I saw a post on ELU about a more general question, Softened pronunciation of consonants, such as “t” or “s” followed by “y”. The question was answered in regard to palatalization, especially for phrases like "but you", where /t/ is followed by a palatal glide, /j/.

But does "palatalization" also explain the affrication of /t/ and /d/ when they're followed by /ɹ/, instead of by a front vowel or a palatal consonant? E.g. in 'tram' or 'address' (but not necessarily in 'string'?)

I could see how the affricates are closer to the palatal position than the stops, but the trigger for the change doesn't seem to be a following palatal consonant. (Typical English /ɹ/'s place of articulation is described as alveolar rather than palatal.)

Is this an example of a case where "tongue-raising typically affects apical and coronal consonants", i.e. palatalizes them?

This question is explored some at http://literalminded.wordpress.com/2010/03/17/chricky-affrication/ and its follow-up blog posts, but I didn't see a definitive answer for what causes the change, or what the process is called.

P.S. See also John Well's post how do we pronounce train ?, which is interesting: the "allophonic rule that retracts t and d when followed by r" - what rule is that and what causes it?

Strangely, when asked about "the shift of any alveolar plosive preceding a rhotic consonant to a post alveolar affricate," Wells concludes "I don’t believe there is any such phonological change in progress." I really don't understand how that conclusion is justified.

  • 1
    Re Wells's conclusion - it seems to follow on from the paragraph "If the answer [as to whether cent'ry is different from sentry ] is no... If it is yes,..." He thus denies it on either ground: if cent'ry is identical to sentry, then it is neutralization--you will have a new archiphoneme covering /t/~/tʃ/ before r, and thus not /tʃrein/ but something like /Trein/; and if cent'ry is distinct from sentry--and he says he usually finds it so--then /tr/ and /tʃr/ are still distinct, so therefore /trein/ has not merged to /tʃrein/ (though phonetically it may be quite similar).
    – Muke Tever
    Dec 3, 2011 at 7:17
  • 1
    @Muke Clouding matters is that he admits in the comments to using a broader definition of 'phonetic' that does not contrast with 'phonemic'.
    – Muke Tever
    Dec 3, 2011 at 7:25
  • @Muke: exactly ('clouding matters'). That discouraged me from trying any harder to follow his logic. Thanks for your 1st comment which helps make that clearer. So even though he denies that there is such a phonological change in process, he is not denying a phonetic (in the strict sense) process, which could be what the asker meant.
    – LarsH
    Dec 5, 2011 at 16:04

4 Answers 4


Phonetics isn't my forte, so I don't follow literature on phonetics and I might be wrong. Here's my guess. An affricate is a stop followed by a fricative, homo- or heterorganic (same or different place of articulation). The tip of the tongue is retracted (moved back) for "t/d" in anticipation of the following "r". In fact, O'Connor 1994 argues that it is on the front of the hard palate. See below what he says:

tr: "the tongue is in position for the friction of /r/."

dr: "there is friction as the voiced air passes over the tongue-tip for the /r/."

  • Thanks for this reference. It seems to confirm what I found elsewhere (retraction).
    – LarsH
    Dec 5, 2011 at 16:08

I think you'll find that it's initial dental stops that "soften" before /r/ ("soft" and "hard" are not phonetic terms, alas). Treat is an example, while street isn't, at least not so much. What's going on in treat is technically called Affrication, i.e, the initial, voiceless, aspirated /t/ and the following voiceless retroflex /r/ are joining together as an affricate, just like /tʃ/ in cheat. And why does treat sound so much like cheat?

The key is the fact that initial voiceless stops in English /p t tʃ k/ are aspirated. This means they get a little puff of air when they're pronounced. This is quite automatic but very easy to perceive, and the aspiration merges with the /r/, which is already voiceless because it's following a voiceless stop, and the result does sound very much like /tʃ/. It's possible to distinguish them, but we often can choose which one to "hear", in context.

Note that English doesn't use /r/ after /tʃ/ or /dʒ/, though it occurs after all other stops; try pronouncing "chree" so that it doesn't sound like "tree", and you'll see. The palatal affricates behave otherwise like normal stop consonants; this restriction on /r/ clusters is because palatals are too close to /r/ in the mouth for us to distinguish easily. This is a common situation in many languages.

In Mandarin, for instance, which has contrastive Palatal and Retroflex stop series (in Pinyin, palatal stops are all single letters j, q, and x, while retroflex stops are two-letter combinations with a second letter h), hearing and pronouncing the distinctions between them is always a challenge for English speakers.

In English, however, we only have the one retroflex consonant /r/ (which you can represent as /ɹ/ if you feel it's required), and /r/ is unique phonetically, and famously variable in different English dialects as well.

  • 2
    Interesting theory about aspiration and the voicelessness of /r/ being important to the affrication. Do you have any references for it? It sounds plausible for /t/, but it falls apart for /d/: initial /d/ is not aspirated, and /r/ after /d/ is not made voiceless by a preceding voiceless stop; yet initial /dr/ is pronounced /dʒr/.
    – LarsH
    Dec 1, 2011 at 22:38
  • You're right, it doesn't account for /dr/. But it does make it more noticeable with /tr/. Depending on individual habits, I would guess it's impossible for some people to transition between a dental stop and /r/ without producing some palatalization effects. I don't know the literature on the phonetics; I'm a general linguist. Sorry.
    – jlawler
    Dec 1, 2011 at 22:43

Can't find a simple online citation for this, but it's discussed in Sociophonetics: An Introduction by Erik Thomas.

As a note about why /r/ might cause retraction in /t/ and /d/, I would say that the American /r/ is most commonly one of two forms - a retroflex approximant or the "bunched-r" where the tip of the tongue is not raised and the body of the tongue is retracted somewhat (and so it would not be at all correct to refer to it as an alveolar approximant). Both of these sounds require the tongue to be further back in the mouth than the typical pronunciation of /t/ or /d/ and so the fact that strong coarticulation results in something like the /tʃ/ and /dʒ/ affricate sounds is not actually surprising. It might not be appropriate, however, to call the resulting affricate "palatalized" in all cases (it may be retroflexed).

I'm not sure of a good source on the effect of bunched-r on coarticulation (as compared to other /r/ variants). It's been a while since I read it, but it might be discussed in Mielke, Jeff, Adam Baker, and Diana Archangeli's paper "Variability and homogeneity in American English /r/ allophony and /s/ retraction" (in LabPhon 10).


I guess I've found a satisfactory answer. Following John Wells's allusion to an "allophonic rule that retracts t and d when followed by r", I did some more searching and found the rule described in a couple of places:

The process is labelled as "retraction." I haven't found an explicit statement of the mechanism that causes this rule to happen, but apparently it's because the tongue has to be further back for the r, so it starts moving that way during the preceding segment (sometimes).

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.