When I chat using romaji, Japanese people will always use si instead of shi, for example yorosiku instead of yoroshiku, siteru instead of shiteru. Do Japanese people actually pronounce shi as si?


4 Answers 4


It's the difference between Hepburn romanization and Wapuro romanizaton. Adults generally use Hepburn system for person names, signboards, posters and such (し = shi, づ = zu, じゃ = ja). But when they need to type ordinary Japanese sentences using romaji, they (almost unconsciously) use wapuro style because it's more efficient and unambiguous (し = si, づ = du, じゃ = zya/jya/ja). You don't have to type shi to get し when you can get the same character by typing si.

Please remember that wapuro style has nothing to do with pronunciation. That's just an input method, how they type hiragana/katakana on keyboards. Don't assume anything about pronunciation based on English alphabet.

  • I'm fairly certain that that is not the case. The reason many native speakers of Japanese type “si” is because that is what they were taught in primary school, not because it's shorter, but because it makes more sense to native speakers of Japanese who perceive <さ> and <し> to start with the same consonant, but <し> and <しゃ> with different consonants, which also explains why <しゃ> is a digraph and <し> is not and placed in a different order in the kana table. They also often use <sya> and not <sha> because, again, that is more intuitive to native speakers of Japanese.
    – Zorf
    Commented Dec 18, 2020 at 12:54
  • @Zorf So-called 訓令式 is not as important as you might think. Everyone starts to use the (modified) Hepburn system after graduating from elementary school. Even the most domestic companies like Mitsukoshi use tsu/shi, and you will almost never see "the Japanese system" in the Japanese media or on the streets of Japan. If "what's taught at school" or "what's natural" was really important, everyone would use chi to type ち, shi to type し, and so on. It can be safely said that the only time when Japanese adults use si, ti or tu is when they type romaji on a keyboard.
    – naruto
    Commented Dec 18, 2020 at 15:49
  • I never said it's what important; I said it's what feels more intuitive to native speakers of Japanese. You'll notice that Hepburn is mostly used when Romanization is used to communicate with non-native speakers, and Kunrei mostly to communicate with native speakers in contexts where the normal script isn't available, such as emaill addresses. The point is that <si>, <ti> and <tu> feel more natural to native speakers of Japanese. It's not as you say about saving keystrokes. If so then why would they use <zya> instead of <ja> where <ja> is shorter? <zya> feels more intuitive to them
    – Zorf
    Commented Dec 18, 2020 at 23:13
  • @Zorf See this survey. The most popular method to type じゃ is ja, and the number of people using it increases as they get used to keyboards (this survey was conducted mainly for young people who are not familiar with keyboards yet). As this page says, Japanese kids learn romaji with kunrei at the age of 8 and then "graduate" to Hepburn at 10. Adults use Hepburn almost everywhere except on keyboard. Keyboard is the only place where the weird mixture of kunrei/Hepburn (eg josi for 女子) is encouraged (hence its own name ワープロ式).
    – naruto
    Commented Dec 19, 2020 at 2:53
  • What does that have to do with anything? You're making an argument about which system is more popular here for the second time. I'm merely pointing out that the existence of <si> and <tu> is not to save keystrokes, but because it's considered more intuitive to native speakers.
    – Zorf
    Commented Dec 19, 2020 at 10:16

For their own perception, yes, from the perspective of English speakers, no.

In languages there is a difference between what is called "underlying phonology", which is the discrete sequence of sounds that speakers perceive, and "surface realization", which is what they're actually factually saying. In every language, a process occurs called co-articulation, where sounds take on elements of the sounds that follow them, such as position of the tongue and lips, effectively meaning that sounds blend more into each other in practice.

The "i" is a palatal vowel, a vowel where the tongue is positioned near the palate of the mouth, so while Japanese speakers are pronouncing what they consider "si" the tongue is already moving towards the palate when pronouncing the "s", this results into the s having an objective quality that places it somewhat in-between the sounds of English "s" and "sh" — the end result being that English speakers hear "shi", whereas Japanese speakers just hear "si".

In reverse, the vowel in "gal" is also a palatal vowel, but the vowel in "girl" is not, the result is that the "g" in both words is slightly different with the "g" in "gal" palatalized. This is almost completely imperceptible to English speakers, but Japanese speakers pick up on this, so "gal" got loaned as "ギャル" ["gyaru"] whereas "girl" became "ガール" ["gāru"] — note the "y" in the former.

This gives rise to two different widely used families of Romanization: the "Japanese styles", which are used in Japan, and sometimes in technical settings outside of it, which are based on the Japanese's own perception of their language, and the "Hepburn styles", which are based on the perception of North American English speakers, and mostly used outside of it.

In practice, in Japan, both styles are used, sometimes often in the same word, because Japanese speakers don't know what is up any more; the Japanese government recommends the Cabinet-style variant of the Japanese-style be used for all intra-Japanese communication, which they are taught at school, but also advises Hepburn for external communication with Anglo-Saxons; one can sometimes see "si" and "shi" in the very same word in the wild, due to sloppiness or really inconsistent Romanization.

It's however important to remember that every language has its own phonology and sounds; Japanese is not pronounced like English, or in reverse, and has different sounds. You may look at the Wikipedia article on Japanese phonology; the transcriptions inside of /slashes/ repræsent the underlying phonology that Japanese speakers perceive and what is inside of [square brackets] repræsents the factual sounds made. So what is written as "Fujisan" in Hepburns-style, and "Huzisan" in Japanese-style is perceived as /huzisaN/, with /N/ being different from /n/, by Japanese speakers, which explains why it is written as "Huzisan" in Japanese-style romanization.

  • 1
    An interesting historical point is that //e//, also a more front-of-the-mouth vowel, previously caused palatalization so that せ was pronounced more like [[ɕe]]. See also the whole "X ANTES DE E" section (that is, "words starting with XE") in the 1603 Nippo Jisho. In Portuguese spelling of the time, "x" was the letter for the [[ɕ]] sound. Google Books here. All of these are words starting in せ, and realized in modern Japanese as [[s​e]] instead of older [[ɕe]]. Commented Feb 10, 2020 at 20:18
  • @EiríkrÚtlendi indeed, Japanese still does not contrast /e/ from /je/ in native words; one can also see that older loans like “gelatin” became “zeratin”, which would have become “zyerat'in” if it were loaned today, to reflect the adaptation Japanese has made to accompany foreign phonology. Japanese speakers still release /Ne/ today as if it were /Nje/ frequently but no longer realize /se/ as if it were /sje/ as they still do with /si/
    – Zorf
    Commented Apr 30, 2020 at 14:15

The usual Japanese pronunciation of し is /ɕi/.

Japanese phonotactics often palatalizes the consonant before i or y (/j/), a feature that is also present in Korean ㅅ and some dialects of Brazilian Portuguese with t. The digraph sh in English usually represents /ʃ/, which is similar to /ɕ/ but differs in tongue position.

So しょ /ɕo/ will be romanized as syo instead of sho in systems that use si, such as Kunrei-siki. In the same way, ち /t͡ɕi/, じ /(d)ʑi/, ぢ /(d)ʑi/ can be found as ti, zi and di respectively, because the palatalization of the consonant is expected to native Japanese speakers. Other "irregular" consonants occur on ふ /ɸɯ/, つ /t͡sɨ/, づ /(d)zɨ/.

When I was a teenager, JICA sent a native Japanese speaker to our school in Brazil. I was surprised to actually hear something very close to /hu/ instead of /fu/ for ふ when she spoke (I was used to Hepburn). My grandmother, a native Japanese speaker, could not distinguish /si/ and /ɕi/ well. That's why Kunrei-siki makes a lot of sense from the point of view of a native Japanese speaker.

Romanizations are tools for easing transcriptions, but they may induce errors because of biases we have due to our native languages. That's why any discussion that involves pronunciation benefits greatly from IPA :)

Kunrei-siki does have a weakness though: how to romanize stuff like スィ, セィ, ティ, ホゥ, ドゥ, combinations that are rare but do exist and may occur in loanwords? The kana here indicate we are supposed to keep our /s t h d/ sounds and not let them become /ɕ t͡ɕ ɸ dz/. Kunrei-siki just ignores the issue and romanizes these as si, si, ti, hu and du. So, in extremely rare cases, your si will indeed be /si/.

And yes, both スィ and セィ exist and are supposed to be /si/!

(some edits for clarity)

  • 2
    In Kunrei-style the romanization of スィ is <s'i> with an apostrophe. Same for <t'i>, <d'i> et cētera as one expects.
    – Zorf
    Commented Feb 11, 2020 at 21:06

Japanese people learn the forms si for し and tu for つ at elementary school. This romanisation system is called Kunrei. If they are chatting in romaji they are likely to use those forms since it also saves typing one character.

You must log in to answer this question.

Not the answer you're looking for? Browse other questions tagged .