So in my opinion, scarves is pronounced as the dictionary has it: with a Short O or /a/.

But I believe that scarf and other "ar" words that are followed by voiceless consonants, are not actually pronounced by most people with a Short O or /a/. I think they are pronounced with a Short U as in fun. (a short U+R, to be exact).

I think we open our mouths less on a "ar" before a voiceless consonant. I can't find any data for this, but I feel it. Am I crazy, or have I been pronouncing scarf, lark, art, harp, etc wrong all these years?

I was born in the American Midwest.

  • 3
    They are the exact same vowel for me. Would you say the fourth pronunciation on this page is like what you hear? forvo.com/word/scarf Commented Aug 23, 2018 at 21:44
  • 1
    Also, I'd suggest you read about the International Phonetic Alphabet, because I'm not sure if /a/ is what you meant (that's the vowel in "hat"). en.wikipedia.org/wiki/International_Phonetic_Alphabet#Vowels Commented Aug 23, 2018 at 21:48
  • 2
    @AzorAhai She’s right: scarf has a slightly different vowel than the one scarves has in her speech. This is normal in the American Midwest. See writer/rider.
    – tchrist
    Commented Aug 23, 2018 at 21:53
  • 1
    @tchrist Oh, I don't disbelieve her, I just wanted to make sure she was talking about the vowel raising you wrote up in your answer. Commented Aug 23, 2018 at 22:05
  • 3
    How come nobody talks about this? Shush!! People are listening!
    – Hot Licks
    Commented Aug 23, 2018 at 22:09

2 Answers 2


No, you aren’t crazy; you have a really good ear. And as a native speaker, you can’t have been pronouncing those “wrong” all these years. Your tart simply has a very slightly different vowel than your tars.

It’s normal to have a “higher” vowel when there’s an unvoiced consonant after it, particularly in North America. What you’re observing is the same raising that happens in writer compared with rider: the main /ɑ/ vowel is raised to /ʌ/ as in cut.

That means scarf comes out as [skʌɹf] but scarves comes out as [skɑɹvz]. It may actually be [ɐɹ] rather than [ʌɹ], but that’s a very technical distinction that you probably won’t be familiar with.

Neither of those has an /o/ sound like wore/war (both either [wɔɹ] or [woɹ]) has, or for that matter like dwarf [dwoɹf] does. Oddly, the plural dwarves [dwoɹvz] is unchanged in its vowel. That’s curious.

  • 1
    Do you have a reference for this allophone? I've always heard that with such minimal pairs a voiced consonant lengthens the duration of the vowel (bid-bit) but not the vowel height.
    – KarlG
    Commented Aug 23, 2018 at 22:04
  • 1
    Thank you for assuring me I'm not crazy! I agree with writer and rider. I would be very interested in your explanation of "it may actually be ___ rather than ____" That's a technical distinction I'm not familiar with but would love to know. Sorry- I don't know how to type these symbols here, but I'm not familiar with the upside down ar in the first blank. I'm very familiar with the distinction between er and short u and short o. If you could compare your sounds to that or forward me to an article that explains it, I would appreciate it! Commented Aug 23, 2018 at 22:19
  • 1
    Canadian raising affects /aɪ/ and /aʊ/, neither of which is present here.
    – KarlG
    Commented Aug 23, 2018 at 23:44
  • 1
    @KarlG It’s the same effect: the same raising of /ɑ/ to [ɐ] or [ʌ], triggered by the same factor of whether it’s followed by an unvoiced consonant. Our arrhotic brethren would consider "ar" in the coda there a "centralizing diphthong", so that's not so different from the "ai" or "au" diphthongs. (Not that I think of "ar" as a diphthong myself.)
    – tchrist
    Commented Aug 23, 2018 at 23:55
  • 4
    If dwarves is an exception, could that have something to do with the fact that the traditional plural is dwarfs, while the form dwarves was invented by Tolkien?
    – bof
    Commented Aug 24, 2018 at 4:25

Basically: If you have a sensitive ear you likely can detect slight differences in the words you mention, but there are also differences in the sound of a single word from one "native" speaker to the next (even if both speakers have the same regional accent), and often differences, based on adjacent words, wider context, audience, time of day, etc, between two uses of the SAME word by the SAME speaker.

Further, in the examples you're considering, the variations are basically "programmed in" to the speech mechanisms of "native" speakers, so they often can't detect the differences.

And notations such as IPA are insufficiently precise to enable computer pronunciation -- ie, they still rely on human "interpretation".


A) The notation is simply not precise enough to record these variations.

B) There is no great advantage in making the notation more precise since the "noise" (due to the factors noted above) is greater than the potential for increased precision.

  • The notation is precise enough to distinguish the vowels in tight and tide, or writer and rider.
    – tchrist
    Commented Aug 23, 2018 at 22:27
  • @tchrist - Precise enough to pretend to make such distinctions. Other potential distinctions are glossed over.
    – Hot Licks
    Commented Aug 23, 2018 at 22:35

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.