Why some English sounds sound the same to you Your ear was tuned to your first language before you could talk. Here's how that happens, and how adults tune it again.

Your ear sorts English into your first language's boxes

Every language divides speech into a small set of sound categories. Japanese has five vowels (a, i, u, e, o), while most accents of English have more than a dozen. Japanese has one sound, the tapped r of ra, ri, ru, re, ro, where English has two: L and R.

When you listen, your brain doesn't keep every detail of a sound. It decides which of your categories the sound belongs to and moves on. That shortcut is what makes listening to your first language fast and effortless. But when two English sounds land in the same Japanese category, they come out the other side as one.

Researchers call this perceptual assimilation. Strange and colleagues (1996) showed it for vowels: Japanese listeners mapped American English vowels onto the five Japanese ones, often two or three English vowels to a single Japanese vowel.

So this isn't a hearing problem, and a hearing test wouldn't find it. It's a sorting problem, and sorting can be retrained.

Why Japanese speakers hear L and R as the same: it starts before age one

Babies start out able to hear differences between sounds in any language. In a classic study, Werker and Tees (1984) found that English-learning babies could tell apart consonants from Hindi and from a Salish language of British Columbia at 6–8 months, but by 10–12 months most of them no longer could.

Kuhl and colleagues (2006) tested English R and L directly, with babies in the United States and Japan. At 6–8 months, the two groups told R and L apart equally well. By 10–12 months, the American babies had improved and the Japanese babies had declined. In the same few months, the sounds of the home language get sharper and the others fade.

By adulthood, the effect is strong. Goto (1971) found that Japanese adults had great difficulty telling English L and R apart by ear. Learning to read and spell English doesn't change it, because the sorting happens before you're aware of the sound.

Which English sounds merge for a Japanese ear

Each row below is one Japanese sound that two or more English sounds fall into. The pair pages have the full word lists, and you can play each word in several native voices.

Korean also has one sound where English has L and R, so Korean speakers share that row. Learners from other languages have their own lists. Which pairs are hard depends on which boxes your first language has.

Hear it for yourself

Tap play and pick the word you heard. Each word is a real clip from the app, in a different voice.

If you got some wrong, that's the merge at work, not carelessness. The free listening test checks L and R, B and V, and TH and S in a couple of minutes.

How adults retrain the ear: high variability phonetic training

The method with the most evidence behind it is high variability phonetic training, or HVPT. It was developed in the early 1990s to teach Japanese learners to hear English R and L (Logan et al. 1991), and it works like this:

• You hear one word, such as right.
• You pick which of two words you heard: light or right.
• You find out at once whether you were right.
• The next word comes in a different voice: men and women, young and old, American, British and Australian.

The many voices are the key part. Every speaker makes L and R a little differently. Lively and colleagues (1993) found that learners trained on one voice improved but couldn't carry the skill over to new voices. Hearing many voices pushes your brain to find the difference they all share.

The evidence has grown for 30 years. Thomson (2018) reviewed 32 HVPT studies and concluded that it is a very effective training method and that the gains last. A 2025 meta-analysis by Uchihara, Karas and Thomson pooled 79 studies and found medium-to-large improvements in perception (g = 0.92 from pretest to posttest). The gains held over time and partly carried over to words learners hadn't trained on.

Two results matter most for learners. First, the skill stays. In Lively et al. (1994), Japanese learners trained for 3 weeks; 3 months later their accuracy had dropped only 2%, and 6 months later it was still 4.5% above where they started. Second, it helps speaking. In Bradlow et al. (1997), Japanese learners did listening training only, and native English listeners still rated their R and L pronunciation better afterwards.

How to train your ear: a routine based on the research

• Work on the pairs your language merges, not all of English. For Japanese speakers, the table above is the list.
• Use many voices, not one teacher's or one podcast host's.
• Check every answer right away, while the word is still in your ear.
• Keep sessions short and regular. A few minutes most days is easier to keep up than one long session a week.
• Start where the sound is clearest, at the start of a word (light · right). Then move to clusters (play · pray), then the middle and end of words (collect · correct).
• Don't wait until you can say the sounds. Hearing them comes first, and better speaking often follows.

How fast you improve varies from person to person, so compare yourself with your own past scores, not with others.

TellApart is built on this method: you hear one word, spoken by one of many native voices, and tap which of two words it was. The levels follow the same order, from the start of the word to clusters, then the middle and end.

When the problem isn't the sounds

Not every listening problem is about sounds. If you can hear each word clearly when it's said alone but lose whole sentences, other causes are bigger: fast speech, words running together, and words you don't know yet. Ear training gives you a solid base for those, but you'll also need plenty of real listening.