Encode deeply to retain durably
Memory is not passive recording. The deeper an item is encoded -- meaning the more semantic, emotional, or sensory processing is involved -- the stronger and more durable the memory trace. This is what cognitive psychologists call levels of processing.
A word simply read on a list is shallowly encoded: the brain processes it as a visual signal without giving it rich meaning. A word encountered in a sentence that surprised you, associated with a vivid mental image, linked to an unexpected etymology -- that word is deeply encoded, with multiple connections in the memory network.
The techniques in this guide do not eliminate the need for spaced repetition -- they increase its efficiency by strengthening the quality of initial encoding. The SRS algorithm manages the review schedule; the encoding techniques strengthen what happens in your mind at each review.
Levels of processing: the scientific basis
Craik and Lockhart (1972) showed that the depth of processing of information at the moment of encoding predicts its subsequent retention better than simple repetition. Processing a word semantically -- what does it mean? in what context would it be used? what does it connect to in my existing knowledge? -- produces far superior retention compared to reading it or repeating it phonetically.
Three depth levels are distinguished: structural processing (how does the word look or sound?), phonological processing (does it rhyme with something?), and semantic processing (what does it mean, how is it used?). Semantic processing produces the strongest memory traces. This is why reading a word 20 times without thinking about its meaning produces much weaker retention than encountering it once in a meaningful context.
Why encoding technique and spaced repetition must work together
Spaced repetition ensures that a memory trace is reviewed at the optimal moment -- just before it fades, which produces the strongest consolidation. But the quality of each review depends entirely on what happened during initial encoding. A poorly encoded word produces a fragile trace; the algorithm will reschedule it repeatedly, but cannot replace the missing quality of encoding.
Conversely, the strongest encoding technique without systematic review leads to the classic forgetting curve: within days or weeks, even deeply encoded items will fade without reinforcement. The combination -- deep encoding at first contact, spaced reviews calibrated by algorithm -- is what produces durable mastery at scale.
Craik and Lockhart demonstrated that depth of processing at encoding predicts later retention better than repetition frequency alone. Semantic processing (meaning, usage, associations) produces memory traces 2-3x more durable than structural or phonological processing.
Craik & Lockhart (1972), Levels of processing: A framework for memory research, Journal of Verbal Learning and Verbal Behavior.Which words to learn first: Zipf's law
Before addressing encoding techniques, a strategic question must be answered: in what order should you learn vocabulary? The answer rests on a robust empirical principle known as Zipf's law.
Zipf's law describes the frequency distribution of words in all natural languages. It states that a small number of very frequent words account for the vast majority of occurrences in everyday texts. In practice: the 1,000 most frequent words in a language cover approximately 80% of everyday texts. The next 2,000 words add only about 10% more. Beyond 5,000 words, each new word learned contributes marginally to general comprehension.
The direct consequence for learning: always start with high-frequency words, not thematic or specialised vocabulary. Mastering the 1,000 to 2,000 most frequent words in a language gives you the ability to understand the essence of everyday conversations and common texts -- long before knowing the names of all fruits and vegetables or legal terminology.
- English: COCA lists (Corpus of Contemporary American English) rank words by real frequency across registers
- French (as foreign language): Lexique lists and A1-B2 CEFR vocabulary provide priority targets by level
- Japanese: JLPT vocabulary and kanji lists by level (N5 to N1) are the reference, built on real corpus frequency
- Mandarin: HSK 1 to 4 cover the 1,200 most useful words for everyday communication
- For any other language: search for 'frequency list [target language]' -- learner communities maintain reliable lists
Zipf's law in practice: what it means for your learning strategy
The frequency distribution of language is steeply unequal. In English, the 100 most common words (the, be, to, of, and, a, in...) account for approximately 50% of all written text. The next 900 words bring coverage to 80%. This is why absolute beginners who focus on these core words make very rapid comprehension progress.
A common beginner mistake is starting with thematic vocabulary: fruits, colours, animals, clothes. These lists are intuitive and easy to find, but they contain many low-frequency words -- hippopotamus, turquoise, kimono -- that will not appear in real usage for months. A frequency-first approach guarantees that every word learned immediately increases comprehension of real texts and conversations.
Recommended frequency lists by language
For English learners: the COCA frequency lists (Corpus of Contemporary American English) rank words by real frequency across different registers (academic, fiction, news, spoken). The first 1,000 are the essential foundation.
For Japanese: the JLPT vocabulary and kanji lists by level (N5 to N1) are structured by frequency and are the reference for all serious learners. For Mandarin: HSK 1 to 4 cover the 1,200 most useful words for everyday communication. For other languages: search for 'frequency lists [target language]' -- learner communities maintain reliable, regularly updated lists.
The keyword method
The keyword method creates a sound-based link between a foreign word and a native-language word that sounds similar, then ties both meanings together in a vivid, often absurd mental scene.
Keyword method in practice: concrete examples
Spanish caballo (horse): it sounds phonetically like 'caballero'. Imagine a horse dressed as a caballero -- feathered hat included -- riding another horse. The absurd image anchors the association. Next time you hear caballo, the image resurfaces, and with it the meaning.
Japanese ki (tree -- character 木): the character visually resembles a tree with roots and branches. But for words where the visual connection is not obvious, create a short story with the pronunciation: ki -- a tree whose trunk makes a 'ki ki ki' sound when struck. The narrative, however silly, creates multiple connection points in memory.
English 'salary': it comes from the Latin salarium -- the salt allowance given to Roman soldiers. Once you know this, salary is no longer an arbitrary phoneme sequence but a piece of Roman history. The etymology surprise creates a memory hook that no flashcard repetition alone can replicate.
Why the keyword method works: the evidence
The keyword method exploits two powerful cognitive mechanisms: phonetic association (a bridge between the native language and the target language) and visual encoding (the mental image creates a distinctive trace separate from the word form).
Meta-analyses by Atkinson and Raugh (1975) and subsequent reviews show retention improvements of 30 to 60% for words processed with this method compared to direct memorization. The effect is most pronounced for words with no phonetic or semantic similarity to native-language equivalents -- which is precisely the hardest vocabulary to anchor.
The method has one limitation: it is not scalable for very large volumes. Creating a vivid, memorable keyword association takes 30 to 60 seconds per word -- acceptable for difficult words but prohibitive for entire frequency lists. Use it surgically on the 20% of words that resist after normal reviews.
Atkinson and Raugh's original controlled study showed that learners using the keyword method retained 72% of Russian vocabulary after one week, compared to 46% for rote repetition -- a 56% relative improvement. Subsequent meta-analyses confirmed the effect across multiple languages and age groups.
Atkinson, R.C. & Raugh, M.R. (1975), An application of the mnemonic keyword method to the acquisition of Russian vocabulary, Journal of Experimental Psychology.Memorize words in sentences, not in isolation
A word learned in context is encoded more richly than an isolated word. Context carries information that translation alone cannot convey: register (formal? informal? technical?), typical collocations (what words normally accompany this one?), syntax (is it a transitive verb? an uncountable noun?), and actual meaning in use.
Why context produces superior retention
Psycholinguistic research consistently shows that context-learned words are not only better retained but also more readily mobilized in production. A learner who memorized 'resilience' with the sentence 'Her resilience in the face of adversity was remarkable' can deploy it in speech; a learner who memorized it as 'resilience = mental toughness' can recognize it but struggles to produce it naturally.
The reason is encoding depth: a word in context is processed semantically (what does it mean here?), syntactically (how is it used grammatically?), and pragmatically (in what kind of situation would this be said?). Three simultaneous processing dimensions create three independent retrieval pathways -- any of which can trigger recall when the word is needed.
How to choose the right example sentence
The example sentence on your flashcard should be short (one line), authentic (extracted from a real text or phrased naturally), and memorable. Not a neutrally fabricated sentence like 'The word X means Y' -- a sentence that speaks to you, that illustrates a typical usage, or that contains a detail that makes you react.
Even better: use a sentence from a text, film, or song you enjoyed. The emotional context of the source material creates an additional memory anchor -- you will remember the word all the better for remembering the context in which you discovered it. If you encountered 'resilience' in a novel that moved you, that emotional association becomes part of the retrieval network for the word itself.
Reading extensively in the target language (novels, articles, subtitles) is a form of implicit vocabulary learning in context -- effective long-term but slow for targeting specific vocabulary. Flashcards with example sentences combine the best of both: explicit targeting of priority words + contextual anchoring of each word.
Dual coding: combine verbal and visual encoding
Dual-coding theory (Paivio, 1971) shows that information encoded simultaneously in verbal and visual form is retained better than single-format encoding. The brain has two distinct representation systems -- a verbal system and an imagery system -- and information stored in both creates mutually reinforcing traces.
In practice for vocabulary: associate every important word with a concrete, distinctive mental image -- even for abstract words. The image does not need to be logical or representative; it needs to be distinctive and personal.
- For action verbs: imagine a dynamic scene -- someone performing the action with exaggerated expressiveness
- For adjectives: visualize an object or person that embodies the adjective to an extreme -- the laziest of the lazy, the most brilliant of the brilliant
- For abstract words: find a visual symbol that captures the essence of the concept -- freedom = a door opening onto the horizon, constraint = a rope around the wrists
- Add sensory detail: texture, colour, sound, smell -- the more multisensory the image, the richer the memory trace
Dual coding for concrete words
Concrete words (umbrella, bridge, scaffold, ignition) are straightforward to dual-code: visualize the object with an unusual, distinctive detail. Not a generic umbrella -- a giant neon-pink umbrella that someone is using as a walking stick. The unusual detail is more memorable than a realistic generic image because the brain is wired to notice and remember what deviates from expectation.
For action verbs, imagine a dynamic scene: someone performing the action with exaggerated expressiveness. 'To streamline' -- visualize a worker with scissors manically cutting through layers of bureaucratic paperwork while smiling. The exaggeration and emotional coloring make the image sticky.
Dual coding for abstract words: visual metaphors
Abstract words (resilience, momentum, nuance, deterioration, leverage) require visual metaphors because they have no concrete referent. Create a personal symbol that captures the essence of the concept.
Examples: resilience = a reed bending in the wind without breaking and snapping back to vertical (borrowed from the original French metaphor). Momentum = a boulder rolling downhill, accelerating as it goes, impossible to stop. Nuance = two similar paint colour swatches side by side with only the subtlest difference visible.
The image need not match the formal definition -- it needs to trigger the meaning reliably. A personal or slightly absurd image is more memorable than a textbook-accurate one. Add multisensory detail: texture, colour, sound, smell. The richer the sensory content, the more retrieval pathways the image creates.
Emotion and etymology as memory anchors
Memories linked to emotion -- curiosity, surprise, humour, mild embarrassment, admiration -- are significantly better consolidated than neutral memories. This is why words discovered in an emotionally compelling context (a novel you were captivated by, a film that moved you) stick far better than words from a cold list.
Etymology: the pleasure of understanding origin
Etymology is an inexhaustible source of memory anchors. Knowing the origin of a word transforms an arbitrary phoneme sequence into a logical narrative -- and narrative is infinitely more memorable than raw sequence.
Examples that illustrate why etymology sticks: salary comes from the Latin salarium -- the salt allowance given to Roman soldiers (hence the expression 'not worth his salt'). Disaster comes from the Latin dis-astrum -- 'bad star', the ill-omened heavenly body. Quarantine comes from Italian quarantina -- the forty days of isolation imposed on ships arriving in Venice in the 14th century to prevent plague transmission.
These facts are memorable precisely because they surprise and tell a story. Note interesting etymologies directly on your flashcards -- they anchor meaning while creating a network of associations that extend far beyond the word itself. The word is no longer isolated; it is connected to history, culture, and logic.
Humour and personal associations
A funny, slightly absurd, or personally significant association is more memorable than a neutral one. If you are learning the Japanese word 危険 (kiken = danger) and you imagine 'ki ken' as a person named Ken enthusiastically shouting 'ki ki ki!' whenever he sees something dangerous -- that absurd, slightly ridiculous image is exactly the kind of anchor your brain will retain.
Self-deprecating humour works particularly well: an embarrassing mistake you made while using a word, a comical situation in which you encountered a word for the first time -- these personal memories are exceptionally durable anchors. The emotional charge (even mild embarrassment) activates the amygdala, which amplifies memory consolidation in the hippocampus.
Use this deliberately: when you encounter a stubborn word that refuses to stick, look for something funny, surprising, or personally resonant about it. A word connected to a story is a word you will not forget.
From first encounter to mastery: the optimal workflow
The encoding techniques presented here do not replace flashcards -- they amplify them. Here is the complete workflow for turning every encounter with a new word into durable memorization.
- Encounter the word in authentic context (reading, listening, conversation). Do not immediately look up the translation -- try to infer the meaning from context. This active inference attempt is already a first deep encoding.
- Verify the translation and encode actively: apply an encoding technique proportional to the word's difficulty. Easy to anchor (similar to native language, concrete meaning): example sentence is enough. Resistant word: add a mental image or phonetic association.
- Create the flashcard immediately (or within 24 hours): front = word in target language (+ pronunciation for non-Latin scripts), back = translation + example sentence + optional mnemonic note or etymology. Do not wait -- the emotional context of the discovery fades quickly.
- Let the SRS algorithm manage reviews: do not plan yourself when to review a word. The FSRS algorithm calculates the optimal interval for each card based on your performance history. Only review the cards presented each day.
- Re-evaluate cards that resist: if a word returns regularly as 'hard' or 'forgotten' despite several reviews, reformulate the card or add a stronger encoding technique (phonetic keyword, absurd mental image). A poorly encoded card will stay difficult regardless of review frequency.
When a card resists: reformulation and strengthening
If a word returns repeatedly as 'hard' or 'forgotten' despite several reviews, the problem is usually in the card, not in your memory capacity. A poorly formulated card will remain difficult regardless of review frequency. The solution is not more reviews -- it is reformulating the card.
Signs a card needs reformulation: the example sentence is too abstract or forgettable; the front is ambiguous (multiple possible answers); the back contains too much information to hold in working memory during review. Fix the weakest element first, then add a stronger encoding hook -- a keyword association, a vivid image, an etymology note.
One practical rule: if a card has been marked 'hard' or 'again' more than 4 times, it is almost certainly a card design problem. Spend 2 minutes reformulating it rather than continuing to review a broken card.
Reserve advanced techniques (keyword method, absurd associations, etymology research) for the 20% of words that resist after 2 to 3 normal reviews. For the remaining 80%, a well-formulated card with a strong example sentence is sufficient. Investing advanced techniques on every word is counterproductive -- time spent creating elaborate associations for easy words is time lost.
Memia: enriched encoding and spaced repetition combined
Memia combines spaced repetition (FSRS algorithm) with AI flashcard generation that automatically produces cards with contextualised example sentences -- the primary vector of deep encoding. Import a vocabulary list or a text in the target language, and the AI generates well-formulated cards, ready to review.
For resistant cards, Memia lets you add mnemonic notes, images, and associations directly to the card -- enriching encoding without leaving the application. The FSRS algorithm adapts review intervals to your individual memory profile: easy words return less often, difficult words more often, without you having to think about it.
In 15 to 20 minutes per day, you can actively maintain a deck of 300 to 500 words -- with measurable retention rates visible in the application's statistics. This is far more efficient than the same time spent rereading vocabulary lists, which produces recognition without retrieval and fades within days.
AI generation for vocabulary decks: what it produces
When you import a word list (JLPT N3, HSK 4, TOEIC business English, or any custom list), Memia's AI generates cards with translation, register note, example sentence, and -- for languages that use non-Latin scripts -- phonetic transcription (furigana for Japanese, pinyin for Mandarin).
The generation is adapted to the exam register where relevant: business English cards include professional context examples; JLPT cards include vocabulary in sentence patterns that reflect the actual exam format. This saves 3 to 5 hours of card creation per 200 words compared to manual creation, while producing cards that are at least as well-formulated.
Import the 500 most frequent words in the language you are learning. The AI generates cards with translations and example sentences. Review for 15 minutes per day. In 2 months: 500 well-encoded words -- and comprehension that starts to take off.
Frequently asked questions about vocabulary memorization
How many words do you need to know to be functional in a foreign language?
Approximately 1,000 to 2,000 words cover 80 to 90% of everyday texts and conversations in most languages -- the threshold for functional survival. For comfortable fluency in speaking and writing, 3,000 to 5,000 active words are needed. A native adult speaker commands between 20,000 and 40,000 words, but the vast majority are only needed for specialised domains. The key insight: each additional word learned beyond the first 2,000 contributes less than the previous one to general comprehension.
Is the memory palace (loci method) useful for vocabulary?
The loci method is highly effective for memorizing ordered lists -- speeches, process steps, sequences. For vocabulary at scale, it is less practical: each new word requires a distinct location in the mental palace, which becomes unmanageable beyond a few hundred words. For everyday high-volume learning, SRS flashcards remain the most efficient approach. The loci method can be useful for short thematic vocabulary lists before an exam or trip, or for memorizing a specific set of words where order matters.
Should you write words by hand to remember better?
Evidence is nuanced. Handwriting activates additional motor processing areas and can strengthen encoding -- particularly for languages with complex writing systems (Japanese, Mandarin, Arabic, Korean). For these languages, manual character writing practice is strongly recommended alongside digital flashcards. For Latin-alphabet vocabulary, the advantage of handwriting over typing is more modest and less decisive. If you are time-constrained, prioritise well-formulated digital flashcards with example sentences over handwriting -- the quality of the card matters more than the medium of review.
How do you handle false friends (words that resemble your native language but mean something different)?
False friends are among the most difficult words to memorize because they create interference with the native language. The most effective strategy: create a specific flashcard for the false friend with an explicit note on the difference. Example for English: 'eventually does NOT mean eventuellement (French) -- it means in the end, ultimately'. Add an example sentence that clearly illustrates the real meaning. A contrastive image (the false meaning crossed out in red, the true meaning highlighted in green) can also help anchor the distinction.
Is it better to learn words by theme or by frequency?
By frequency first, by theme afterwards. Starting with thematic vocabulary (animals, colours, clothes) is intuitive but suboptimal: these lists often contain low-frequency words -- hippopotamus, turquoise, kimono -- that will not appear in real usage for months. Frequency-first lists guarantee that every word learned immediately increases comprehension of real texts and conversations. Once high-frequency vocabulary is mastered, thematic lists serve to go deeper in specific domains relevant to your needs or interests.
How do you memorize prepositions and grammar words, not just lexical vocabulary?
Prepositions and grammar words (articles, conjunctions, auxiliaries) are particularly difficult to memorize in isolation because they have little meaning out of context. The most effective method: memorize them always in fixed expressions or recurring syntactic patterns. For English: 'interested in' (never 'interested of'), 'responsible for', 'depend on', 'regardless of'. One flashcard per pattern, with several examples on the back. Repeating patterns in context is more effective than memorizing abstract rules -- the pattern becomes automatic through repeated retrieval, not through conscious rule application.
How fast can you reasonably acquire vocabulary with flashcards?
With 10 to 15 new cards per day and regular review, you can solidly consolidate 300 to 450 words per month. Over 6 months of regular practice: approximately 2,000 words -- the comfortable comprehension threshold in most languages. This pace is sustainable long-term and produces durable retention, unlike intensive learning over a few weeks which fades quickly without spaced review. The bottleneck is not daily card volume but consistency: 10 cards reviewed daily for 6 months outperforms 50 cards per day for 3 weeks then stopping.