Japanese pronunciation: 41 minimal pairs to listen to

Japanese has few sounds an English speaker cannot make. What it has is a rhythm English does not use: every beat — the mora — takes about the same time, and length alone can separate two words. That is where most of the 9 groups below are: long vowels, the short silence, the n that is a beat of its own, vowels that are whispered rather than dropped.

Each group first says what an English speaker does instead; the words below are the evidence. The pitch-accent pairs follow the Tokyo standard and include only well-established cases. Phonetic symbols are given for readers who use them; nothing depends on them.

What happens here: English times its speech by stress. It hurries over the weak syllables to reach the next strong one, so a word of four syllables can take barely longer than a word of two. Japanese times it by the mora, roughly one beat per kana, and every beat gets about the same time — a full syllable, the second half of a long vowel, a ん and the silence of っ alike. A small ゃ, ゅ or ょ is not a beat of its own: it fuses with the kana before it, so きょ is one beat and きよ two. Count 病院(びょういん) as four beats, びょ・う・い・ん, and 美容院(びよういん) as five, び・よ・う・い・ん. Tapping the beats on the table while you say a word is not a game for children here; it is the quickest way to hear what an English rhythm leaves out.

What happens here: Japanese has five vowels, and each keeps its full colour in every position, at the end of a word as much as at the start. English does two things to vowels that Japanese does not. It lets e and o glide, so that め comes out as ‘may’ and と as ‘toe’ — and to a Japanese ear the glide is a second vowel, which is how 目(め) (eye) turns into 姪(めい) (niece). And it weakens the vowels around a stressed syllable, so that the end of 酒(さけ) drifts towards ee: the familiar ‘saki’. Hold each vowel still, keep it short, and give the last one as much care as the first. う [ɯ] is made with the lips relaxed or only slightly drawn together, not pushed forward as for ‘food’; the rounded oo is understood, but it is one of the clearest signs of an English accent. Three kana are said differently when they are particles: は is wa, へ is e, and を is o, exactly like お.

What happens here: In Japanese a vowel held for two beats is a different vowel from the same one held for one, and the difference is as real as a difference of consonant: おばさん is an aunt or any woman of middle age, おばあさん a grandmother or an old woman. English has long and short vowels too, but they also differ in quality, as in ‘ship’ and ‘sheep’, so an English ear listens for the quality and lets the length go. Here only the length changes. Hold the vowel twice as long, without a break, without a second push and without letting it glide. The spelling helps once the pattern is known: in hiragana a long vowel is written with a second vowel, ああ, いい, うう, but a long o is mostly written おう and a long e mostly えい, and in ordinary speech both are said as one held vowel, not as o plus u or e plus i. Katakana usually marks a long vowel with a bar: ビール.

What happens here: The small っ is not a sound but a beat in which the mouth gets ready for the next consonant and holds it. Before k, t or p the mouth closes and stays closed for a whole beat; before s or sh the hiss simply lasts a beat longer. English writes double consonants but does not say them — the p of ‘hopping’ is no longer than the p of ‘hoping’ — and the only place English holds a consonant is across two words, as in ‘black cat’ or ‘bookcase’. That is exactly the feel: 切手(きって) is like ‘kit’ and ‘te’ run together, with the k held. Said with an ordinary short closure, 切手(きって) (stamp) becomes 来(き)て (come). The fault also runs the other way: a learner who has just met っ starts to hear it, and to say it, where there is none. Count the beats.

What happens here: English n is always made with the tip of the tongue behind the top teeth, and it joins the vowel after it whenever it can: ‘an egg’ becomes ‘a-negg’. Japanese ん is a beat of its own and never joins the next vowel, and its sound is set by what comes next. Before p, b or m it is an m (散歩(さんぽ), 新聞(しんぶん)); before t, d, n or r an n (本当(ほんとう)); before k or g the ng of ‘sing’ (銀行(ぎんこう)). At the end of a word it is made far back, with the back of the tongue raised towards the uvula and the lips apart. Before a vowel, y, w, s or h the tongue closes nothing at all, and the nasal sound simply carries on through the beat. That is where English goes wrong: run the n into the vowel and 禁煙(きんえん) (no smoking) turns into 記念(きねん) (commemoration). Keep the beat, let the n hum, and start the next vowel fresh.

What happens here: In standard Japanese, い and う usually lose their voice when they stand between two voiceless consonants — k, s, sh, t, ch, ts, h, f or p — and often at the end of a sentence after one. The vowel is not dropped: the mouth still shapes it and it still takes its beat, but it is whispered, so です sounds close to ‘dess’, した close to ‘shta’ and 好(す)き close to ‘ski’. English speakers go wrong in two opposite ways. Reading from kana, they voice every vowel and say de-su and shi-ta in full, which is understood but sounds laboured. Having heard ‘dess’, they cut the beat as well, or replace it with an English word that has a different length. The whispered vowel also leaves a trace: the consonant before it is still shaped for it, which is how two words can stay apart when the vowel that separates them is not voiced at all.

What happens here: Japanese has a single sound where English has r and l, and it is neither of them. The tip of the tongue taps once, lightly and quickly, on the ridge behind the upper teeth and comes straight off again [ɾ]. It is the tap many North American speakers make for the t or d in the middle of ‘water’ or ‘ladder’, and one many Scottish speakers use for r. The English r, with rounded lips and the tongue drawn back without touching anything, is understood but sounds strongly foreign; an English l is understood too. The confusion that changes words is a different one. The tap is so close to a very quick d that an English ear hears はら as hada, and a speaker who makes the contact too long or too firm says 肌(はだ) (skin) for 腹(はら) (belly). Keep the contact as brief as you can; the tongue touches and is gone.

What happens here: English has the sound ts, but only at the end of a word or syllable — ‘cats’, ‘lets’ — and an English speaker who meets it at the start of a word drops the t: ‘tsunami’ is said sunami. In Japanese that turns one word into another, because す is a syllable of its own. つ is a t and an s fused into one consonant and released together; start it exactly as you end ‘cats’. ふ has a different trap. Japanese has no f: ふ is made by blowing gently between the two lips, as if cooling a spoonful of soup, with the teeth taking no part [ɸ]. An English f, with the lower lip against the upper teeth, is understood without trouble but heard at once as foreign. Only the ts changes meaning, which is why the pairs here are all about つ.

What happens here: English marks a word by stress — one syllable louder, longer and higher than the rest. Japanese marks it by pitch alone: each beat is high or low, and loudness and length stay as they were. In the Tokyo standard the pattern is simple to state. The first and the second beat differ, so a word starts either high then low or low then high. Once the pitch has fallen it stays down for the rest of the word, and some words never fall at all. That last point is why some pairs only separate when a particle follows: 花(はな) (flower) and 鼻(はな) (nose) are both low–high, but the pitch drops on the が of 花(はな)が and stays up on the が of 鼻(はな)が. An English speaker turns the high beat into a stressed one and lengthens it, and that is often the more serious fault, because a lengthened vowel is heard as a long vowel. A wrong pitch on its own rarely stops anyone understanding; context does most of the work. Other regions, Osaka among them, use other patterns, so the pairs here are Tokyo only.