How a Language Grows: Input, Practice, Feedback & Time

You can know three thousand words and still freeze when the waiter answers. Memory is half of a language; the other half is using it. Words have to be heard in context, pushed out of your own mouth, corrected, and given enough hours to settle.

This piece covers that second half: what kind of input helps, why speaking and feedback matter, what the research says about grammar, pronunciation and age, and how long a language honestly takes. It ends with a weekly routine you can build in a minute. It is the last method piece before the Spanish ones.

Input, interaction and output: three ideas, one recipe

Stephen Krashen argued in the 1980s that we acquire a language mainly by understanding messages slightly beyond our current level, which he called comprehensible input. He also claimed that conscious study cannot turn into fluent use and that anxiety blocks learning (the “affective filter”). What the field kept: input is essential, there is no substitute for a lot of it, and it has to be understandable. Studies of extensive reading, for instance, find real gains in vocabulary and fluency. What it disputed: that input alone is enough, and the sharp split between “acquisition” and “learning”, which is hard to test.

Two other ideas filled the gap. Michael Long’s interaction hypothesis says conversation helps because, when you do not understand, you and your partner adjust, which makes input more understandable and draws your attention to forms. A meta-analysis by Alison Mackey and Jaemyung Goo (2007) found interaction does help learning. Merrill Swain’s output hypothesis comes from Canadian French-immersion students who understood plenty but spoke with many errors: producing language forces you to notice what you cannot yet say, test guesses, and get a reply.

No single idea wins. The combination that fits the evidence: lots of comprehensible input, a smaller but real amount of speaking and writing, and feedback on what you produce. Apps are strongest at the first small piece, so the recipe has to be completed outside them.

Grammar and feedback: explicit help works

Should adults be taught rules at all? Two meta-analyses say yes. Norris and Ortega (2000) pooled 49 studies and found that explicit instruction (teaching the rule) produced large gains, usually more than implicit approaches. Spada and Tomita (2010), looking at 41 studies, found explicit instruction helped both on tightly controlled tests and in freer use, for simple and complex rules alike. A caution: many of the tests favoured conscious knowledge, so the question of how much of it turns into automatic speech is still debated.

Feedback matters too. In 1997 Roy Lyster and Leila Ranta studied French-immersion classrooms and found that teachers mostly used recasts, quietly restating the learner’s sentence correctly, and that learners often did not notice them as corrections. Prompts that push the learner to fix the error themselves (“Say that again?”, “How do we say that in the past?”) led to more repair. Researchers still argue over how well recasts work in other settings, and a later meta-analysis found prompts at least as effective. Either way, learners need a human or a tool that tells them what they got wrong.

Ears first: pronunciation and listening

You cannot say reliably what you cannot hear. Learners often struggle with sound contrasts their first language does not use, so listening comes first. In experiments from the early 1990s, Japanese speakers learning English were trained to tell /r/ from /l/ using many different voices and contexts, which is called high-variability phonetic training. They improved, the gains carried over to new speakers and later to their own pronunciation, and they lasted months. A narrative review by Thomson and Derwing (2015) found pronunciation teaching generally effective.

Shadowing means listening to a recording and repeating it aloud a moment behind, copying rhythm and sound. Studies, mostly small and mostly of English learners in Japan, suggest benefits for listening and pronunciation, but the evidence is modest and still preliminary, so treat it as a plausible, cheap practice rather than a proven one. Finally, a strong accent is not the same as being hard to understand: Munro and Derwing (1995) found that listeners could understand heavily accented speech well. Aim to be clear, not to sound native.

Age: what children have and what adults have

In 2018 Joshua Hartshorne, Joshua Tenenbaum and Steven Pinker analysed an online grammar test taken by about 670,000 people. The ability to learn a new language’s grammar stayed high until about age 17 and then declined. So the “critical period” for reaching a native-like level is real, and it is a long window that ends late in adolescence, with most adult learners past it. But native-like is rarely the goal, and the same research tradition shows that adults and older children usually learn faster than young children in the first months and years (Snow and Hoefnagel-Höhle, 1978; Krashen, Long and Scarcella, 1979). Adults read, reason about rules, plan and choose their method. Children win in the long run because they get thousands of hours of immersion. The adult plan is to buy those hours deliberately.

How long will it take?

It takes longer than apps imply, and it depends heavily on the language you start with and the one you choose. The US Foreign Service Institute (FSI) sorts languages into four categories by how long its diplomats need to reach professional working proficiency in speaking and reading. These are class hours for full-time students who are native English speakers with good aptitude and other languages already, in small classes, in class about 23 hours a week with several more hours of directed self-study each day.

Category	Class hours	Languages in this library
I	552 to 690 (24 to 30 weeks)	Spanish, French, Italian
II	about 828 (36 weeks)	German
III	about 1,012 (44 weeks)	Russian, Hebrew
IV	about 2,200 (88 weeks)	Arabic, Mandarin, Japanese, Korean

FSI marks some languages, Japanese among them, as harder than others in their category. The ratings are one institution’s experience with one kind of student, and the list is for English speakers. If your first language is Spanish, Italian and Portuguese are far closer to you than they are to an English speaker, while Japanese or Korean are about as far. Treat a category list as a guide to your own pair of languages.

The Common European Framework of Reference (CEFR) describes ability in six levels. Cambridge English publishes rough guided-learning hours to reach each level starting from zero, for English learners; they exclude self-study and are guidelines, not promises.

Level	In plain words	Guided hours (Cambridge)
A1	Simple everyday phrases; introduce yourself	90 to 100
A2	Routine exchanges: shopping, directions, family	180 to 200
B1	Most travel situations; describe plans and experiences	350 to 400
B2	Talk with native speakers without strain; follow complex texts	500 to 600
C1	Flexible use for work and study; implied meaning	700 to 800
C2	Understand virtually everything; precise, nuanced expression	1,000 to 1,200

The planner below combines the two crudely: it takes the Cambridge midpoint for a level, scales it by each FSI category’s hours relative to Category I, and divides by the time you spend. The band is deliberately wide, from 0.7 to 1.6 times the central figure, because people vary at least that much. It counts any time you actively study or listen, not only time in an app.

What apps do well and what they cannot do

Apps and spaced-repetition tools (Duolingo and Anki are the well-known examples) do three things well: they make the daily habit easy, they force retrieval, and they handle the schedule for you. Streaks and points can help a habit stick, though they can also keep you doing easy review to protect a number. Independent evidence on any one product is thin; many of the published studies are funded by the companies. What no app replaces is the large volume of real input, conversation with an actual person and feedback on what you produce. Think of an app as the review layer, not the whole course.

How these Default Library language paths work

Each language is a path of interactive pieces in three stages: Start (sounds and spelling, the present tense, irregular verbs, and the first 150 words), Build (past and future, the grammar map, the first 1,000 words in frequency bands) and Read and listen (short graded texts written for the library, and where to go next). Drills live inside the pieces and work offline. A reader’s review schedule is saved in the browser, with no account, and can be exported and moved to another device.

The daily session is ten to twenty minutes: review the cards that are due, do one drill, then read or listen to one short thing. Do that most days. Then add what no page can supply: conversation with a tutor or language-exchange partner, comprehensible-input video and podcasts, graded readers, and a frequency dictionary for when you want to go beyond the deck. Build your week below.

What sticks

• Input is essential but not enough: add some speaking, some writing and feedback on both.
• Explicit grammar teaching helps adults, and being prompted to fix your own errors helps too.
• Train the ear first; shadowing is plausible and cheap, with modest evidence.
• Adults start faster than children; the child’s long-run advantage is hours of immersion, which adults can buy.
• Timelines are hundreds to thousands of hours, and the language you start with and the one you choose set the scale.
• An app is the review layer. A week needs review, input, output and feedback.