Punic in Proto-Germanic.

Robert Mailhammer at The Conversation (Phys.org) writes about a study of contact between the early Germanic peoples and the Carthaginian empire:

The presence of the Carthaginians on the Iberian Peninsula is well documented, and it is commonly assumed they had commercial relations with the British Isles. But it is not generally believed they had a permanent physical presence in northern Europe. By studying the origin of key Germanic words and other parts of Germanic languages, Theo Vennemann and I have found traces of such a physical presence, giving us a completely new understanding of the influence of this Semitic superpower in northern Europe.

Identifying traces of Punic in Proto-Germanic languages tell an interesting story. Take the words “shilling” and “penny”: both words are found in Proto-Germanic. The early Germanic people did not have their own coins, but it is likely they knew coins if they had words for them. In antiquity, coins were used in the Mediterranean. One major coin minted in Carthage was the shekel, the current name for currency of Israel. We think this is the historical origin of the word “shilling” because of the specific way the Carthaginians pronounced “shekel,” which is different from how it is pronounced in Hebrew.

The pronunciation of Punic can be reasonably inferred from Greek and Latin spellings, as the sounds of Greek and Latin letters are well known. Punic placed a strong emphasis on the second syllable of shekel and had a plain “s” at the beginning, instead of the “esh” sound in Hebrew.

But to speakers of Proto-Germanic—who normally put the emphasis on the first syllable of words—it would have sounded like “skel.” This is exactly how the crucial first part of the word “shilling” is constructed. The second part, “-(l)ing,” is undoubtedly Germanic. It was added to express an individuating meaning, as in Old German silbarling, literally “piece of silver.” […] Similarly, our word “penny” derives from the Punic word for “face,” panē. Punic coins were minted with the face of the goddess Tanit, so we believe panē would have been a likely name for a Carthaginian coin. […]

One area of Carthage leadership was agricultural technology. Our work traces the word “plow” back to a Punic verb root meaning “divide.” Importantly, “plow” was used by Proto-Germanic speakers to refer to a more advanced type of plow than the old scratch plow, or ard. […] The Old Germanic and Old English words for the nobility, for example æþele, are also most likely Punic loanwords. We found Punic also strongly influenced the grammar of early Germanic, Germanic mythology and the Runic alphabet used in inscriptions in Germanic languages, until the Middle Ages.

I grew more and more skeptical as I read, but it’s interesting stuff, and I’ll be interested to see what people say. Thanks, Dmitry!

Update. David Marjanović cites this post by “The Linguistic Physicist,” which begins:

Vennemann’s at it again and with no more evidence than the last time. It’s overblown surface level similarities with no real evidence.

There’s very little evidence of Punic accentuation, but certainly none to demonstrate that Punic lost or reduced the first vowel in šeqel (in fact, Punic seems to have preserved the first vowel in the plural of segolates, whilst Hebrew lost it, i.e. Punic had šiqelim where Hebrew has šqalim). The evidence that Punic had merged š into s before their domination by the Romans is also lacking (although the literature does frequently claim such a merger). Early Phoenician and Punic inscriptions are pretty consistent with which words they spell with s and which š suggesting the distinction was still present, they just don’t always agree with Hebrew […]

And summarizes thus: “It’s nonsense built on nonsense.” Pretty devastating.

Comments

  1. David Eddyshaw says:

    I really must publish on Scandi-Congo.

  2. J.W. Brewer says:

    Before I even clicked through I thought to myself “this sounds like one of those theories that what’s-his-name has. You know, um, that guy.” Took me a little bit before I reconstructed who that guy was, viz. https://en.wikipedia.org/wiki/Theo_Vennemann. Then I clicked through.

  3. PlasticPaddy says:

    Dwds traces pflücken to a late Latin *piluccare “einer Iterativbildung zu spätlat. *pilūcāre ‘aus-, ablesen’ (woraus aprov. pelugar ‘ausklauben, rupfen’), das von lat. pilāre ‘enthaaren’ (zu lat. pilus ‘Haar’) abgeleitet ist.” I do not really understand why plow cannot be derived from the same or a related source with reference to the “bald” aspect of a plowed field.

  4. Lars Mathiesen says:

    I liked the comparison of the first four letters of the Punic abjad and the futhark (if you leave one out from one of them). The shapes are very similar indeed, you just have to ignore the totally different values and the well-documented development of the Greek and Roman alphabets that led to the runes. I think there’s also a gap of a few hundred years to bridge.

    But if the claim is just that the futhark was ordered to make it look like a North Semitic abjad, it does make you go hmm. As far as I know there is no other theory about why that order was used.

    I found the link at the bottom of the article more interesting: A 2017 story summarizing Iversen and Kroonen on a possible Funnel Beaker (Yamnaya? Old European?) substrate in Proto-Germanic, but it’s still paywalled.

  5. Mailhammer

    What sort of a moniker is that? Google/wp seems to know only this dude.

    Is it derived from hammering (chain)mail? (Chain) mail is Latinate; hammer is Germanic; did the Carthaginians bring metalworking northwards as well?

  6. Took me a little bit before I reconstructed who that guy was, viz.,

    A “Vasconic” language family ancestral to Basque is a substratum of European languages, [wp]

    Ah, I see what you mean. It’s Edo Nyland in disguise.

    Edit: (before Hat accuses me of merely sneering): I’m not saying he’s wrong; neither am I saying he’s right; just that this is no more falsifiable than the Faculty of Language or Merge.

  7. squiffy-marie von bladet says:

    Well I for one had been wondering why German had such an unusually rich lexical stock (compared to its geographical neightbours) for describing the manoeuvres of battle elephants, so this is certainly timely

  8. Bathrobe says:

    German had such an unusually rich lexical stock (compared to its geographical neightbours) for describing the manoeuvres of battle elephants

    Care to elaborate? 🙂

    Or is this tongue in cheek! 😀

  9. David Marjanović says:

    and had a plain “s” at the beginning, instead of the “esh” sound in Hebrew.

    That’s completely irrelevant to Germanic, which had no distinction between these until Old English developed one.

    it would have sounded like “skel.” This is exactly how the crucial first part of the word “shilling” is constructed.

    I must say this seems to work better than the traditional derivation from “shield”, which doesn’t explain, as far as I’ve noticed, where the *d went.

    “plow”

    Works beautifully as far as I understand, but I’ve also read somewhere that the plough-as-opposed-to-ard is an Early Medieval innovation that spread, with the word, from Slavic. Unfortunately there were no details about the Slavic side.

    We found Punic also strongly influenced the grammar of early Germanic

    I haven’t read this article yet, but in Vennemann’s own work (on academia.edu) all I’ve seen is the idea that SVO is VSO with the subject fronted for emphasis. Fair enough, but SVO is also all over Romance (notably unlike Latin), and it’s the default in Slavic…

    Dwds traces pflücken to a late Latin *piluccare “einer Iterativbildung zu spätlat. *pilūcāre ‘aus-, ablesen’ (woraus aprov. pelugar ‘ausklauben, rupfen’), das von lat. pilāre ‘enthaaren’ (zu lat. pilus ‘Haar’) abgeleitet ist.”

    That would be the only case in all of West Germanic where *i…u, as opposed to *u…i, underwent umlaut. Unless of course we take the unattested *piluccare and assume an even more unattested spontaneous metathesis to *puliccare… not impossible by any means, but.

    But if the claim is just that the futhark was ordered to make it look like a North Semitic abjad, it does make you go hmm. As far as I know there is no other theory about why that order was used.

    Seconded.

    What sort of a moniker is that?

    Probably folk etymology of -hamer, from a place name in -ham (Bavarian for Heim and very common). On the Mail- part I’m out of suggestions, though, beyond Meile “mile” which isn’t terribly convincing.

    A “Vasconic” language family ancestral to Basque is a substratum of European languages,

    That part is completely reasonable: agriculture arrived in Europe in a massive immigration from Anatolia, which can be expected to have spread a single language family all over the place; the most parsimonious assumption is that Basque is the last remnant of that, and indeed the agricultural vocabulary of Basque is not IE (…but instead looks cognate to those of Caucasian and Burushaski). In other words, the closest living relative of whatever was spoken in the Funnel Beaker culture is most likely Basque.

    Where Vennemann runs into trouble is his implicit assumption that Basque has changed extremely little in the last several thousand years so that hydronyms and other words all over Europe can be explained from basically the modern language.

  10. Haribo delenda est?

  11. Jen in Edinburgh says:

    I was reading at the weekend about an elephant which Charlemagne took with him to a war against the Danes, which died in what is now Germany on the way…

  12. Big chunks of the book this article previews, The Carthaginian North, the Semitic influence on early Germanic, are in google books.:
    https://books.google.com/books?id=iDG2DwAAQBAJ

    I’d have guessed aethele and Latin aedile were cognates. I see that’s not accepted. Is it implausible? Aedile is said to go back to P. It. “aeth” = fireplace related to OE ad and OHG eit = pyre. I don’t know the sound change laws. Do they not allow aethele from a root that could also give ad and eit, so that aethele would be a hearth-official or keeper of a sacred flame?

  13. Lars Mathiesen says:

    tongue in cheek — it’s Des, his cheeks are well developed.

    Getting bored in Corona times, are we, Des? You seem to be popping in at magnitudes over your recent once-a-year rate. You’ll be updating dairyland next, I shouldn’t wonder.

  14. Lars Mathiesen says:

    I usually see shilling derived from *skiljanaⁿ, not *skelduz, Modern Danish has the concept of skillemønt = ‘change,’ small denominations to make up a difference, forskel (Sw skillnad) — but in the 10th the shilling was the largest coin minted, so that’s probably not relevant to the etymology no matter how attractive it is to a Dane.

    So can anybody get at the Iversen and Kroonen and see if their substratal plant names look Basque?

  15. AJP Crown says:

    Des is on Twitter, most days.

  16. before Hat accuses me of merely sneering

    Oh, I have no objection whatever to your sneering at Theo Vennemann or Edo Nyland!

    Des is on Twitter, most days.

    You know who else is on Twitter, most days? Come back to the real world, Des!

  17. Lars Mathiesen says:

    Does Twitter do RSS feeds?

  18. I have a review of the book this is based on that should be appearing at some point. It’s very interesting, but the overall thesis is not convincing — to the point that, given that this is the refinement of decades of research by Vennemann coupled with the considerable linguistic expertise of Mailhammer, and so about the ‘best case’ argument we could reasonably expect, I’m happy to firmly rule out any meaningful Punic influence on Germanic fairly definitively. Which is kind of a shame, really — I think this would have been rather a cool thing to be true.

    I must say this seems to work better than the traditional derivation from “shield”, which doesn’t explain, as far as I’ve noticed, where the *d went.

    I’m not sure I particularly believe the ‘shield’ etymology, but in terms of strict phonological development that’s not an issue — there’s a regular assimilation of *dl (as it was post-Verner) to *ll, the traditional example of which is stall nearly = stabulum (and Middle Welsh ystadl) *skillinga-, granting that preform.

    Of course, there’s the question of the *l in the suffix in the first place, but Mailhammer & Vennemann’s own explanation for the double *ll is itself not really watertight. In this piece, Mailhammer seems to just assume a suffix *-ling-, but in the book they explicitly reject this in favour of gemination to close a light open syllable — which is conceivable, but not paralleled in Germanic (*kuningaz ‘king’, etc.), and so a bit ad hoc.

    On the whole, derivation directly from the root *skel- ‘split, divide’ seems better. The idea of a coin as a ‘piece’ is a standard semantic origin for currency, and might be helped along by the widespread use of hack silver among early Germanic speakers. The *ll is still a problem though, and one I don’t have a perfect answer for. A suffix *-linga- or *-ninga- would do the trick formally, but that’s only pushing the issue into the derivational morphology.

  19. Thanks, that all sounds reasonable.

  20. David Marjanović says:

    a regular assimilation of *dl (as it was post-Verner) to *ll

    Oh yes, I forgot. And I should have mentioned that I only know -ling (somewhat productive in modern German) as a much later reanalysis of -ing.

    I’d have guessed aethele and Latin aedile were cognates. I see that’s not accepted. Is it implausible?

    Yes; Germanic *a and Latin ae are not compatible (you’d expect Germanic *ai > Old English ā), Germanic *þ [θ] aren’t compatible either (you’d expect Germanic *d or *t).

  21. AJP Crown says:

    You know who else is on Twitter, most days?

    Des wants him to relocate to the Low Countries; something about cloud cover, skin cancer and prince of Orange. The late and former keiser would be a precedent for the president.

  22. David Marjanović says:

    America first – the Netherlands second.

  23. John Cowan says:

    Five ways to get Twitter feeds as RSS. I haven’t tried any of them.

  24. AJP Crown says:

    The beginning of this piece by Nailhammer reminds me of the Oxford Professor of Ancient History Josephine Quinn’s book, In Search of the Phoenicians, and a damned good book it is:

    The Phoenicians had a second wind as well, out in the Atlantic. Just as nation states were beginning to emerge in the sixteenth century CE, they become unexpected national heroes, first in Britain and then in Ireland. These were islands that as far as we know ancient Phoenician sailors never reached, but they were lands looking for origin myths. Some British scholars looked to King Arthur, other to the Trojans. And some hit on the notion that it was actually the Phoenicians who originally colonised Britain, or that Phoenician settlers in the South combined with Germanic settlers in the North: this explains, one author tells us, why the Scots are so much larger and fiercer than the English. The connection with the Phoenicians was particularly useful for the British in the seventeenth century because it differentiated them from their great enemies the French, who were always more associated with the Romans as a land-based, territorial power, while the Phoenicians, like the British, were famous sailors with a maritime empire. In Ireland by contrast an imagined relationship constructed with the Phoenicians in the eighteenth century was all about a very real relationship with Britain: the idea that the Irish were descended from Phoenicians cast the British occupation of Ireland in terms of the great struggle during the Punic Wars between sophisticated, noble Carthage and the savage imperial power of Rome.

    More here:
    https://press.princeton.edu/ideas/all-at-sea-the-maritime-lives-of-the-ancient-phoenicians
    and it’s available from Amazon at the very reasonable price of £9.99 or something like that, in hardback.

  25. David Marjanović says:

    Nailhammer

    I see what you did there. 🙂

  26. @David Marjanović: While Woody Guthrie was not so much known for his wordplay, he did have this one number, which I just love the end of:

    So I’m a gonna tell you people: If Hitler’s gonna be beat
    The common working people has got to take the seat
    In Washington, Washington

    And I’m gonna tell you workers, ’fore you cash in your checks
    They say “America First,” but they mean “America Next”
    In Washington, Washington.

  27. And I should have mentioned that I only know -ling (somewhat productive in modern German) as a much later reanalysis of -ing.

    Indeed. The usual view is that -ling often incorporated the diminutive *-l- as well, especially given the typical semantics of -ling later on. This would work rather nicely for shilling semantically, except that this reanalysis seems to be wholly post-Proto-Germanic. There are a few other words that might be early formations (Gothic gadiliggs ‘cousin’, OHG gatiling ‘relative’, etc., is the only option with a Gothic cognate), but a really productive -ling only seems to be extracted rather later. So either shilling is one of the very earliest examples of derivation of this type, or it has a different etymology.

    Or, I suppose, it could be a late-ish formation that spread around. In Gothic, it’s only found in the Naples and Arezzo deeds, which only date from the middle of the 6th c. I wonder if this could have been coined (heh) after Proto-Germanic, and spread with the adoption of more systematic coinage technology (or with the popularity of some particular type or view of currency) between differentiating languages and dialects. Gothic otherwise uses skatts for units of currency. It’s only a vague possibility, but it would at least allow the word to be much younger (probably as young as the 5th c., if we wanted), and so make the presence of a -ling suffix less problematic.

    Looking at Casaretto, I’m also reminded that when we do find it, -ling seems very distinctly denominal in earlier Germanic. So that would be an objection to derivation straight from the root *skel-. Maybe we could propose some sort of (later completely lost) nominal derivative as an intermediate step: *√skel- ‘split’ -> *skel-laz ‘??divided thing, chunk’ (perhaps *skell-ingaz ‘piece’, or the like. This could also be a way to smuggle in the second *l without needing such an early *-ling- suffix.

    Difficult word, anyway.

  28. David Marjanović says:

    Interesting.

    I wonder if the amount of loans from West to East Germanic, which is of course generally difficult to investigate, has been underestimated.

    The example that comes to mind is… there’s the argument that word-initial *g was [ɣ] in Proto-Germanic because the word “Greek”, plausibly a very early loan from Latin, has a *k throughout the Old Germanic languages (and still in Middle High German). But the vowel in that word is *ē₂, which most likely didn’t exist, but is a Northwest Germanic innovation (not counting “here”, whose [eː] was an allophone of *i in PGmc: Hill 2017). So maybe it’s a loan from Latin into West Germanic, and from there into North and East Germanic…

  29. An additional consideration for PGmc *[ɣ-] is that this must have been the sound-value at least in some West Germanic dialects (clearest in Dutch, of course, but usually also assumed for early OE and OFris. where we get early palatalization of ‘*g’ to what eventually becomes [j]: plausibly, though not certainly, *ɣ > ʝ > j). This could of course be a secondary development instead.

    Interesting idea on Krekos.

    Btw, I should be more careful about checking my posts while they’re still editable. I meant to say that the hypothetical *skellaz could potentially be from *skel-no-, since *no was used to form similar derivatives in PGmc (e.g. OE swefn ‘dream’ and cognates), but I suppose my use of an angled bracket to show derivation made this get taken as code and eaten.

  30. Trond Engen says:

    There are several known derivations from the *skelH- root. What would it take for the “shell” word (< *skal-jo-) to work? Two rounds of umlaut and a derivation with -ling? It was borrowed into Italian as scaglia “chip”, says EtymOnline. This would support the semantics but make the timing of the sound changes very difficult.

    Back to the original post. How can he say that Germanic speakers would treat a final-stress she’kel as **skel-? If borrowed directly from Carthaginian, it must have happened before Germanic got fixed initial stress, and maybe before both Grimm and Verner — or at least when they were synchronic processes. Both syllables should be preserved, and the unstressed *e would become *i. That would yield something like Late PGmc **si[ɣ]él, immune to the loss of unstressed schwa.

    I was going to suggest that without any archaeological evidence of Phoenician presence in Northern Europe, any Semitic borrowing in Germanic should be assumed to be a Wanderwort. It’s not unreasonable to think that it the coin was known as **skel in Central Europe, it would have been identified with the root *skel by Germanic speakers. The most likely mediators of both Wörter and Sachen from the Phoenician sphere to Early Iron Age Germania would have been the Celts of Gaul, or maybe the Ligurians. What would that do to the word?

  31. Umlaut won’t work, both because this wouldn’t be a factor in Gothic at all (unless the borrowing idea actually is right after all), and because *a won’t umlaut up to -i-. There could be a root connection, but the stem formations must be different.

    M&V’s idea is that the *ə was simply elided out by Germanic speakers, who took a sequence *səqél, and interpreted it as /skel/ in their own phonology. If we grant everything else, I think this is reasonable enough for the post-Verner period. By that time, Germanic would not have had a /ə/ in the phonology, and would be used to aligning stress with root initial position.

    (A pre-Verner *səqél would probably have come out as ˣsakil-. A pre-Grimm *səqél- would have become *saɣil-. Or *sakel, *saɣel- if you don’t believe in the raising of unstressed *e. Inherited schwas were retained until sometime after the fixing of intial stress, at which point they either became *a under stress, or disappeared — *ph₂tḗr > *fəþḗr [never ˣfaþḗr, though you can find this anachronistic form in handbooks] > *fádēr; probably pre-PGmc *sokʷ-ə-tó- ‘said’ > *soxəþó- > PGmc *sagda-.)

  32. David Marjanović says:

    this must have been the sound-value at least in some West Germanic dialects

    Absolutely. My point is that I can’t tell if that was an innovation of Proto-West Germanic, or goes all the way down to Proto-Germanic as Moulton thought.

    use of an angled bracket

    I always write &lt; and &gt;.

    If borrowed directly from Carthaginian, it must have happened before Germanic got fixed initial stress, and maybe before both Grimm and Verner —

    Last time I checked, Vennemann assumed Punic influence on Germanic before, during and after Grimm (not sure about Verner).

    Inherited schwas were retained until sometime after the fixing of intial stress, at which point they either became *a under stress, or disappeared —

    Oh, awesome. That seems like it explains a number of things…!

  33. PlasticPaddy says:

    Re celtic borrowing:
    PIE *skey is attested in Proto-Celtic with identical reflex. The modern Irish word sciollán means “flake”, and the án is only a suffix. There is also scealpa, but this looks to me like a borrowing from Latin. Proto-celtic reconstructions I have seen do not have ll for words like coill “forest” and mall “slow” but I would have to check more of these. Old Irish did have ll, as it did for Latin borrowings like cella > cill.

  34. ktschwarz says:

    early OE and OFris. where we get early palatalization of ‘*g’ to what eventually becomes [j]

    Oh, here’s where I can ask this! In the first line of Beowulf,

    Hwæt! We Gar-Dena in geardagum

    I’ve heard an argument that at the time of composition, gear must have been pronounced with an initial [g] in order to alliterate with gar. (Or alternately, both pronounced with [ɣ]?) But when I go to listen to various recordings of Beowulf, everybody seems to pronounce gear with initial [j]. And Wiktionary derives it “from Proto-West Germanic *jār, from Proto-Germanic *jērą (“year”), from Proto-Indo-European *yōro-, *yeh₁ro- (“year, spring”), *yeh₁-”. So is the alliteration argument wrong? Was the Dena/dagum alliteration sufficient, and the g-alliteration unnecessary? Or are [g] and [j] considered close enough for alliteration?

  35. Old English had two allophones of initial *g, probably (at least in early OE) [ɣ] and [ʝ], though at some point the latter became [j]. There was also a phoneme [j]. All three could alliterate with one another — the first two, because of their phonemic relation; the latter two, because of their phonetic similarity (or perhaps identity), and the first and last (as here: ɣɑ:r — jɑ:r) because of the bridge provided by the middle one as part of the poetic conventions.

  36. ktschwarz says:

    Thank you, that makes it clear! So the [g] pronunciation of gear is wrong. And just to see how common this alliteration is, I opened Beowulf to a random page and saw an example with all three:

    1948 gyfen gold-hroden geongum cempan ‘given, adorned in gold, to the young champion’

    where if I understand correctly, gyfen (< PGmc *gebaną) and geongum (< PGmc *jungaz) would be pronounced with [j] and gold (< PGmc *gulþą) with [g], assuming Beowulf is after “some point”.

    And I’m surprised to see “champion” in there, since it’s from French, but I guess English was not completely airtight against French before 1066!

  37. ktschwarz says:

    D’oh, sorry, that’s not “champion” from French, it’s cempa, a doublet, borrowed into Proto-West Germanic from Latin. Its Modern English descendant would be kemp if it had survived. PGmc has way more Latin in it than I thought.

  38. Rodger C says:

    Chemp?

  39. No, kemp is indeed right — palatalization in Old English took place before umlaut, and this was from earlier *kampijōn- > *kampjan- (or the like). It’s the same lack of palatalization as in unkempt (related to comb) or the place-name Kent.

  40. John Cowan says:

    Its Modern English descendant would be kemp if it had survived.

    As it has, of course, in the surnames of all the people, some of them still alive, named Kemp and Kempe. William Kempe was a profit-sharer in the theatre company known as “The Lord Chamberlain’s Men” along with Will Shakespeare, but he left the company around the time the Globe was built. He is best known for his “Nine Days’ Wonder”, in which he morris-danced from London to Norwich in nine days (spread out over several weeks, to be sure).

  41. Lars Mathiesen says:

    Da/Sw kæmpe/kämpe = ‘fighter’ and in Danish now = ‘giant’ and general intensifier via ‘strong fighter’. Listed as a loan from OE into ON, for a change.

    Also the network load balancer at work is a KEMP appliance from Kemp Inc, but I haven’t found who or what it would be named for. It could be a person or maybe somebody snuck in an allusion to Beowulf, in either case it seems to have been written out of corporate history.

  42. Trond Engen says:

    Norwegian too (kjempe), also as an intensifier. I think the intensifier was generalized from kjempestor “giant big” and kjempesterk “strong as a giant/champion”. The relation to fighting isn’t completely lost — there’s a word slåsskjempe “person fond of fighting/prone to getting into fights”. I think I’ve told before that it’s one of those loanwords that can be both masculine and feminine in the dialects in spite of having a prototypically male (not to say masculine) referent. I’ve wondered if that goes back to the Lat. ending -ion.

  43. John Cowan says:

    Punic placed a strong emphasis on the second syllable of shekel

    Wait, what? How could that possibly be known? In Hebrew it bears (and always has borne, as far as I can make out) initial stress.

  44. John Cowan says:

    Was the Dena/dagum alliteration sufficient, and the g-alliteration unnecessary?

    No. The way OE alliterative verse works is that the third stressed syllable sets the alliteration pattern: the first or the second stress or both must alliterate with it, the fourth stress does not. Which is why Richard Wilbur’s “Junk” epically fails as alliterative verse for all its alliteration. It quotes two and a half lines from Waldere:

    Huru Welandes || worc ne geswiceð
    monna ænigum || ðara ðe Mimming can
    heardne gehealdan.

    The first line is a 2-3 alliteration on /w/, the second a 1-3 alliteration on /m/, and the third is going to be a 1-2-3 alliteraton on /h/, even though the second half of the line is not given by Wilbur. Weland is the smith of the Aesir, Mimming a named sword. See Laudator Temporis Acti on this bit.

    But alas! Look at Wilbur’s own lines:

    An axe angles || from my neighbor’s ashcan;
    It is hell’s handiwork, || the wood not hickory,
    The flow of the grain || not faithfully followed.
    The shivered shaft || rises from a shellheap
    Of plastic playthings, || paper plates,
    And the sheer shards || of shattered tumblers
    That were not annealed || for the time needful.

    He alliterates on a vowel, /h/, /f/, /ʃ/, /p/, /ʃ/, /n/ — but his Reimgeber is the fourth rather than the third syllable, with the exception of the penultimate line! What is more, triple alliteration is normally a special effect: it is not to be drummed into the audience’s head like this. “Such epithets, like pepper / Give zest to what you write / And if you strew them sparely / They whet the appetite / But if you lay them on too thick / You spoil the matter quite!”

    Tolkien, of course, knew exactly how to do it:

    Out of doubt, out of dark, to the day’s rising
    he rode singing in the sun, sword unsheathing.
    Hope he rekindled, and in hope ended;
    over death, over dread, over doom lifted
    out of loss, out of life, unto long glory.

    with alliterations on /d/, /s/), /h/, /d/, and /l/. This one does have a lot of triple alliterations, but it’s short and a minstrel’s showpiece. His translation of Sir Gawain has 2532 lines of which 2027 are proper alliterative lines, and not always the same alliterations as in the original. Take the first two lines:

    Siþen þe sege and þe assaut watz sesed at troye
    þe borȝ brittened and brent to brondez and askez

    Tolkien’s equivalents have the same alliteration in the first line, but are completely different in the second:

    When the siege and the assault had ceased at Troy,
    and the fortress fell in flame to firebrands and ashes,

    where the loss of burgh/borough in the sense ‘fortified place’ and brit(tened) ‘broken into pieces, shattered, demolished’ has forced Tolkien to find all-new words. He has, however, preserved the grace-note (of no metrical significance, but noticeable and an especial feature of Sir Gawain) that one of the unstressed syllables also alliterates: borȝ is replaced by its quasi-equivalent fell.

    English was not completely airtight against French before 1066

    By no means. A few such late-OE words that may be from pre-OF are castle, clerk, mantle, purse, trail, turn. Etymological dictionaries tend not to mention French in this connection, just saying “Vulgar Latin”, but they have a more Frenchy flavor than most of the direct Latin borrowings, not to mention a different subject matter for the most part.

  45. David Marjanović says:

    Kämpe is found as an archaism in very literary/poetic German.

    Kent used to be Cantium.

    PGmc has way more Latin in it than I thought.

    At least West Germanic does.

  46. The way OE alliterative verse works is that the third stressed syllable sets the alliteration pattern

    This is a common description, but not really true linguistically (and I do not believe true metrically, though views differ on this issue). The most accurate way of putting it is that the first primary-stressed syllables in each half-line must alliterate with one another. The first line of Beowulf only has two primary stresses (gār- and ġeār-), and the hūru of Waldere is not stressed (or at least not enough to count for metrical purposes). These are rounded out with a secondary stress in each half-line (which in this case bear ornamental secondary alliteration and d-), but this is not necessary and plenty of lines only have a single potentially alliterating stress. Others have up to three primary stresses, or up to four primary-plus-secondary stresses (though the four-stress lines always involve stress clashes). The idea of a four-stress line is broadly true in the average, but very frequently untrue in detail, and at least very plausibly epiphenomenal as far as metrical structure goes (though again, this is a point of debate).

    There’s still good reason to believe that the first stress of the second half-line was compositionally key, the hǫfuðstafr in Norse terminology. It’s just that this isn’t necessarily the third stress in the line as a whole. And double- and triple-alliteration in the second half-line is strictly prohibited in Old English (Middle English poets are often a bit more willing to violate this principle).

  47. PlasticPaddy says:

    @ng, jc
    I think better with examples
    1 Nū scylun hergan hefaenrīcaes Uard,
    2 metudæs maecti end his mōdgidanc,
    3 uerc Uuldurfadur, suē hē uundra gihwaes,
    4 ēci dryctin ōr āstelidæ
    5 hē ǣrist scōp aelda barnum
    6 heben til hrōfe, hāleg scepen.
    7 Thā middungeard moncynnæs Uard,
    8 eci Dryctin, æfter tīadæ
    9 firum foldu, Frēa allmectig

    1 h (4th stress, 1st after caesura)
    2 m (3rd stress, 1st after caesura)
    3 uu (as 2)
    4-5 (interpolation – because original too pagan?)
    6 h (as 2)
    7 m (as 2)
    8 as 4-5
    9 f (as 2)

  48. Both nū and scylun are metrically unstressed, so the first line is a typical illustration of the ‘first stress in each half-line’ rule: nū scylun HERġan || HEfanrīċaes UARD (all main stresses are capitalized; -rī- bore secondary stress). (This kind of rhythm, where only a final trochaic word is stressed and alliterating, is called type A3 in the taxonomy devised by Eduard Sievers, and is only possible in the first half-line of a verse.)

    Lines 4, 5, and 8 all have vocalic alliteration: Ēċi DRCtin || ŌR ā-STElidae; hē Ǣrist SCŌP || AELda BARnum; Ēċi DRYCtin || ÆFter TIAdæ. I can’t see any reason to think these lines are interpolations. That all vowels alliterate with all other vowels is a general rule of alliterative verse. One reasonable idea is that vowel-initial words really began with an allophonic glottal stop, so that we’d have, for instance:

    ˈʔeːcɪː ˈdryxtɪn ˈʔoːr ʔɑːˈstelɪdæː

    (Note that ā- in āstelidae is invisible to the alliteration since it’s unstressed — otherwise we’d have illegal double alliteration in the second half-line.)

    We happily find this rule applied to names from Christian tradition:

    þā wæs Ēve || Ādames brȳd (Genesis A 186)

    The basic rules of Old English alliteration are these:

    1) The first stressed syllable in each half-line must take part in the alliteration.
    2) More syllables can (and sometimes must) take part in the alliteration in the first half-line.
    3) No other syllables may take part in the alliteration in the second half-line.
    4) Only stressed syllables count for alliteration; un- and weakly stressed syllable are invisible.
    5) Alliteration usually involves only the first consonant of the stressed syllable in question (this goes even for things like hn-, hr-, etc. which some have been tempted to take as digraphs for single consonants — these just alliterate as h-, so that heben til hrōfe has double alliteration).
    6) The clusters sp-, st-, and sc- (later becoming a fricative [ʃ]) are exceptions, and count as distinct units for alliteration.
    7) /j/ counts as alliterating, as a special licence, with [ɣ] and [ʝ].
    8) All vowels alliterate with all other vowels.

    The only thing this leaves out is exactly when double alliteration is mandatory in the first half-line (basically the details of rule 2), since this depends on specific issues of rhythm and word-structure. This is also all basically true for Old English and Old Saxon verse, and for the few fragments of Old High German alliterative poetry that have survived, with minor adjustments (e.g. substitute sk for sc; rule 7 does not hold for Old High German, and is not relevant for Norse where all apparent initial j‘s are rather the first elements of rising diphthongs).

    If anyone wants to know more, there’s a very good book on the subject: https://books.google.be/books/about/Old_English_Metre.html?id=ml2c1X1VS9IC

    It doesn’t get into all the controversial questions of metrical theory (hardly a word on the word-foot theory, for instance), but it’s very good as a descriptive overview and introduction to most of the important issues.

  49. PlasticPaddy says:

    Wow, thanks!

  50. John Cowan says:

    Thanks for the detailed precisions; I was writing in haste, as is too common in my corona-era life, and just trying to get the basics down.

    I would point out that English has always had a huuuuge tolerance for mucking with the syllable counts, even in what appears to be foot-verse: the “unstressed syllables don’t count” has carried through even into Romance-ified pre-20C Modern English verse. In Sayers’s translation of Dante, we get the line “Then the terror of alighting seemed worse than the terror of soaring”, which has only five stresses but 12 unstressed syllables instead of four, five, or six, given the iambic pentameter context. On the othe hand, stress demotion is a big feature too, as in Pope’s well-known lines: “When Ajax strives some rock’s vast weight to throw /The line too labours, and the words move slow”, where the meter has deprived vast and move of their normal status as stressed monosyllables. (In performance, I slow down on these lines, and then go uptempo with “Not so, when swift Camilla scours the plain / Flies o’er the unbending corn, and skims along the main” before returning to normal tempo.) Still more demotion, of course, appears in Pope’s demonstration of crappy iambic pentameter: “And ten low words oft creep in one dull line.”

  51. Lars Mathiesen says:

    ek hleu̯aɣastiʝ holtii̯aʝ | hornã tau̯iðo

    Deyr fé | deyja frændr
    Deyr sjalfr | it sami
    Ek veit einn | att aldregi deyr
    Dómr of | dauðan hvern

  52. David Marjanović says:

    (Note that ā- in āstelidae is invisible to the alliteration since it’s unstressed — otherwise we’d have illegal double alliteration in the second half-line.)

    I suppose it’s also possible that, because it was unstressed, it didn’t begin with [ʔ]. In the northern half of German today, [ʔ] is not inserted before vowel-initial words, but before vowel-initial stressed syllables: [ˈʔ]Astero[ˈʔ]iden und Ko[ˈ]meten.

    The first 5 pages of this pdf of a conference handout contain a neat description of the situation in Old English, and what became of it in Middle English: [ʔ] was mostly lost, and poets started using word-final consonants for alliteration if the following word began with a vowel.

    (In southern German, [ʔ] is inserted only before vowel-initial utterances; it’s a postpausal voice onset. Unlike in French or some Englishes, it’s obligatory in this last position, though.)

  53. Or quite possibly still:

    ek xlewaɣɑstiz xoltijaz xornã tawiðoː

    The length of the final vowel is pretty likely still. The exact chronology of the weakening of initial *x and the beginnings of palatalization/rhoticization of *z are unclear. I’m inclined to think by the 5th century the sound had developed to something beyond *z, while *x might still be realized as such, but it’s awfully hard to be sure of the details.

    Hávamál is probably old (or at least made of old parts — I can’t imagine it was transmitted in anything like its Frankenstein’s-monster-esque final form in an oral tradition), and that famous stanza is echoed in the Hákonarmál, in a way that’s pretty clearly adapting well-known words, so we’re probably looking at the late Viking Age at least for these lines. Here’s an impression of what that might have sounded like at that time (using ʀ for the outcome of *z that was widely perceived as /r/ by non-Norse speakers, but was distinct from etymological *r):

    døyʀ feː døyjɑ̃ friændʀ
    døyʀ si̯ɑlvʀ hɪt sɑmɑ
    ek wæit æinn ɑt ɑldræiɣɪ døyʀ
    doːmʀ ʊ̃v dɑuðɑn xʷern

    One of the nice things about alliterative verse is that the alliterations, at least, tended to survive the kinds of sound changes during these periods pretty well. Some exceptions with cluster simplifications, but a lot of this poetry can be appreciated in its manuscript form reasonably well (or even in modern Icelandic), though in some cases the overall phonology will have changed a bit between the probable dates of composition and the creation of the surviving MSS.

    The caesuras in the odd lines are not at all clear — maybe a minor point, but possibly important for the overall structure of this verse type. This is ljóðaháttr, which has a normal-ish set of paired half-lines, followed by a ‘full verse’, which is less long than a typical line, but longer than most half-lines. This has internal alliteration, but its internal breaks may not be the same — in my view, full-verses are more akin to half-lines than to lines (they are similar in some important ways to Old English hypermetric lines, though there are important differences). Insofar as there is an internal boundary, I’d put it before sjalfr rather than after (among other things, the run of unstressed syllables in the initial upbeat can be expanded fairly significantly, while unstressed syllables between the two alliterating peaks are much more strictly regulated).

    Possibly more representative of the style of the earliest Germanic verse would be something like Atlakviða:

    Eldi gaf hon þá alla, er inni vóro (vocalic)
    ok frá morði þeira Gunnars komnnir vóro ór Myrkheimi; (M)
    forn timbr fello, fiarghús ruko, (F)
    bœr Buðlunga, brunno ok skialdmeyiar (B)
    inni, aldrstamar, hnigo í eld heitan. (vocalic; the double H in the second half-line is artistic flourish)

    Or, with the same late-Viking-Period facelift (though it’s really hard to know if the poem is in fact actually that old):

    ʔældɪ ɣɑf hʊn θɑː ʔɑllɑ, eʀ ʔinnɪ woːʀʊ̃
    ok frɑ̃ː morðɪ θæiʀɑ ɣunnɑrs komnɪʀ woːʀʊ̃ oːʀ myrkhæimɪ
    forn timbr fellʊ̃ fi̯ɑrɣhuːs rukʊ̃
    bœːʀ buðlʊŋgɑ brunnʊ̃ ok ski̯ɑldmøyjɑʀ
    ʔinnɪ, ʔɑldrstɑmɑʀ hniɣʊ̃ ĩː ʔæld hæitɑn

    (Possibly some mistakes there — did this fairly quickly.)

  54. I suppose it’s also possible that, because it was unstressed, it didn’t begin with [ʔ].

    Oh, absolutely. There’s a lot we don’t know, or don’t know very confidently, about glottal stops in earlier Germanic.

    In this case, just to be clear (though I don’t think you were suggesting otherwise), that’s not an issue for alliteration. Other prefixes such as be- and for- show the same behaviour and are invisible to alliteration as unstressed verbal prefixes.

    Thanks for the link to Ricardo’s handout — his work on these things is always worth reading!

  55. Lars Mathiesen says:

    What is the difference between {j} in døyjɑ̃ and {i̯} in si̯ɑlvʀ?

    The thing about carrying over consonants to vowel-initial words reminds me of the underreported phenomenon of enlace in Spanish — संधि by another word; even the compendious WP article on Spanish phonology gives it only passing mention and links to the only page on the Internet with an explanation. (At least that’s the page where I always end up if I try looking for more details). Regardlessly it’s the feature of spoken Spanish that most trips me up.

  56. It’s phonological: /j/ is a consonantal phoneme, ‘i̯’ only occurs as the first element of what were rising diphthongs in later Norse. Exactly how this played out in real pronunciations at various times is a bit difficult to know. Certainly the diphthongs continued to alliterate as vowels, so jǫrð alliterates as a vowel, not as a consonant /j/ (which would only be from rising diphthongs anyway, since Germanic initial *j was lost in North Germanic in general). It’s possible that phonetically i̯ and j are indeed exactly the same (with the former preceded by glottal stops when fully initial), but it’s also possible that the dynamic movement in the diphthong made for a realization of the initial element that was noticeably different from /j/. Hard to say.

  57. David Eddyshaw says:

    Agolle Kusaal has a contrast between word-initial /j/ (ya “houses”) and /i̯/ (ia “seek”) which I think is realised by /j/ being both tenser and shorter in duration. The /i̯a/ actually patterns like a short monophthong (which is what corresponds to it in the Toende dialect) and I did myself play at one stage with the idea that the contrast was /ja/ versus /ʔja/, but that proved very difficult to make work consistently; there was no other reason to posit a contrastive /ʔ/ word-initially, and no real phonetic support for it anyway.

    Romanian has a similar initial /j/ ~ /i̯/ contrast. Interestingly, neither Agolle Kusaal nor Romanian has a contrast of initial /w/ ~ /u̯/; I came across a paper which stated that this is cross-linguistically usual, as there isn’t the phonological leeway there at the back of the mouth to maintain a similar contrast easily.

  58. David Eddyshaw says:
  59. David Eddyshaw says:
  60. David Eddyshaw says:

    (In my 7:56 comment I shouldn’t have said initial contrasts in Romanian; I was falling into a Kusaal-centric viewpoint, where there are no word-internal /Cj/ consonant clusters in any case, so the issue of contrast only arises word-initially. Not so in Romanian.)

  61. Kusaal-centricity is a long and honored tradition in these parts.

  62. John Cowan says:

    संधि by another word

    I can’t read Devanagari for beans, but context told me that was a (harmlessly) exoticized version of sandhi. Nevertheless, I ran GT over it, which told me it was Hindi and meant ‘the treaty’, which is indeed one of its Skt meanings. Apparently it is a cultismo.

    the only page on the Internet with an explanation

    Fortunately, that page links to a page on enchaînement in French, about which there is a good deal more available. It is clearly the same process, which probably means it is of at least Proto-Romance age and possibly goes back to Classical Latin or even earlier.

    English is another story, though. One of Wikt.en’s definitions of this word is resyllabification, and that led me to this paper, which studies the intelligibility of resyllabified words like my by *kizz for my bike is. It takes longer for anglophones to understand the resyllabified version than the original, and indeed phonetically the unexploded [k̚] at the end of bike is very different from the aspirated [kʰ] at the beginning of *kizz. (The effect disappears when the words are nonsense in the first place.)

    I have always thought that structuralist (small-s) phonology was correct to talk about a “juncture phoneme” in English, something that generative phonology discarded but has not really replaced. There really is a phonemic difference between nitrate /naitrate/, night rate /nait+rait/, and the Nye trait of pottery, something better known to linguists (of a certain age) than to archaeologists, it seems.

  63. David Eddyshaw says:

    I have always thought that structuralist (small-s) phonology was correct to talk about a “juncture phoneme” in English

    Definitely, and not just in English. I think you always can avoid setting up juncture phonemes with sufficient ingenuity; but why should you, when they often lead to a much simpler and clearer description? (I suppose that if you’re committed to linguistic subPlatonic realism, as some people are, it’s not so simple.)

  64. David Marjanović says:

    That’s the main subject of the same handout I linked to, actually. Basically you have to be careful to distinguish a ~ morphophonemic and a ~ phonemic level: the former with word boundaries, the latter with aspiration, active deaspiration measures and pre-fortis clipping and all that jazz.

    why should you

    To be able to interpret it from the hearer’s point of view. The hearer can’t hear word/morpheme/syllable boundaries except by the mentioned kind of cues, which are therefore phonemic…

  65. Prosodicists are happy with juncture marking of various types (including the prosodic word) — these are indeed ‘phonemic’ under a general enough definition (though I suppose really it’s the hierarchical prosodic layers that are meaningful, and the boundary effects are cues not for the junctures as such, but the whole unit well beyond the point of juncture), but whether they are ‘phonemes’ or not is, I think, a question linguists mostly just haven’t been terribly interested in for the past few decades.

  66. Daniel Briggs says:

    AntC, thank you for reminding me of Edo Nyland’s name. I remember seeing his website back around 2001, and it was a lot of fun.

  67. Old English had two allophones of initial *g, probably (at least in early OE) [ɣ] and [ʝ], though at some point the latter became [j]. There was also a phoneme [j]. All three could alliterate with one another — the first two, because of their phonemic relation; the latter two, because of their phonetic similarity (or perhaps identity)

    Another option that comes to mind is that before the [ʝ] > [j] merger, also /j/ may have had [ʝ] as a possible allophone or a conditionally split reflex (with some overlap modulo theoretical approaches).

    I’m reminded of the similar case in oldest Finnish — a [ʝ] allophone of /ɣ/ (before the unrounded front vowels /i e æ/) can be inferred from its eventual development to /j/, but also original etymological /j/ before /i e æ/ is in a few occasional cases spell’d g(h). This actually matches quite well with how the sequences /ji je/ were natively forbidden word-initially (also so was /ɣ/ though, giving e.g. Kiesus as the colloquial reflex of Jesus).

  68. Lars Mathiesen says:

    exoticized — we have three writing systems from the same family in fairly common use here, and Hebrew pops up too, so is नागरी really that exotic? Nobody objects to kana.

  69. I doubt JC was objecting.

  70. Lars Mathiesen says:

    Remarks, then.

  71. Trond Engen says:

    I’m reminded of the similar case in oldest Finnish

    I’m reminded to ask if there are traces of influence from ancient Germanic poetics in Finnish (or Finnic) verse. I know that there’s alliteration in the poetry in the Kalavala tradition, but that’s too widespread to be diagnostic of anything on its own.

  72. David Marjanović says:

    The principle of Internet circularity compels me to link to this take on the proposal.

  73. Well, that’s pretty devastating. I’ll add it to the post.

  74. David Marjanović says:

    I do have to add, though:

    The evidence that Punic had merged š into s before their domination by the Romans is also lacking (although the literature does frequently claim such a merger). Early Phoenician and Punic inscriptions are pretty consistent with which words they spell with s and which š suggesting the distinction was still present, they just don’t always agree with Hebrew

    Who cares? There’s no reason to think either of them would have been borrowed as anything other than /s/ into the Germanic of the time (or long after).

    Some of these words have plausible Indo-European etyma (e.g. shilling < *(s)kelH “to cut, to separate” which given the hacksilver bullion economy of early Germanic peoples makes sense as a source of a name for a coin)

    Oh, that reminds me of Müller’s law: *RH > *RR – that way we can have a shill-ing without having to assume an anachronistic -ling. This is actually perfect.

    or fail to take into account variants within Germanic or in other IE families (e.g. penny appears and panding in many early Germanic languages, and is borrowed into Slavic as *pě̀nę̄dzь again with a d that isn’t accountable for with Vennemann’s claim)

    There is no /d/ in *pě̀nę̄dzь; there is a /d͡z/ recently created from /g/ by one of the palatalizations (I think the Second). In other words, it’s borrowed from *penning.

    I’m not aware of any “panding” form, but this is one of the words with unexplained inner-Germanic variation (irregular dissimilation?) between *-ing and *-ig (penny, German Pfennig, but place name Pfenningberg etc. etc.); another is king (German König).

    His claims about the runes also rely on an unusual method of transmission: that the letter names were translated and the acrophonic principle retained (i.e. aleph “ox” > fehu “ox/wealth” and therefore the f rune looks like a Punic ʔ) but this ignores the much better correspondences with Old Italic scripts if letters were borrowed for their sound value

    In his most recent paper on the subject, Vennemann compromised, saying the runes have most of the proposed origins at once, so that the place of f at the beginning is Punic, but other things are Old Italic or whatever.

  75. David Marjanović says:

    Müller’s law

    *RH > *RR specifically after the accent.

    Müller’s thesis (in German) is here on Google Books; I haven’t read it yet, and probably won’t very soon.

  76. David Marjanović says:

    I lie, of course; I’ve already started. But I might not finish soon; it does have some 300 pages, and they all seem to be included in the preview.

  77. John Cowan says:

    My point was that writing the English borrowing sandhi in the writing system used by the source language is exoticizing, in the same way that writing “I just went out and bought a самовар to make my tea” or “I like to put საწებელი on my shish kabob” would be. It’s harmless because this is Languagehat, where such things are expected.

  78. Lars Mathiesen says:

    Point taken.

    But still many Greek and Russian words (and paragraphs) get quoted here without transliteration, whereas Sanskrit words are almost always transliterated. (Hebrew, Georgian, Runic, Glagolitic, even Gothic pop up on occasion, to stay in the family, but generally with transliteration. I don’t know how to sort them by exoticness).

  79. David Marjanović says:

    they all seem to be included in the preview

    Of course not, LOL. And the author isn’t on academia.edu (though several people with the same name are); I haven’t yet searched ResearchGate.

  80. Халаад says:

    Which reminds me: Hat, is there any way you could set your blog up to show Mongolian traditional script (ᠮᠣᠩᠭᠣᠯ ᠪᠢᠴᠢᠭ᠌) in connected form? It doesn’t need to be vertical, just connected.

  81. Bathrobe says:

    Which reminds me: Hat, is there any way you could set your blog up to show Mongolian traditional script (ᠮᠣᠩᠭᠣᠯ ᠪᠢᠴᠢᠭ᠌) in connected form? It doesn’t need to be vertical, just connected.

  82. Bathrobe says:

    I see that your blog doesn’t accept user names in either Mongolian script (ᠬᠠᠯᠠᠳ᠋) or Cyrillic (Халаад).

  83. It just puts them into moderation; I freed one of them and axed the other, figuring we didn’t need identical comments with varying usernames.

    I can see ᠮᠣᠩᠭᠣᠯ ᠪᠢᠴᠢᠭ᠌ in your comment just fine; isn’t it in connected form?

  84. David Eddyshaw says:

    Seems to be a font issue: it’s all joined up comme il faut when I look at it on my Android tablet, but not on my laptop.

  85. By the way, last time I caught a glimpse of the post title out of the corner of my eye I read it as “Panic in Proto-Germanic.” Which is a great title.

  86. Rodger C says:

    Occasional loanwords via trade can come from relatively distant sources, especially if transmitted thru some intermediaries. (There are, for example, some half a dozen loanwords originating from Latin in Proto-Finnic.)

    Hence, no doubt, my favorite palindrome, saippuakauppias ‘soap merchant’.

  87. Bathrobe says:

    @Hat

    You’re right. It shows up fine on my iPhone but not on my Macbook.

    Apple has always been crappy at accommodating Mongolian script. iOS regularly breaks it.

  88. Lars Mathiesen says:

    saippuakauppias: that was the longest palindrome in the Guinness book of World Records that I had almost 50 years ago — before it got all glitzy and non-nerdy. I remember that and the heaviest twins in the world on their motorbikes. (The McCrary twins, stage name McGuire, but I had to look that up right now).

  89. John Cowan says:

    There is this Ancient Custom whereby historical linguists (but not other linguists) are assumed to be able to transliterate Greek in their heads, whereas other scripts they (are assumed to) need transliterations for.

    Of course, running text as opposed to single words in any language shouldn’t be transliterated at all. Unfortunately, WordPress has changed so that it doesn’t like posts with mixed scripts unless one script thoroughly dominates, so people posting comments like that need to transliterate where they normally would not.

  90. Lars Mathiesen says:

    I thought the HLs were supposed to be able to read Greek, sencillamente. Like three (3) pages of Homer before bed innit.

  91. David Marjanović says:

    HLs were also supposed to be able to read Devanāgarī and Avestan, and maybe Armenian too, but then the European publishers lost the ability to print those (probably in WWI) and transcriptions have been used ever since.

  92. David Eddyshaw says:

    were supposed to be able to read Greek

    Πῶς γὰρ οὔ;

  93. Lars Mathiesen says:

    So never mind the script, how many HLs are working now who can just read Latin, Greek and Sanskrit for sense? (Beside English, French, German, Polish, Russian and Italian for reference works). I mean, that used to be a barely acceptable starting point.

  94. David Eddyshaw says:

    Polish was probably pushing it (at least for non-Poles) even in the Heroic Age; maybe even Russian. The essential modern languages were probably German, French, English (in that order …)

    I’d say German and French are still necessary. (French, at least, is certainly still very necessary for African comparative work, for obvious reasons.)

    The big loser seems to have been German. As I’ve moaned before, linguistic works which a generation ago would have been in proper highfalutin scholastic German are now published in not-quite-correct badly-proofread EU English instead. (I’m looking at you, Brill.)

  95. David Marjanović says:

    On the IE side, Jerzy Kuryłowicz wrote in French (and published a lot).

    At least half of Tungusic linguistics has of course been done in Russian, to the extent that just two years ago a fellow named Eero Talvitie wrote a thesis on the Tungusic long vowels (I think it was) in Helsinki in Russian and even followed some of the literature in using Cyrillic letters with diacritics to represent Proto-Tungusic. The rest of Tungusic linguistics is in Chinese, Japanese, some French, and then English and maybe German.

  96. David Eddyshaw says:

    Russian is certainly necessary for that whole area.

    Chinese, Japanese and even Russian differ from the others in that they have never been international languages of modern scholarship (however worthy of it they are in themselves), so the effect of publishing in those languages exclusively is academic ghettoization. For some scholars, I suspect this is a feature rather than a bug.

  97. ə de vivre says:

    I have a theory that Assyriology remains so stubbornly multi-lingual (not just because it’s a small, out-of-the-way discipline and so more inclined to methodological conservatism) because so much of the primary sources are housed in specific museums: If you want to study Ebla, you have to go to Italy. Lagash was mostly excavated by the French. German, English, French, and Italian are the only languages you’d ever need because those were the languages of the only European countries that got their hands on Mesopotamian antiquities in any appreciable quantity. The governments of Turkey and Iran encourage archeological research in their territory, and English wins out there because lord knows you’d never confuse an Oriental source language for a real European tool language. But the political situation in Iraq has prevented this from happening for the study of sites in its territory.

  98. John Cowan says:

    Were Europeans on the dexter side of the Iron Curtain publishing in Russian at all?

  99. I’ll add it to the post.

    The quote you’ve added, though, is not by me, it’s reblogged from Tristan “The Linguist Physicist” from Cambridge/Londonish. My own comments on @possessivesuffix begin with “Far from the worst nonsense…” after the link to the article.

    (Yes, tumblr has been doing something weird with the order of comments in reblogs lately for some reason…)

  100. Woops, sorry about that — I’ll fix the addition!

  101. (Thanks, Hat.)

    the longest palindrome in the Guinness book of World Records that I had almost 50 years ago

    The Finnish lists for that usually give instead saippuakivikauppias ‘soapstone merchant’, but this already starts going down the chasm of words that have only been coined for their palindromic properties instead of having ever been used in authentic communication. A vivid example from a bit deeper down that I recall is tomaattimatokotamittaamot ‘tomato worm ?? measuring stations’, which parses fine syntactically but where it’s not even clear what sense of kota would be intended (something akin to siemenkota ‘seed pod, apple core, etc.’?)

    just two years ago a fellow named Eero Talvitie wrote a thesis on the Tungusic long vowels (I think it was) in Helsinki in Russian

    Slight extenuating circumstances: Russian is his first language and IIRC he adopted the name Eero Talvitie only after moving to Finland. But that Russian was thought of as an option at all does count for something.

    Basically every Uralic group other than Hungarian or Sami also has a long tradition of research published in Russian. For a few particular languages (Karelian, Komi, Selkup?) I think it has not even been the second fiddle to German or English. French is meanwhile way down the list, maybe at seventh place behind Estonian and falling.

  102. AJP Crown says:

    J Pystynen, Soapstone is greenish, quite like granite in texture but with veining like marble and good for kitchen counters (where marble erodes over time) and fireplaces. I’ve needed a soapstone merchant occasionally. In that case, granite or marble from any old mason or stoneyard wouldn’t do. When you need soapstone, you REALLY need it. I’ve also worked in Finland as an architect. The two have never coincided; however, that’s just chance. I’m sure other people are desperate to find for soapstone merchants in Helsinki and Turku all the time. So it’s a perfectly straightforward palindrome, in my opinion.

  103. David Marjanović says:

    Chinese, Japanese and even Russian differ from the others in that they have never been international languages of modern scholarship

    Russian has been, though only behind the Iron Curtain (so international more in name than otherwise), and even there it was never as exclusive as English is today.

    In the mid-20th century, the Chinese journal Vertebrata PalAsiatica occasionally had papers in Russian.

  104. David Eddyshaw says:

    Russian has been, though only behind the Iron Curtain

    Fair point.

  105. David Marjanović says:

    Müller’s law

    *RH > *RR specifically after the accent.

    ….and even more specifically only when a vowel followed, because *-CHC- > *-CC- happened earlier, finds Müller. In short, *ˈ(-)VRHV- > *(-)VRːV-.

  106. Bathrobe says:

    In Japanese 英仏独 ei-futsu-doku ‘Britain, France, and Germany; English-French-German’ is a kind of mantra, although it’s probably becoming anachronistic now. It reflects the world that Japan modernised into.

  107. AJP Crown says:

    Bathrobe, have you seen the film The Wind Rises, Japanese 風立ちぬ ? It’s a beautifully drawn animation on that very subject, the interwar-period world that Japan modernised into. (I watched it via Netflix.)

  108. Bathrobe says:

    I missed it. Ensconced in China at the time.

    Wouldn’t mind watching it if I can get hold of it. Miyazaki is one of my favourites.

    But the actual period Japan modernised into was the late 19th century when Britain, France, and Germany were at the heart of European power and provided the model for Japan’s modernisation. (The Japanese were, of course, also very much into the US.)

  109. AJP Crown says:

    Do find it; mostly because of the graphics and also because of the way he decided to depict technology and life in Japan during that period. It’s one of the best films I’ve ever seen. It’s set then because it’s a biography of a famous aircraft engineer, the man who designed the Mitsubishi Zero. He takes trips around the world, stopping off to observe skeptically the Junkers plant in Dessau and admire some early Italian designed planes that feature a lot in the film (Britain & America don’t come up much, France not at all). You wouldn’t expect that to make a great movie. There are no war scenes, it’s all butterflies & Alpine wildflowers, aircraft being drawn by oxen taking a day to get to the runway.

  110. When you need soapstone, you REALLY need it

    Yes, that’s enough for the veneer of seeming like a plausibly real word. Too bad that kauppias means ‘merchant’ in the sense of ‘person specializing in the trade of a particular product’, not in the more general sense of ‘vendor’.

    We could indeed also think of a context where let’s say saippuakamariviramakauppias ‘soap chamber virāma merchant’ would be useful; perhaps an Indian guild of soap merchants gets really into fridge-magnet-esque rearrangeable sculpted Devanāgari letters with diacritics sold separately?

  111. David Marjanović says:

    with diacritics sold separately

    Day saved.

  112. AJP Crown says:

    let’s say saippuakamariviramakauppias

    Well, give me five minutes to practise and I can try. I agree that diacritics should always be sold separately in any fridge decorations. I don’t really understand your specialist and vendor distinction but my immediate image of a soapstone Merchant is a statue or monument to Antonio or to the stage set of the Merchant of Venice. You must decide if that could be described by saippuakivikauppias. They all seem reasonable to me, even your very long one. Now do it in iambic pentameter.

  113. Shouldn’t it be viraama? If it isn’t, I’m disappointed in Finnish. Languages with phonemic vowel quantity should stick together.

  114. John Cowan says:

    That’s like saying inflected languages should stick together, but they don’t do that. When someone speaking inflected language A learns inflected language B informally, without grammar instruction, they often end up speaking it without inflection or with minimal inflection. See the extreme case of Norfolk traditional dialect, where the uninflected foreigner-talk spread to the whole native population.

  115. Best I can do:
    http://accidentalbear.com/natalie-merchant-giving-up-everything-in-new-music-video/

    Alas, the mask is probably just plaster.

  116. David Eddyshaw says:

    That’s like saying inflected languages should stick together, but they don’t do that

    Phonology is another matter, though: tone languages do preserve tones in borrowing (admittedly adapting them to fit their own tonemics) and languages with length contrasts do keep them in loanwords.

    So if the Finns borrowed the word orally from actual Sanskritophones …

  117. David Marjanović says:

    There are even English-based creoles with tones, where stress was interpreted as tone… but apparently no French-based ones.

  118. David Eddyshaw says:

    English loanwords in Kusaal tend to represent English stress with a high tone that doesn’t conform to the usual tone sandhi rules, as in lɔ́r(ɛ̀) “car”, plural lɔ́yà, contrasted with gɛ́l “egg”, plural gɛ̄lá, which shows the expected regular pattern.

    It’s interesting (but not at all unexpected for an African tone language) that the borrowed tone pattern is actually more resistant to analogical remodelling than the segmental form is.

    Unsurprisingly, the tones of loanwords from Hausa are a lot more integrated, and have regularly undergone word-internal tone sandhi changes in parallel with Kusaal native vocabulary (in fact, if there were a good way of dating the adoption of the Hausa loans, you could use this to date the Kusaal tone changes.)

    The matter is further complicated by the fact that some English loans have evidently been transmitted via Hausa … (like wādá “law”, from Hausa óódà, English “order.”)

  119. Stress interferes in Finnish too: foreign non-initial vowel length in loanwords is fairly consistently overriden by a prescribed rule that non-initial vowel length is to be used for rendering foreign contrastive stress in open syllables (following the example of Swedish and perhaps Vulgar Latin where long vowels only occur in these cases).

  120. David Marjanović says:

    The best evidence that certain borrowings into Middle High German had non-initial stress comes from the ones that were passed on to Hungarian, a language that retains merciless word-initial stress to this day: upon borrowing, Hungarian used to harmonize the vowels to the one that was stressed in the source.

  121. On languages used in historical linguistics – on one hand, there is a school of thought that any IE linguist should be able to read and be aware of anything written in any IE language. I still fondly remember the late Erich Neu handing back a seminar paper of mine on Slavic endings with a note indicating that I had overlooked a relevant paper on the topic written in Lithuanian. On the other hand, my impression is that e.g. Russian isn’t read much by IE linguists outside of specific sub-fields like Balto-Slavic or IE accentuation, echoing a complaint by Lev Gumilev from the 60s that Rossica non leguntur. E.g. I have read some good ideas on Lachmann’s law by Otkupshchikov from the 50s or 60s that I haven’t seen mentioned in any Western language articles on that question.

  122. David Marjanović says:

    Lachmann’s law is always intriguing, so please tell me what they are!

  123. David Marjanović says:

    But of course there are examples of Russian literature being overlooked in IE linguistics by people who can apparently read it. Here’s a paper from 2014 or later that trounces Dybo’s law (shortening of pretonic long vowels after the postvocalic laryngeals were gone), explicitly stating in the abstract and the introduction that it’s only supposed to apply in Italo-Celtic and not e.g. in Germanic, citing Dybo’s original publication of 1961 and his follow-up work on Italo-Celtic from 2007, but not this openly accessible paper from 2008, again by Dybo himself, which is explicitly about Dybo’s law applying in Germanic. Somebody with a lot of time and dictionaries should check if the criticisms still apply to the 2008 paper…

  124. Lachmann’s law is always intriguing, so please tell me what they are!
    In a nutshell, his idea is that Lachmann doesn’t apply to cases where the derivation of words that results in a contact of plain voiced and unvoiced stop already happened in PIE, but only where the derivation happened later or the contact is due to later apocope. Specifically, he claims that the in cases where the participle in -to- causes Lachmann indicates that PIE didn’t have a verbal adjective in *-tó-, but in *-nó- for these specific verbs / roots.

  125. David Marjanović says:

    That’s similar to pp. 405–416 in this book from 2004, where Jasanoff presented about the same idea (most of the input to Lachmann’s law was generated by post-PIE analogy, and the rest by post-PIE syncope) without citing Otkupshchikov or mentioning *-nó-. Jasanoff cited Andersen (1969), which is about lack of certain voicing assimilations in Ukrainian and does not cite Otkupshchikov either – the striking similarity to Lachmann’s law is mentioned in a footnote, only to be dismissed by a claim that Lachmann’s law didn’t exist, citing Watkins’s work that had not yet appeared and has meanwhile been trounced by Jasanoff (2004)…

  126. I need to find some time to read that and re-read Otkupshchikov. I remember reading Jasanoff’s take on the matter and finding it different to what Otkupshchikov was saying. But it’s a couple of years since I read both.

Speak Your Mind

*