Learning a Language

Duolingo gives you XP streaks. It's not teaching you a language. The dopamine loop is real; the acquisition isn't. After a year of daily Duolingo, most users can produce a handful of disconnected phrases and fail to understand a native speaker at normal pace. The app is optimised for engagement, not fluency.

Classroom instruction moves slowly and often teaches in your native language, which means you spend most of your learning time in English. Textbooks front-load grammar rules before your ears are remotely ready to use them. Both approaches produce people who can explain subjunctive mood and still not hold a five-minute conversation.

The most robust framework in second language acquisition research is Stephen Krashen's Input Hypothesis. The core idea: you acquire language by understanding messages slightly above your current level. Not by studying rules, not by drilling vocabulary in isolation, not by translating from English. You acquire language the same way children do: massive exposure to meaning, in context, over time.

This has been developed practically by people worth knowing: Steve Kaufmann (polyglot, founder of LingQ, 20+ languages acquired in adulthood), Matt vs Japan (documented his Japanese acquisition from zero to near-native over years of intensive CI), and Paul Noble (audio-only courses that never use your native language as a crutch). The research and the practice align unusually well.

The principle is called i+1: your input should be at your current level plus one step. Comprehensible, but not trivially easy. The 20% you don't understand is acquisition in progress. Your brain is doing the work.

Two approaches work. Choose one before you start.

The first is the comprehensible input method: from day one, you focus on massive listening and reading in the target language, keep grammar study minimal, and trust that patterns emerge through exposure. Progress feels slow in the first few months: you understand very little. But from around month four to six, comprehension accelerates. Learners who take this route tend to have more natural-sounding output and better long-term retention. The CI method is harder to measure and harder to stick with early on.

The second is the structured method: a textbook (or structured course) plus grammar study, vocabulary flashcards, and speaking practice from early on. Progress feels more visible in the first few months. The risk is fossilisation: locking in errors because you're producing output before you've built enough input. Done well, the structured method works. Done lazily, it produces people who are technically correct and impossible to understand.

Both paths lead to fluency. The CI method is slower at first and faster long-term. The structured method requires more discipline about output quality. Most successful learners end up doing a blend, but choose a primary method before you start. Mixing randomly is how you end up with a shelf full of resources you've never finished.

Pick based on what you'll use. The most common mistake is choosing a language because it seems impressive and then abandoning it because there's no real reason to keep going. A language you speak on holiday, with family, or because the culture interests you will outlast a language you chose to seem interesting.

Practical considerations for UK learners: French, Spanish, and Italian are the easiest starting points: closest to English grammatically, with enormous resource libraries and millions of speakers in Europe. German is harder (case system, word order, compound nouns) but highly practical. Japanese, Korean, and Mandarin require learning a new script as well as a different grammatical structure and tonal/pitch system. They take roughly three times as long as European languages and require stronger initial motivation to survive the early plateau.

Pick one language. Don't switch for at least six months. The grass is always greener in another language when yours is hard.

Free: Anki (spaced repetition flashcards), YouTube channels for your target language, podcasts made for learners, LingQ on the free tier, Duolingo as a supplement only. Paid, worth considering: Pimsleur (audio-only, around £10/month), italki community tutors (£10–20 per hour for conversation practice), LingQ paid (£10/month). Language exchange apps (Tandem, HelloTalk) are free and useful at intermediate level.