The standard advice for starting Georgian is to spend your first weeks drilling the alphabet, then memorize verb tables, then translate sentences. The research on how languages are actually acquired says that order is backwards. Until now, nobody had written down why, with sources, for this specific language.
So we did. How Georgian Is Learned: An Evidence Synthesis is the newest publication from EasyGeorgian Research, following the spoken-frequency corpus study and the Georgian AI benchmark. It pulls together five decades of second-language acquisition and memory research, works out what it means for Georgian in particular, and shows exactly how we apply it. Like everything from EasyGeorgian Research, it is free under CC-BY, every claim is cited, and corrections go to [email protected] and get applied in public.
Here is the short version.
The five effects that survived 50 years of research
Language teaching is full of famous ideas that fell apart under testing. Learning styles, for one, have no supporting evidence that survives a controlled study. The paper is built on the short list of findings that have replicated across decades, languages, and laboratories:
- You acquire language from input you can mostly understand. Material slightly above your level teaches. Material far above it is noise.
- Speaking is what makes your grammar accurate. Learners with thousands of hours of listening and broken production are a documented phenomenon. Comprehension can skim past grammar. Producing a sentence cannot.
- Retrieval beats rereading. Recalling a word from memory strengthens it far more than seeing it again, and rereaders feel more confident while learning less.
- Spaced practice beats cramming. The forgetting curve has been measured since 1885. Reviews at widening intervals are what make memory durable.
- Word frequency is extremely top-heavy. A few hundred words carry most of everyday speech, so the order you learn words in is a first-class decision, not a detail.
None of this is secret. It is the boring consensus of the field. Georgian just never received it, because none of the big platforms whose scale forces methodological discipline ever built for the language.
Georgian is hard, but not where people think
The US Foreign Service Institute rates Georgian at roughly 1,100 classroom hours, Category III of IV, alongside Russian and Finnish. The paper’s most practical contribution is decomposing that number, because learners consistently fear the wrong parts and budget their effort accordingly.
The script is the easiest part. Mkhedruli is 33 letters, no capitals, and the mapping between letters and sounds is almost perfectly one-to-one. It is a weekend project, which is why we handle it with a free alphabet tool rather than course time. Spending your first month on handwriting drills front-loads the easiest component of the language at the moment your motivation is highest.
The difficulty lives elsewhere. The Georgian verb packs subject, object, direction, and tense into a single form and resists the tidy tables that work for Spanish. The vocabulary shares almost nothing with European languages, so there are no free words: a Spanish learner recognizes thousands of cognates on sight, a Georgian learner earns every word. And the words you need first are not the words written-frequency lists teach. In our corpus study, არა (ara), the word for “no”, ranks 16th in real speech and 8,662nd in the standard written ranking. A learner following a written list would meet the Georgian word for “no” somewhere after 8,000 others.
It gets worse. The most widely copied “Georgian frequency list” online turned out, when we decoded it, to be a Bulgarian frequency list corrupted through a broken character encoding. Flashcard decks built on it are still circulating.
What we do about it, and the part where we admit our stake
The paper ends its argument with a list of commitments you can check against any of our materials: input stays comprehensible and narrow, every lesson demands spoken production from the first session, every exercise is a retrieval prompt rather than a reread, repetition is scheduled by the system at widening gaps, and vocabulary follows our measured spoken-frequency ranking rather than topical chapters. If you have taken the audio course, you have felt each of these decisions, probably without noticing them.
We sell courses built on this synthesis, and the paper says so in the introduction, in the operationalization section, and again in the limitations. It also draws a line we think more of the industry should draw: the design is evidence-based, not evidence-proven. No randomized trial of our specific curriculum exists. Publishing the reasoning is what lets anyone check whether we read the literature honestly or in our own favor.
The level test methodology lives there too
One more thing moved into the paper: the complete methodology of our free Georgian level test, from the adaptive ability estimation down to why the “I don’t know yet” button is mathematically engineered so that honest skipping never costs you anything relative to guessing. If you ever wondered whether an 8-minute click-only test can place you fairly, Section 5 is the full answer, including a blunt list of what the test does not measure.
The full paper, with every source and exhibit, is at easygeorgian.com/methodology. If you read one section, read Section 3: it will probably change where you spend your next hundred hours of Georgian. And if you find something wrong in it, tell us. Corrections are applied in public, with a changelog, which is exactly how we would want everyone else’s teaching claims to work.
