LINGUIST List 9.333

Fri Mar 6 1998

Sum: Japanese pronunciation of English

Editor for this issue: Elaine Halleck <>


  1. MARIE MELENCA, Japanese pronunciation of English

Message 1: Japanese pronunciation of English

Date: Thu, 05 Mar 1998 10:42:19 EST
Subject: Japanese pronunciation of English

On November 18, 1997, Marie Melenca sent the following
message out to LINGUIST and TESL-L lists. I received quite a
large number of responses, and I thank each and every one of
you who participated. As promised, I am posting a summary of
the responses; in it, I avoided repetition.

>Respected colleagues:

>Re: Accent Production and Linking Rules of Japanese
>Speakers of English

>I am in the process of writing my thesis in the M.A. of
>Applied Linguistics at Concordia University in Montreal and
>need some help. Of course, any assistance you provide would
>be acknowledged, and I would be very grateful.

>I am looking for the following resource materials:
>1) The existence (or not) of linking rules in spoken
>2) The teaching (or not) of linking rules in English to
>Japanese learners and its success rate.
>3) Research performed on Japanese speakers of English which
>looks at whether their production of non-linked items
>causes lack of comprehension. I would be happy to give you
>a summary of responses sent to me personally.

>Thank you.
>Marie Melenca, Concordia University, Applied Linguistics
>McGill University, Montreal, Pronunciation
Date: Thu, 20 Nov 1997 10:15:39 +0100
...there is a new book issued by Lincom called "Japanese
Phonetics" (sorry, can't get at the author's name from where
I'm sitting). It handles a lot of the details of Japanese
pronunciation, often seen from the standpoint of Western
[second e-mail:]
here is the complete reference:
Japanese phonetics : theory and practice / Tsutomu Akamatsu. -
Muenchen [etc.] : LINCOM Europa, 1997. - XVII, 412 p. : ill. ; 21
cm. - (LINCOM studies in Asian linguistics ; 03)
This is a very readable and accurate work!
Paul Boersma
Institute for Phonetic Sciences, University of Amsterdam
Herengracht 338, 1016CG Amsterdam
phone (31)20-5252385
Date: Thu, 20 Nov 1997 09:15:31 -0600
...For an earlier (but very reliable) study of the Japanese
system, including the distribution and status of glottal
stop see: Bernard Bloch. Studies in Colloquial Japanese IV
Phonemics. This piece appeared originally in Language
26.86-125. 1950, but was reprinted in: Readings in
Linguistics, ed. by Martin Joos. New York: American Council
of Learned Societies. 1958. pp. 329-348. Your library may
not have RIL, given the eclipsing stance, but it will have
Language and you'll find it helpful by way of data if not
As to the second part, Bates Hoffer has done some work on
contrastive English and Japanese. He is at Trinity
University in San Antonio, Texas. And also Winfred Lehmann
at the University of Texas might have some references...
James E. Copeland, Chair
Department of Linguistics
Rice University
Houston, TX 77252-1892
Office: (713) 285-5150
Home: (713) 666-9582
Fax: (713) 527-4718
Date: Thu, 20 Nov 1997 11:25:47 +0100 (MET)
I think the answer is pretty simple: Japanese has no closed
syllables (see note at bottom) , i.e. very simply, all Japanese
syllables have the structure (C)V. (There are two qualifications
to this which both are irrelevant to the present issue: there are
long vowels, which can be analyzed as VV, i.e. they follow the
CVCV pattern where the second C is zero, which for several
reasons (mostly morphological and prosodic - japanese vowels have
pitch, high or low, and "long" vowels can have different pitches
in the beginning and the end) is a better analysis than analyzing
them as (C)V with a long V); and there is a "syllabic" nasal /n/,
which forms a syllable of its own, sometimes rendered N in
transcription; thus there is a difference between juNi (ju-n-i)
and juni (ju-ni).) Thus, a Japanese speaker will perceive
something like English got up as go-tX u-pX, where X corresponds
to some Japanese vowel, usually i or u (which both have voiceless
allophones in Japanese in exactly these positions). Since neither
[ti] nor [tu] are phonetically possible combinations in Japanese
(/t/ has an allophone ch before i and ts before u), after t the
vowel inserted is usually an o (similarly after /d/). This
is exactly what happens when English words are incorporated
into the Japanese language as loans. Cf. poketto beru
'pocket bell' (these small gadgets - some people call it a
beeper - which you can have in your pocket, they have a
phone number and they make a beep sound when you call them,
but you can't talk to them like to a mobile phone), beddo
'bed', rabu 'love (in tennis or movies)'. There is a
Japanese brand of cigarettes, Seven Stars, which is
pronounced somewhat like 'sebun staasu', and (my favorite
example), a tourist bus service in Kobe city, the 'City
Loop', 'shiti ruupu'. You should have a look at Japanese
renderings of English loans; Bernard Bloch's classical paper
on Japanese phonology (Studies in Colloquial Japanese IV, in
Language 1950) has a lot of data and there is also a book by
Harald Haarmann published a few years ago about the use of
foreign languages (mainly English) in advertising that
should have an ample list of examples. [Marie found: (1989)
Symbolic Values of Foreign Language Use: From the Japanese Case
to Socioliguistic Perspective. Berlin: Mouton de Gruyter]
{second e-mail}
[Of course there are syllables that look as if they were closed,
but they are restricted to two types: syllables followed by 'n'
(which really is to be counted as a separate syllable or rather,
mora), and words which contain 'long' consonants (like gakkoo
'school'), which also have to be analyzed as containing four
morae: ga-k-ko-o. In neither case, we have real closed syllables
like in English. But this is just an aside; as a principle, there
_are no_ closed syllables in Japanese.]
My web page is, if you're
interested. Regards, Hartmut Haberland
Date: Thu, 20 Nov 1997 22:30:02 -0400
Subject: from LingList; RE: Japanese English Pronunciation
I'm a syntactician/semanticist, not a phonologist /
phonetician, so I don't have a list of reading on such a
topic. I reply to your posted message, though, just because
I'm a bit interested in the topic, as a Japanese native
speaker struggling to speak English well (without hope).
> I noticed that when a Japanese
> student speaks English, s/he often inserts a glottal stop
> or some temporal pause between C(C)VC to VC(C) (in either
> word or syllable boundaries, such as "got up", "biology").
I don't really get the mean, between C(C)VC
AND VC(C)? It's true that Japanese people tend to insert
a glottal stop between "got" and "up" (like [go?app] or
[go?t'app] where [?] denotes a glottal stop and [t'] an
unreleased stop). I don't know what you are trying to
say about "biology"...I think Japanese may insert a glottal
stop between "bi-" and "ology", like [bai?olodjii] where
[dj] indicates a voiced postalveolar affricate.
Then let me guess what's going on. My feeling is there are
two characteristics in Japanese phonology affecting this
First, the existence of so-called "mora obstruent", which is
often transcribed phonemically as "Q". This is a
phonologically neutralized obstruent that occurs before
another obstruent.
Some examples:
 /baQtari/ -> [battari]
 /baQsari/ -> [bassari]
 /baQchiri/ -> [bacchiri]
All of these are onomatopoeic adverbs indicating some
action or state. (/Q/ are not restricted in onomatopoeic
What is interesting is, in colloquial speech, this "mora
obstruent" may appear in word-final position. One example
is an interjective expression /aQ/ (corresponding to English
"Oh!"). There is no obstruent following /Q/ in this case.
What happens in this case? Below is the answer:
 /aQ/ -> [a?]
As you may know, Japanese is a "mora language" where all
syllables have the equal rhythmic status regardless of
emphatic accents. Because of this, when Japanese imports
foreign words, a mora obstruent is sometimes added to
accented syllables, in order to make the syllable "heavy".
 Eng. "battery" -> Jpn /baQteri:/ [batteri:]
When we, Japanese, hear the Eng word "battery", we feel
the first syllable is significantly emphatic and heavier
than other two syllables (of course, due to the accent).
Note that there is no accent in Jpn [batteri:] (yes, there
are many words without accent in Japanese!). The only thing
that makes the first syllable relatively "heavy" is the
existence of /Q/. (This strategy, though, is not applied to
all loan words).
Another interesting fact in borrowing words. This fact is
related to the constraint in Japanese that all the
consonants should be followed by a vowel. Let's see what
happens in the case of borrowing Eng. word "bat".
Due to the constraint, the last segment [t] in "bat" will be
followed by an epenthetic vowel: the result is /bato/
[bato]. However, we, Japanese, feel something uneasy in this
transcription. The reason is this: in English, [a] is
syllabic while [t] is, of course, not; in this sense, the
[a] sound sounds heavier (or longer) than the [t] sound. We
can feel the difference. However, our tentative
transcription, [bato], does not reflect this feeling. [a] is
transcribed as [a] and [t] is transcribed as [to]: both are
To make the difference, we add a mora obstruent to the first
syllable. The result is:
 Eng "bat" -> Jpn /baQto/ [batto]
Here, the first syllable is a bit heavier than the last one.
We feel this transcription is closer to the rhythmic pattern
of the original English word "bat".
This strategy is frequently (but not always) applied. Here
some of other numerous examples:
 "get" -> /geQto/ [getto] (rather than [geto])
 "bed" -> /beQdo/ [beddo] (rather than [bedo])
 "pit" -> /piQto/ [pitto] (rather than [pito])
 "rock" -> /roQku/ [rokku] (rather than [roku])
 "at" -> /aQto/ [atto] (rather than [ato])
(This strategy won't be applied to the (C)VC words where
the final C is not a stop: "pass" -> [pasu] not [passu])
 "got up" -> /goQto aQpu/ [gotto appu]
When speaking English (not speaking imported loan words in
Japanese), Japanese people try to omit the epenthetic vowel.
Though Japanese transcription of "get" would be [getto], in
speaking English, Japanese try to omit the final vowel [o]
and say [get]. But people sometimes fail to omit the "mora
obstruent": what people have in mind is /geQt/ rather than
/get/. In this case, people may pronounce "get" as [gett]
or [ge?t].
Then "got up" is now [go?t'app] or [go? app]
Sorry the story is getting really long...
My second point is, Japanese, being a mora language, tries
to preserve the "independence" (sorry, such a fuzzy word)
of syllables. Therefore, when pronouncing Eng. sentence
"It is an apple", Japanese people tend to pronounce each
word separately, like [it,iz,an,apl], rather than to combine
the word-final C with the following word-initial V
We can observe this tendency when a vowel is followed by
another vowel. Thus, "I am" may be pronounced as [ai am]
rather than [aiyam]. But, you know, it's sometimes hard to
pronounce [ai] and [am] separately, so people sometimes put
a glottal stop between the vowels in order to separate the
succeeding vowels ([ai?am]).
This strategy is common in Japanese language.
 /ai/ "love" + /o/ "Accusative case marker" -> [ai?o]
 /daioo/ "great king" (/dai/=big, /oo/=king) -> [dai?o:]
 /jooo/ "queen" (/jo/=woman, /oo/=king) -> [jo?o:] or
[jo:?o:] /toooo/ "East Europe" (/too/=east, /oo/=Europe)
-> [to:?o:] /ion/ "allophone" (/i/=different, /on/=sound)
-> [ion] or [i?on] not [iyon]
In a careless speech, glides may be used instead of glottal
stops. /ai/+/o/ -> [aiyo]. When there is no morpheme
boundary between the vowels, glides are often used. (/io:/
"sulphur" -> [iyo:])
Therefore, I think
> I also wonder if glides
> exist in Japanese,
this question of yours involves a good point.
Actually, there ARE phonemic glides in Japanese.
 /yoi/ [yoi] "good"
 /warui/ [warui] "bad"
But my feeling is, non-phonemic (epenthetic) glides are not
usually used in Japanese. Well, this is the end of my
(informal) story. Hope it helps. Best, -ken-
Date: Fri, 21 Nov 1997 07:52:41 -0500
From: "Karl V.(van Duyn) Teeter" <>
Subject: Japanese pronunciation
 I wish I were able to give you references to
literature, but I am no longer "au courant". Your
observations make good sense, however. Japanese is a
syllable-timed language requiring separation of vowels, much
like Italian, so the syllable structure in CVCVCV.... An
Italian trying to say "biology" will accordingly come up
with biology, just as a Japanese will use a glottal for the
separation. There are no "glides" in the language, though h
with variants [x] and [phi], y, and w before a all occur as
syllable onsets. Yours, kvt, sorry I can't be more
concrete! Good luck.
Date: Wed, 19 Nov 1997 17:02:57 -0600 (CST)
I am not a specialist in Japanese linguistics or phonology,
but I have an MA in linguistics, lived in Japan for 11
years, my wife is Japanese and we speak mostly Japanese at
home. I think the phenomenon you are observing is almost
entirely due to the fact that Japanese has syllable-unit
stress, as opposed to English, which is timed stress. This
means every syllable receives (roughly) equal stress and is
(roughly) the same length, whereas in English not every
syllable receives equal stress, and only the time between
stressed syllables is equal. This may make it sound like
there are glottal stops between vowels. However, the glottal
stop is not a regular phoneme of Japanese and does not
appear in the pronunciation of the language.
> ... Japanese or English by the Japanese. I also wonder if
>glides exist in Japanese ...
Not as such. Where English has a diphthong consisting of a
full vowel and a glide, Japanese has two full vowels (and
two syllables)...
Stuart Luppescu Ph.D.
University of Chicago
It's not whether you win or lose but how you played the
game. -- Grantland Rice
Date: Thu, 20 Nov 1997 08:41:47 +0000
...the phenomena you describe have a lot to do with
differences in the syllable structures of the two languages
involved. There are at least two articles I know that may
provide some help or even explanation. There is a very nice
paper on "The optimal syllable in L2" in one of this year's
issues of _Studies in second language acquisition (which
deals however mainly with breaking up consonant clusters
by deleting consonants or inserting vowels), and there is a
paper by Ito and Mester on the syllabic structure of English
loanwords in Japanese. I can look up the full reference of
the latter paper, if necessary. A fascinating topic! Best,
Dr. Ingo Plag
Institut fuer Anglistik und Amerikanistik
Philipps-Universitaet Marburg
Wilhelm-Roepke-Str. 6 D
D-35032 Marburg Germany
Tel: 06421-285560
Fax: 06421-285799
Date: Mon, 24 Nov 1997 17:32:00 -0000
Regarding your query about Japanese pronunciation on the
linguist list, what you've noticed is not a glottal stop,
but insertion of an identical consonant before voiceless
obstruents (i.e. geminate consonants). Geminate consonants a
phoneme, often symbolised as /Q/. Phonetically it is an
identical sound to the following consonant. In other words,
it has a long closure time.
e.g.) Here "-" denotes a syllable boundary, and "." (dot)
denotes a mora boundary.
/sa.-ka/ (slope) -- /sa.k.-ka/ (novelist)
/i.-ta/ (was present) -- /i.t.-ta/ (went)
/ni.-si/ (west) -- /ni.s.-si/ (dairy note){note
/si/ is [shi] esh}
(*Sorry to confuse you, but all geminate 'moraic' consonant
[with X mark] can be written as /Q/), So the above example
will be :
/saQka/, /iQta/, /niQsi/ respectively.
I'm doing the opposite job of yours (teaching Japanese to
English speakers) though I'm a phonetician and not an
applied linguist. And I've noticed that even speakers of
advanced level often cannot tell the difference in words
with and without geminate consonants. It is obvious in
As you might know, Japanese syllable structure is
(C1)(C2)V1(V2)(C3). The /Q/ can occupy only in C3 position.
It can form an independent mora, but cannot form an
independent syllable.
When to insert a voiceless geminate consonant is a complex
issue. This is actually my next research topic. I guess it
is both phonetic and phonological factors behind it. e.g.
phonetic duration of an entire word, and a mora, the point
at which the F0 starts to fall, etc. Phonological
acceptability of accent pattern, etc.
Japanese has glide /w/ and /j/. /w/ can occur in C1
position, or occasionally in C2 in some dialects and in
archaic pronunciation. The other glide /j/ occurs in either
C1 or C2, making other consonants palatal. The heavy
syllable C1C2V occurs only when C2 is a glide (usually /j/
but occasionally /w/ in some dialect and archaic
pronunciation as I said above). (e.g. /kja/, /kju/,
/kjo/, /sja/, /sju/, /sjo/, /tja/, /tju/, /tjo/,
/mja/,/mju/, /mjo/, etc.)
(Though in my recent paper, I'm arguing that it might be
possible for Japanese to take any voiceless consonants other
than glides in C2 position.)
If you don't read Japanese, so far the best reference written
in English would be: Timothy Vance "An introduction to
Japanese phonology"(1987) State University of New York Press
Sorry for incomplete answers/suggestions in haste.
if you have any specific query, please let me know.

 Mariko Kondo
 Scottish Centre for Japanese Studies,
 University of Stirling,
 Stirling, FK9 4LA, Scotland, U.K.
 Tel. +44-1786-466080 (Departmental Office)
 -467576 (Direct)
 Fax. +44-1786-466088
Date: Fri, 21 Nov 1997 18:16:51 -0500 (EST)
From: Natsuko Tsujimura <>
... First, words like "got up" could be pronounced with a
glottal stop even by some native speakers. It's natural to
put a glottal stop at the beginning of a vowel-initial word.
Furthermore, the final /t/ in "got up" is somewhat
glottalized to begin with, so I don't think what you
observed is peculiar to Japanese speakers. If, however, you
notice a distinct pause between these two words, I think
it's because non-native speakers tend to pronounce each
word separately: I know I used to do that until I became
more fluent in English. Second, I am not absolutely sure
which syllable boundary you are talking about, but assuming
that you refer to the boundary between "i" and "o" in the
first two syllables, again this is simply because the second
syllable starts with a vowel, which tends to be accompanied
by a glottal stop even in native speaker's speech although
some speakers have a glide [y]. I hope this helps.
Date: Wed, 19 Nov 1997 18:17:31 -0600 (CST)
The answer to your problem probably lies in the simple fact
that the japanese syllable is limited to...
Thus if you ask a Japanese student how many syllables in the
word "apple" for example, they will usually say four (unless
they have been taught otherwise) i.e. a + p(V) + p(V) + le ,
where (V)is a schwa like vowel, because that is the only
permissable syllable structure in Japanese. Naturally this
sounds odd to our ears, the english syllabic laterals and
nasals are quite difficult as well. Vowels in Japanese are
often "closed" with a glottal stop in word final position
which lends to the rather "clipped" sound sometimes
mimicked. Now consider all the consonant clusters in
English which has a syllable pattern something like...
 C(0-3)V C(0-4)
and you'll know why extra vowels appear so often. Try
looking up something about the syllable structure in
Japanese for more detail.
In my experience as an EFL teacher, once you've gone over
the basic sounds you need to teach a little about English
syllables... this later helps in dealing with consonant

Exercises such as the ones using the three allomorphs of -ed
(past tense) often prove enlightening... some students find
it hard to except that both "walk" and "walked" and "rain"
and "rained" are all only one syllable! Of course problems
with english consonant clusters are not restricted to the
Japanese, it is an odd system. good luck and have fun!
c. aortas
Date: Wed, 19 Nov 1997 18:17:31 -0600 (CST)
I am a linguistics graduate student at Ohio University
formally tutored pronunciation for Ohio Program of Intensive
On temporal pause in "got up," I agree with your suggestion
that it is a transfer from Japanese. Japanese has what is
called mora-phoneme /Q/ (Ohye, 1967; Vance, 1987), which is
like a glottal stop, as shown in the minimal triplets below:
 Long vowel V+/Q/
 huuko 'wind direction' hukko 'revolution' huko
 [hu:ko:] [huk:o] [huko:]
 kooshi 'lecturer' kosshi 'gist' koshi
 [ko:Si] [koS:i] [koSi]

 Note: [u]: high back *unrounded* vowel
 [h]: voiceless *bilabial* fricative
 [S]: voiceless postalveolar fricative
I would suspect that Japanese speakers unconsciously insert
the /Q/ when they come across words you pointed out, such as
- > "gotto" /goQto/, "big" --> "biggu" /bigQu/. Vance (1987)
covers this issue very well, it seems to me.
Phonetically, "got up" [gaRAp] example is very interesting,
because many things happening in there. ([R]:voiced alveolar
*trill*, [A]: open-mid back unrounded) First linking (a.k.a.
liaison, resyllabification) takes place at the word
boundary. Then, flapping of /t/ takes place. Moreover,
because of word-boundary effect, the first vowel is a little
shortened. On the contrary, Japanese speakers tend pronounce
it as /goQto aQpu/. Spectrographic analysis shows that the
first vowel length is significantly longer of Japanese's
utterance than the native speakers. The similar point was
found by Bond and Fokes (1985).
Let me finish up by providing some examples that might
undergo /Q/ insertion by Japanese speakers:
 mi/Q/d August
 the hea/Q/d office
 spo/Q/t announcement
 job-relate/Q/d exhaustion
 Tagame/Q/t HB 200
 Oh, he di/Q/d i/Q/t again.
 Sometimes they ge/Q/t embarrassed.
 Wha/Q/t are the things that raise mankind above
 The job marke/Q/t is so depressed.
 I agree with Davi/Q/d, I think.
Hope this will be of use!
 Bond, Z.S., and Fokes, J. (1985) Non-native patterns of
English syllable timing. "Journal of Phonetics, 13,"
 Ohye, S. (1967) The mora phoneme /Q/ in English loan
words in Japanese. "Study of Sound, Vol. XIII," 111-121.
Tokyo: The Phonetic Society of Japan.
 Vance, T.J. (1987) "An Introduction to Japanese
Phonology." NY: State University of New York Press.
Yutaka Anraku
Dept. of Linguistics
Ohio University
Date: Wed, 19 Nov 1997 23:39:32 -0500 (EST)
Subject: Re: Japanese English Pronunciation
Hello, I also teach ESL to adults.
I hope this will help you.
Very generally: Japanese sounds, represented by the Hiragana
writing system (one of three used in Japan), occur in
prescribed combinations. There are five different vowel
sounds, with corresponding written symbols, which first
occur alone, and then in combination with specific consonant
sounds. As far as I know, consonants never occur alone.
There are (prescribed) variations in pronunciation of
consonants as well as 'weakening' of specific vowel sounds
under specific conditions. The students repeat the sounds
of English to the nearest Japanese sounds. The teacher must
then find a way to explain and illustrate in an effective
way how this 'nearest' sound differs from what they know. I
have been fairly successful at it. It takes a determined
student to overcome the existing ingrained habits of
speech. Of course aural discrimination must also be
developed. I find that this type of problem is readily
identifyable as based in L1. It doesn't matter what the L1
is. It is then interesting to learn a bit about the
language in order to find the reason for the problem...
Mon, 13 Oct 97 13:30:58 EDT
Hello, Marie: I work with NNS who want to reduce their
accents. The approach we use is not based on any particular
L1. I have little specific knowledge of Japanese so can't
address #1. I am assuming your use of linking means the
smooth co-articulation between phonemes rather than chunking
which I think of a way to break of the flow of speech into
parts-of-ideas units.
2) The teaching (or not) of linking rules in English to
Japanese learners and its success rate. Since I don't design
teaching approaches to single L1, I can't answer. And any
text on pronunciation addresses linking (e.g., Gilbert,
Prator & Robinett, Celce-Murcia, etc). I tend to think there
is really only one rule though this doesn't make it easy for
NNS to execute. The rule "All sounds are linked within and
between words until there is a pause. If you must break
between words, make it light, easy & short". (This last part
is tagged on because some speakers can't link unless they
delete a final consonant or simplify a final cluster or
added a vowel off-glide.) We measure our speakers'
intelligibility both from a specific word-level task and
from a four-minute narrative using a taped sample played to
a naive listener. We do this this before and
after training in order to compare changes. We don't,
however, have any specific measure of linking or chunking.
3) Research performed on Japanese speakers of English which
looks at whether their production of non-linked items causes
lack of comprehension. Are you making a distinction between
intelligibility (understanding the word the speaker intends)
and comprehensibility (understanding the ideas the speaker
wishes to convey)? I tend to think linking is more related
to intelligibility and chunking to comprehensibility.
Karen A. Carlson, Communicative Disorders
University of Wisconsin-Madison

End of responses sent to me although I am in correspondence
with a few in order to clarify a few items. I will
continue my studies and hope to gain more insight as
time goes on. So, if you have any more information or
any questions concerning this subject, please
contact me at the address at the top of this text.
Marie Melenca
Mail to author|Respond to list|Read more issues|LINGUIST home page|Top of issue