I just noticed that Tinker Tailor Soldier Spy (the unforgivably disencommaed movie adaption of Tinker, Tailor, Soldier, Spy has the Japanese title Uragiri no sākasu 裏切りのサーカス, "Circus of betrayal".

The original novel has always been simply Tinkā, teirā, sorujā, supai, so this represents a change in expected audience, naming conventions, or both. My money's on the former, based on the hypothesis that Japanese people reading books by John le Carre are doing so at least partly out of Anglophilia, and thus welcome exposure to English nursery rhymes they wouldn't otherwise know, while the movie promoters are afraid that leaving the first half of the title completely opaque (even English-speaking people barely remember what a tinker is) will discourage potential viewers.

Sidebar: The Spy Who Came in from the Cold is translated Samui kuni kara kaette kita supai 寒い国から帰ってきたスパイ, "The spy who came back from a cold country." This is a bit unfortunate in a couple of ways.

First, the fact that you can't use adjectives as substantives in (contemporary) Japanese means that the delightful ambiguity of "the cold" has to become "a cold country" (I suppose samusa could have been used, but combining that with kaeru that feels unidiomatic to me — "the spy who came in from the coldness").

Second, and related, the Japanese calque for "cold war" was 冷戦 — different character for "cold." I'm sure most readers figured it out, but the upshot is that no way was found to recreate the original title's balance of simplicity and depth.

Completely unrelated and extremely nerdy: An interesting (if unfortunately unfriendly on one side) exchange took place on japanese.stackexchange.com yesterday. To summarize, the publisher Shōgakukan prefers to render its name, in both kana and Roman characters, as shown at left — that is, in a way that preserves the etymology: shōgaku 小学 "elementary school[ing]" + kan 館 "house" (they started out publishing materials for elementary school-age kids). But the name is actually pronounced shōgakkan, same way that gaku 学 + 校 = gakkō 学校, "school." So we see citations like this on WorldCat: the Shogakukan progressive English-Japanese dictionary, published by "Shōgakkan."

On the other hand, it's not hard to find another entry for the book on WorldCat which gives "Shogakukan" as the publisher, and we can also find the Shogakukan Random House English-Japanese dictionary, published by "Shogakukan" (note no macron either!) but attributed to "Shōgakkan," along with Random House. So clearly WorldCat entries are not going to prove anything.

It seems to me that this is basically a style guide issue: do you romanize a publisher's name (as opposed to a book's title, or a personal name) strictly according to pronunciation, or do you use their preferred spelling? I had a quick look in the MLA Style Manual and couldn't find anything on this issue specifically, although we are advised to "follow known preferences" when romanizing Chinese and Japanese personal names. Do other commonly-used style guides prescribe more specific rules for "Shogakukan/Shōgakukan/Shōgakkan" situations?

Popularity factor: 19

Leonardo Boiko:

I’d go with the publishing house’s official spelling, but only because that’s presumably what they’ll use in their own books if they ever print romanized bibliographic information. I generally try to preserve the ortography of the original editions. In an ideal world, databases like WorldCat, Librarything etc. would then have extra information on alternative spellings (as aliases) pointing to the same publisher, author etc. (which is also a desirable feature to relate various transcriptions in the original writing system and romanizations).


For my dissertation, I mostly refer to Japanese books by made up English titles where possible, so I'm probably going to end up doing Japanese bibliography entries three times: English, romanized, Japanese. So, in this case, the English would be Shōgakukan, romanized Shōgakkan, and Japanese 小学館.

Now ask me where I put spaces and a title case caps in romanized titles.


Even English-speaking people barely remember how to spell 'tailor' ;)

L.N. Hammer:

The Wikipedia guideline is to use the romanization preferred by the person or corporation, where it is possible to determine, otherwise use modified Hepburn (with macrons).


Leonardo Boiko:

Carl: I once found this cute little paper called « On Word Boundaries in Japanese » http://www.jstor.org/stable/488773 .

The author observes informally that Japanese natives, when writing spaced kana for children, often do not insert spaces before most joshi: にほんの こころは うつくしいです。 One could run with this and propose an analysis where taigen (“nominals”) are inflected, after all: oishi-i, kirei-na, hadashi-no; oishi-i, kirei-da, hadashi-da, etc.

(Notice spoken Japanese generally doesn’t allow pauses before joshi, though I think Martin managed to come up with one or other context when it does. But visual spaces are not necessarily representations of spoken pauses or junctures, anyway.)


Morgan: Oh God. Now we're all just going to pretend that that never happened.

Carl/LNHammer/Leo: Interesting, thanks. The distinction between "English name" and "Romanized name" is one that makes sense to me.

(I am actually curious about how you decided to handle the capitalization-and-spacing issue. Are there any "best practices" for this in academic writing, e.g. some influential journal's policy that others have adopted?) What does that paper's author conclude?

Leonardo Boiko:

Concluding paragraphs:

“Even though Westerners naturally tend to regard Japanese nouns as uninflected, lacking as they do the familiar inflectional categories of number and gender, they would not have treated them thus had it not been for the low degree of fusion between noun stems and their suffixes. […] When it comes to determining word boundaries through syntax, the concept of immediate constituents is at best difficult to apply. If we are to regard /e/ ‘to’ as one of the immediate constituents of /anotatémono-e/ ‘to that building’, and then to call it a word because /anotatémono/ is two words, what is to prevent us from deciding that /ta/ ‘past tense’ is not an immediate constituent of /takusantábe-ta/ ‘ate a lot’ (one immediate constituent being the tense and the other being the action ‘eating a lot’), and then calling /ta/ a word because /takusan tábe/ is two words? Of course we don’t do this. To begin with, there is considerable fusion in Japanese between verb stems and tense suffixes. But quite apart from this, we subconsciously treat tense morphs as affixes. We do it even in the case of the Japanese adjectives, where the fusion is little greater than for the noun.

As a result of such analysis, Japanese appears to have a rather lopsided structure. The major word classes of verb and adjective are inflected, but the other major word class consists of the uninflected nouns. In addition there is a class consisting of one inflected word—the copula.

Have we not,in effect, taken a language which has three major classes of inflected words—nouns, verbs, and adjectives—and stripped one class of its inflection, and then called that inflection the copula and the particles? If so, then it follows that when we ask unsophisticated Japanese to do the same, they may not always be sure when to write ‘suffixes’ as independent ‘words’.”

I don’t know if any major grammar ever tried to pursue this idea, or what Japanese linguists think of it, or whether it would stand the typical criteria for “wordness” in modern linguistics.


Hm, but I suspect if you ask native speakers of inflected languages to pull out one word from a sentence, they'll pull out the inflection on the noun as well. I haven't run this experiment (I was stuck on WH-movement blockage in contextually negative sentences), but I think if you ask a native Japanese speaker to repeat the noun as it was included in the sentence hon o yomu, you won't get hon-o. Which to me indicates that what we call participles is not as inseparable from the host word as you'd expect in inflection.

But experimental morphology was a long time ago, and in another country, and besides the wench is dead.

Fun fact! Worldcat shows that the text 御産部類記 is parsed as Osanbu/Gosanbu ruiki. Which is just so, so wrong. (It's a buruiki--a genre--about births-that-require-honorifics. No idea what they thought a sanbu was.)


Okay, on second read I see he sort of addresses this. Although, and I'm not certain, he's treating たくさん食べた as one word because of the pitch contour, which I think is a bit mistaken--this was only current knowledge in 1997, but you couldn't determine words by pitch or inflection. We were only modeling on sentence-level, and there was a lot of hand-waiving about how to get around it (which I sort of made sense of by using optimality theory and clever distractions).

But basically, the impression I always had from lab work was that phonological data did not provide useful information for morphological categories. There was too much brain involved in categorizing things.

(And really, why should Japanese be thought of as weird by Indo-European categorization, requiring a re-analysis of the grammar? Indo-European's weird by other language standards, so I'm certainly not inclined to look for noun inflection because it "must" be there.)


"I'm certainly not inclined to look for noun inflection because it "must" be there".

And of course there is Mongolian, where all the case markers are treated in the old Uighur script as separate particles, but in the modern spoken language are fused with the noun like Indo-European inflections...


Hey, if it's good enough for Portuguese missionaries...

One problem is that there's no absolutely 100% reliable cross-linguistic definition of "word". I think we can all agree that particles are freer than Latin noun declension endings but less free than, say, the nouns they attach to; the question of whether they end up on the "word" or "inflectional ending" side of the line is kind of academic. (Although I can't resist: they're definitely closer to "word" than verb/adjective endings are. There are lots of situations in spoken Japanese when は, が, を, and even に are optional ("食べ行く?"), but た or て are always required. Bringing up conjugation is just pointless water-muddying IMHO.)

Leonardo Boiko:

Well one could argue that 食べ occurs alone as well, and the fact that the supposed “noun inflections” can be ommited doesn’t in itself prove that they’re less inflections than the “verb inflections” (in spoken Brazilian Portuguese we often omit verbal inflections). But yes, I’m not saying I subscribe to Chew 100%. I agree with you that (to use school grammar terms) the “bunsetsu” unit of taigen+joshi (りんご+を) seems to be less solid than the yougen+jodoushi unit (食べ+た). Still, it’s not pointless to notice the facts that there’s no pause allowed between りんご and を, and that natives tend to write them without spaces, and that the entire bunsetsu works as a unit for the standard pitch accent, etc. Whether to call this phenomenon “inflection” or “words with particles” is just a matter of labels; but no matter what the label, I agree with Chew that がっこう+へ forms a kind of unit that behaves differently than English “to school”, and that fact should be recognized.

“Word” is a troublesome label IMHO. There are words as delimited by spaces in an orthographic tradition; there are words felt as such by natives (since most languages have a word with a role similar to “word”); and there are words as defined by objective criteria by linguists (phonological contours, free ocurrence, anaphoric behavior etc; Haspelmath lists some). I think the three kinds of “word” seldom match. And then you realize each category actually has a lot of competing alternatives (ortographic traditions will vary with time and place and style guides, and illiterate natives differ widely on what they call “a word”, and don’t let me get started on all the linguistic schools of thought). I think I read in John Miles Foley that oral traditions tend to use “word” as a very elastic unit, which can describe short expressions or whole verses or even entire passages. And then there’s the psychological tangibility that a spaced writing system gives to its words, which many (Sampsom, Rogers) have described as a kind of soft logography; we learn to recognize word shapes visually at a glance, and this influences the “wordness” for everyone who can read. E.g. in Portuguese “em cima” = “on top” is two (typographical) words but “embaixo” = “below” is one, even though they’re totally parallel in morphological makeup, pronunciation, syntactical behavior etc. If you ask a literate Brazilian he’ll say “em cima” is two words and “embaixo” one, but I suspect an illiterate native will choose one or the other.


Good points. My main objection to the verb ending = word thing is that while it may work for "tabe ta", it's harder to justify "it ta" (breaks phonotactics, although I concede that this may be because the standard description of J. phonotactics are based on the wrong assumptions) or "ton da" (phonotactically okay, but we are now arguing that these two separate words change in unison).

IIRC, in some Ryukyu languages there IS actually fusion between nouns and (some?) "particles", to the point where the ibflection analysis makes sense. And on the mainland there are examples like "bokaa" (< boku wa) from modern Japanese and "konnittaa" (< konnit(i)? wa) from early modern Japanese which make me more comfortable with a continuum model than a binary word/ending distinction.


I propose we call any such ambiguous situation "ibflection" now, and perhaps "Treyvaudian ibflection" when we want to be formal.

(It's just a typo, but it's such a fun typo to say! "Ibflection.")

L.N. Hammer:

Yes, but what with English phonetic changes, it'd become "imflection" within a generation.


And with contemporary English lexical change, it would be MORPHOLOGY FAIL LOL a generation after that.

(Just my like to make a typo incorporating the one VC syllable that isn't some kind of Latin root...)


Housemaker Kobori Juken changed its name many years ago to SxL (see sxl.co.jp).

Overseas, this is invariably read as SXL, in the same manner as TDK, NEC or JVC. The middle "x" is frequently capitalized.

In fact, the name is a building reference. The x shouldn't even be in lower case, it's actually a multiplication sign. The name in Japanese is エス・バイ・エル. From the company's website:



Speaking of lower case, when Horiemon was at his peak, many people went out of their way to refer to his firm as livedoor.

Once he started to be dragged through the courts, however, everyone seemed to give up, and Livedoor it was. I noticed this again when Kazuhiko Shimokobe was named as the new TEPCO chairman. All references to his involvement in the investigation of the internet company call it Livedoor.


Didn't know that about livedoor -- that'd make an interesting anecdote in a sociolinguistics paper.

Re SxL, that's some good trivia. Mind you, with that domain name and that logo I don't feel like they have anyone to blame but themselves for the misunderstanding.

Comment season is closed.