Methodology

How Kilo Lingo works

Fair question to ask of any language course: prove this is grounded in something sensible. This page sets out the method behind the courses, the evidence for it, and the parts of it you can try right here.

The short version: teach words in order of how often the language uses them, only ever ask learners to read or practise language built from words they have been taught, and enforce both rules with software rather than good intentions.

Where the method comes from

I learned Spanish, French, Mandarin and Russian as an adult, and I spent time teaching English in Spain and France. Two classroom memories shaped everything on this site.

I once watched a child learn the word "giraffe" in a lesson about animals, before that child could express hunger, or describe what they wanted.

Giraffe is a wonderful word and a hopeless priority. Teaching it early spends the learner's scarcest resources, memory and attention, on a word they may never need, while the hundred words that carry half of every conversation go untaught. Most courses and apps do a politer version of the same thing: vocabulary organised by theme, not by usefulness. Grouping by category has a second cost: words from the same set, such as eye, nose and ear, learnt together get in each other's way, while words linked by a scene (frog, green, pond) are learnt more easily (Tinkham, 1997). So categories never set the teaching order here, and themes come back as revision sessions built around a short read.

The second memory: hearing "immersion" invoked to justify talking at learners in language they had no way to parse. Being surrounded by noise doesn't teach you much. The research points to comprehensible input: language you can understand, pitched a little above what you already know, so you follow the meaning and the new language attaches itself to it (Krashen, 1982). Chasing that idea properly is what led to the graded readers and the vocabulary ceilings below.

Frequency first: chapters of 100, blocks of 10

Each course is a sequence of chapters, and each chapter is the next 100 words of the language by real-world frequency, taught in blocks of ten. Chapter one is the hundred most common words; chapter two is words 101 to 200; and so on into the thousands. You always know why you are learning a word: it is the next most useful one. Giraffe is in no hurry. It can wait its turn.

of spoken English is made up of the top 1,000 word families: 80%+
words per chapter, in strict frequency order: 100
words per learning block: 10
lemmas in the deepest course, every one with its own page: 5000

The most frequent 1,000 English word families (lists built from the British National Corpus) make up about 81-84% of the running words in spoken English (Nation, 2006). Counted as lemmas, the top 1,000 make up about 88% of spoken Spanish (Davies, 2005). Every language has a steep frequency curve, but the exact figure varies by language, and coverage is not comprehension: see the comprehensible input section below.

How we build the frequency lists

A frequency list is only as honest as the text it is counted from, so we chose the text for each language on purpose and say which is which (the full list of sources is below):

Film and TV subtitles

The closest large corpus there is to everyday speech, and the base of every list except Latin: people learn a living language to talk in it.

Published books

Public-domain literature is the whole of the Latin corpus, and part of the Esperanto blend, where it keeps the written register represented.

Sentence corpora

Translated everyday sentences from Tatoeba go into the Esperanto blend, sitting between the spoken and written registers.

Raw counts are where the work starts. A raw list counts word forms, pads itself with names (unfiltered, the name "Tom" ranks in the top twenty of one popular corpus) and subtitle noise, and has no idea which form belongs to which word. So every list is cleaned of proper nouns and then lemmatised form by form: for Spanish we reviewed the top 13,000 word forms, for French 12,000 and for English 9,000, mapping each one to its dictionary headword. Each headword is then ranked on the strength of all its forms, which is why tener sits at 14 in our Spanish list even though no single form of it gets into the raw top 50. "Top 1,000 words" therefore means the 1,000 most frequent lemmas. "Speak", "speaks", "spoke" and "spoken" count once, not as four separate achievements. We reviewed those mappings rather than trusting automatic lemmatisers, which both miss merges they should make and merge words they should not. Where a language needs special handling it gets it: Mandarin has no inflection to collapse, and the Latin corpus deliberately excludes Neo-Latin science, because a learner heading for Caesar should not have their core vocabulary shaped by Newton's optics.

Sources and credits

Spanish, French, English, Mandarin, German, Italian, Portuguese and Russian: word counts from film and TV subtitles, the OpenSubtitles 2018 corpus (P. Lison and J. Tiedemann, 2016, OpenSubtitles2016: Extracting Large Parallel Corpora from Movie and TV Subtitles, LREC 2016), distributed via OPUS from opensubtitles.org, as counted by Hermit Dave's FrequencyWords lists (licensed CC BY-SA 4.0). We lemmatised, cleaned and re-ranked those counts as described above.
Esperanto: a weighted blend of public-domain books from Project Gutenberg, sentences from Tatoeba (CC BY 2.0 FR) and OpenSubtitles 2018, reduced to roots.
Latin: public-domain Latin texts from Project Gutenberg, with Neo-Latin science excluded.
GCSE courses: the vocabulary lists published in the 2024 AQA and Pearson Edexcel GCSE specifications, which set the word lists; the order we teach them in is ours.
Mandarin levels: the HSK 1 to 6 vocabulary lists.

The anatomy of a chapter

A chapter is not a hundred flashcards in a row. Between the word blocks sit different kinds of lesson. Each has its own job and its own icon, the same one everywhere on the site, so you can read the shape of a chapter at a glance.

Chapter opener

A short welcome that frames what the next hundred words will unlock.

Word lessons

The next ten words by frequency, with audio, illustrations and typed recall.

Affix lessons

Esperanto's word-building machinery: pieces that snap onto hundreds of roots.

Measure words

Mandarin's counting machinery, drilled as tappable chips.

Grammar lessons

A rule explained at your level, with a bespoke drill built from that lesson's own content.

Sound Lab

The mouth, drawn and animated. Produce the sounds, not just hear them.

Review games

Crosswords, bubble shooters and the rest of the arcade, drawing only on words you have been taught.

Mistake clinics

Two things learners keep mixing up, told apart.

Themed revision

Words you met in frequency order, regrouped into the clusters you will use them in.

Phrase sessions

Whole sentences assembled from words you already hold separately.

Conversations

A short scene between you and one of the story cast, every line voiced, made only of words you have been taught.

Songs

Real songs with lyrics, unlocked once you have the vocabulary to follow them.

Community lessons

Optional lessons that plug you into the real community: Reddit, Discord, YouTube and beyond.

Graded readers

The chapter's milestone: an original story using only vocabulary you have already met.

Chapter quiz

The summit. Pass it and the chapter's badge is yours.

Practice that varies, and never cheats

The drills between the learning blocks pull the same word out of memory in different ways (typed recall, crosswords, bubble shooters, morphology drills), moving from recognising it to producing it, with spaced repetition underneath deciding what comes back and when. Pulling a word out of memory does far more for it than reading it again (Karpicke and Roediger, 2008), and spacing those retrievals out beats cramming them (Kim and Webb, 2022). Practice testing and spaced practice are the two study techniques a major review rated most useful (Dunlosky et al., 2013).

The rule that makes it fair: practice only ever draws on words the course has already taught. No drill, game or quiz reaches past the point you have reached. Getting something wrong should mean you need more practice, never that the site asked you something it had not taught.

Rather than take our word for it, the box alongside is the real drill engine with real course words. No sign-up.

Try it now - no sign-up

Quick taster

3-word session

A quick mix of exercises - recognise, recall, type and build each word. Miss one and it comes back round until it sticks.

Comprehensible input, enforced by software

Every chapter ends with a graded reader: an original story written using only the vocabulary from the chapters you have already completed, followed by a comprehension quiz. It means real connected prose, at length, with nothing in it you have not been taught. Grammar examples and worksheets respect the same ceiling.

Why insist on every word? Because five unknown words in a hundred is enough to lose most readers. Adequate comprehension needs somewhere between 95% and 98% of the words known (Hu and Nation, 2000; Laufer and Ravenhorst-Kalovski, 2010). Strictly, that makes the readers fluency and consolidation reading rather than a stretch. The new words and grammar arrive in the chapter's lessons, and the story is where they settle in. A balanced course needs both, alongside speaking, writing and focused study (Nation, 2007).

Taught is not the same as known, either. In studies of flashcards, word lists and similar exercises, a good share of what learners had learnt was gone by a delayed test (Webb, Yanagisawa and Uchihara, 2020), which is why spaced review runs underneath the course and every word in a story can be tapped for its meaning. (The research behind it, and its limits, is in what comprehensible input is.)

An automated check enforces the "only". It runs over every story before anything ships, and one word over the chapter's vocabulary ceiling fails the build:

$ npm run check

[vocab-tier] checking 100+ graded stories against their chapter ceilings...

FAIL story uses a word beyond the learner's current chapter

The story is rewritten. The learner never sees the slip.

Human authors drift; scripts do not. Every course on the site is held to this ceiling on every commit.

As your vocabulary grows, the course opens up

A frequency-ordered course earns compound interest: every hundred words you bank makes more real material readable. So as you progress, the course introduces themed revision sessions (food, travel, work, weather) that regroup words you met in frequency order into the clusters you'll use them in, each with its own short read. Alongside them come songs with lyrics: real songs, unlocked only once you have the vocabulary to follow them, because a lyric you can follow is the most pleasant spaced repetition there is. Phrase sessions do the same for conversational set pieces, built entirely from words you already hold.

Every word is a door

Behind each course sits a full dictionary of lemmas, and every target-language word shown anywhere on the site, whether in a lesson, a story, a grammar example or a song lyric, can be clicked for an instant definition and a link to that word's own page: example usage, illustrations to hang a mnemonic on, etymology, recorded audio, and a list of everywhere the word appears across the site.

Here is that machinery, live. Tap any word:

"The dog drinks water", in Esperanto.

The definition appears here.

The course points outwards

No website is a language community, and a course that pretends otherwise is keeping you captive. So the curricula include optional community lessons whose entire job is to plug you into the wider world of the language you are learning: the subreddits worth joining, the Discord servers where learners and native speakers actually talk, the YouTube channels producing input at your level, and email newsletters worth the inbox space. The measure of the course is how confidently you can leave it.

For teachers

Every drill, grammar lesson, word page and story has downloadable PDF worksheets designed to be printed and used in a classroom, and each language has a teacher's hub collecting them chapter by chapter:

If you teach and something here almost works for your classroom but not quite, tell us at hello@kilolingo.com. The worksheets exist for you, and your complaints are the roadmap.

The evidence, and its limits

Why grammar lessons at all? Krashen argued that consciously learnt rules play only a limited role in real language use (Krashen, 1982). The later evidence is against him: two large reviews of the research found that explicit instruction produces large, durable gains, and beats leaving learners to work the rules out for themselves (Norris and Ortega, 2000; Spada and Tomita, 2010). So the grammar is taught, in step with the words it needs.

Comprehensible input is a principle, not a proven theory. Krashen's wider model has been examined critically since it appeared (Gregg, 1984), and we do not lean on it. We rely on the part almost nobody disputes: you learn from language you can understand, and very little from noise.

Input is not the whole job. A well-designed course gives roughly equal time to four strands, and one of them is speaking and writing (Nation, 2007). A website is weak at real conversation, which is why the community lessons send you out to where they happen.

The coverage numbers are language-specific. The 81-84% figure is for English word families in speech (Nation, 2006). Counted as lemmas, the top 1,000 words make up about 88% of spoken Spanish but only 75-80% of written Spanish (Davies, 2005). Every language front-loads its most common words; the exact share differs.

There is no magic review moment. The best gap between reviews depends on how long you want to remember (Cepeda et al., 2008). Our scheduler is sensible engineering on top of a robust effect, not a precise science.

What it costs

All of the instruction is free to read: every chapter, grammar lesson, story and word page, with no account needed, plus the audio for the first two chapters of every course. The Kilo Lingo Plus subscription covers the practice and the audio beyond that, and paying for the drills is what keeps the teaching free. More on who is behind the site on the about page.

Pick your language Meet the author

References

Cepeda, N.J., Vul, E., Rohrer, D., Wixted, J.T. and Pashler, H. (2008) 'Spacing effects in learning: a temporal ridgeline of optimal retention', Psychological Science, 19(11), pp. 1095-1102. Available at: https://doi.org/10.1111/j.1467-9280.2008.02209.x.
Davies, M. (2005) 'Vocabulary range and text coverage: insights from the forthcoming Routledge frequency dictionary of Spanish', in Eddington, D. (ed.) Selected proceedings of the 7th Hispanic Linguistics Symposium. Somerville, MA: Cascadilla Proceedings Project, pp. 106-115. Available at: https://www.lingref.com/cpp/hls/7/paper1091.pdf.
Dunlosky, J., Rawson, K.A., Marsh, E.J., Nathan, M.J. and Willingham, D.T. (2013) 'Improving students' learning with effective learning techniques: promising directions from cognitive and educational psychology', Psychological Science in the Public Interest, 14(1), pp. 4-58. Available at: https://doi.org/10.1177/1529100612453266.
Gregg, K.R. (1984) 'Krashen's monitor and Occam's razor', Applied Linguistics, 5(2), pp. 79-100. Available at: https://doi.org/10.1093/applin/5.2.79.
Hu, M. and Nation, P. (2000) 'Unknown vocabulary density and reading comprehension', Reading in a Foreign Language, 13(1), pp. 403-430. Available at: https://hdl.handle.net/10125/66973.
Karpicke, J.D. and Roediger, H.L. (2008) 'The critical importance of retrieval for learning', Science, 319(5865), pp. 966-968. Available at: https://doi.org/10.1126/science.1152408.
Kim, S.K. and Webb, S. (2022) 'The effects of spaced practice on second language learning: a meta-analysis', Language Learning, 72(1), pp. 269-319. Available at: https://doi.org/10.1111/lang.12479.
Krashen, S.D. (1982) Principles and practice in second language acquisition. Pergamon Press. Available at: https://sdkrashen.com/content/books/principles_and_practice.pdf.
Laufer, B. and Ravenhorst-Kalovski, G.C. (2010) 'Lexical threshold revisited: lexical text coverage, learners' vocabulary size and reading comprehension', Reading in a Foreign Language, 22(1), pp. 15-30. Available at: https://hdl.handle.net/10125/66648.
Lison, P. and Tiedemann, J. (2016) 'OpenSubtitles2016: extracting large parallel corpora from movie and TV subtitles', in Calzolari, N. et al. (eds) Proceedings of the Tenth International Conference on Language Resources and Evaluation (LREC'16). Portorož: European Language Resources Association (ELRA), pp. 923-929. Available at: https://aclanthology.org/L16-1147/.
Nation, I.S.P. (2006) 'How large a vocabulary is needed for reading and listening?', The Canadian Modern Language Review, 63(1), pp. 59-82. Available at: https://doi.org/10.3138/cmlr.63.1.59.
Nation, P. (2007) 'The four strands', Innovation in Language Learning and Teaching, 1(1), pp. 2-13. Available at: https://doi.org/10.2167/illt039.0.
Norris, J.M. and Ortega, L. (2000) 'Effectiveness of L2 instruction: a research synthesis and quantitative meta-analysis', Language Learning, 50(3), pp. 417-528. Available at: https://doi.org/10.1111/0023-8333.00136.
Spada, N. and Tomita, Y. (2010) 'Interactions between type of instruction and type of language feature: a meta-analysis', Language Learning, 60(2), pp. 263-308. Available at: https://doi.org/10.1111/j.1467-9922.2010.00562.x.
Tinkham, T. (1997) 'The effects of semantic and thematic clustering on the learning of second language vocabulary', Second Language Research, 13(2), pp. 138-163. Available at: https://doi.org/10.1191/026765897672376469.
Webb, S., Yanagisawa, A. and Uchihara, T. (2020) 'How effective are intentional vocabulary-learning activities? A meta-analysis', The Modern Language Journal, 104(4), pp. 715-738. Available at: https://doi.org/10.1111/modl.12671.