Phonemic initial /ð/ in English (and other Germanic)

[Dewrad]

This has been bugging me for, like, ever.

What causes the irregular voicing of initial /θ/ in some words (primarily function words) in English? For example, we have Proto-Germanic *þankōjanan 'to thank', with an initial /θ/. This gives English thank, again with an initial /θ/. We also have PGmc *þū, again with initial /θ/. However, this gives English thou, with an initial /ð/. What causes this?

And it's not just English. It also happens in (continental, at least) Scandinavian: the Norwegian cognates to the two words above are takke and du, with corresponding voiceless and voiced reflexes.

I did think at one point that the conditioning factor was that the words in question were unstressed proclitics, but now I'm not so sure: I don't think subject pronouns in Germanic can be analysed this way. And why only some function words: why doesn't it apply to prepositions like "through"?

[Morrígan]

I did think at one point that the conditioning factor was that the words in question were unstressed proclitics, but now I'm not so sure: I don't think subject pronouns in Germanic can be analysed this way. And why only some function words: why doesn't it apply to prepositions like "through"?

My naive guess would be that it could be based sentential stress patterns, and the relative frequencies of certain items causing them to be re-analyzed as having a voiced onset. This may not be viable, but should be easy enough to falsify.

[Thry]

afaik, the "official" hypothesis is that PG *th- stayed /th/- in Old English and then voiced to /dh/- in function words afterwards - through must be an exception to this; imo, it's the only word I'm aware of that is not followed by a vowel but has /r/ instead, maybe /dhr/- is verboten by the Germanic Gods. Is there any other exception? If not, maybe an unfalsifiable [out of minimal sample set] phonetic environment is the reason xD.

I'm not aware of any theory that relates this to North Germanic, but I see the analogy you point to. Dutch and German are useless for comparison because they leveled everything to /d/-; and looking at the Scandinavian languages I searched for more function words and came across this [yea, continental, but then look at insular]:

Cognates of "then":
Faroese: tá
Icelandic: þá
Norwegian Nynorsk: då
Swedish: då

Cognates of "that/the":
Old Norse: sá
Icelandic: sá
Faroese: tann, tað, sá (obsolete)
Swedish: den, det, de
Danish: den, det, de
Norwegian:
Bokmål: den, det, de
Nynorsk: det, dei

Cognates of "thou":
Old Norse: þú
Icelandic: þú
Faroese: tú
Elfdalian: du
Swedish: du
Danish: du
Norwegian: du

So... two coincident innovations of initial voicing of /th/-? What if North Germanic brought the innovation to English [would that be historically coherent]? Take a look at this:

Cognates of "they" [which is a borrowing]: [...] to Old English þā (“those”) (whence Modern English tho), Scots thae, thai, thay (“they; those”), Icelandic þeir (“they”), Faroese teir (“they”), Swedish de (“they”), Danish de (“they”), Norwegian de (“they”)...

PS: wtf? Why is the alleged reflex of "there" her in Faroese and not **ter or sth?

I don't think subject pronouns in Germanic can be analysed this way.

Well they mostly are so in most modern Germanic languages, so at some dialectal point in time that has to have made sense.

[ol bofosh]

afaik, the "official" hypothesis is that PG *th- stayed /th/- in Old English and then voiced to /dh/- in function words afterwards - through must be an exception to this; imo, it's the only word I'm aware of that is not followed by a vowel but has /r/ instead, maybe /dhr/- is verboten by the Germanic Gods. Is there any other exception? If not, maybe an unfalsifiable [out of minimal sample set] phonetic environment is the reason xD.

/Dr/ would be weird for me, and I can't think of a single example. /Tr/, in the other hand, is very common: three, thread, threat, thrift, throttle, throw, threw, Thry.

[KathTheDragon]

Just browsed a dictionary, and there are no words I would pronounce with initial /ðr/ that I could find.

[Dewrad]

That's because there aren't any: that's not the issue at hand. We know how many words with initial /ð/ there are in English already. The question is why those words?

[communistplot]

My two cents is that initially /T/ became /D/ but some speakers retained /T/ and then, at least in non-function words, those speakers who retained /T/ became the prestigious ones. Trask talks about changes like that in Historical Linguistics (Meat/Mate split in EME). It's also been shown that more commonly used words are more like to undergo changes than less commonly used words.

[kodé]

A relatively simple OT approach could go like this: start with a general lenition constraint favoring [D] over [T] in connected speech, and rank a faithfulness constraint favoring preservation of [T] indexed to content words over it. This would get lenition in function words but not content words. I suspect that prosody is really what's at work, since most function words are extraprosodic. There is some good evidence that there is more pressure to lenite in extraprosodic environments, and more pressure to remain faithful within prosodically parsed material (i.e., positional licensing). Does this make sense?

[kodé]

The prosodic account actually has the nice effect of explaining <through>: /Tru/ is typically parsed into a foot, and thus doesn't get affected by lenition.

[jmcd]

I found this very short pdf: https://www.academia.edu/1720172/Zounds ... le_English
which confirms kodé's idea that it is prosodic-related.

As for Faroese, T>h occurs sporadically in Faroese, as it does in some Norwegian dialects and in Norn. It occurs regularly in some Scots dialects as well as all the Goidelic languages.

[Thry]

That's because there aren't any: that's not the issue at hand. We know how many words with initial /ð/ there are in English already. The question is why those words?

Because they are function words, and because through was the only such word word that would've yielded /dhr-/ and consequently didn't or did and was so unstable it shifted back to /thr-/ - the exception is understandable. If you're asking why function words, or words in a specific prosodic context voiced, ... well, no fucking idea. The possibility is there, and it operated, I doubt the ultimate conditionant is available for us to know for this sound change or for any sound change at all, but it is true that grammatical categories can be a conditionant like any other; cf. Portuguese that keeps high vowels in some pronouns but not in determiners: isto vs este/a(s); tudo vs todo/a(s); aquilo vs aquele/a(s); isso vs esse/a(s)...

My other concern, though: you brought up the fact that *some* North Germanic languages do it too, which I hadn't realized, so now, from my own historical ignorance, I ponder two a priori viable theories: 1) it is monogenic [and as such began in North Germanic and through the historical influence of Old Norse in English it was transferred, and that explains why other West Germanic languages don't exhibit the trait] or 2) it has two unrelated coincident origins in North Germanic and in Old English... I'm waiting for someone to falsify 1) historically; does anybody know whether this is a chronological impossibility? If not, then I'd tend to believe that.

[Dewrad]

That's because there aren't any: that's not the issue at hand. We know how many words with initial /ð/ there are in English already. The question is why those words?

Because they are function words, and because through was the only such word word that would've yielded /dhr-/ and consequently didn't or did and was so unstable it shifted back to /thr-/ - the exception is understandable. If you're asking why function words, or words in a specific prosodic context voiced, ... well, no fucking idea.

Of course that's what I'm asking. If we reformulate the initial question to "why do (most) function words have an initial voiced /ð/?", then the answer "because they're function words" is unhelpfully circular. (It is unfortunate that the only English preposition with an initial reflex of PGmc *þ is "through", because yes, I agree that the likely reason why this didn't become voiced is because of the following rhotic.) I think Kodé's prosodic theory above is the most likely explanation from what we've seen.

The possibility is there, and it operated, I doubt the ultimate conditionant is available for us to know for this sound change or for any sound change at all, but it is true that grammatical categories can be a conditionant like any other; cf. Portuguese that keeps high vowels in some pronouns but not in determiners: isto vs este/a(s); tudo vs todo/a(s); aquilo vs aquele/a(s); isso vs esse/a(s)...

Here is where I disagree with you. Soundchange operates on sounds, not grammatical categories: there must be some phonetic difference which is the conditioning factor. The Portuguese example you cite above isn't a counter-example: it's an example of metaphony conditioned by a no longer distinct final segment. (It is really cool though. Portuguese vowel reflexes are fascinating.)

My other concern, though: you brought up the fact that *some* North Germanic languages do it too, which I hadn't realized, so now, from my own historical ignorance, I ponder two a priori viable theories: 1) it is monogenic [and as such began in North Germanic and through the historical influence of Old Norse in English it was transferred, and that explains why other West Germanic languages don't exhibit the trait] or 2) it has two unrelated coincident origins in North Germanic and in Old English... I'm waiting for someone to falsify 1) historically; does anybody know whether this is a chronological impossibility? If not, then I'd tend to believe that.

This is tricky. If it were monogenetic, we'd expect to see it in the Insular Scandinavian languages as well: ON was introduced to England around 875 CE, roughly the same time as it was introduced to Iceland. On the other hand, the ON spoken in England was (mainly) a East Norse variety, while that introduced to Iceland was a West Norse variety. So, it is possible that this is an East Norse phenomenon, which was brought to England and also spread to continental West Norse (i.e. Old Norwegian), but not to the Insular varieties.

[Dewrad]

Further to Kodé's statement that "there is some good evidence that there is more pressure to lenite in extraprosodic environments", Celtic provides plenty of evidence for this principle. The initial consonant of most prepositions in the Brythonic languages are frequently "permanently lenited": for example the expected outcome of Proto-Celtic *dū 'to' in Cornish would be something like **de, while the actual form encountered is the lenited dhe (Welsh goes a step further and loses the initial consonant altogether, giving i).

[Thry]

Hm. So the possibility exists. For some reason, two unrelated origins don't really convince me.

Here is where I disagree with you. Soundchange operates on sounds, not grammatical categories: there must be some phonetic difference which is the conditioning factor. The Portuguese example you cite above isn't a counter-example: it's an example of metaphony conditioned by a no longer distinct final segment. (It is really cool though. Portuguese vowel reflexes are fascinating.)

I stand corrected then, can I read more on this somewhere you know? That explanation sounds awesome. Okay, I know you're right in that we can break down grammatical sound changes to real sound changes upon which analogy in grammatically-correlated/conditioned environments gives the apparent result of a "grammatical sound change", which isn't a real category, but other than the proclisis thing here I don't have anything more to add. I'd like to point out again, though, that your two main objections [through and subject pronouns not being clitics] can be overcome. Moving on to the second one, why not? Aren't English, Danish, Swedish and Norwegian all non-pro-drop with clitic subjects? Isn't it historically coherent that, by the time /th/ developed the [dh]- allophone in those environments, ik/ek, dú, wiz... and their reflexes were already atonic in North Germanic, and so qualified?

[Radius Solis]

Sound changes are not supposed to be grammar-sensitive, in theory, but it's a bit like the principle that sound changes apply without exception: it does work that way, except when it doesn't. Analogical forces can interfere in either. Sound changes are generally known to spread from less-common to more-common words by force of analogy, usually but not universally make it through all of them; and then force of analogy can act against what it itself wrought and level out grammatical paradigms. Both contribute to potential for the results of a sound change to be sensitive to grammatical category.

English, then, has so many grammatical words in <th>, most of which are extremely frequent, that it would be not at all surprising if they were last to get a sound change affecting initial <th>; and so many fall into the general species "determiner" that it would be not at all surprising if speakers somewhere along the way treated the initial sound as something like a grammatical paradigm. And in fact it was part of a paradigm with w- and h-, in the hither/thither/wither sets, so there's extra reinforcement of it. Put all together and the voicing discrepancy seems less surprising to me.

[zompist]

I was wondering myself about a kind of Trojan horse scenario. (One word can slip into a category and drag a bunch of other in; this often happens with gender.) The/that might be a likely culprit, as they're historically related.

I looked in the OED to see if it had any suggestions, but no. Though perhaps it's relevant that as late as Middle English, you'd get abbreviations like hastow = hast thou, artow = art thou... so, perhaps, thou was still voiceless?

[jmcd]

I see nothing unexpected in those abbreviations pointing to voiceless th. In the first example, the th assimilates in voicing to two preceding voiceless sounds and in the second, it assimilates to one preceding voiceless sound.

[vec]

I'm pretty sure this is a very old variation in Germanic languages.

Icelandic þú becomes -ðu (or -du or -tu) when cliticized and pronouns þeir, þær, þau, það etc. all become voiced intervocalically.

Icelandic is only marginally analyzed as even having phonemic distinction between /þ/ and /ð/. There are no available minimal pairs but a near minimal pair is baðkör /baðkör/ "bathtubs" and maðkur /maþkur/ "worm" and all the cases of unvoiced intervocalic /þ/ are in Greek loan words such as Aþena.

I think this distinction became grammaticalized to various degrees in languages and it seems nearly certain that analogy is one of the key factors at play here.

Also note that they, them etc. are borrowed from Norse and thus may have been voiced all along. This may have helped spread/set in stone the grammaticalization of function words as having /ð/. Since Icelandic has unvoiced þar "there" but Danish has voiced der, I wonder if Norse influence played a part in English there as well.

[Nortaneous]

Sound changes are not supposed to be grammar-sensitive, in theory, but it's a bit like the principle that sound changes apply without exception: it does work that way, except when it doesn't.

They can be prosody-sensitive in ways that make it look like they're grammar-sensitive, as in Burmese, which split some Proto-Lolo-Burmese roots into two different words, a verb with one tone and an adjective/noun with another.

[Particles the Greek]

Further to Kodé's statement that "there is some good evidence that there is more pressure to lenite in extraprosodic environments", Celtic provides plenty of evidence for this principle. The initial consonant of most prepositions in the Brythonic languages are frequently "permanently lenited": for example the expected outcome of Proto-Celtic *dū 'to' in Cornish would be something like **de, while the actual form encountered is the lenited dhe (Welsh goes a step further and loses the initial consonant altogether, giving i).

Another is the second person singular pronoun in Scottish Gaelic, which is thu beside Irish tú. (Or has Irish reborrowed from Spanish, complete with diacritic? Hmmmm.)

[Hallow XIII]

Scottish Gaelic in General is ridiculously lenition-happy. My favorite is the past tense particle do (that caused lenition in the first place) being lenited by analogy where it appears: the past tense of falbh is dh'fhalbh (from earlier do falb or something).