Cognitive Psychology
Memory Encoding, Storage, Retrieval
Three stages of how memory works — and how each can fail
Memory is not a single act but a sequence. Encoding converts perception into a memory trace (acoustic, visual, semantic codes). Storage maintains the trace across time, distributed across hippocampus and cortical regions. Retrieval reactivates the trace, often reconstructively rather than as playback. Atkinson and Shiffrin (1968) modeled three stores: sensory (≤1s), short-term/working (15-30s, ~7±2 items, Miller 1956), and long-term (potentially permanent, declarative + procedural). Craik and Lockhart (1972) refined this with levels-of-processing — deeper semantic processing produces stronger memories than shallow rehearsal. Failure can occur at any stage: inattention blocks encoding, decay or interference erodes storage, retrieval cues may be missing. Each stage is a distinct lever for learning, testimony, and clinical care.
- Three stagesEncoding, storage, retrieval
- Multistore modelAtkinson & Shiffrin (1968)
- STM capacity7±2 items (Miller 1956)
- Levels of processingCraik & Lockhart (1972)
- Key brain regionHippocampus (consolidation)
- Famous patientH.M. — bilateral hippocampus removal, no new long-term memories
Interactive visualization
Press play, or step through manually. The visualization is yours to drive — try it before reading on.
Watch the 60-second explainer
A condensed visual walkthrough — narrated, captioned, under a minute.
Why memory stages matter
- Studying. Locate failures — encoding (attention), storage (sleep), or retrieval (cues).
- Eyewitness testimony. Reconstructive retrieval makes confident recall unreliable.
- Therapy. Trauma memory work targets reconsolidation windows.
- UX and instruction design. Respect working-memory limits — chunk, label, scaffold.
- Aging and dementia. Different conditions impair different stages; diagnosis depends on which.
- Marketing. Recall of brands depends on encoding depth, not exposure count alone.
- Skill learning. Procedural memory uses different systems from declarative facts.
Common misconceptions
- Memory is recording. It's reconstructive — every retrieval edits the trace.
- Forgetting equals failure. Forgetting is adaptive; it filters irrelevant detail.
- Re-reading works. Passive review feels productive but underperforms self-testing.
- Photographic memory exists. True eidetic memory is rare and fades by adulthood.
- Confidence implies accuracy. Vivid, confident memories can be entirely fabricated.
- STM and working memory are identical. Working memory adds active manipulation.
Frequently asked questions
What is encoding?
Transforming perception into a storable trace. Codes vary. Acoustic — sound; dominant in short-term memory. Visual — imagery. Semantic — meaning; dominates long-term memory. Encoding is selective; only attended information becomes a robust memory. Distraction during input is the most common reason something feels "forgotten" — it was never encoded. Depth of processing matters more than rehearsal time.
What is short-term vs working memory?
Short-term memory (Atkinson-Shiffrin) is passive storage of ~7 items for ~20 seconds without rehearsal. Working memory (Baddeley & Hitch 1974) extends this — it actively manipulates information. Components include a phonological loop (verbal), a visuospatial sketchpad (imagery), an episodic buffer, and a central executive. Working memory predicts academic achievement and reasoning ability better than short-term capacity alone.
What is consolidation?
The transition from labile short-term to stable long-term memory. Hippocampus binds elements during initial encoding; over hours to years, traces gradually become cortex-dependent (systems consolidation). Sleep — especially slow-wave and REM — promotes consolidation. Disruption during the consolidation window (concussion, anesthesia) can erase recent memories without affecting older ones. Reactivation reopens this window; reconsolidation can edit the trace.
How does retrieval work?
Reactivating a trace via cues. Encoding specificity (Tulving & Thomson 1973) — best recall when retrieval cues match encoding context. State-dependent (mood) and context-dependent (place) effects follow. Recall reconstructs rather than replays; gaps are filled with schemas and inferences. The act of retrieval itself strengthens memory (testing effect, Roediger & Karpicke 2006), making practice tests more effective than re-reading.
Why are memories distorted?
Storage and retrieval are reconstructive. Bartlett (1932, "War of the Ghosts") showed retellings shift toward cultural schemas. Loftus's misinformation paradigm — leading questions implant false details. Reconsolidation makes retrieved memories temporarily editable. Eyewitness testimony is unreliable for this reason; confidence does not track accuracy. Memory functions for behavioral guidance, not historical archive — distortion is a feature, not a bug.
Who was patient H.M.?
Henry Molaison (1926-2008). Bilateral medial temporal lobectomy in 1953 to treat epilepsy removed his hippocampi and surrounding tissue. Result: severe anterograde amnesia — could not form new declarative memories — while procedural learning, working memory, and old semantic knowledge remained intact. Brenda Milner's studies established the hippocampus as essential for long-term declarative memory. H.M. is the most studied patient in neuroscience history.
What improves memory?
Evidence-based methods. (1) Spaced repetition — distribute practice across days. (2) Retrieval practice — self-testing beats re-reading. (3) Elaboration — link new content to existing knowledge. (4) Sleep within 24 hours of learning. (5) Interleaving topics rather than blocking. (6) Generation effect — produce answers before checking. (7) Dual coding — verbal plus visual. Mnemonics (method of loci) work by adding semantic and spatial structure to arbitrary content.