Why these hide for years
The majority of Japanese consonants are close enough to sounds you already produce that no correction is needed. That very ease is what conceals the exceptions: a learner who is right about twenty consonants assumes they are right about all of them, and nobody corrects a sound that is comprehensible.
None of these errors will stop you being understood, which is both reassuring and the reason they persist. They belong in the category of things that make an accent audible rather than things that break communication, so they are worth working on after timing rather than before.
The h that is not an h
Before the vowel u, the Japanese h is produced with both lips rather than in the throat, giving a soft sound with no friction against the teeth. Learners usually substitute either an English h or an English f, and both are audibly wrong.
The way to find it is to shape your mouth as if blowing out a candle and let the air pass between both lips. Your top teeth should not touch your bottom lip at any point.
| Japanese | Romaji | Meaning |
|---|---|---|
| ふじさん | Fujisan | Mount Fuji: both lips, no teeth |
| ふゆ | fuyu | Winter |
| ふたつ | futatsu | Two things |
| ふろ | furo | Bath |
| ふうとう | fūtō | Envelope |
| ふく | fuku | Clothes |
Check with a mirror. If your top teeth are touching your bottom lip, you are producing an English f rather than the Japanese sound.
The s and sh split
Japanese s becomes a sh-like sound before the vowel i, systematically. This is why the kana chart has no plain si, and why writing it as shi in romaji is not an arbitrary convention.
Learners who force a plain s before i produce a sound that does not exist in Japanese. The same pattern applies to the t series, where the sound before i is a ch-like consonant, and the sound before u is ts.
| Row | Before i | Before u | Note |
|---|---|---|---|
| s | shi | su | No plain si exists |
| t | chi | tsu | No plain ti or tu |
| z | ji | zu | Merged in most accents |
| h | hi | fu | The fu is bilabial |
These gaps are why the kana chart has irregular rows. They are features of the sound system rather than exceptions to be memorised.
Word-initial ts, the genuinely hard one
English has the ts sound but only at the end of a syllable, as in cats. Japanese uses it at the beginning of words, and English speakers find this position surprisingly difficult, usually producing a plain s instead.
The trick that works for most learners is to build it from a word you already say. Say a word ending in ts, then say it again and stop just before the final vowel, then start from that consonant.
| Japanese | Romaji | Meaning |
|---|---|---|
| つくえ | tsukue | Desk |
| つぎ | tsugi | Next |
| つめたい | tsumetai | Cold to the touch |
| つづける | tsuzukeru | To continue |
| いつつ | itsutsu | Five things |
| みっつ | mittsu | Three things |
The z and j pair
The Japanese z and j sounds vary by position and by speaker, and the distinctions that exist in some historical spellings have merged in most modern accents. Learners often over-differentiate them, producing a distinction native speakers do not make.
The practical guidance is to follow what you hear rather than what the romaji suggests. At the start of a word the sound often has more of an initial closure than the plain z of English.
| Japanese | Romaji | Meaning |
|---|---|---|
| じかん | jikan | Time |
| ざっし | zasshi | Magazine |
| かぞく | kazoku | Family |
| みず | mizu | Water |
| じゅうしょ | jūsho | Address |
| ずっと | zutto | The whole time |
Two of the kana in these rows are pronounced identically to two others in modern standard Japanese, which is a spelling matter rather than a pronunciation one.
Drilling contrasts you cannot hear
The recurring problem with consonant work is that you cannot reliably hear a distinction you do not yet produce. Your own output sounds correct to you, because it matches what you intended, and a polite listener will not raise it.
This is the case for objective scoring rather than self-assessment. Unihongo's pronunciation lab plays a native model line, records your attempt and scores it, so a bilabial h that is still coming out as an f shows up as a number rather than as a vague feeling that something is slightly off.
Keep the sessions narrow. One pair at a time, six words, ten minutes, and the same six for a week; mixing several contrasts in one session slows all of them.
These are worth fixing and they are not urgent. Nothing here will stop you being understood, which is exactly why it is comfortable to leave them for years.