Via paperpools, this interesting tidbit about Japanese, from Senko K. Maynard’s Expressive Japanese: A Reference Guide for Sharing Emotion and Empathy:

Emotion words and expressive strategies cannot, as a rule, be used in reference to persons other than the speaker. In cases where they do refer to persons other than the speaker, most must either be used in quotation or must go through grammatical manipulations that explicitly mark them.
This is particularly true in the case of adjectives of emotion, for example, kanashii, ‘sad’. Now, Kanashii ‘I’m sad’ is grammatically correct, but *Ano hito wa kanashii ‘He is sad’ is not acceptable under ordinary circumstances. Instead, what is acceptable is Ano hito wa kanashi soo da ‘He appears to be sad’. In Japanese it is necessary to mark the sentence if it is not about the speaker’s but about someone else’s emotion. The reason for this is that, although one can experience one’s own emotions directly, someone else’s emotions are not so accessible.
The distinction in Japanese is not necessarily required in other languages. In English, for example, it is possible to say “I’m happy” and “I think so” as well as “Yamada is happy” and “Yamada thinks so.” It is important to remember that in Japanese, the distinction is obligatory, and extra attention must be paid when referring to one’s own or another’s thoughts and feelings.

Interesting if true, of course. Can any of my Japanese-speaking readers verify this?


  1. michael farris says:

    You haven’t come across this before? I tought it was old new. I first read about this in a book on evidentials and numerous Japanese speakers I’ve asked have confirmed it.
    It also helps to disambiguate things in a language where over subjects are not required as ‘kanashii desu’ can only refer to the speaker….

  2. I’ve read it, too, back in the 1970s or early 80s. Kuroda, I think it was. But I think it needs to be qualified. You can tell someone to ‘have a nice day’, or ‘avoid sad things’, for instance, in which case it is a nice day or sad things from that person’s perspective. But it is still ‘subjective’ in its meaning.
    Apart from kanashisō da there is the form kanashigatte iru ‘acting sad’. The implication is that the person ‘looks sad’ or ‘acting sad’, but the true internal feelings of the person can only be judged by the person himself.

  3. Not being a native speaker, I can’t really confirm this, but I agree with the general intuition. However, I’m having trouble validating the thesis in google hits.
    私は悲しいです。 “I am sad.” 85,800
    あの人は悲しいです。 “That person is sad.” 0
    あの人は悲しそうです。 “That person seems sad.” 0
    彼は悲しいです。 “He is sad.” 10
    彼は悲しそうです。 “He seems sad.” 4
    *さんは悲しいです。 “*-san is sad.” 1,580,000
    *さんは悲しそうです。 “*-san seems sad.” 318,000
    私はうれしいです。 “I am happy.” 15,100,000 000
    あの人はうれしいです。 “That person is happy.” 1
    あの人はうれしそうです。 “That person seems happy.” 0
    彼はうれしいです。 “He is happy.” 9
    彼はうれしそうです。 “He seems happy.” 8
    *さんはうれしいです。 “*-san is happy.” 10,600,000
    *さんはうれしそうです。 “*-san seems happy.” 284,000
    Searches made with quotes in I resorted to the politeness marker (desu) as a way to try to fish whole sentences and filter out embedded clauses and other grammatical trickery. It didn’t go so well—in that last search, for example, many hits seem to be of the form “tanaka-san wa ‘ureshii desu’ to itta” —“Tanaka-san said ‘[I am] happy’”. The fact that google doesn’t index punctuation really hurts here.
    What these results suggest to me instead is that forms like “<pronoun> wa <adj> desu” that we learn in textbooks are artificial and little used in real Japanese (in many ways the very concept of “pronoun” feels like a calque). Japanese will often just omit the subject altogether (and then how can we test the hypothesis?). Many of the hits we get are from language-related websites (learning, translation &c.)—thus, not representative.
    One more thing—AFAIK things like “Tanaka-san wa ureshii.” (“Tanaka-san is happy.”) are appropriate for fiction with an omniscient narrator. This makes it even harder to test it in google. I wonder if a real linguist would be willing to go through an appropriate corpus.

  4. The reason why “tanaka-san wa ‘ureshii desu’ to itta” isn’t yielding any hits is that in Japanese, the main verb following quoted speech must be in the plain form.
    うれしいと言った has 818000 hits.

  5. Er, “Tanaka wa ureshii” feels more like adult fiction (to refer to a character as “Tanaka-san” in prose would be like calling him “Mr. Tanaka”, I suppose.)

  6. Oh but “*さんは「うれしいです」と言った” does yield hits. The problem is that these hits interfere with what I want to search for, which is “*さんはうれしい[end of sentence].”
    …Wait, we can do better if we simply filter out 言った and いった。 I see some うれしいですかs so let’s also remove ですか。
    «”さんはうれしいです” -言った -いった -云った -話した -ですか» → 887,000 hits.
    We still get lots of quotations and other false positives though.

  7. And what if adjectives of emotion are applied to animals? Can a dog be happy, despite never having a chance to tell about it in a human language?

  8. You haven’t come across this before?
    Japanese, sadly, has not been one of my languages since I was four.

  9. A 楽しい人 tanoshii hito is a person who one likes to be with. Not a person who is enjoying themselves.

  10. David Boyk says:

    There’s a similar thing in Hindi, though not as strict and not only for third persons. Sensations in general can hypothetically be referred to as facts, but it’s much more normal to refer to them as impressions. Examples:
    “Mujhe bhukh lagi hai” (मुझे भूख लगी है), “Hunger appears to me,” i.e. “I am hungry.”
    “Mujhe film achchhi lagi” (मुझे फ़िल्म अच्छी लगी), “The film seemed good to me,” i.e. “I liked the film.”
    The verb lagna लगना, “to seem or appear,” literally means “to apply or be attached.”
    But like I say, the rule is weaker than it apparently is for Japanese.

  11. michael farris says:

    “Japanese, sadly, has not been one of my languages since I was four”
    Oh, I thought you’d kept up with it more (don’t know why…)
    It also occrs to me that something like this might happen in Polish Sign Language. And idea like “He’s said” seems more natural as HE LOOK SAD, or HE SAY HE SAD (or ‘I SAD’ HE SAY) or I THINK HE SAD though this is impressionistic and not based on real data or anything.

  12. Michael, you make a great point in your first post. I think the tendency among many Westerners writing about Japanese would be to leap to the orientalizing conclusion that Japanese people refuse to deal with emotions directly – but another interpretation is that the “-soo da” construction simply removes ambiguity, and we can’t make grand conclusions about “the Japanese mind”.

  13. @vanya
    I’m not sure many Westerners would jump to the conclusion that Japanese people refuse to deal with emotions directly. In this instance, it would be more obvious to jump to the conclusion that they regard the contents of other people’s minds as less knowable than the contents of their own – or at least there is always an element of doubt.
    @ David
    I don’t think that quite same point can be made about Hindi. The gross generalisation about Hindi expressions of emotion/affect/etc is that Hindi speakers regard themselves as less in control, less the protagonist of their own lives, than do speakers of many other languages – so emotion is described as something that happens TO someone, rather than being something that the person is the originator of. With Japanese, the gross generalisation is that the Japanese regard other minds as unknowable in a way that one’s own mind is not.
    Disclaimer: Unlike most commenters here, I know nothing about languages whatsoever

  14. I, too, can verify it as a general rule. I’m not sure that it’s about inaccessibility of emotion (per se) in particular, though, more just regular old evidentiality: “I am happy (and I know it because I am experiencing it)” vs “Tanaka is happy (or so I deduce from what I see/what I hear/what she says/etc.)”, with different ways to indicate each example information source.
    And the thing is that you use the same markers to say things like “The food there is great (or so I deduce from what I see/what I hear/etc.)” and “Sales have been in a nose dive (or so I deduce from etc.)”, as opposed to “The food there is great (which I know because I ate it)” etc.
    So “inaccessibility of emotion” is indeed the ultimate cause for always needing these markers for discussing other people’s subjective states, but it’s not a special quirk in the language so much as an unremarkable result of applying the general rules consistently.

  15. michael farris says:

    “I’m not sure that it’s about inaccessibility of emotion (per se) in particular, though, more just regular old evidentiality”
    My vote is for evidentiality, a common enough linguistic category including IINM many of the languages that Japanese just may possibly be related to. Can anyone say anything about Korean, Mongolian and/or Turkic?
    The Hindi stuff looks more like a distinction between voluntary action and involuntary states, also a common enough distinction, including in Slavic languages.

  16. In Irish one says that hunger, sadness, etc. is on a person. Some people try to make this indicate some difference in outlook from languages that allow people to be hungry, sad, etc. I’m not so sure.
    My wife speaks several languages, but she always gets Irish and Hindi mixed up, and she maintains there’s a underlying resemblance between them, more than with other European languages.
    The theory that old features are preserved best at the fringes might apply, as Hindi and Irish are on the eastern and western edge of the Indo-European areas.

Speak Your Mind