How 「えい」 should be pronounced in the words like 英語, 先生, etc?

Question

P.S.: I heard that 「生」 (せい) in the word 先生 can be pronounced either as [sei] or as [see]. If it is so, is there any semantic difference between these variants?

Answer 1

Short answer: The allowed pronunciations depends somewhat on the word origin.

• For Sino-Japanese words (漢語), such as 英語<えいご> or 先生<せんせい>, the underlying vowel sequence is always ええ, but can be pronounced as either えい or ええ (despite its native orthography being <えい>).
• Most Yamato (和語) words are the same as the Sino-Japanese words, but in some cases the えい pronunciation has been discarded leaving only ええ as the correct pronunciation. For example, the Yamato お姉さん<おねえさん> has underlying ええ and its only acceptable pronunciation is ええ (and not えい). I'm not sure how many of the Yamato words have the singular pronunciation ええ (anyone know?).

When a word affords both ええ and えい pronunciations, えい is more formal.

Long answer: The pronunciation of both vowel segments [ei] and [eː] begin with [e]. The difference between the two is that [ei] has a pronunciation that shifts from the starting [e] quality to a final [i] quality whereas [eː] retains the [e] quality throughout its entire articulation. Vowel segments that shift in quality mid-articulation but are still treated as one single vowel by natives are called diphthongs (jp 二重母音). When you encounter a segment like [ei] in the speech stream of any language you have to ask whether it is a diphthong (a single vowel whose ending quality differs from its starting quality) or if it is a succession of two distinct vowels (jp 連母音).

Japanese phonologists stipulate a 3way classification of possible vowel sequences:

1. Intrinsically long vowels such as こう (as in こうする), transcribed phonetically as [koː]. Here there is no change in vowel quality during articulation. Long vowels like this one are phonemically transcribed as /VR/ where V is any of the five vowels and /R/ is a special 'vowel lengthening' phoneme which is a segment whose only purpose is to extend the articulatory duration of the preceding vowel (while preserving vowel quality). So, こう is phonetically [koː] and phonemically /koR/.
2. A succession of two identical vowels, called a double vowel, such as こお, which has two possible phonetic transcriptions [koː] or [ko^ʔo] where [^ʔ] represents a light glottal stop or pause, called "vowel rearticulation" which represents as a "brief dip in intensity" (The Sounds of Japanese (Vance, 2008)) of the speech signal The first phonetic transcription then is the same as that for intrinsically long vowels, namely [Vː] and the second transcription represents an alternative pronunciation. That is, in the case of two identical vowels, a speaker may articulate a [^ʔ] between the two, in which case the speaker is said to rearticulate the final vowel. To demonstrate vowel rearticulation for yourself, pronounce the words 「塩.を」/ɕio.o/ and pay attention to what you do at the '.' marker. Considering that vowel rearticulation is optional I should mention that the speech style for which [^ʔ] is most frequent is that of careful pronunciation (no contractions, weak forms, or other verbal shortcuts)
3. A succession of two distinct singular vowels /V₁V₂/ where V₁,V₂ ∊ {a, e, i, o, u} and V₁≠V₂. All possible 5² - 5 permutations (/iu, ie, io, ia, ui, ue, uo, ua, ei, eu, eo, ea, oi, ou, oe, oa, ai, au, ae, ao/) are found in the Japanese lexicon. Each singular vowel in /V₁V₂/ is distinctly articulated in normal speech. However, /V₁V₂/ sequences are not distributed uniformly across the lexical strata (和語、漢語、外来語、擬声語・擬態語、外国語). For example, the strata of native words [+Yamato] (和語) and Sino-Japanese words [+SinoJap]　(漢語) are observably limited to only 10 permutations (/ai, oi, ui, ie, ae, oe, ue, io, ao, uo/ excluding /ei/) when you consider /V₁V₂/ sequences contained within a morpheme. This stringent restriction on morpheme-internal vowel sequences might not be believable when considering [+Yamato] words like 払う where /V₁V₂/ = /au/. But, the morphemic composition in this case is actually 払う = 払 + う, meaning the /au/ sequence is interrupted by a morpheme boundary. The 10-permutation constraint (/ai, oi, ui, ie, ae, oe, ue, io, ao, uo/) only applies to morpheme-internal /V₁V₂/. However, The 外国語 strata, being the least restrictive, contains all 20 permutations.
   
   The question of whether vowel sequences in this class are true diphthongs as opposed to two successive vowels depends on whether or not the vowel sequence, say /ai/, resides within a word as in(敗者 = [hai]+[ɕa]), or if the sequence spans a word boundary as in(歯医者 = [ha]+[i]+[ɕa]). Labrune (The Phonology of Japanese, 2012) mentions that several leading phonologists in Japan agree that vowel sequences /Vi/ (including /ei/) within a single morpheme must be diphthongs, for example /ai/ in 貝 /kai/, soley because they are contained within a single morpheme.

The intrinsically long vowels, /VR/, and the corresponding identical vowel sequences, /VV/, are the same phonetically, [V:], it is only theorized that the underlying phonemic forms are different.The unit of prosody in Japanese is the mora, and these 3 vowel sequence classes have the following mora types:

1. Intrinsically Long Vowels /VR/ : 2 mora
2. Two Identical Vowels /VV/ : 2 mora
3. Two Distinct Vowles /V₁V₂/ : 2 mora

In other words, the 3 types of vowel sequences have equal prosodic weight, namely 2 mora.

These are the 3 classes of vowel sequences, but there is still the complication of word boundaries that must be accounted for. For example, the words 歯医者 and 敗者 both phonetically [haiɕa] but the constituent morphemes differ: 歯医者 = [ha]+[i]+[ɕa] whereas 敗者 = [hai]+[ɕa]. So you can see that with 歯医者 there is a morpheme boundary breaking the /ai/ sequence. Or in other words, we have a /V₁V₂/ sequence, but it is not contained within a single morpheme. There is more to say about this, but since your question is about vowel sequences within morphemes, I'll stick to that, just note that in the general case, you have to consider the possibility of morpheme boundaries breaking vowel sequences in the speech stream.

It is an etymological artefact that Yamato words (the native stratum of the lexicon) such as エイ <えい> 'stingray', 鰈<かれい>, and 姪<めい>, contain only the intrinsically long vowel /eR/ and never the sequence /ei/ at the phonemic level (as listed above the 10 possible disctinct vowel sequences for the Yamato words include only /ai, oi, ui, ie, ae, oe, ue, io, ao, uo/ but not /ei/ ). All the other strata, however, tolerate /ei/ vowel sequences. It may be worth it to reiterate this kind of unexpected last point: even though phonetic forms for Yamato words can in general be [eː] or [ei], the underlying phonemic representation is always /eR/, the long vowel and not /ei/. /ei/ is absent from the Yamato words. For most Yamato words then, /eR/ can then realize as either [ei] or [eː], but for a small subset this is not true and only [eː] is acceptable. For example, the Yamato word お姉さん /oneRsaN/ with necessarily underlying intrinsic long vowel /eR/ is always pronunced as [oneːsaɴ] with the long [eː], and never with [ei].

The kana orthography does not in general consistently represent /eR/ between the Yamato and Sino-Japanese words. For example, for Yamato お姉さん with transcription <ねえ> the pronunciation matches the kana, but for Sino-Japanese 先生 there are two pronunciations [eː] and [ei] with only one kana form <せんせい>. Also like the Sino-Japanese 先生, the Yamato word 姪 has kana <めい> but either pronunciation [eː] or [ei].

The difference, as pointed out by user1205935, is that [ei] is more frequently observed than [eː] in formal conservative registers, in addition to one form or the other being more prominent in certain dialects ([ei] is more comman than [eː] in Tokyo Japanese but such relative frequencies do not hold for some dialects in the Kyushyu area.)

Answer 2

There is no semantic difference. The pronunciation varies with local dialects, and with the level of politeness.

As for politeness, [sei] is a pronunciation sometimes used by people to emphasise formality (e.g. in conjunction with 敬語), but this is nowhere near a necessity. I would say that [see] is the common pronunciation. Try sticking in an almost silent [i] when you are almost stopping your voice at the end of a long [see] and you'll sound practically native ;)

EDIT: In response to some of the comments, a related phenomenon is the (slight) pronunciation of the otherwise silent [u], like in です [desu] (rather than the usual [des]) in formal situations (e.g. in conjunction with the use of 敬語).

Answer 3

The せい of 先生 is a good example of 長音{ちょうおん} (a long vowel). While it is written as せい , in reality it is pronounced as セー with a エー sound (not a エイ sound).

Other examples include:

Kanji   hiragana   prononciation
-----   --------   -------------
映画　   えいが      エーガ
英語　   えいご　    エーゴ
時計　   とけい　    トケー
丁寧　   ていねい    テーネー

Another example of a 長音 that is pronounced differently to the way it is written is the う in お父{とう}さん which is pronounced オトーサン with an オー sound instead of a オウ.