/t/ versus /d/ and /tS/ versus /dZ/ alternation in NAE

[zompist]

tl;dr - I enjoy the feeling of being superior, thank you very much.

Which is different from the rest of you how...?

It seems like a silly thing to argue about... I think Travis has a point: if you aspire to be linguists you shouldn't be frightened of a phonetic transcription. And we are all over-influenced by phonemic analyses, which can get in the way when talking about phonetics per se.

I think it's true enough, though, that a lot of details are presented that don't seem relevant to the question... e.g. I'm looking at the transcriptions of 'congratulate' and the only relevant bit seems to be [tʃ] vs. [dʒ]. At least, if something else is relevant it's not clear what it is.

And as a minor annoyance, I can't see a couple of the characters on my Mac (e.g. d̥ʒ̊ appears as d-box-ʒ-box), while generating IPA is only easy on the Mac.

[Dewrad]

tl;dr - I enjoy the feeling of being superior, thank you very much.

Which is different from the rest of you how...?

The comment refers to the entire argument (which has, after all, been hashed out several times before): not just Travis' post.


Also, Travis: yes, this is true.

[Nortaneous]

I think the problem here is that people don't know when to use phonemic transcriptions. There's nothing wrong with including a phonetic transcription, but it can get confusing without a phonemic one. ("What's that [M] doing there?"/"What the flying fuck is [M\_-r\`)_G_-_w]?"/"Where did those syllabic retroflex nasals come from?"/whatever)

In this case, there are, as far as I know, no relevant details that only show up in phonetic transcriptions. All the relevant information is there in the phonemic transcription, so why not just use that?

edit: But still, Travis has a point. I've seen a lot of really bad "phonetic" transcriptions of English. If you find yourself simplifying things to the point where your phonetic transcription is almost exactly the same as your phonemic one, just use the phonemic one.

[Travis B.]

I think the problem here is that people don't know when to use phonemic transcriptions. There's nothing wrong with including a phonetic transcription, but it can get confusing without a phonemic one. ("What's that [M] doing there?"/"What the flying fuck is [M\_-r\`)_G_-_w]?"/"Where did those syllabic retroflex nasals come from?"/whatever)

In this case, there are, as far as I know, no relevant details that only show up in phonetic transcriptions. All the relevant information is there in the phonemic transcription, so why not just use that?

There are two other reasons why I personally avoid phonemic transcriptions to describe synchronic forms.

The first is that I have strong doubts as to the applicability of the phoneme in its classical sense as a synchronic psycholinguistic construct, as opposed to a diachronic linguistic abstraction. The problem here is that if one presumes that the phoneme is an actual synchronic psycholinguistic construct and not merely a convenient abstract, it becomes very hard to explain observed sound change phenomena such as lexical diffusion and cheshirization. Likewise, many synchronic phonological phenomena that take place in an irregular or lexically-dependent fashion are hard to explain adequately if one assumes a underlying phonemic model.

The second is much more specific to North American English varieties, particularly those that have lost historical phonemic vowel quantity*, and is one I have discussed here before. Namely, it is that what had been allophonic vowel quantity is better preserved than many historical obstruents are, and readily undergoes cheshirization when those obstruents are somehow modified or elided. This is to the point that, synchronically, there are word pairs distinguished only by vowel quantity** in at least some dialects such as my own, and if one were to transcribe said dialects without vowel quantity, one would not be able to reassign vowel quantity properly in many cases. Yet, if one were to say that what had been allophonic vowel quantity has become phonemic, one runs into another problem - in the vast majority of cases, vowel quantity still acts as if it were allophonic, and one cannot construct new words with arbitrary vowel quantity. Thus vowel quantity acts as if it were both phonemic and allophonic simultaneously. In a traditional phonological model, this is untenable; either it is phonemic or it is allophonic. However, there are fundamental problems with both analyses here that cannot be adequately resolved***. Hence I prefer to avoid the question by not directly representing such dialects phonemically in the first place.

* and not just my own dialect, despite my using it commonly as an example of this; I have heard such cheshirization phenomena combined with a lack of apparent historical phonemic vowel quantity in NAE dialects from at least around the Inland North, and I know in less detail that many other NAE dialects do not actually preserve historical phonemic vowel quantity

** aside from minor associated differences in vowel height and phonation

*** the only way to resolve this that I can think of is that vowel quantity is not phonemic, but rather is derived from consonant phonemes that may be silent; the problem here is that there are some words in which the conditioning consonant phonemes are always silent except when attempting spelling pronunciation, and it is hard to justify consonant phonemes that are always silent

[linguofreak]

[M\r\`_-)_G_-_w]?

I would tend to transcribe it (IPA) [ɚ] (X-SAMPA) [@`]. Technically, it's a vowel symbol, so should only be used for syllabic /r/, but it's still the most convenient symbol I can think of for the non-syllabic variant.

[finlay]

And as a minor annoyance, I can't see a couple of the characters on my Mac (e.g. d̥ʒ̊ appears as d-box-ʒ-box), while generating IPA is only easy on the Mac.

Me too. Annoyingly, it used to be fine, and I swear it's the latest version of Firefox that's the problem. It shows up fine if you copy it into textedit, and more of it shows up fine on my other computer. Gah.

As for the narrow phonetic transcriptions, I quite agree with Maknas – it's a skill to be able to do them, but it's an even better skill to know when not to do them. There's too much superfluous information there, and lesson number one of the ZBB is that for every ZBBer who's autistic, there's at least twice as many who are borderline, or plain neurotic (I include myself in this count, FWIW). And they'll pick up on the superfluous information and ignore the original question (I can't count the number of times this has happened when I ask questions about the pronunciation of a word). So really don't put the superfluous information in in the first place – you could have just phrased it orthographically as zompist did, by spelling them 'congradulate' or 'sangwidjes'.

Also, in Travis's case I'm convinced there's observer's bias, as of the limited sample of his speech that I've heard, it never sounded like he had any form of [R]. Lesson number one of phonetic analysis is that you don't do it on yourself. I've listened to myself so much that I don't have a fucking clue what features are "native" to my idiolect and what are introduced, and which ones I've tried to hear and therefore exaggerated. (eg: I'm fairly sure i don't have a FORCE/NORTH merger, which would be normal for a Scottish accent, but I'm not 100% sure if I always did. And then NORTH can be pronounced /or/ instead of /ɔr/. etc)


And as for the original question, we don't tend to get congradulated in britain, kindergarten is a foreign word that yanks use, but sandwidjes is sometimes heard – usually it's more 'sangwidjes', though, to rhyme with 'languages'. I don't know what the geographical restriction is on that one, but I've heard my Scottish father saying it.

[Radius Solis]

If we're having a nice big cathartic mutual tell-off today, count me in, it's been a terrible day.

[Travis B.]

And as a minor annoyance, I can't see a couple of the characters on my Mac (e.g. d̥ʒ̊ appears as d-box-ʒ-box), while generating IPA is only easy on the Mac.

Me too. Annoyingly, it used to be fine, and I swear it's the latest version of Firefox that's the problem. It shows up fine if you copy it into textedit, and more of it shows up fine on my other computer. Gah.

Maybe I am too used to Firefox (aka Iceweasel) under Debian testing with the font setup I have at home being able to just eat whatever IPA I throw at it and not only display it but display it well. I have seen my transcriptions on other machines and with other browsers, and sometimes they are practically unreadable (e.g. whenever I have seen it under Safari), but I have figured the user could always copy and paste it need be. I could use X-SAMPA, but the X-SAMPA for what I want to transcribe is generally far too lengthy and very, very ugly.

As for the narrow phonetic transcriptions, I quite agree with Maknas – it's a skill to be able to do them, but it's an even better skill to know when not to do them. There's too much superfluous information there, and lesson number one of the ZBB is that for every ZBBer who's autistic, there's at least twice as many who are borderline, or plain neurotic (I include myself in this count, FWIW). And they'll pick up on the superfluous information and ignore the original question (I can't count the number of times this has happened when I ask questions about the pronunciation of a word). So really don't put the superfluous information in in the first place – you could have just phrased it orthographically as zompist did, by spelling them 'congradulate' or 'sangwidjes'.

Normally when speaking about historical linguistics I only talk about abstract phonemes; in this thread I had included phones as largely supporting information for one of the cases I was asking about. Note that I prefer not to imply phonetic or phonemic values with orthography, as a matter of personal taste, and rather tend to leave orthography as merely symbolic representation of lexemes, hence why I did not use anything like "congradulate" or "sandwidjes".

Maybe I just assume far too much from the reader with respect to being able to understand what is being asked while not being distracted by supporting information. In my posts I usually provide about the same amount of detail as I have seen in academic linguistic papers on the web, sometimes a bit more (due to marking variation and alternate forms more), and to me it seems appropriate. Yet here it seems that this level of detail and elaboration is far beyond the capabilities of some readers here to think about and respond to rationally, and instead invite completely off-topic comments and what are essentially complaints about having to read anything detailed at all.

Also, in Travis's case I'm convinced there's observer's bias, as of the limited sample of his speech that I've heard, it never sounded like he had any form of [R]. Lesson number one of phonetic analysis is that you don't do it on yourself. I've listened to myself so much that I don't have a fucking clue what features are "native" to my idiolect and what are introduced, and which ones I've tried to hear and therefore exaggerated. (eg: I'm fairly sure i don't have a FORCE/NORTH merger, which would be normal for a Scottish accent, but I'm not 100% sure if I always did. And then NORTH can be pronounced /or/ instead of /ɔr/. etc)

This is why I prefer to use other individuals' speech to back up anything that I observe in my own speech, before I make too many conclusions about it. This was much easier to do when I was living in Milwaukee, due to there being many people there who speak quite similarly to myself, but for things like what the original post was about, other speakers from the northern US are often suitable as well for specific points of detail that I suspect to be more widespread.

In the case of my /r/, I made a point of observing, to the best of my ability, just how other people from southeastern Wisconsin articulated it, for this very reason. I could not get the same level of detail as I could get about my own articulation, but what was obvious was that what other people had certainly did not have coronal articulation except after another coronal and certainly was not rounded aside from highly inconsistent "compressed lips" in prevocalic positions. Sound-wise, it sounded roughly the same as what I have, making it more likely that I had not simply picked up some idiosyncratic pronunciation somewhere.

And as for the original question, we don't tend to get congradulated in britain, kindergarten is a foreign word that yanks use, but sandwidjes is sometimes heard – usually it's more 'sangwidjes', though, to rhyme with 'languages'. I don't know what the geographical restriction is on that one, but I've heard my Scottish father saying it.

Okay, this is useful information, particularly because it indicates that the /t/ > /d/ in congratulate is something that happened in North America and not something inherited from dialects in the British Isles. On a somewhat off-topic note, it is interesting that you have /nd/ > /ŋ/ or /ŋɡ/ (I cannot tell which you mean from your post) in the British Isles; I had not heard of this before.

[brandrinn]

carded = carted != carton

When you said "post-stress /t/ often merges with /d/," I immediately thought of the word "sentence," and all the other instances of glottal t, which for me always occur before a coronal nasal. Is this the only environment that blocks your dialect's post-stress merger? And how is this "port-stress" merger any different from the general "non-stressed" merger exhibited by most American dialects?

Also, people say sandwidge? That blows my mind. The things you learn on the internet...

[zompist]

I was being sloppy... I was originally going to say "post stress medial t/d" but you know, r is kind of vocalic...

So, t in clusters tends not to merge with d. "Sentence" is [sɛ~ʔn̩s], but it's the syllabic n that turns t into ʔ here. "Center" is [sɛ~tr̩], "sender" is [sɛ:~dr̩].

[Travis B.]

I was being sloppy... I was originally going to say "post stress medial t/d" but you know, r is kind of vocalic...

So, t in clusters tends not to merge with d. "Sentence" is [sɛ~ʔn̩s], but it's the syllabic n that turns t into ʔ here. "Center" is [sɛ~tr̩], "sender" is [sɛ:~dr̩].

These pronunciations are interesting, as you are from as close to Milwaukee as Chicago, yet there are notable differences here between what we have. The closest one of these that I have to you is:

sentence: [ˈsɜ̃ʔn̩ts] (X-SAMPA: ["s3~?n=ts])

However, I have notably different forms in the cases of center and sender, having:

center: [ˈsɜ̃ɾ̃ʁ̩ˤ(ː)] (X-SAMPA: ["s3~4~R=_?\(:)]) ~ [ˈsɜ̃̂ːʁˤ] (X-SAMPA: ["s3~_F:R_?\])
sender: [ˈsɜ̃ːndʁ̩ˤ(ː)] (X-SAMPA: ["s3~:ndR=_?\(:)]) ~ [ˈsɜ̃ːnːʁ̩ˤ(ː)] (X-SAMPA: ["s3~:n:R=_?\(:)]) ~ [ˈsɜ̃̂ːɜ̯̃ʁ̩ˤ] (X-SAMPA: ["s3~_F:3~_^R=_?\])

In these cases, the /r/ does not seem to be treated as part of a cluster by my dialect but rather as a consonant separated by a vowel from the /nt/ or /nd/, which being between vowels and followed by an unstressed one are very susceptible to lenition and outright elision.

One thing that must be taken into account here, though, is that there may actually be underlying differences in language varieties that are not simply a matter of place here. What I speak is essentially traditional Milwaukee dialect as picked up by suburban kids and with more progressive features added on top, whereas even in the Milwaukee very many middle-aged or older middle and upper-class people speak something that is essentially just localized General American, and you can very well find many forms transitional between the two in practice. Note that the distance between the two seems to be greater than that between, say, traditional Milwaukee dialect and traditional Chicago dialect, at least from the working-class people from Chicago I have heard speak. Hence there could be another dimension of distance between language varieties here aside from merely place.

[Alces]

Okay, this is useful information, particularly because it indicates that the /t/ > /d/ in congratulate is something that happened in North America and not something inherited from dialects in the British Isles. On a somewhat off-topic note, it is interesting that you have /nd/ > /ŋ/ or /ŋɡ/ (I cannot tell which you mean from your post) in the British Isles; I had not heard of this before.

I think this is just an elision of the medial /d/ followed by POA assimilation of the /n/ to /ŋ/. In my dialect /ŋ/ is not phonemic, only appearing due to POA assimilation of /n/, and I say 'sandwiches' as [saŋwɪtʃʰəz] (although both [tʃʰ dʒ] sound normal to me).

[Travis B.]

Okay, this is useful information, particularly because it indicates that the /t/ > /d/ in congratulate is something that happened in North America and not something inherited from dialects in the British Isles. On a somewhat off-topic note, it is interesting that you have /nd/ > /ŋ/ or /ŋɡ/ (I cannot tell which you mean from your post) in the British Isles; I had not heard of this before.

I think this is just an elision of the medial /d/ followed by POA assimilation of the /n/ to /ŋ/. In my dialect /ŋ/ is not phonemic, only appearing due to POA assimilation of /n/, and I say 'sandwiches' as [saŋwɪtʃʰəz] (although both [tʃʰ dʒ] sound normal to me).

Okay, so this is a regular assimilation of /nw/ (post-elision) by POA? Never heard of that in English dialects, but by no means surprising at all once you think about it. I myself am used to /nw/ > [w] or [wː] between vowels with the preceding vowel being left nasalized, even across word boundaries, in everyday speech, which applies to how I normally pronounce sandwich(es).

Your and some of the preceding comments by Britons are also interesting because they indicate that the variation between /dʒ/ and /tʃ/ in sandwich(es) is found not only in North America but also in the British Isles. Also, it makes the analysis of sandwich(es) more complex, as it raises the possibility that there has always been dialectal variation between /dʒ/ and /tʃ/ in the name of the food item and the place name which its name originally comes from, and that forms containing both /dʒ/ and /tʃ/ could have arrived independently in North America, without requiring any later spelling pronunciation to change /dʒ/ to /tʃ/.

[dhok]

I think the reason people are annoyed is because a lot of the characters show up as boxes.

[Boşkoventi]

Your and some of the preceding comments by Britons are also interesting because they indicate that the variation between /dʒ/ and /tʃ/ in sandwich(es) is found not only in North America but also in the British Isles. Also, it makes the analysis of sandwich(es) more complex, as it raises the possibility that there has always been dialectal variation between /dʒ/ and /tʃ/ in the name of the food item and the place name which its name originally comes from, and that forms containing both /dʒ/ and /tʃ/ could have arrived independently in North America, without requiring any later spelling pronunciation to change /dʒ/ to /tʃ/.

From the Online Etymology Dictionary, via dictionary.com:

cabbage
mid-15c., caboge, from M.Fr. caboche "head" (in the Channel Islands, "cabbage"), from O.Fr. caboce "head," from L. caput "head" (see head). Introduced to Canada 1541 by Jacques Cartier on his third voyage. First written record of it in U.S. is 1660s. The decline of "ch" to "j" in the unaccented final syllable parallels the common pronunciation of spinach, sandwich, Greenwich, etc.

-- Linked from the entry on "sandwich":

The family name is from the place in Kent, O.E. Sandwicæ, lit. "sandy harbor (or trading center)." For pronunciation, see cabbage.

So it looks like the form(s) with /dZ/ is/are old enough to have come over from England. And in the case of "spinach" and "cabbage", there seems to have been variation even in French (perhaps under influence of the suffix "-age"?).

Also, I've never heard of pronouncing "sandwich" (or any of the other words) with /dZ/. Pure madness.

[Travis B.]

Also, I've never heard of pronouncing "sandwich" (or any of the other words) with /dZ/. Pure madness.

In my dialect, spinach always has /dʒ/ and as stated before sandwich normally has /dʒ/ when dialect borrowing and spelling pronunciation do not apply, but Greenwich has /tʃ/. In the case of Greenwich, though, it is certainly a learned pronunciation not native to my dialect, which really has no "native" pronunciation of Greenwich.

[AnTeallach]

Okay, this is useful information, particularly because it indicates that the /t/ > /d/ in congratulate is something that happened in North America and not something inherited from dialects in the British Isles. On a somewhat off-topic note, it is interesting that you have /nd/ > /ŋ/ or /ŋɡ/ (I cannot tell which you mean from your post) in the British Isles; I had not heard of this before.

I think this is just an elision of the medial /d/ followed by POA assimilation of the /n/ to /ŋ/. In my dialect /ŋ/ is not phonemic, only appearing due to POA assimilation of /n/, and I say 'sandwiches' as [saŋwɪtʃʰəz] (although both [tʃʰ dʒ] sound normal to me).

Interesting that you assimilate /n/ to [ŋ] before /w/; I have [m] (and as far as I can tell phonemic /m/) in this word: /samwɪdʒəz/.

[finlay]

And as for the original question, we don't tend to get congradulated in britain, kindergarten is a foreign word that yanks use, but sandwidjes is sometimes heard – usually it's more 'sangwidjes', though, to rhyme with 'languages'. I don't know what the geographical restriction is on that one, but I've heard my Scottish father saying it.

Okay, this is useful information, particularly because it indicates that the /t/ > /d/ in congratulate is something that happened in North America and not something inherited from dialects in the British Isles. On a somewhat off-topic note, it is interesting that you have /nd/ > /ŋ/ or /ŋɡ/ (I cannot tell which you mean from your post) in the British Isles; I had not heard of this before.

That's rather misleading, since it's not a general sound change; it's limited to a single word – presumably the influence of the [w] and/or the word 'language' – and isn't said by everyone. I still say it fairly normally (at least, with /tS/), although it's actually something like /samwItS/ – again, the influence of the [w], and predictable shortening of the consonant cluster, I think.

(On the subject of weird pronunciations like this, badminton is often pronounced /badmɪŋtən/. I have no idea where this pronunciation comes from – again, it's isolated to this word and quite idiosyncratic.)

[Travis B.]

And as for the original question, we don't tend to get congradulated in britain, kindergarten is a foreign word that yanks use, but sandwidjes is sometimes heard – usually it's more 'sangwidjes', though, to rhyme with 'languages'. I don't know what the geographical restriction is on that one, but I've heard my Scottish father saying it.

Okay, this is useful information, particularly because it indicates that the /t/ > /d/ in congratulate is something that happened in North America and not something inherited from dialects in the British Isles. On a somewhat off-topic note, it is interesting that you have /nd/ > /ŋ/ or /ŋɡ/ (I cannot tell which you mean from your post) in the British Isles; I had not heard of this before.

That's rather misleading, since it's not a general sound change; it's limited to a single word – presumably the influence of the [w] and/or the word 'language' – and isn't said by everyone. I still say it fairly normally (at least, with /tS/), although it's actually something like /samwItS/ – again, the influence of the [w], and predictable shortening of the consonant cluster, I think.

I was meaning to suggest that this was a general change, and yes, I suspected /w/ had something to do with it.

(On the subject of weird pronunciations like this, badminton is often pronounced /badmɪŋtən/. I have no idea where this pronunciation comes from – again, it's isolated to this word and quite idiosyncratic.)

I had never heard of this before either. In this case, other than a very strange isolated change the most plausible path I can think of is that the /n/ was originally elided with the vowel being left nasalized, and then the vowel nasalization was reconstituted as [ŋ]...

[Radius Solis]

(On the subject of weird pronunciations like this, badminton is often pronounced /badmɪŋtən/. I have no idea where this pronunciation comes from – again, it's isolated to this word and quite idiosyncratic.)

I had never heard of this before either. In this case, other than a very strange isolated change the most plausible path I can think of is that the /n/ was originally elided with the vowel being left nasalized, and then the vowel nasalization was reconstituted as [ŋ]...

Seems doubtful. I'd rather look to all the English placenames in -ington as a source of analogic pressure.

[Magb]

(On the subject of weird pronunciations like this, badminton is often pronounced /badmɪŋtən/. I have no idea where this pronunciation comes from – again, it's isolated to this word and quite idiosyncratic.)

That's funny, badminton is usually (but not necessarily) pronounced with a [ŋ] in Norwegian too. I remember having been puzzled by that alternation before.

[Alces]

Okay, this is useful information, particularly because it indicates that the /t/ > /d/ in congratulate is something that happened in North America and not something inherited from dialects in the British Isles. On a somewhat off-topic note, it is interesting that you have /nd/ > /ŋ/ or /ŋɡ/ (I cannot tell which you mean from your post) in the British Isles; I had not heard of this before.

I think this is just an elision of the medial /d/ followed by POA assimilation of the /n/ to /ŋ/. In my dialect /ŋ/ is not phonemic, only appearing due to POA assimilation of /n/, and I say 'sandwiches' as [saŋwɪtʃʰəz] (although both [tʃʰ dʒ] sound normal to me).

Interesting that you assimilate /n/ to [ŋ] before /w/; I have [m] (and as far as I can tell phonemic /m/) in this word: /samwɪdʒəz/.

To be honest, I'm not actually sure whether it's [m] or [ŋ], and they could easily be in free variation. [mw ŋw] sound virtually identical, anyway, when not pronounced very carefully.

[finlay]

And as for the original question, we don't tend to get congradulated in britain, kindergarten is a foreign word that yanks use, but sandwidjes is sometimes heard – usually it's more 'sangwidjes', though, to rhyme with 'languages'. I don't know what the geographical restriction is on that one, but I've heard my Scottish father saying it.

Okay, this is useful information, particularly because it indicates that the /t/ > /d/ in congratulate is something that happened in North America and not something inherited from dialects in the British Isles. On a somewhat off-topic note, it is interesting that you have /nd/ > /ŋ/ or /ŋɡ/ (I cannot tell which you mean from your post) in the British Isles; I had not heard of this before.

That's rather misleading, since it's not a general sound change; it's limited to a single word – presumably the influence of the [w] and/or the word 'language' – and isn't said by everyone. I still say it fairly normally (at least, with /tS/), although it's actually something like /samwItS/ – again, the influence of the [w], and predictable shortening of the consonant cluster, I think.

I was meaning to suggest that this was a general change, and yes, I suspected /w/ had something to do with it.

(On the subject of weird pronunciations like this, badminton is often pronounced /badmɪŋtən/. I have no idea where this pronunciation comes from – again, it's isolated to this word and quite idiosyncratic.)

I had never heard of this before either. In this case, other than a very strange isolated change the most plausible path I can think of is that the /n/ was originally elided with the vowel being left nasalized, and then the vowel nasalization was reconstituted as [ŋ]...

It's really not a general change, either of them. They're isolated changes by analogy. Really nothing too complicated.

In any case, your explanation is implausible since vowels aren't nasalised in British English nearly as much as they are in American English, nor do final n's get dropped. That's a French Thing. It would work for NAE but it's very contrived for BrE. Furthermore, you'd have a case if it happened anywhere else at all, which it doesn't. Go for the analogical route... the kind of analytical/mathematical sound change that you're clearly more used to just doesn't always work.

In the case of sandwich, -ŋgw- is on analogy with language, I'm fairly sure, but then I think neither of us would have an easy time proving it since there aren't really any other words where this would apply that I can think of. -mw- (having elided the d) might be a possibly-regular sound change, but note that in some other words with -nw- like Greenwich (in fact, with quite a few words with -wich or -wick this seems to happen in other contexts), the w is elided. (/grɛnɪtʃ/)

[Travis B.]

It's really not a general change, either of them. They're isolated changes by analogy. Really nothing too complicated.

I really had not been familiar with this sort of analogical change in English varieties; this particular sort of thing does not occur in any North American English varieties that I know of, and I had not heard of it before with regard to English varieties in the British Isles, with which I tend to be less familiar.

In any case, your explanation is implausible since vowels aren't nasalised in British English nearly as much as they are in American English, nor do final n's get dropped. That's a French Thing. It would work for NAE but it's very contrived for BrE. Furthermore, you'd have a case if it happened anywhere else at all, which it doesn't. Go for the analogical route... the kind of analytical/mathematical sound change that you're clearly more used to just doesn't always work.

This is as opposed to the English I am familiar with, where nasals and /n/ in particular have taken a route similar to that in Middle French or Old Galician-Portuguese, with vowels unambiguously marking /+nasal/ and a following directly realized /n/ essentially being redundant in most cases and being extremely frequently elided, but still being preserved before plosives other than /t/ and /tʃ/, nasals including /n/, and sibilant fricatives (normally undergoing POA assimilation in the process and, before sibilant fricatives, insertion of an epenthetic stop), before stressed vowels in the same word, utterance-finally, and in careful speech; this applies across word boundaries, where what follows will frequently cause final /n/ to be elided or to undergo POA assimilation and/or insertion of an epenthetic stop.

In the case of sandwich, -ŋgw- is on analogy with language, I'm fairly sure, but then I think neither of us would have an easy time proving it since there aren't really any other words where this would apply that I can think of. -mw- (having elided the d) might be a possibly-regular sound change, but note that in some other words with -nw- like Greenwich (in fact, with quite a few words with -wich or -wick this seems to happen in other contexts), the w is elided. (/grɛnɪtʃ/)

All of this is definitely useful information for me, as it is essentially outside the English I am familiar with or have contact with. I really wish we had more people on here with an in-depth knowledge of the linguistics of Anglic varieties in the British Isles who are willing to post on the subject.

[finlay]

I sometimes elide the n, especially before [ʔ] (so [kãʔ] for can't), but it's variable, and I'm not sure whether I've adopted this from American varieties or if it is present otherwise in Britain. When I was young, 'can't' was [kanʔ] (whether [a] had nasalisation i'm not sure), and was very difficult to distinguish from 'can' in fast speech, annoyingly. I suspect this is why so many scots adopt 'cannae' instead of 'can't', even if they might otherwise not have (m)any 'Scots' forms. (I never did, but meh)

Putting excessive nasalisation on vowels is one way that we start to sound American, FWIW. It's not the case that nasalisation is completely absent, but I just think it's a lot stronger in American English. I can't claim that no Brits drop [n] but I don't know of any such varieties that do it consistently.