis (Psychological Bulletin, 132, 354-380) — Nicholas J. Cepeda, Harold Pashler, Edward Vul, John T. Wixted, Doug Rohrer; publication page, York University
  16. Similarity matters: A meta-analysis of interleaved learning and its moderators — Matthias Brunmair, Tobias Richter; Psychological Bulletin 145(11), 1029-1052, 2019; PubMed record 31556629
  17. How People Learn II: Learners, Contexts, and Cultures, Chapter 4: Processes That Support Learning — National Academies of Sciences, Engineering, and Medicine; The National Academies Press
  18. Planting misinformation in the human mind: a 30-year investigation of the malleability of memory — Elizabeth F. Loftus; Learning & Memory 12(4), 361-366, 2005; PubMed record 16027179
  19. Neuromyths in education: prevalence and predictors of misconceptions among teachers — Sanne Dekker, Nikki C. Lee, Paul Howard-Jones, Jelle Jolles; Frontiers in Psychology 3:429
  20. Do we use only 10 percent of our brain? — Rubina Veerakone, McGovern Institute for Brain Research, Massachusetts Institute of Technology (26 January 2024), quoting graduate researcher Mila Halgren

EliExplains lessons are original prose written from the open, credible references above. See Copyright & Licensing.

Researched 2026-08-18

Educational content only. It is not medical, legal or professional advice. Found an error? Tell us.