When intonation overrides tone, part 2

« previous post | next post »

Richard Warmington has a deep interest in the relationship between tone and intonation, especially in Mandarin.  He has made a number of penetrating observations and asked a series of probing questions on this phenomenon.  Since this is also a subject that has come up numerous times on Language Log (see below for a several previous posts), I will list here a few of Richard's remarks about tones and intonation, with an eye toward encouraging further discussion.

To get started, listen to the audio for the first six sentences on this page.

To play the audio, you click the red "play" button (which confusingly has a "Download" link underneath it that looks like a label for the "play" button; you do not need to download the file).

The question we're interested in is whether the second syllable of péngyǒu 朋友 ("friend") is always pronounced in the third tone (or sometimes in neutral tone), as it is marked in dictionaries.  The problem is that, in more than half of the ChinesePod recordings, it comes out as fourth tone, which seems to be rather shocking and dismaying for someone who tries their hardest to pronounce the tones correctly.  Third tone, the prescribed tone for yǒu 友, is low and dipping (at the bottom of a speaker's register).  It is understandable how, through lack of stress or emphasis, the third tone might turn into a neutral tone, but the fourth tone is a high, falling tone (starts at the top of a speaker's register), so something else must be going on here that has caused the speaker to pronounce so many of the third (or neutral) tone yǒu 友 syllables as yòu.

The first set of video clips has péngyou 朋友 ("friend") in the context of xiǎopéngyou 小朋友 ("little friend").

There are other audio clips, such as those for lǎo péngyou 老朋友 ("old friend"), in which you can hear the same sort of thing.

Those who favor péngyou over péngyǒu for 朋友 might argue that, for the sake of consistency, 老朋友 should be lǎo péngyou rather than lǎo péngyǒu.  But it turns out that, as evident from the just cited ChinesePod recordings, neither lǎo péngyou nor lǎo péngyǒu is necessarily obligatory, because the speaker actually says lǎo péngyòu 老朋友 rather than lǎo péngyǒu or lǎo péngyou in many of the examples.

In any event, it would seem that "consistency" with the pronunciation of another word (péngyǒu 朋友 ["friend"]) is not a sufficient reason to say that 老朋友 has a neutral tone on the last syllable. Rather, one would need to check how people actually pronounce 老朋友 in isolation and in various circumstances. (After all, to take an example from English, the pronunciation of "continent" is different when it's part of "continental", since the stress is on a different syllable.)

In fact, 友 in both 朋友 and 老朋友 is pronounced sometimes with 3rd tone and sometimes with neutral tone, depending on the speaker, locale, and other factors (and even, as in the audio clips, with something like a 4th tone).

I sent the ChinesePod recordings to a number of native speakers of MSM, and they had to admit that, although they previously never thought that it existed, the speaker really did use the 4th tone in many of the examples.  They didn't have a good explanation for why she did so, but one of them averred:

The fourth tone only appears when it is at the end of a question where this term "朋友/老朋友" being emphasized.

But that's not true, since the 4th tone version also occurs in the middle of sentences that are not questions.

One of the respondents gave this rather convoluted explanation:

It’s normally pronounced in neutral tone, especially when spoken by 北京人 [VHM:  Beijingers], as shown in those recordings. Neutral tone generally is an unstressed falling tone, thus sounds more like a 4th tone. However, a 4th tone must be stressed.

But this begs the question of how or why the speaker was stressing the instances when she saying the 友 syllable in the fourth tone.  She clearly says yòu (4th tone) in many cases, yet there is no obvious reason why she would be emphasizing that syllable in all of those cases.

Finally, we have these very helpful observations from a senior lecturer In Mandarin who is from Taiwan:

The 友 in 朋友 is always pronounced and marked as third tone in Taiwan. But in Mainland China, it is pronounced and marked as neutral tone. However, even in Taiwan, since it's usually pronounced as half third tone (2-1) instead of full third tone (2-1-4), it actually sounds very similar to a neutral tone. Still, In the textbooks published in Taiwan, it's always marked as third tone.

The girls in these recordings apparently have northern accent and all speak quite dramatically. I think in a few cases they overly exaggerated the word 朋友 which is why the 友 sounds a little like a fourth tone. As far as I know, 友 is either pronounced as 3rd or neutral, but never as a 4th tone.

Richard proceeds to even more difficult aspects of tonal transformations when he observes:

1) I'd be interested to learn whether a "full" tone is any less susceptible to being overridden than a neutral tone.

2) I believe I have heard spoken Mandarin words in which a syllable starts off with its proper tone, but ends up being overridden by intonation, probably being prolonged as a result.

A few earlier posts:

"When intonation overrides tone" (6/4/13)
"'Ni hao' for foreigners" (10/11/16)
"Dissimilation, stress, sandhi, and other tonal variations in Mandarin" (8/26/2014)
"Where did Chinese tones come from and where are they going? " (6/25/13)

There are many other posts on the origins of tones, their canonical forms, departures from the canonical forms, tones in the various topolects, and the evolution of tones.

[Thanks to Jinyi Cai, Fangyi Cheng, Maiheng Dietrich, and Melvin Lee]


  1. Jenny Chu said,

    May 11, 2017 @ 8:57 pm

    One thing I have been wondering: are there variations, depending on language, that impact when intonation wins vs. when tone wins? I always found Vietnamese tones to be "stronger" – in other words, you don't change them as much based on intonation. Hence, they are also more distinct / easier to hear than Cantonese tones … Cantonese tones seem to be more easily able to be adjusted on the fly, depending on the sentence. Is my impression borne out by any proper study? Or am I fabulating?

  2. Lai Ka Yau said,

    May 12, 2017 @ 12:03 am

    I suspect it may have something to do with tone spreading as well. If one listens to the 小朋友也在看我! recording, it sounds like that the high pitch of the 朋 was spread rightward to the beginning of the 友 to produce something like the forth tone…

  3. Bob Ladd said,

    May 12, 2017 @ 12:29 am

    It seems to me that the question of whether the tone on the second syllable of péngyǒu is "really" tone 3 or tone 4 is completely analogous to the question of whether the nasal sound at the end of ten in a fast informal rendition of ten past two is "really" /n/ or /m/. The issue here is not the interaction of tone and intonation, but rather the relation between phonological abstractions and actual connected speech. The example in the post shows that, just as in the segmental domain, so also in the tonal domain it is difficult to describe phonetic detail in terms of a string of categorically distinguishable units.

    The resistance of native speakers (and dictionaries, and so forth) to the idea that we are dealing with tone 4 here is related to the difficulty of convincing beginning phonetics students to transcribe ten past two with /m/.

  4. Will M said,

    May 12, 2017 @ 2:01 am

    More examples:
    猴子 "hóuzi" http://chinesepod.com/tools/glossary/entry/%E7%8C%B4%E5%AD%90
    闲着 "xián zhe" http://chinesepod.com/tools/glossary/entry/%E9%97%B2%E7%9D%80
    完了 "wán le" http://chinesepod.com/tools/glossary/entry/%E5%AE%8C%E4%BA%86
    (compare: 玩乐 "wánlè" http://chinesepod.com/tools/glossary/entry/%E7%8E%A9%E4%B9%90 )

    For what it's worth, I've been corrected more than once by (Northern mainland) native speakers on my pronunciation of the neutral tone: it seems that after second tone, the neutral tone always falls slightly. Without this fall, native speakers would insist I was using first instead of neutral, no matter how light and "neutral" I tried to make it. Does anyone know of an example that contrasts purely by intonation? A tone sandhi rule 5 → 4 / 2_ seems to explain this adequately, from my experience.

  5. Philip Taylor said,

    May 12, 2017 @ 4:47 am

    My first lessons in Mandarin Chinese were given by a visiting professor from SISU using Kan Quian's Colloquial Chinese as the foundation for the lessons. In Lesson 3, the protagonists are at a company party, and towards the end of Dialogue 1 Rachel says "这是我的好朋友舒兰 (Zhè shì wǒde hǎo péngyòu Shūlán)" with a definite fourth tone on the 2nd element of 朋友 (I have just listened to the CD to confirm my recollections). But she then goes on to say "舒兰,这是我的男朋友斯图亚特 (Shūlán, zhè shì wǒde nán péngyou Stuart)", and in that 朋友, the second element is absolutely neutral. I think that in the first utterance the speaker is placing emphasis is on "good friend", rather than on the name "Shūlán", whereas in the second it is the name "Stuart" that is key rather than the phrase "boy-friend".

  6. Philip Taylor said,

    May 12, 2017 @ 5:46 am

    "Quian" -> "Qian", of course …

  7. Mark Liberman said,

    May 12, 2017 @ 9:10 am

    See Wei Lai, Mark Liberman, Jiahong Yuan, and Xiaoying Xu, "Prosodic strength intrinsic to lexical items: A corpus study on tone reduction in Tone4+Tone4 words in Mandarin Chinese", ISCSLP 2016.

  8. Jonathan Smith said,

    May 12, 2017 @ 10:23 am

    A way to look at this is the relationship between the various effects of phonetic neutralization and the tonemic paradigm. Observing that the second syllable of péngyǒu/péngyou "sounds like Tone 4" is anecdotally interesting, but more interesting is if phonetic studies can demonstrate (close) identity in certain contexts and/or if the syllable actually shifts tone membership in certain contexts. Compare Beijing huìyìshǐ, jiàoshǐ etc. for a case where the best analysis might be rephonemicization (>3) after neutralization.

    @Phillip Taylor, in nánpéngyou we may be up against a 2+2 sandhi which would give different phonetic effects…

  9. Guy_H said,

    May 12, 2017 @ 11:20 am

    The comment about Taiwan Mandarin speakers sometimes making a "half" third tone rather than a "full" third tone certainly rings true.
    Just as an anecdote, I remember as a kid in Taiwan, my teacher admonishing me to "round out" my third tone in words like 走 and 友 and that only making half the tone was being lazy. I remember her making me practice by saying 有沒有 a few times too.
    This would have the early 1990s, so I see things haven't changed!

  10. Neil Kubler said,

    May 12, 2017 @ 2:08 pm

    Interesting, but I'm not convinced this is a case of a genuine Tone 4! It still sounds to me like a (somewhat variant version of) neutral tone, no doubt influenced by the rising Tone 2 on péng-. Certainly standard Northern Mandarin is always péngyou and NOT péngyǒu (which is common in Southern China because of reading pronunciations and also because neutral tone is less common in southern Chinese topolects to begin with). Have any of you ever noticed that shémme "what?" (with officially a Tone 2 on the first e) is actually pronounced by many native speakers as shĕmme? And then there is the case of Taiwanese Mandarin băbá for bàba "daddy" (and same Tone 3 + Tone 2 for mama, gege, jiejie, didi, meimei), but that seems to be a morphological, meaning-related tone change that is not purely on a phonological level as with pengyou or shenme.

  11. Chris Button said,

    May 15, 2017 @ 1:59 pm

    This discussion reminded me of an interesting comment by Kratochvil (1998) that I managed to dig out regarding the well-known supposed shift of tone 3 to tone 2 before another tone 3 syllable: "If any modification in sequences of Tone 3 syllables takes place at all in normal speech, which is rare and limited to recurrent strings, the resulting shape is not that of Tone 2 but an idiosyncratic form distinct from Tone 3 and Tone 2, as noted in Zee (1980) and Kratochvil (1987a)."

RSS feed for comments on this post