The enigmatic language of the new Windows 8 ads

« previous post | next post »

Everybody has been puzzling over the language of the series of online ads for Windows 8 that it recently released in Asia.

Native speakers of Mandarin, Cantonese, Japanese, and Korean declare that it is not any of those languages.  The first time I listened to them, the ads sounded as though they contained elements of some Wu topolect, a bit like mangled Shanghainese, but I could also definitely hear bits of Mandarin, albeit with unusual tonal contours and slurring.  What was most perplexing of all to me was that, although I was certain that the ads contained Chinese phrases and sentences, every Chinese person to whom I showed them emphatically maintained that they could not understand a single word!  In contrast, several non-native speakers of Mandarin said they could pick out a word of Chinese here and there.

Here is a sampling of the scores of replies I received from Chinese, Japanese, and Korean speakers (all native except where otherwise noted):

Mandarin speaker:  It's a made-up language.

Speaker of many Chinese topolects:  It's gibberish.  My wife and I together know all the major non-Mandarin dialects. We listened to the ads, but did not understand any. Who are Microsoft's consumers? Who does it target at?

Mandarin speaker:  The kids have invented their own language.

Mandarin speaker:  They do not sound like any Chinese language.

Shanghainese speaker:  I don't understand any of the words in these ads. They're not like any Wu languages that I know. I guess they speak Microsoftish.

Korean speaker:  I'm not sure but I don't think it is Korean. It's funny that MS declined to say what language these commercials are in!

American speaker of Mandarin and a bit of Shanghainese:  It does have hints of something Wu but I can't place it.

Speaker of Mandarin and Cantonese:  I am too busy to be bothered with those strange ads.

Fluent American speaker of Japanese who also knows a little Korean and Mandarin: First one is Chinese, second one is Korean, third one is very mumbly, probably Japanese based on the actors.

Korean speaker who also knows some Mandarin and Japanese:  I have no idea. They aren't speaking in Korean for sure, and it doesn't sound like Japanese either. I only can think that they are speaking in a Chinese dialect (including Taiwanese) or South Asian language.

Korean speaker who also knows some Mandarin and Japanese:  I don't recognize the language either. But it will be definitely one of Chinese dialects. Funny commercial though.

Fluent American speaker of Japanese who also knows a bit of Mandarin:  Sounds like make-believe Chinese for non speakers. In Italy there's a big tradition of comedians "speaking" various regional dialects by catching just enough of their sounds and rhythms to sound plausible. What was that poem? Mimsy are the borogoves….

Korean speaker who also knows some Mandarin and Japanese:  They surely are funny, and I don't know what language they are speaking either. ^^; Microsoft is being weird I guess.

Fluent American speaker of Japanese who also knows some Mandarin:  No idea whatsoever.

Speaker of Taiwanese and Mandarin:  I could not understand them, either. However, the tones and pronunciation indeed remind me of Korean language. I am not sure whether they are trying to imitate Koreans or not.

Japanese speaker: If you, an outstanding polyglot, can't tell what language they are speaking, how would I know!?????  Clever ads.  The boys in the 3rd ad are bad pianists but terrific pingpong players!

Fluent American speaker of Korean who also is very advanced in Japanese and Mandarin:  No kidding, Victor. Just crazy. A real mystery! Well, the language is definitely not Japanese or Korean, or Mandarin Chinese or any other variety of Chinese I've ever heard.  Still, here's my guess: Although it's not Chinese, in the first video there's a ma at the end of a clause, isn't there? And I hear yihou (with the wrong tones) introducing a clause in a set of instructions. That's why I think Microsoft just got some Chinese speakers to start making things up. What I'm saying is that I think it's made-up language, gibberish. Oh, and by the way, the styles of clothing, etc. look Chinesy, at least to me. They don't really look Japanese or Korean.

Speaker of Mandarin and a Hunan topolect:  I asked several friends about the three ads. We think the first one may be a Wu topolect 吴方言 (maybe Shanghainese 上海话), the second one like a southwest topolect 西南方言 (maybe Sichuanese 四川话). And we really don't know about the third one. However, we can not confirm with the guess.

Speaker of Mandarin and Hangzhounese:  All three ads speak Korean.

Speaker of Shanghainese and Mandarin:  Are they foreign students trying to speak Shanghainese?  I can't understand any of it.

Fluent American speaker of Mandarin who also know Shanginese and other Wu languages:  The ads are entertaining. But I think the language is fake. Though it seems intended to sound like some incomprehensible Chinese dialect. I guess that there is a remote chance that it is real. But I will be really surprised if someone identifies any of it as a real language. The ads were probably filmed in the U.S., maybe the west coast, all in the same room, which if you look closely has hazard signs all printed in English and no Chinese at all.

Speaker of Taiwanese and Mandarin:  I do not understand any of them. Quite creative!

Fluent American speaker of Mandarin who also is conversant with Min topolects:  There's only enough to listen to in the first one. I hear 人 and 一點 in what sound like Wu forms. I'm surprised, however, not to hear the ubiquitous Wu form of 不.

Speaker of Mandarin and several Wu topolects:  It is none of the Yangtze Delta dialects I understand. Check with someone from Fujian or native Tawanese?

Fluent American speaker of Mandarin who also knows a lot about many of the topolects:  I do not speak any Wu dialect, but it should be fairly easy to find someone who does and ask them about it. There are Shanghai speakers all over the place in the US now, and Philly must be full of them.  I can understand some words and phrases in it, which leads me to suspect that it could perhaps be some form of Mandarin. Southern Mandarin of the Yangtze watershed would be a possibility, it seems to me. Or even some sort of Central Plains Mandarin (中原官話).  My wife says she cannot understand a single word of it, which I frankly think is a gross exaggeration due to a psychological block of some kind. (Some people totally shut down when they hear a speech form that is even slightly unfamiliar to them.  My wife is one of those people. For example, she also says she can't understand a word of Chaozhou/Shantou, but even I can sometimes catch entire phrases of it when I hear it, and I am not a native speaker of Southern Min. I think she just doesn't want to understand these things.)

A couple of people from China suggested to me that the language might be that of Ruian (near Wenzhou) 瑞安的温州话, which has about 5,000,000 speakers.  But I only think they said that because the speech of Ruian is famous for being virtually impossible for outsiders to comprehend in the slightest, so much so that (along with other Wenzhou topolects) it has supposedly been used in wartime as a secret language (e.g., when the Chinese fought against the Vietnamese in 1979).

For references, see here, here, and here.

And, in this Wikipedia article, there is a section about the legendary incomprehensibility of Wenzhou topolects:

Due to its long history and the geographical features of the region on which it is located, Wenzhou Chinese is so eccentric in its phonology that it has the reputation of being the "least comprehensible dialect" for an average Mandarin speaker. It preserves some vocabulary from classical Chinese lost elsewhere, and has noticeable grammatical differences from Mandarin.

For those who wish to hear what Wenzhounese sounds like, here are some YouTube videos: 1, 2, 3, 4, 5.

One of my Chinese friends told me that someone online stated that the female speaker in the first video was from a place in Shanxi called Lishi 離石.  But I think that he said that only because that area — like Ruian in Wenzhou — has a reputation for being hard for outsiders to comprehend.

But what do the ads really say, and in what language are they spoken?  Here I concentrate only on the first ad because the other two ads have such a small amount of speech, and it is sotto voce.

The first expression spoken in the ad is one of the most ubiquitous utterances in Chinese:  zàijiàn 再見 ("goodbye").  It is strange to me that no Chinese speakers could recognize it.  Perhaps it is because it comes right at the beginning, is totally out of context, is spoken very quickly, and is not Modern Standard Mandarin.

A bit later comes yǐhòu 以後 ("after"), though with altered tones.  The same tonal alteration of yǐhòu 以後 ("after") is heard soon again and then once or twice more later on.

Other words that I hear are the following:

wúlùn 無論 ("regardless")

néng bùnéng wǎn yīdiǎn 能不能晚一點 ("can you [come] a bit later?")

zhèyàng ba 這樣吧 ("like this")

nǐ tīngjiàn 你聽見 ("did you hear?")

pò zhège zhōngguó màozi ("ruining this Chinese hat" [?])

nǐ gàn ma 你幹嘛 ("what are you doing?")

shénme dōngxi 什麼東西 ("what thing?")

yǒu zhème hǎo fǎzi 有這麼好法子 ("there's such a good way" [not sure of the next-to-last syllable])

tèbié 特別 ("special")

wǒ gěi nǐ nòng 我給你弄 ("I'll do it for you")

N.B.:  the order of what I've written down here is not necessarily that which is the actual sequence in the video, since I just quickly jotted down what I heard during several passes through the video. I probably can grasp twice again the amount of what I've given here, but this should afford an idea of the nature of the speech.  I'm fairly certain that it is basically some variety of Mandarin.

Of  course, what the voice is saying is totally unrelated to the actions in the video, so I suppose — in addition to the fast speed and muttered, altered quality of the speech —  that is a major factor in causing native speakers of Chinese to aver that they cannot understand anything that is being said.  Another problem is that at a number of points the pronunciation is electronically modulated in an unnatural way.

My question to Language Log readers is this:  why do foreign speakers of Chinese languages seem to pick up more of these ads than the native Chinese speakers for whom they were intended?

[Thanks to Mark Liberman, Ben Zimmer, W. South Coblin, Bob Ramsey, Richard VanNess Simmons, Sanping Chen, Cecilia Segawa Seigle, Xu Wenkan, Zhu Qingzhi, Minglang Zhou, Jidong Yang, Feng Shengli, David Spafford, Frank Chance, Kellen Parker, Nathan Hopson, Grace Wu, Haewon Cho, Gianni Wan, Cheng Fangyi, Rebecca Fu, Daniel Sou, Yunu Song, Sophie Wei, Chin Yi Young, Frank Lin, Summer Hu and her parents, and Stefan Krasowski]


  1. Carl said,

    May 14, 2013 @ 11:35 pm

    I used to work in two schools in Japan, a school for the hearing impaired and a regular high school. I showed my mother a video of the English students from one of them and she asked, is this from the deaf school? It was from the regular high school, but the pronunciations seemed so distorted it was hard for her to understand.

    I think native speakers have trouble understanding non-natives unless they have a lot of prior exposure to their style of mispronunciation. On the other hand, non-natives don't have that trouble.

    One more anecdote: I took a Japanese language class while in Japan. I noticed that the Swedish student spoke Japanese with a Swedish accent, the Pole with a Polish accent, etc. with one curious exception: the American students all sounded like they had great Japanese accents to me!

  2. JS said,

    May 15, 2013 @ 12:38 am

    Try watching "Beautiful and Fast" without sound, preferably at a slowish playback speed. It seems the actress was indeed reading from a meaningful script and in a Mandarin dialect. Many people are bound to be better lip-readers than me, though; during the tight shot all I can guess at is … 眼睛 … 不可以 …, for example. The manipulated audio track applied above all this, though, doesn't at all match what's detectable from lip-reading… no idea if the track might contain any of the phrases you suggest, represents some manipulation of the original audio (at points this seems possible), or what…

  3. michael farris said,

    May 15, 2013 @ 1:25 am

    Possible reasons might be that any kind of distortion will render Mandarin more unintelligible more quickly than most languages.

    Related to a commenter on another thread, I read once that the number of potenial but unused syllables in Mandarin is extremely small. Combined with the general tendency for syllable-morpheme equivalence the result is that almost any syllable possible within the phonological system is used with some meaning (so hearers can't rule out any possible syllables). This was cited as one reason that mutual intelligibity breaks down so quickly in Mandarin (more so than other Chinese languages).

    A colleague who is fluent in (Taiwanese) Mandarin related that he had occasion to listen to a conversation in a variety of Wu(?) whose phonetics were really close to Mandarin and it was unsettling because it sounded like Mandarin words put together in bizarre nonesense ways.
    That is he didn't hear anything like : blah jahd do elw wioe (a string of sounds he couldn't assign any meanng to)
    or: The blivvish agmoration on his rettlesome plockthurp (a string of pseudo words in a broadly recognizable grammatical framework)
    but more like : bicycle top sing, spleen? they parrot underside!

  4. Kevin Wong said,

    May 15, 2013 @ 1:53 am

    Maybe the quickest way to solve this problem is for some person or organisation of note to accuse Microsoft of casual racism (e.g. "It's basically equivalent to running up to a Chinese person and yelling 'ching chong ching chong!', but a lot more sophisticated"). Microsoft will then be compelled to address the issue, for PR reasons, and presumably reveal the truth in doing so.

    I'm really only joking in suggesting this, but it _could_ work, right?

  5. Peter Taylor said,

    May 15, 2013 @ 2:05 am

    Are your instructions conditioning people to expect an unusual topolect fluently spoken? If you instead say, "Please transcribe this voice-over by a non-native speaker," would the native speakers understand more?

  6. Terence said,

    May 15, 2013 @ 2:51 am

    Are you sure the first thing she says is 再見? Doesn't really fit in the context. To me sounds like 姐姐 which would make more sense. I am sure I hear sionn tiam laa ("I am too tired" in Minnan) at 3 seconds. I think maybe they used some software to combine a number of Chinese topolect readings.

  7. Sven Sahle said,

    May 15, 2013 @ 4:27 am

    Is it possible that the videos were produced in a language neutral version for later adaption to different asian languages? Perhaps someone found it then funny or cool to release them as they are.

  8. AntC said,

    May 15, 2013 @ 5:15 am

    @Victor, please explain A bit later comes yǐhòu 以後 ("after"), though with altered tones.
    In a tonal language, if you alter the tones, don't you end up with a different 'word'? Or something that isn't a word?
    Do you mean that there are no such tones in the putative language (like speaking Mandarin with Cantonese tones)?
    Do you mean that yǐhòu is the closest in sound to what's said?
    I suppose if someone said "ufti" I might describe that as "after" with altered vowels(?)

  9. Brendan said,

    May 15, 2013 @ 5:38 am

    I'm not sure about this starting out with "再见" — it could be a quick, muffled "在这里," which would make sense if the audio actually is intended to be related to what's happening onscreen. (Though it looks to me as if the audio has been dubbed in.) I think I can hear a "不懂事" in there as well around 0:22, and at 0:48 a couple of words sound as if they've been reversed. At 0:29, there's something that might be "…不知道怎么上," though I don't think I've ever heard 上 used for makeup.

    For what it's worth, this sounds more like one of the Wu topolects to me than anything else, but only in that it's weird and buzzy and incomprehensible. My wife (native speaker of Beijing Mandarin) didn't have any guess more specific than "something Southern, probably," which is vague enough not to really count at all.

    As for why non-native speakers of Chinese are more willing to venture guesses: I'd guess that it's just because we're more used to having to strain than native speakers are, and more used to reassembling sentences based on partially understood input. Sometimes this doesn't do us any favors — Chinese grammar is so poorly taught that plenty of students often end up scanning sentences for words that they know and then making up stories involving those words. Hard to know whether or not that's what happening here, in the absence of any script or clear ID of the language being spoken.

  10. KK said,

    May 15, 2013 @ 6:06 am

    In response to your question at the end of the post: my hypothesis is that this native/non-native speaker effect arises because the acoustic stream in these adverts is, in fact, more noise than signal.
    Phonological processing is an interaction between the language-specific templates in memory which ongoingly set up a prediction, and incoming acoustic stream. If the incoming stream does not match the phonological prediction, an error signal ensues. Considering humans as acoustic signal processing units, non-native speakers have larger "error tolerance" than native speakers because their language-specific templates are more fuzzy. Thus, non-native speakers are more likely to continue parsing attempts in situations where native speakers have already abandoned processing (as the accumulative error is too large) and consider the stream as pure noise. This would explain why non-native speakers appear as better signal detectors in such case.

    Incidentally, I believe that the adverts are quite likely intended not for Asian market, but for viral spreading in western social networks, intended to capture interest due to "exotic wackyness" or puzzlement about the language.

  11. richardelguru said,

    May 15, 2013 @ 6:29 am

    Kevin Wong is definitely right!!

  12. spherical said,

    May 15, 2013 @ 6:32 am

    @Kevin Wong: Bad boy. :)

  13. jfruh said,

    May 15, 2013 @ 6:46 am

    Another odd and perhaps relevant aspect of the ad: it seems pretty clear to me that the speaker isn't actually speaking the pseudo-language. Her mouth movements don't really match up and something about the quality of the sound is off. So presumably she's speaking her actual, real native language, and Microsoft dubbed in this fake language later?

  14. Victor Mair said,

    May 15, 2013 @ 7:17 am


    You are certainly right about that. There is no attempt at lip-synching. The actress does makes no effort to match her words to what is being said by the voice-over. Moreover, the content of the voice-over is not only disconnected, it is intentionally garbled in places and completely unrelated to the actions in the video.


    Brilliant analysis! Just brilliant! Thanks!


    If you are a speaker of MSM (Modern Standard Mandarin) and you travel across the Mandarin-speaking areas of China, you will be astonished at the enormous variety of tonal configurations for words. My wife was from Shandong and grew up in Sichuan, so I got a liberal dose of the tonal patterns of both areas. I became so conditioned to them that, if I were upstairs and someone called who spoke Mandarin with Sichuan or Shandong tones, I could tell very quickly to which my wife downstairs was speaking, since — even though she spoke beautiful MSM — she would switch into the very different tone patterns of her interlocutor. I often said that Sichuanese tones were "upside down" in relation to MSM (though they are consistent within their own phonological system).

    From Bob Ramsey:

    I can't believe I didn't notice the zaijian in the beginning, but yeah, there were plenty of garbled and out-of-place Mandarin words there for those willing to hear them. The real puzzle is just what you point out: why did fluent Mandarin speakers not hear any of these words and phrases? (XXXX and his wife [VHM: cited in my original post as knowing all of the main non-Mandarin topolects] told me they were pretty sure it was Korean! It sounds like "Korean tones", he said.) This little sampling of yours actually says a lot about cultural differences and how understanding of language takes place.

  15. Victor Mair said,

    May 15, 2013 @ 7:30 am

    @Peter Taylor

    I just asked my correspondents what language they thought was being spoken in the videos — totally neutral question. Some of the responses were viscerally dismissive and almost angry that they couldn't make any sense of what was being said. This reminds me a bit of the reaction of viewers of the artist Xu Bing's made-up characters. Westerners who knew some Chinese tended to try to make some sense of possible meanings that might be derived from the assembled elements of the characters, whereas character-literate Chinese tended to get annoyed or upset that they couldn't understand Xu Bing's texts made up entirely of concocted characters. Westerners who knew no Chinese at all just admired the beauty of the woodblock-printed pages. It's as though Xu Bing and Microsoft were "messing with people's minds".

    I forgot to mention Minkyung Ji in the acknowledgements.

  16. JonHughes said,

    May 15, 2013 @ 8:24 am

    I'm a native English speaker who speaks some Japanese. I don't know any Chinese or Korean.

    This is just a thought. Perhaps they're doing something like what is happening in this video ( to make it sound Chinese-y without being understandable to Chinese speakers. Would that then be easier for non-native speakers to pull out individual words while remaining unintelligible to native speakers?

  17. Victor Mair said,

    May 15, 2013 @ 8:37 am

    From the director of a major university Chinese language program:

    Incredible. I played it several times and thought it sounds like Korean, Taishan, or a minority language.

  18. Mark Liberman said,

    May 15, 2013 @ 8:48 am

    All of the comments make it sound like this is a skillful instance of doubletalk. An English version with a few common words stuck in is discussed here; more here.

  19. Theophylact said,

    May 15, 2013 @ 9:06 am

    Perhaps Microsoft was trying to suggest a "Panasian" language?

  20. Nathan said,

    May 15, 2013 @ 9:13 am

    Yup, just Asian-style yaourter.

  21. julie lee said,

    May 15, 2013 @ 10:55 am

    Charlie Chaplin, the great Hollywood mimic and comedian, could speak gibberish to make it sound like any language (e.g., italian-sounding, French-sounding, Chinese-sounding, Japanese-sounding, gibberish) and the effect was hilarious.

  22. Ben Zimmer said,

    May 15, 2013 @ 1:46 pm

    I like Sven Sahle's theory that the ads were "produced in a language neutral version for later adaption to different asian languages." Supporting the idea that the ads were not initially intended for public release is the discussion on after the tech blogs noted the ads on Microsoft's YouTube channel, the videos were set to private, but then made public again. That suggests the original release was in error, but then someone at Microsoft realized the potential for getting people intrigued by the "mysterious" pseudo-language of the ads.

  23. JS said,

    May 15, 2013 @ 2:08 pm

    ^ except that "Beautiful and Fast" is pseudo-Chinese, while the other two ads are pseudo-Korean-and-or-Japanese (maybe someone can make a distinction here; I'm fortunate if I'm able to tell apart the genuine versions…)

  24. Flex said,

    May 15, 2013 @ 2:09 pm

    A bit off-topic, but many of the responses reminded me of the interviews from from the crowd in Poe's Murders in the Rue Morgue.

    A wide varieties of nationalities were in the crowd, and each person thought that the orangutan was speaking a European language that the individual was unfamiliar with.

  25. HP said,

    May 15, 2013 @ 2:14 pm

    I don't know if it was intentional or not, but the way Prof. Mair structured this post reminds me of Poe's Murders in the Rue Morgue, e.g.,

    Alfonzo Garcio, undertaker, deposes that he resides in the Rue Morgue. Is a native of Spain. … Heard the voices in contention. The gruff voice was that of a Frenchman. Could not distinguish what was said. The shrill voice was that of an Englishman—is sure of this. Does not understand the English language, but judges by the intonation.

    Alberto Montani, confectioner, deposes that he was among the first to ascend the stairs. Heard the voices in question. The gruff voice was that of a Frenchman. Distinguished several words. The speaker appeared to be expostulating. Could not make out the words of the shrill voice. Spoke quick and unevenly. Thinks it the voice of a Russian. Corroborates the general testimony. Is an Italian. Never conversed with a native of Russia.

    I can only conclude that Windows 8 must be an orangutan.

    But seriously, I suspect that this is a Sinitic-flavored version of Simlish. It's meant to suggest speech and emotion, while getting around the need for costly localization.

  26. Victor Mair said,

    May 15, 2013 @ 2:42 pm

    Amazing that Flex and HP both noticed the resemblance to Poe's "Murders in the Rue Morgue" within five minutes of each other.

  27. Victor Mair said,

    May 15, 2013 @ 2:57 pm

    Even after I called this post and its comments to the attention of many Chinese friends who speak Taiwanese, Cantonese, Shanghainese, Hangzhounese, etc., as well as Mandarin (all of them), they still say they cannot understand anything that is spoken in the videos and that they have no idea what language is being spoken — despite the fact that there is an awful lot of Mandarin in there (albeit somewhat mangled, altered, and slurred or mumbled, as well as interspersed at times with nonsense syllables, and spoken very rapidly).

  28. Avinor said,

    May 15, 2013 @ 4:42 pm

    The discussion of language versions of reminds me of this Swedish commercial from about a decade ago:

    "Mr Yamamoto", samurai by profession and a single father, slices things with his sword all day. This is exhausting, so when making dinner at home, he enjoys using pre-sliced frozen meat from Scan.

    The only confusing thing is that "Mr Yamamoto" and family actually speak MSM.

    After complaints, they actually re-dubbed it into Japanese.

  29. Theodore said,

    May 15, 2013 @ 5:15 pm

    I do not speak or understand any dialect of Chinese, Japanese or Korean, but regarding the reaction of native speakers, I wonder what the results would be if they heard audio only. Since the videos appear to have dubbed voices, are the unsynchronized lip movements disrupting natives' comprehension?

  30. Bob said,

    May 15, 2013 @ 5:41 pm

    I think I hear the music "from Wenzhou with love" in the opening (defining) notes… or, it can be "you live only twice, one for your dream and one for your codes….." a certain/non-certain financeer has used microsofts to issue a dump all your gold order…….
    btw, it was not Cantonese, Taishanese, non any Ghangdong dialects/topolects, based; another attempt from the centralists to stump down Ghangzhouese for sure….

  31. Victor Mair said,

    May 15, 2013 @ 6:25 pm

    From a highly educated Chinese acquaintance who speaks MSM and Hangzhounese:

    We were trying to find out what really happened behind those visual ads. One person's educated guess is that MicroSoft has flubbed this project by assigning it to a regional contractor in Asia, and that whoever was charged with this task did not know Mandarin, Korean, or Japanese but claimed to know in order to get the assignment. What seems to have happened is that this dude found some local individuals who are either illiterate or undereducated (therefore cheap to recruit for oral enactment) who only spoke highly isolated topolects (due to the lack of formal training in standardized versions of their mother tongue). Personally, I knew similar cases in the US, when I listened to some audio tapes of "Mandarin" which turns out to be its substandard version bearing heavy regional accents.

  32. William Steed said,

    May 15, 2013 @ 6:35 pm

    It certainly sounds Southern Wu to me, but not exactly Wenzhou. It sounds something like the Qingtian (inland from Wenzhou) that I've heard. It has the distinction of being the language of lots of European Chinese ex-pats. I wonder if any of them would recognise it. That said, given the rise of Putonghua, I'm sure there's lots of code mixing, so hearing Mandarin in there isn't remotely surprising.

  33. Jonathan Dushoff said,

    May 15, 2013 @ 9:55 pm

    The original question: "why do foreign speakers of Chinese languages seem to pick up more of these ads than the native Chinese speakers for whom they were intended?" touches on what I think is an interesting phenomenon.

    As a reasonably functional, but not strong, speaker of Mandarin, I hear pieces of Chinese _all the time_, especially when I'm not really paying full attention. I think that's just how the brain works. My brain subconsciously hears oral communication, recognizes right away that it's not English, and tries for a while to make it Chinese — whether it is or not. Probably the Chinese speaker sitting next to me in the bus is subconsciously trying to make the French into English, while I'm trying to make it into Chinese.

    In the clip, I hear lots of Mandarin, and a bit of Minnan — but I don't believe any of it. I think the reason "we" hear it is that our brains are programmed to grope for it; and the reason "they" don't hear it is that it's not there. For example, the zai jian at the beginning: I picked up on it right away, but I don't know anyone who says it like that, and I don't know whether most native speakers would even make the connection.

  34. Victor Mair said,

    May 16, 2013 @ 6:50 am

    @Jonathan Dushoff

    "As a reasonably functional, but not strong, speaker of Mandarin, I hear pieces of Chinese _all the time_, especially when I'm not really paying full attention."

    Well, I am a fully functional, "strong" speaker of Mandarin, and I listened to the video very, very carefully dozens of times, playing it at various speeds. The Mandarin really is there.

    You've missed the whole point of the excellent discussion in the comments to this post. I think you need to go back and read what your fellow Language Log readers have been saying, and think about it "carefully".

  35. Faldone said,

    May 16, 2013 @ 11:32 am

    Not much language in it but I like this one:

  36. Guy said,

    May 16, 2013 @ 11:59 am

    From what I've seen on the Chinese blogosphere, the ads were produced by advertising agency JWT's Beijing office. Its supposedly a made-up language based on a mix of pseudo-Shanghainese, pseudo-Wenzhou dialect and general gibberish.

    It's interesting that several native Chinese respondents guessed correctly that it was invented or gibberish. Even throwing in a few vaguely Chinese words or phrases did not make it sound "real" to them.

  37. Jonathan Dushoff said,

    May 16, 2013 @ 12:25 pm

    @VictorMair: I have read the comments. Instead of asking me to read more carefully, could you briefly say what you think "the whole point" that I've missed is?

    The consensus seems to be that the majority of native Mandarin speakers don't hear these snippets as Mandarin. Why are you convinced that they are?

  38. Jonathan Dushoff said,

    May 16, 2013 @ 12:30 pm

    @VictorMair: to be more specific. If you heard a similar dialogue in Wenzhou or Shanghainese (without knowing what it was, and assuming you are not familiar with those languages), how sure are you that you wouldn't find real Mandarin in the dialogue?

  39. Jamie said,

    May 16, 2013 @ 3:23 pm

    To take Flex's comment a step further (and, I guess, agree with Jonathan Dushoff) how do you know that the non-native speakers are recognising things that are really there? It could just be an example of pareidolia similar to the "electronic voice phenomenon". (Although this hypothesis is rather undermined by Guy's comment…)

  40. Jongseong said,

    May 16, 2013 @ 4:31 pm

    I was going to comment on the resemblance to "The Murders in the Rue Morgue", but Flex and HP beat me to it. But if we're alluding to Poe's story, we should at least use his spelling, "Ourang-Outang".

    As a native Korean speaker, it amuses me to no end that a number of Chinese speakers are guessing that this sounds Korean, especially these couple of comments—"the tones and pronunciation indeed remind me of Korean language" and "It sounds like 'Korean tones'"—since the Standard dialect of Korean is not tonal at all. The Southeastern dialect of Korean is tonal (well, it uses a pitch accent), but how many of these Chinese speakers would have had much exposure to this dialect?

    But maybe lots of Korean speakers do use, shall we say, expressive intonations that make us sound like we're speaking a tonal language. I have very little intuition of what my native language would sound like to those who don't understand it.

    The language in the video sounds nothing like Korean to me, of course. I would merely have guessed that it was some Sinitic variety (yep, to me, the tones sound Chinese) and not suspected it was made up. But the other two videos did sound a lot like mumbled gibberish. Clearly some people are better at making gibberish sound like real language than others.

  41. KWillets said,

    May 16, 2013 @ 5:26 pm

    My Korean must be pretty good, because I could barely hear any resemblance to Korean in the second video. But I reach the same conclusion with my father-in-law (Taegu dialect).

  42. Jonathan Dushoff said,

    May 16, 2013 @ 7:42 pm

    This is funny. I was just watching with a native Mandarin/Minnan speaker, and trying to get her to focus on the stuff that I pick out subconsciously as Mandarin. I picked out "a little later (wan yidian)" (I was not surprised to see this was one of the things flagged by the original poster). She said she heard that phrase as "what an idiot".

  43. Nathan Myers said,

    May 16, 2013 @ 8:13 pm

    If you took regular, MSM Chinese speech and ran it through a program that first removed all tonal inflections, and then applied random inflections in their place, properly registered to syllable boundaries, would it sound just like this?

  44. JS said,

    May 16, 2013 @ 8:58 pm

    ^ No; that would produce Shandongnese.

    Kidding… but what you describe would retain a very high degree of comprehensibility — especially to us foreigners!

  45. Victor Mair said,

    May 16, 2013 @ 9:32 pm

    @Jonathan Dushoff

    "'what an idiot'"

    I don't think anyone, even your Mandarin / Minnan speaker friend, really believes that they're speaking English in that first video. She was just joshing you.

    And what do you mean by "pick out subconsciously as Mandarin"? How subconscious can it be if you are able to pick it out as Mandarin?

  46. Victor Mair said,

    May 17, 2013 @ 8:28 am

    From a Chinese colleague who is a specialist on Chinese cinema:


    Thank you for your great deciphering! Microsoft certainly got what they intended with these ads: attention, even if only a few seconds. Interestingly, it used to be a texts versus images competition, now, it's an acoustic versus visual one. And the appeal of these ads is the dissonance between audio and visual, instead of accordance.

  47. Jonathan Dushoff said,

    May 17, 2013 @ 8:41 am


    My interpretation is that "what an idiot" is what her language center picked out; whether she believes it or not is a separate question. I think it's likely that some non-English speakers would conclude that there are "real" English snippets in that dialogue, but that's a separate point.

    I'm not "joshing" when I say I pick out "zai jian", "yihou" and "wan yidiar" – but picking them out doesn't mean I think that's what they're really saying.

    Not sure how to clear up the subconscious thing: what I mean is there's a part of my brain that interprets those sounds that way – just like there's a part of my friend's brain that hears "what an idiot".

    @JS @NathanMyers That would be a fun experiment. I agree it would be a lot more comprehensible to me. But I really wonder how comprehensible it would be to native speakers – not only are the tones deeply imprinted, but a lot of the vowel and consonant sounds are highly variable.

  48. Jonathan Dushoff said,

    May 17, 2013 @ 11:03 am

    @VictorMair. I now think that when you said "the whole point", you were referring to KK's excellent post, which indeed I should have read more carefully.

  49. chh said,

    May 17, 2013 @ 12:27 pm


    A number of people have reported on a relatively recent change in Seoul Korean from a primarily VOT-based contrast for aspirated-lenis-fortis voiceless stops to a contrast primarily in f0. This looks like the emergence of a tone system, at least a system where pitch is already a major cue to a contrast. If you aren't consciously aware of this you might find the data pretty cool to look at.

    I agree that the recording sounds nothing like Korean :)

    [(myl) For an alternative take on the role of F0 in Seoul Korean, see Jonathan Wright's 2007 dissertation "Laryngeal contrast in Seoul Korean", especially Chapter 8 on Tonogenesis, which begins

    There are two tonal phenomena in Korean: the assignment of phrasal tones, and segmentinduced tone. The change in progress in Seoul Korean suggests that segment-induced tone could be an early stage of tonogenesis. In this chapter I walk through several possibilities for tonogensis in Korean, before concluding that modern Seoul Korean has had lexical tone since the time of Middle Korean.


  50. Stephen Zweig said,

    May 17, 2013 @ 6:29 pm

    I'm not actually sure but I think I heard some variety of Chinese mixed with English, e.g. "eyeline 线"

  51. Jongseong said,

    May 17, 2013 @ 7:34 pm

    chh & myl: Thanks for the reading suggestions. They look really interesting.

    As a younger speaker of the Seoul dialect myself, I agree that VOT is not the sole basis of contrast. The difference between the aspirated series and the lenis series, which is also lightly aspirated, is not likely to be just about degrees of aspiration. Relying only on introspection, I wouldn't be able to say much more than that the aspirated series is "strong" and the lenis series is "weak". But I've taken enough occasional glimpses at the literature to know that F0 plays a role in the contrast. I've seen this applied to an examination of how stops and affricates in Chinese loanwords are adopted by Korean speakers in China—the tone in the original language has an influence on what series it is mapped to.

    But I think it's possible to read too much into this. F0 cues are only auxiliary and would be meaningless by themselves—it is trivial to produce each series with the "wrong" pitch, for example. And any "tonal" effects coming from this would be extremely minor compared to the "phrasal tones" used for expressive purposes. I can't see any learner listening for "tones" to distinguish sounds in Korean doing much better than chance.

    I'll have to read Jonathan Wright's paper, but I'm very skeptical if the conclusion is that modern Seoul Korean has always retained lexical tone. To my knowledge, the lexical tone of Middle Korean was preserved as lexical length, lexical tone, or a combination of the two in the different dialects, with pretty regular correspondences between dialects (e.g. high tone in one dialect corresponds to long vowel in another). Standard Korean itself is theoretically supposed to have a lexical length distinction, with a long initial syllable corresponding to one of the three tones in Middle Korean and a short initial syllable corresponding to the other two.

    But I can attest that any lexical length distinction is completely, utterly lost in my generation. This has led to a collapse in distinctions that would have been meaningful in the previous system. Korean orthography doesn't mark vowel lengths, so you have to know which words are pronounced with long initial syllables. My peers would do no better than chance if we were tested on this unless we memorized those words as you would with a foreign language.

    To summarize, the lexical tone contrast of Middle Korean evolved into the lexical length contrast of modern Standard Korean, which has disappeared in the present generation. But the "lexical tone" being claimed must refer only to the tonal effects due to the 3-way contrast of the stops and affricates. In other words, all that the "lexical tone" is distinguishing must be between the lenis series, the fortis aspirated series, and the fortis unaspirated series—a contrast that is fully encoded in the orthography and fully preserved in all dialects I'm aware of. So I don't get what the connection is supposed to be between the historical lexical tone of Middle Korean and what seems to be emerging in Seoul Korean today. If anything, the generational and regional differences seem to point to the latter being a recent innovation.

  52. JS said,

    May 17, 2013 @ 8:34 pm

    Kim and Duanmu (2004) reanalyze the three-way voiceless stop contrast of traditional treatments of Korean (aspirated/tense/lax) as underlyingly aspirated/voiceless/voiced, with the third set devoicing and triggering low vowel tone when in initial position… don't know if this was the first suggestion of emergent "tonogenesis" in Korean, but the notion of an underlyingly voiced segment would bring it into line with what is known of earlier developments elsewhere in E/SE Asia, as the authors points out.

  53. Nathan Myers said,

    May 17, 2013 @ 10:15 pm

    One reason I asked about programmed transformation is that it seems to me it would be very difficult to utter nonsense syllables at length, fluently. Some people can do it, but if you listen to people "speaking in tongues" at church services (right before snake-handling, I think), even those who must have had years of practice can't do it convincingly. In particular, you hear a lot of excess repetition.

  54. Jongseong said,

    May 18, 2013 @ 6:10 am

    JS: Kim and Duanmu (2004) reanalyze the three-way voiceless stop contrast of traditional treatments of Korean (aspirated/tense/lax) as underlyingly aspirated/voiceless/voiced, with the third set devoicing and triggering low vowel tone when in initial position

    I really like this analysis. I've never really been convinced by the traditional treatment which makes Korean look highly unusual, and in particular makes the tense unaspirated series something unique to Korean when I can't hear much difference between it and the typical voiceless unaspirated series you find in many languages.

    It's much easier to explain as a typical aspirated/unaspirated/voiced contrast with interesting allophonic behaviour in initial position, namely the devoicing of the voiced series. This nevertheless maintains a contrast with the other two, and in modern Seoul Korean low F0 looks like it plays a large part in the perception of this contrast. I'm confident though that it hasn't progressed to the point that it is the only cue—voice quality and VOT continue to play roles, and I suspect the contrast doesn't depend on a single numerical measure that you can pluck from the spectrogram.

    However, my point in the previous post was that if modern Seoul Korean does end up developing a tonal system based on this F0 contrast, this will have nothing to do with the tonal system of Middle Korean, which after evolving to a length distinction system in modern Standard Korean, has subsequently been irretrievably lost for all but the oldest speakers.

  55. catx said,

    May 19, 2013 @ 8:00 am

    This reminds me of a tape of 'Golden Oldies' that was produced in Taiwan — the singers were very clearly singing english according to some phonetic pronunciation guide, rendering it generally recognizable as 'English', but completely impossible to understand.

  56. McJohn said,

    May 19, 2013 @ 7:01 pm

    Is it maybe that the producers took a recording of speech in Mandarin, played it backwards, and had their voiceover artists learn it phonetically? That's been done VERY occasionally in English-language pop, and it always sounds spooky as hell–just enough familiarity to drive you nuts trying to figure it out.

  57. Kristi Jarvis said,

    May 19, 2013 @ 8:24 pm

    As a monolingual English speaker, I have no idea what the language is being spoken. However, the meaning of the commercials is still very clear, and humorous.

    I'd consider a deliberately invented "almost" gibberish language a clever play on the universally accessible physical comedy with a universally exclusionary "non-language". However, I'm not sure I'd credit Microsoft marketing with such wit.

  58. Daniel C said,

    July 15, 2014 @ 2:14 pm

    It's a definitely a variety of Jin Chinese, and not made up. That myth is actually quite stupid.
    In the following video, the Fenyang Jin dialect is being spoken. It sounds remarkably similar to the woman in the video:

  59. David Marjanović said,

    August 27, 2014 @ 5:52 am

    Charlie Chaplin, the great Hollywood mimic and comedian, could speak gibberish to make it sound like any language (e.g., italian-sounding, French-sounding, Chinese-sounding, Japanese-sounding, gibberish) and the effect was hilarious.

    He does that with (one very particular kind of) German in The Great Dictator.

RSS feed for comments on this post