Third tone

In the iPad app, at least, the third tone (at least on single characters) sounds different than I’ve heard elsewhere. There’s a break or stop in the middle, like “yee-eh” for 也.

Is that correct? I’d hate to be learning these wrong.

When pronouncing the third tone as a dipping tone, which normally only happens in isolation (including in Skritter when it’s a single character) or when heavily stressed, it’s perfectly normal to have a creak in the middle. That being said, the current recording we use for 也 does sound a bit off, maybe because of compression or some other part of the post-processing of the audio. We’ll get it rerecorded!

If you want to listen to a reasonable creek, check the pronunciation of some other third-tone syllable. I checked a few myself and these sound good, for example 马 or 奶. Especially the latter has very clear creaking in the middle, but still sounds good.

Tl;dr: Yes, the particular example you brought up sounds a little off, but also yes, this type of creak is normal.

That’s funny; I thought I had picked a representative one!

Well, I only checked a handful to try to find a good example of a creak that is normal. However, I manually listened to every single-syllable audio upload when they were first recorded, so I’m fairly certain that the weirdness in 也 is not typical. I also use Skritter myself with lots of single characters with audio on and haven’t noticed this myself. If you do find other example that you think sound odd, do let me know and I’ll be happy to verify them for you (and for all other users, of course).

I think that it’s more that I couldn’t tell the difference. They all sound more-or-less the same to me, like (to take a random example) 匕—an almost-but-not-quite glottal stop in the middle. But if you say that’s normal, and I’m not screwing my pronunciation up by memorizing them like this, I believe you!

Yes, the 匕 pronunciation is perfectly fine; you can learn that without worries. People differ a bit in how creaky their voices are and how low they go, so expect some variation among native speaker here. An interesting side note is that speech-to-text features on smart phones are rather bad at handling single third tones for this reason. I conducted a non-scientific experiment with speech-to-text a while ago and often third tones were transcribed as two different syllables. That’s of course because actually there kind of are two parts with a visible break in the spectrogram. If you want to read more about this, I wrote two posts about it, the first here: