Lingua
Pronunciation is the part of a language most learners practise least and worry about most. It gets skipped because it is awkward, and because the usual advice, "listen to natives and imitate", leaves out the awkward middle: what to do when you imitate and it still comes out wrong. There is a method for that middle, and it starts with the ear rather than the mouth.
First problem: you cannot hear the difference
Every language sorts the sounds humans can make into its own set of categories, and adult ears are tuned to the categories of their native language. A Japanese learner of English often hears right and light as the same word; a Spanish learner hears ship and sheep as the same word; an English learner of Korean hears three different consonants as one. You cannot reliably produce a contrast you cannot perceive, so pronunciation training that starts with the mouth is starting in the wrong place.
The training that fixes perception is called minimal pairs: two words that differ by exactly one sound. You listen to many examples of both, spoken by several different voices, and choose which you heard. The research on this, known as high-variability phonetic training and dating back to Logan, Lively and Pisoni, found that with a few weeks of short daily sessions, adult learners measurably improved at hearing contrasts their native language does not have, and that the gain transferred to their own production.
Second problem: the mouth has habits
Once you can hear the contrast, production is a motor skill. Three tools do most of the work.
- Know where the sound is made. A thirty-second description of tongue position is often the difference between guessing and getting it. For a few difficult sounds per language, look it up.
- Exaggerate, then relax. Overdo the new sound for a week. It will feel ridiculous and sound, to a native, merely correct.
- Drill in phrases, not in isolation. A sound that works alone and fails in a sentence is a sound that has not been learned yet. Practise the pairs inside short sentences at natural speed.
Third problem: nobody tells you
Self-study has a feedback gap. You say the sentence, it sounds fine to you, and it is wrong in a way you cannot detect, because the detector is the same ear that made the mistake. Closing the gap needs an outside listener, and there are three kinds.
- A recording. Record yourself, play the native model, play yourself. Even an untrained ear hears differences in a side-by-side comparison that it misses in real time.
- A person. A partner, a teacher, a tandem. Ask for one correction at a time and repeat it immediately.
- Speech recognition. An engine that turns your speech into text is a blunt but honest judge: if it wrote the wrong word, a native listener might have heard the wrong word too. Repeat until the text is right.
A ten-minute daily plan
- Two minutes of minimal pairs for the one contrast you are working on this week.
- Three minutes of shadowing a short native recording, matching rhythm and melody.
- Three minutes reading sentences aloud that contain the target sound, recorded, compared.
- Two minutes on the sentences a checker rejected yesterday.
One contrast a week, ten minutes a day. In three months you have covered the sounds that were marking you as foreign.
In LinguaPair
The reading section gives you the models: every text has a clear AI narration with word highlighting and a slower speed, made for shadowing. The courses give you the check: speaking exercises play a sentence and then listen to you say it, and a sentence that was not recognised comes back for another try, so each lesson is a few rounds of exactly the feedback loop above. And your partner, if they speak the language, is listener number two: the translator's word cards carry pronunciation, so you can hear the model and then hear them.