Background, problem statement

Very often, I find myself in situations where I have to build structurally complex sentences in Japanese, and find myself struggling, trying to put all I want to say in a single sentence. As far as the other languages I know well enough go, it's not an issue since

  1. they have relative pronouns that resolve many ambiguities (that, which, who, whose, qui, que, dont, auquel…),
  2. their grammar allow incremental stacking of relatives, starting with the base of the sentence (see example).

I guess there are two viable solutions to my problem, but I never really paid attention to which was usually chosen in spoken (nor, in fact, written) Japanese.

Solution 1, the most likely

Break your sentence in many small chunks, make a sentence of each chunk, and convince yourself that unlike French or English, it's not awkward to have a train of sentences like "Aです。Bです。AとBの関係はCです。Dです。CとDの関係はEです…"

Solution 2, the "wished" one

It is possible to express unambiguously sentences like

On Monday, the dog that ate the pudding that I cooked and whose owner's sister I met yesterday will be castrated.

My attempt at this sentence would be like:


But even though I feel quite satisfied with this simple, quite linear one, I don't think it's likely to be heard… (FWIW, the sentences I build are often describing inter-related complex mathematical relations, which makes thing even harder…)

The recent remark on "invertion" makes me wonder even more if this common to have such sentences, because inversion may cause ambiguities to arise:


The person with a Japanese interlocutor? The Japanese with an interlocutor?


The partner with a Japanese person? The person with a Japanese partner?


  • Do you have issues expressing complex relations in Japanese?
  • How do you get round this issues?
  • Are there relation patterns in English that you will definitely break into several Japanese sentences?
  • Do you have trouble understanding the aforementioned kind of Japanese sentences?

And subsidiary question, if ambiguity is definitely a major issue to all: how could the language not evolve to avoid ambiguities?

  • 2
    You cannot always do word-to-word / clause-to-clause translation. I learned it in the other way around; some sentences in Japanese cannot be translated into English without breaking them into several sentences. Aug 31, 2011 at 15:39
  • 2
    You might find the "「どっちにてん?」" and "絵にぴったりの文を作ろう" games here of interest, as examples of how minor changes affect interpretation of sentences: nhk.or.jp/school/bangumi/wakaru-tubo/2-hp3-game.html To be honest, on first glance of your example I thought Monday was the name of the dog. (British English speaker here - I'd put "On Monday...")
    – nkjt
    Aug 31, 2011 at 16:49
  • @Dave: The questions are short, but not the preamble :)
    – Axioplase
    Sep 1, 2011 at 2:33
  • @Tsuyoshi: I quite guess that's how it works, but I wonder if there are some constructions that are known to be problematic (like those using "of which" or "whose") when going from English-like language to Japanese.
    – Axioplase
    Sep 1, 2011 at 2:34
  • 2
    Another common case where you cannot translate an English sentence to a Japanese sentence literally is what is called 無生物主語 in English classes at schools in Japan, such as “A toothache kept him awake last night.” (This example was taken from Wikipedia.) Sep 1, 2011 at 14:39

3 Answers 3


This happens a lot in patent translations, so you might get some hints by searching for these terms: 特許 請求項 翻訳

This site has this example:

  1. A dynamic random access memory including at least two banks, each of said banks including memory cells arranged in rows and columns, said memory cells storing data provided by at least one bit line and by at least one data line, the dynamic random access memory comprising: first switching means for selecting one of said at least two banks; and second switching means connected to said first switching means for selecting one of said columns, wherein said first and second switching means couple one of said bit lines to one of said data lines, enabling data to be written into or read out of memory cells common to said selected bank and to said selected column.

[請求項1] 少なくとも2個のバンクを含み、前記各バンクが行と列に配列されたメモリ・セルを含み、前記メモリ・セルが少なくとも1本のビット線と少なくとも1本のデータ線から供給されるデータを記憶するダイナミック・ランダム・アクセス・メモリであって、


This site has many more translation examples.
The wording in patent translation is obviously very formal and it's probably not what you want, but I guess if you want to investigate whether one language is somehow better at unambiguously expressing sentences with lots of relative propositions, this could be an interesting corpus.

  • 1
    Very good idea!
    – Axioplase
    Sep 2, 2011 at 10:24
  • 3
    This is a very good idea. However, it might be worth pointing out that patents are written in the “patent Japanese” which is not exactly the same as Japanese outside patents. For example, I think that ~ランダム・アクセス・メモリであって、~ランダム・アクセス・メモリ is usually considered incorrect outside patents (the correct form would be ~ランダム・アクセス・メモリであって、~もの). “Specialized languages” like this exist also in other fields such as mathematics and law. Sep 2, 2011 at 12:28
  • 1
    Since the OP did mention mathematics, it's probably worth mentioning that ci.nii.ac.jp sometimes has titles or abstracts in both Japanese and English, searching for a mathematical term in Japanese might turn up something of interest.
    – nkjt
    Sep 2, 2011 at 14:01
  • 2
    Also one cautionary tale.. Many patent translations are really, really bad so you need to be careful. Sep 3, 2011 at 3:44

I don't see anything wrong with solution 1, but not because it makes it easier for the speaker. Breaking up complex ideas also makes things easier for the listener to digest, piece by piece. Of course it's ridiculous to take it to the level of "Here's this. Here's that. That relates to this in a certain way. . . ad nauseam". But you can and probably should limit how much information you put into one sentence, then logically work in more complex relationships in separate sentences. Even in English, I don't think I would ever think to put all the information contained in your example into a single sentence. I'd be more likely to say something like "Yesterday I met the sister of the guy whose dog ate that pudding I made. She said the dog will be fixed on Monday." Or even more simply "You know that dog that ate the pudding I made? I met his owner's sister yesterday. Apparently, the dog's getting fixed Monday." In Japanese, maybe 「昨日、私が作ったプリンを[Edit-see comments (X食われたX)]食った犬の飼い主さんのお姉さんに出会った。犬は月曜日に去勢されるよ。(へへへへ)」

  • 2
    +1 for the general idea. However, 私が作ったプリンを食われた犬 is unnatural because of the combination of passive and relative clause. 私が作ったプリンを食った犬 is the usual way to say it. (I might also change 食う to the more common 食べる, but that is a different story.) Sep 1, 2011 at 13:37
  • I see. Thanks for the correction. Would 「犬に食われたプリン」 be OK (not in this particular sentence, but hypothetically), or does that not work either?
    – rdb
    Sep 1, 2011 at 14:27
  • 1
    It works. I do not think that it is possible to extract the ~に part of a passive sentence as a relative clause, but ~が part is fine. Sep 1, 2011 at 14:30

You’ve probably noticed this but English is a right-branching language, meaning subordinate clauses generally get deeper as you go to the right. Japanese in contrast is a left-branching language. Language is always spoken from left to right (forgive the conflation of speaking and writing), so naturally a right-branching language lets you deeply nest subordinate clauses with little confusion about how the clause relates to the utterance as a whole. If you were to deeply nest subordinate clauses in Japanese, the listener would hear the deepest subordinate clause first.

When speaking Japanese, you already have to keep a sort of stack (in the programming sense) of components in your head as you hear the sentence, and delay interpretation until hearing the predicate (or until you’re confident of what the predicate will be, given the flow of the dialogue). Deeply nesting to the left, hence, greatly elongates this stack; so unless the speaker can keep more than 7 components in mind at once, they are forced to make an interpretation without full information, which can in the worst case lead to them having to re-hear the sentence if their interpretation turned out to be irreparably wrong.

Those sorts of garden path sentences can be used for comedic effect by utterly changing the meaning of an entire long utterance with a のではない, らしい, or some other ending. In those cases, one can simply negate or slightly alter there existing interpretation, so there isn’t a great mental burden or a need to rehear the sentence. This also applies to relatively semantically-light function words like こと、よう, etc. that don’t give the listener much mental burden when transitioning from a long relative clause.

At the same time, this is what makes deeply nested sentences like those found in English unnatural in many cases. Moderately long sentences with one, rarely two, layers of nesting can be fine. The subordinate clauses should be kept brief in most cases if they modify something that is crucial to their interpretation, but you’ll find some flowery novels that push them to the limit.

When translating an English sentence with deep nesting or lengthy subordinate clauses, though, you’ll just have to break up the sentence a bit. “I met a guy who knows a guy x 100 who knows Ariana Grande.” would not translate well to Japanese without breaking it up. A good test may be to reread what you wrote in Japanese two days later and see if you can easily interpret it without recalling the English.

Another tactic is to translate relative clauses by putting them in parentheses after the word they describe. For example, “which is located in Shimbashi, Minato-ku Tokyo-to” could be rendered as “(東京都港区新橋にある)” immediately following the noun it describes (preceding any particles). This induces some mental burden to interpretation but not as much, and since they are parenthetical the reader may choose to ignore them on first pass.

You must log in to answer this question.

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