Thursday, 4 February 2010


Dominik Lukeš writes
I've always assumed that [ts] is one of the lesser affricates of English on the cusp of phoneme status given the single minimal pair in pizza/Pisa. I've never studied English phonetics (only general [phonetics] having Czech as the starting point) so when I started preparing some talks on pronunciation for literacy teachers, I was amazed to find out that [tr] and [dr] are considered affricates and [ts] is at best affricate-like. All the samples on the pronouncing dictionaries' CD-ROMs are distinctly pronounced as affricates and yet the Cambridge one even splits the syllables of pizza as /piːt.sə/.
Is there a phonetic reason for this? Am I being misled by my Central European ears? Or is this just a result of convention? I've asked a few non-expert native speakers to contrast pizza/Pisa and Nazi/*nasi and they were all able to do so effortlessly. When asked to reflect on what sound they were pronouncing, one speaker said: 'it's a sort of squished soft sound' - surely a folk definition of an affricate.

The definition of “affricate” has always been a bit tricky when we are discussing English phonetics. (That’s why we set it as a topic for student essays.) Do we define it in purely articulatory terms, interested only in the manner of articulation? Or are we concerned with English-specific phonology? (E.g., do these entities have phonemic status?)
I don’t understand why Dominik adduces Pisa to go with pizza: the consonant(s) in the middle of the latter are voiceless (ˈpiːtsə), while that in the former is voiced (ˈpiːzə). If we’re looking for minimal or near-minimal pairs, there’s more mileage in blitz blɪts vs bliss blɪs, curtsey ˈkɜːtsi vs cursing ˈkɜːsɪŋ, Betsy ˈbetsi vs Bessie ˈbesi. But then who has ever doubted that ts was in contrast to s in English?

Except for a handful of proper names and recent loanwords and the single item curtsey (derived from trisyllabic courtesy), English ts always involves a morpheme boundary between the plosive and the fricative: cat#s, put#s, hint#s, drift#s, list#s, bat#s#man, eight#some. (Please don’t start quibbling about flotsam.) No word begins with ts (ditto: no quibbling about zeitgeist or tsetse). No single-morpheme word ends with ts (except recent loanwords).

In this respect ts is like , just an adventitious consonant sequence straddling a morpheme boundary. On the other hand occurs readily in initial, medial, and final position in single-morpheme words: chin, kitchen, touch. It even contrasts with the corresponding plosive-fricative sequence: ratchet vs rat shit. The same is true of tr, at least in initial and medial position.

I would even say that to English ears (or brains) ks is probably more of a single unit than ts (because of box, six, exit etc). (Articulatorily, of course, ks is obviously not an affricate, because the articulation is not homorganic.)


  1. Can I quibble about 't's unusual and friends?

  2. No, you can't. That is a derived surface phenomenon.

  3. Boris Blagojević4 February 2010 12:34

    English /t#s/ (as in lists, bits, sheets, shoots...) sounds very different from the affricate /ts/ in Croatian to me, and I would never consider it to be affricate even phonetically. Of course, this could partially be due to the priming by spelling (for example, it takes a bit of convincing to see that 'Hrvatska' too contains that affricate, since it's usually spelled ), but I doubt it could be the whole story.

  4. Boris Blagojević4 February 2010 12:39

    *"since it's usually spelled )" should of course be "since it's usually spelled c-in-angle-brackets)".
    I forgot how that would turn out.

  5. Does the prefix fitz count? Or does that come under your 'handful of proper names'?

  6. Well, in my Russian phonetics class I was taught that the affricate [ts] has the same duration as any other stop, while in sequences (as in e.g. the Russian cognate of Hrvatska, though I don't think there are many, if any, examples of tautomorphemic t-s clusters) both segments have more or less their normal duration. (This also seems to be reflected in netspeak where people can write d/t-s as [cc], e.g. "adskij" lit. 'hellish', from "ad" 'hell', as "acckij") I'd expect the English cluster to have somewhat similar durational properties.

  7. I'm with Boris on this one. I think the /s/ part is sublty longer in English /ts/ than in Polish, and the synchronisation is subtly different, too. Of course if the /t/ is glottalised, the whole thing sounds totally different.

    For zeitgeist and tsetse things might be different but I'm not going to quibble.

  8. More phonetic evidence is that "rat shit" can be [ɹaʔʃɪt], but I think the affricate in "ratchet" would always be [tʃ] (or [ʔ͡tʃ]) rather than [ʔʃ]. "Pizza" patterns with "rat shit" (how's that for a sentence quite possibly never before uttered by a native speaker?) in that it can be [piːʔsə], which presumably reflects the syllabification /piːt.sə/.

  9. Just thought I'd note that /ks/ has been assumed as an affricate in Blackfoot pretty much since the language was first studied. The [s] is a bit retracted so that it's closer to homorganic, but it's certainly not [kx]. There have even been claims of /ps/ as an affricate as well, though that is not generally accepted.

  10. Boris: if ever you need to cause an angle bracket to appear in a comment, you must write it in HTML, as &lt; (<) or &gt; (>). (How did I manage to get that on-screen?!)

  11. My wife, a native speaker of Russian, claims not to feel a distinction between the affricate ц and the cluster -тс- in хорватский.

    Daniel Jones and Denis Ward write otherwise. They describe (and, indeed represent in diagrams) the tongue position of the affricate as further back from the teeth. And they state that

    Roughly speaking each of the sounds t and s in t-s occupies the same length of time as the whole of the affricate ts.

    The Pronunciation of Russian Jones & Ward, CUP 1969.

  12. Frindt Levente4 February 2010 16:01

    That might be because the [t+s] of хорватский coalesces into the affricate [t:s] which then gets shortened automatically to [ts] because another consonant follows. All this might happen inside the lexicon. Well, this proposition was a bit bold on my part, because I don't speak Russian and can't even read Cyrillic...

  13. &amp; evidently.

    Can I have "tsatske", or have I been playing too much Scrabble?

    Other words that spring to mind:
    1) catsup -- its variant "ketchup" shows that it's a real affricate in the middle.
    2) patsy -- I'd consider "-y" the suffix here, but you might well argue that we have two suffixes on top of one another.
    3) teentsy -- as a variant of "teeny", "teensy" there's the same case to be made as for "catsup".
    4) ditsy -- I'm disappointed by the shallowness of the citations on this one -- surely it goes back further than 1970.

  14. Well David, I could hardly believe that your wife didn't even "feel" a distinction, as I took that to mean she doubted the possibility of one, but there must be a lot of people who feel the same way about it. Google hits for Хорвацкий: 10,100 Russian pages. And that's just people who can't spell!

    But why is it _not_ Хорвацкий in Russian, given the Croat? Can anyone enlighten me on that?

  15. IMO, except for spelling, the grounds *not* to consider /tS/ as a cluster rather than an affricate in English are no stronger than those not to consider /tr/ the same way, and yet practically any dictionary has /tS/ but not /tr/ in the phoneme list. (Yeah, spelling... Does anyone know any other reason why the labiodental nasal in Italian is considered as an allophone of /n/ rather than of /m/?)

  16. "That sit" vs "that's it", anyone?

  17. Interesting examples, James, but
    1) catsup -- So does Tsar's variant Czar show that it's a real affricate at the beginning?? OED has tsɑː(r), zɑː(r) for the first and only tsɑː(r) for the second. (Not that OED's making any more sense than your original statement - I only intend a reductio ad absurdum of going by variants!)
    2) patsy -- I'd consider "-sy" the suffix here, and Pats the abbreviation. Think 'Absolutely Fabulous'. My mother was Mumsy before she was Mums, and the same process happened to many other names of my acquaintance. If Bugsy Malone was ever Bugs, I bet it was after he was Bugsy!
    3) teentsy -- well here we go again with the variants, but I would say that was more marginal than schmaltzy, for example. Extra-systematic, even.
    4) ditsy -- or ditzy: either as 2 or 3

    But reductiones ad absurdum (or absurda) aside, I really really do have to go along with the non-affricators even for your examples: for me there is all the difference in the world between the middle of catsup and the middle of ketchup, and I use your expression because the difference extends to vowel length, tempo, and a few other things I think are in there somewhere!

  18. My intuition (as a native speaker) about "/tr/" has always been that it starts with an affricate phonetically similar to my realisation of /tʃ/ which is then followed by a normal "English r". (Similarly for the voiced version.) So I've always found the descriptions of /tr/ and /dr/ a bit puzzling; presumably I realise them differently from the traditional RP way.

    (If I designed a spelling reform based on my own idiolect I'd probably want to write things like "jrip".)

    As for /ts/, I have to say that in some of these loanwords like "pizza" (and indeed names like "Fitzwilliam") it does indeed feel like it's on the way to becoming a single unit. I wouldn't rule out the possibility that English might borrow a /ts/ phoneme from German and Italian in much the same way that some forms of English have marginal nasal vowels borrowed from French.

  19. This comment has been removed by the author.

  20. Mallamb

    Хорватский is to Хорватия as шотландский is to Шотландия.

    The suffix is -ский no matter what consonant precedes.

    Now you know how to spell Tibetan in Russian.

  21. The distinction can indeed be blurred, especially before consonants (so Frindt is essentially right). Prevocalically, however, it is a very clear one (there are lots of [ts] prevocalic clusters because they pop up before the reflexive suffix -sja) Interestingly, Belarusian opts (as is often the case) for a phonetic spelling, and they spell "Croatian" with one <c> (харвацкій), but all the reflexive verbal forms are spelt with two.

  22. Yes I see that, but why? Why should Russian not have a form that is more cognate with Croat? Perhaps it does, and Хорватский is an etymologizing spelling, and its standard pronunciation a spelling-pronunciation! Your wife may be saying Хорвацкий out of justifiable contempt for all that! And all the misspellers on the internet writing it for the same reason!

  23. Russian has хорват--Croat and хорватский--Croatian, perfectly cognate with Croatian Hrvat, hrvatski. I don't see your problem, mallamb.

  24. @Paul (13.42)
    In my pronunciation 'ratchet' and 'rat shit' are homophones. In both cases, both /t/s are realised as glottal stops - same applies to hatchet, crotchet.
    There's a pub in Huddersfield, W. Yorkshire called the Rat and Ratchet.

  25. I'd like to offer a few more instances where the 'ts' sequence distinguishes meaning inspired by comments on this blog: ts/s: curtsey/cursey, ts/z: bitsy/busy, itsy/Izzy; and ts/tʃ: bitsy/bitchy, itsy/itchy.

    Now the questions is? Are they articulatorily and acoustically more like box/boss or rich/ridge.

    BTW: As a native speaker of Czech, I could go either way on Chovatský or Chorvacký. Funnily enough, my arguably limited intuition about Russian makes me feel that there the 'ц' would be much more commonplace - mostly because I can't spell Russian properly anymore.

  26. I have a hard time conceiving of [ts] as an affricate in English - AFAIK, English speakers find it quite unnatural to use [ts] at the beginning of a syllable, unless they've adopted affected xenophilic pronunciations for things like "tsar" and "tsunami". I fail to see how it's different from any other plosive+fricative sequence, like [ks] or [ps].

  27. Mallamb

    I think I can hear a longer stream of sound when my wife says хорватский. She just says 'Who cares?'

    This would be unsurprising but for the fact that for some years she taught a Russian Phonetics course to undergraduates. Before you get too excited, this was really a course in phonetic transcription as an aid to better pronunciation.

    My wife is an excellent teacher with a clear 'intelligentny' Petersburg accent, but she felt the course was doing more harm than good, as students were struggling with two new alphabets. So she abandoned IPA in favour of some system used in Russia with Cyrillic symbols.

    Dennis Ward also composed Russian Pronunciation Illustrated, in which he offers contrasting uses of ts and t-s.

    Подсо́лнечник [should be подсо́лнух says Lena] цветёт на со́лнце
    under a picture of a sunflower which is indeed growing under the sun

    And under a cartoon sketch of two metal workers bashing metal in a factory's 'auxiliary workshop'.
    Кузнецы́ рабо́тают в подбсо́бном це́хе

    I can't reproduce Ward's pre-Kiel IPA transcription, but you can gather that he makes the sunflower word start with pʌˈt-sоl. The 'auxiliary' word is transcribed pʌˈt-sоbnəm. He transcribes the sound spelled ц as [ts].

  28. Your posts are always so interesting Mr. Wells! I am a fan of your dictionary. As a Spanish speaker (Arg) I've always had trouble with the sequence verb+er= sing+er, verb+ing+in for example, quoting the Beatles...'sing+ing+in the dead of night'. I was told that -ing is more or less like our cinco.

    At college I was taught RP English, therefore I say 'I can't' as (it won't let me copy and paste IPA here) a RP speaker should, but I'm always catious since I'm afraid I'm making it sound like another word that could be embarassing in public. I should pick the AmEng version of can't /kent/ or /keint/, I guess.

    Keep up the good work, Mr. Wells!

  29. @luke
    Thank you. I was going by Boris Blagojević's argument:

    "…'Hrvatska' too contains that affricate … since it's usually spelled … c-in-angle-brackets".

    Which I misread because it got garbled by html problems. I regret to say I thought he was saying it was spelt Hrvacka. But obviously he means the affricate ts in Hrvatska is usually spelt c. So yes, хорватский and hrvatski are perfectly cognate, morphologically speaking. But it still seems they both refuse to consistently behave morphologically! For now I see that Boris is telling us emphatically that it is in spite of the spelling that 'Hrvatska' too contains that affricate.

    But once again, we find that people scorn to spell it in a way that does not visibly contain that affricate, just as they write хорвацкий for хорватский! Googling again:

    1,930,000 Croatian pages for hrvacka televizija.

    Etc. etc.

    I did realize that Russian was rigidly morphological about this, and always spelt –ский, but David's wife's merger with ц made me sit up, and I reported the evidence I had found that she was far from being alone in being, as I thought, in line with this Hrvacka.

    Dominik Lukeš himself now tell us "As a native speaker of Czech, I could go either way on Chovatský or Chorvacký. Funnily enough, my arguably limited intuition about Russian makes me feel that there the 'ц' would be much more commonplace - mostly because I can't spell Russian properly anymore."

    And neither can all those Russians on the internet! It's a laugh a minute, isn't it?

    But this time I am on the qui vive against ignorant misreadings, and I now take him to be talking not about the spelling, but about the pronunciation of Chorvatský as if it were Chorvacký. There are not nearly so many misspellings around in this case. Either Czechs must be better spellers, or there must be less merger and more morphology in Czech!


    "I think I can hear a longer stream of sound when my wife says хорватский." She just says 'Who cares?'

    Well played! Hang on in there! Dare you ask her again, and listen for palatalization of с or both с and т ? That would eliminate хорвацкий, wouldn't it?

    My wife was like Mr Polly's, who the moment she was married to him stopped laughing at his jokes. Mine stopped pretending to care about the job as an informant I had assigned her. I guess we've all been informants at one time or another, and it _is_ a terrible experience, isn’t it?

    So yes, David, it is unsurprising.

    Dennis Ward's examples are unnecessarily uncontentious, it seems to me.


    Thank you, fellow non-affricator. It is of course necessary to mention the 'affected xenophilic pronunciations for things like "tsar" and "tsunami"'. Pace the OED, I think tsɑː is a gross affectation, whether its t-sɑː (which is what is usually affected by the xenophiles) or tsɑː, when zɑː is so well established, but yes, I have an affricate in tsunami, and I don’t think it's an affectation to abhor sunami!

  30. Just as well I didn't make that post any longer, you might think, but I meant to comment on Pavel's point that the distinction between тс and ц can indeed be blurred, especially before consonants (so Frindt is essentially right), but prevocalically it is a very clear one:

    Pavel, you say "there are lots of [ts] prevocalic clusters because they pop up before the reflexive suffix -sja."

    And when they do, is that not another egregiously morphological spelling, since -тся and -ться are pronounced ца? I suppose you think this too obvious to mention! But your point must be that here it is definitely not the sequence т, с. That would be palatalized and the whole point about the pronunciation ца is that it is not! Thus in spite of the spelling, this can only be the affricate.

    And Pavel, I am so gratified that you tell us that Belarusian just throws morphology out of the window and opts for харвацкій and all the reflexive verbal forms spelt with two cs. Is that a geminated affricate tts? Would you like тца for Great Russian?

  31. 1,930,000 Croatian pages for hrvacka televizija.

    That's because Google autocorrects hrvacka to... HR, it seems (for whatever reason). When you turn that off by googling +hrvacka televizija, there's only 2,760 pages left.

  32. Thanks for that corrective intelligence too. Must learn to search properly. I had used the minus to eliminate other problems with these searches, but it would never have occurred to me that I needed to use the plus, especially for such an arcane reason.

    I did wonder at the level of insubordination I thought I had found! Still, 2,760 cases of insubordination is enough to make my point.

  33. Boris Blagojević5 February 2010 13:57

    @Mallamb: Sorry for causing misunderstanding.
    I wouldn't bet my life it's exactly the same sound as in "packa" or similar words (although I'm quite sure it is), but I can say that it's different from English or for example Finnish ts, and even from Croatian "podsjetiti" which is often enough misspelled as "potsjetiti", but almost never "pocjetiti".

    btw. I find that googling words with quote marks often helps get better results and bypass various autocorrections.

  34. @Mallamb: no, the distinction between -тся and -ться is neutralized in this position, so -тся/-ться is the Russian equivalent of your/you're. And it is not pronounced as if spelt -ца: the cluster has a much longer closure phase (around the length of a full stop segment). You do find the spelling -цца in netspeak (and Belarusian), but note that it's still doubled.

  35. Well Boris, in podsjetiti again you've got the morpheme boundary, as in David's подсолнечник from Dennis Ward's Russian Pronunciation Illustrated. I'm amazed it even gets spelt "pocjetiti" at all! That speaks even more volumes, it seems to me.

    I thought I did put hrvacka televizija in inverted commas, but now find I can't replicate that search. I don't remember why I picked that particular combination to search, and can't imagine why it should have fallen foul of the Google software.

    BTW David, why does your wife say it should be подсoлнух?

  36. Oh dear, Pavel, I said all that in my post of 12.07.

    I said "-тся and -ться are pronounced ца", so of course I meant the distinction between -тся and -ться was neutralized. And I suggested the spelling тца to parallel the admirable Belarussian cc, so of course I approved of the idea of a spelling reflecting the compensatory lengthening of the closure phase.

    Did I really need to say "-тся and -ться are pronounced _with_ ца"?

    Thanks for mentioning the netspeak -, but that parallels Belarussian spelling a little too closely for comfort! Both cc and цца give a funny impression of the longer closure phase, don't they? We don't really want to imply tsts, and that's why I suggested тца .

  37. Mallamb

    Lena tells me that подсoлнух is the only Russian word for 'sunflower'. Dennis's подсолнечник is presumably a back formation from подсолнечное масло 'sunflower oil'.

    She used to work with Dennis and knew him well. She describes his Russian as 'quite good'.

    My ear is not sufficiently trained to judge the presence or absence of palatalisation in the middle of хорватский. I'm trying to set up a blog where I can post a sound file recording of her so that you can judge.

  38. No, подсолнечник does exist, even with the meaning of 'sunflower' (next to 'parasol'), it's just not the default word today.

  39. David! What heroism to go back to your wife AND to now be fixing to post a sound file!

    Thanks for the sunflower. I've got it now. The dictionary says (BOT), so presumably she thinks подсолнечник is too technical for this book, but I don't see why she thinks it's a back formation from подсолнечное масло, rather than the latter being a regular derivative, morphologically as well as oleaginously!

  40. Wow! And a picture too! I've fallen in love! What are we going to do? Didn't you foresee this hazard? But never fear, I'm not so easy on the eye as either of you!

    After one hearing: definitely palatalized.
    After two: ditto.
    After 3 4 and 5: dittox3.

    So by my reckoning хорвацкий is exploded.

    Fun while it lasted!

    Another thing that struck me was that she obviously is indeed "an excellent teacher with a clear 'intelligentny' Petersburg accent". You can practically hear the твердый знакъ at the end of хорват!

    So you can hear that the much maligned morphology with -ский really does impact on the т. Or is it the spelling? ʕ8-ʃ

  41. The evidence that I'm as hard on the eye as on the mind.

  42. So how do I display my userpic on here?

  43. Affricates are consonants that are formed by stopping the flow of air somewhere in the vocal apparatus and then releasing the air relatively slowly so that the friction sound can be produced.

  44. An aside: as a native, monolingual SAE speaker, I learned the most basic of Japanese phrases and then spent a little over a week in Tokyo. Much to my surprise, even that short bit of exposure seems to have been enough to make a phoneme out of /ts/ for me. I find that I'm now suddenly conscious of the distinction between /s/ and /ts/, where I certainly wasn't before. "Sunami" now sounds very, very wrong to me--I wouldn't have even noticed before. It's a creepy feeling.

    I'd certainly have to anecdotally argue that /ts/ is most certainly NOT an SAE phoneme. The morphemic boundary explanation for its surface appearance makes a lot of sense. I also don't think you'd find many native SAE speakers claiming that "peanuts" ended in a different sound than "pears," though you could certainly embarrass them by belaboring the point.

  45. I have found myself wondering and asking about what the real affricates are, if it is correct to put it in this way, because as you said the definition is a little tricky.