Laudator Temporis Acti presents a quote from Josephus’ Jewish Wars (5.272, tr. H. St. J. Thackeray):
Watchmen were accordingly posted by them on the towers, who gave warning whenever the engine was fired and the stone in transit, by shouting in their native tongue, “Sonny’s coming”; whereupon those in the line of fire promptly made way and lay down, owing to which precautions the stone passed harmlessly through and fell in their rear.
σκοποὶ οὖν αὐτοῖς ἐπὶ τῶν πύργων καθεζόμενοι προεμήνυον, ὁπότε σχασθείη τὸ ὄργανον καὶ ἡ πέτρα φέροιτο, τῇ πατρίῳ γλώσσῃ βοῶντες “ὁ υἱὸς ἔρχεται.” διίσταντο δὲ καθ’ οὓς ᾔει καὶ προκατεκλίνοντο, καὶ συνέβαινε φυλαττομένων ἄπρακτον διεκπίπτειν τὴν πέτραν.
He also provides Thackeray’s note on “Sonny’s coming”:
Probably, as Reland suggests, ha-eben (“the stone”) was corrupted to habben (“the son”); compare similar jocose terms, such as “Black Maria,” “Jack Johnson,” used in the Great War.
That’s wonderful, and I really hope it’s true.
Mark 13:26 And then shall they see the Son coming in clouds with great power and glory [flung from a catapult?]
[Verse slightly edited . . . I’m sure those words weren’t necessary . . . They were probably interpolations anyway . . . I R SMRT biblical scholar . . . ]
Buth and Pierce, Hebraisti in Ancient Texts: Does Ἑβραϊστί Ever Mean “Aramaic”? (here):
Very interesting stuff — thanks for that quote!
What I don’t understand, though, is how this would work, considering that bgdkpt spirantization already existed. In other words, הַבֵּן בָּא [haˈbeːn baː] ‘the son is coming’ and אֶבֶן בָּאָה [ˈʔɛːvɛn ˈbaːʔaː] (I think) ‘a stone is coming’ are not such a good match, even in the “clipped manner” which Buth and Pierce casually propose.
I am not particularly fond of Orwell’s quip about there being some ideas so stupid only an intellectual could believe them. However, Jesus couldn’t speak Hebrew, would seem to be in that category.
The Gospels themselves attest that Jesus could read a haftorah portion from Isaiah, although of course there are in our own time those with enough reading knowledge to read such a lesson aloud in a confident tone of voice but not be able to carry on a more free-floating conversation in Biblical Hebrew. But granted that Jesus could speak both Aramaic and Hebrew (and perhaps infinite additional tongues) as he wished, he presumably would have still generally chosen to speak in the language that made contextual and pragmatic sense given who he was speaking to and what they would or wouldn’t understand. But it may be significant that toward the end of Matthew 21 Jesus isn’t speaking to a generic crowd or even his own apostles, but to “the chief priests and the elders of the people” (or “the chief priests and the Pharisees,” depending on which verse you focus on), who might plausibly have had better Hebrew fluency than the general run of Judeans of the day and thus be good targets for wordplay that worked in Hebrew but not Aramaic.
@Y:
The bgadkpat spirantisation is very nearly totally predictable* even as late as the Masoretic pointing (if you adopt a distinction between vocalic and silent schwa, which certainly cannot have been lost very long before that, and which I reckon was still extant even at that point); it seems to me very possible indeed that at the date of the Jewish Wars, while it existed phonentically, it was actually still subphonemic.
* This is actually true even of word-initial consonants phrase-medially, but you need to look at the cantillation marks (specifically, their division into major and minor types) to see the predictability. In fact, the rules are so regular that I suspect that the Masoretes were consciously following them. I wrote a long screed on this years ago on LH.
I think the proposed puns in Matthew/Luke are plausible, even if there is a phonetic mismatch. Those are exegeses to be read at leisure. The mismatch is more problematic when explaining what people are yelling, as Buth and Pierce point out, at a dangerous and excited moment.
“What is that? What do you mean, the son is coming? Oh, a stone is coming. Right. Siege and all that.”
@DE. Maybe. As far as I can tell, there are no good data as to when the spirantization was complete, but people hand-wave it to a couple of centuries BC.
Khan’s reconstruction of the Tiberian system is ingenious (and he knows incomparably more about this than I do), but he is actually unhistorically projecting the system of a couple of centuries later onto the system that the devisers of the Masoretic pointing were trying to represent. The pronunciation within the tradition itself had changed over time. Why wouldn’t it?
people hand-wave it to a couple of centuries BC
My point is not about the date of the spirantisation but about when it become contrastive. I think the evidence is quite clear that this hadn’t happened yet in Josephus’ time.
Even the date of the phonetic changes is probably wrongly antedated. Much of the stuff I’ve seen on it seems to assume that Greek aspirated stops had already become aspirates centuries too early, and misinterprets the transcriptions in the LXX in a way which is frankly silly. Even Origen’s transcriptions show no evidence for the bgadkpat spirantisation at all (strangely enough.)
… had already become fricatives (not “aspirates.”) Gah.
Origen’s transcriptions are odd, because by his time the Greek aspirates almost certainly were fricatives, but he doesn’t use them just for the subsequently fricativised /kpt/ but also the non-fricativised: just like the LXX.
He may have been following an older established transliteration scheme. There’s a similar problem in Coptic, where the Greek aspirates are just used as ligatures for the consonant clusters /kh th ph/ in native vocabulary, which can hardly have been much like the contemporary pronunciation of L1 Greek.
Incidentally, the first vowel of all segolates in Josephus’s time* was short, not long, and “stone” would actually have been [‘ʔaβɪn]. The vowel length thing is absolutely clear from the Hexapla, as is the /a/ for later /ɛ/ in segolates of this pattern.
* And much later. I think there is good evidence these vowels were all short, not long, in the spoken tradition used by the creators of the Tiberian pointing; later, all vowel length distinctions were lost (this is the stage described by Khan.) In “Original Tiberian” (or whatever one calls it) vowel length was still contrastive, but it is not directly represented by the vowel signs at all: the system exclusively marks vowel quality, which only correlates imperfectly with length. (Khan agrees about the nature of the vowel signs, but has assumed that whatever is unrepresented in the script never existed at all; this is a bit like denying that Attic Greek had any long vowels apart from /e:/ and /o:/.)
The reason I think Khan is wrong to think that length distinctions were absent from the beginning of TH is that you can’t make sense of the sandhi stress shifts of TH without vowel length. This whole aspect of the tradition seems to have received remarkably little attention (the most accurate and complete study of it all is from the nineteenth century and is wriiten in Latin.)
I really should write all this up properly some day (so as to appear less of a crank … or, at least, for my own satisfaction.)
so as to appear less of a crank …
emm, you mean round here? To appear a crank, you’d need to be drawing parallels with vowel length in some random other language … Kusaal, for instance. (I take it the spelling double-aa is an indicator of vowel length as phonemically significant(?) Aren’t both vowel length and sandhi effects historically volatile?)
Maybe their were just saying “the son of a bitch is coming”, but Josephus decided that it is too earthy to put in writing.
I take it the spelling double-aa is an indicator of vowel length as phonemically significant
In Kusaal? Yes indeed. (I think you could make a case that the long glottalised vowels. like a’a, are actually /aa/ as opposed to /a:/, but the idea rather runs aground on the fact that Kusaal has short glottalised vowels too, and to magic them away needs stratospheric levels of abstraction from yer actual phonetic facts.) There do seem to be some languages that systematically distinguish long vowels from geminate vowels, but you always wonder about the analyses …
Hebrew vowel length has certainly undergone major changes over the history of the language – everyone agrees on that. My own (not-mainstream) view is that the language reflected in the LXX transcriptions was very similar in its distribution of vowel length to the pronunciation of Hebrew used by the creators of the Tiberian system, and the complete loss of contrastive vowel length happened later.
What I’m on about with stress sandhi:
the Tiberian system marks stress, and also carefully marks the changes of stress undergone by words within phrases; this is in the context of a system which also marks different levels of pause within phrases with quite exquisite care. The system is very unlike modern treatments of these things, but it’s a linguistic wonder of the world. The original traditional pronunciation of the Tiberian system was lost in mediaeval times: modern Jewish readings are not part of the same continuity. But a great deal can be learnt from just examining it on its own terms.
Non-Jewish learners of Biblical Hebrew generally ignore the whole thing, except as marking word main stress. Pity. It’s a thing of beauty, and really shows just how damn accurate the Masoretes were.
The system is very complicated (and differs in different books of the Bible.) The rules for stress sandhi are similarly complex, and the major grammars treat them very cursorily. I think most people just assume that the rules are kinda arbitrary and prone to too many exceptions to provide much information about the Masoretes’ own pronunciation, but this is absolutely not the case: the rules are complicated, but turn out to be remarkably regular. Crucially, they make much more sense (i.e. they are much more phonologically natural) if you assume that the language they apply to had vowel length distributed like the Greek transcriptions in the LXX and in Origen’s Hexapla (which agree closely with each other.) If you assume that the Masoretes’ own pronunciation had no contrastive vowel length, the accuracy with which they apply the stress sandhi rules anyway would be frankly superhuman. They were very clever men indeed, but nobody was ever that clever.
Incidentally, the first vowel of all segolates in Josephus’s time was short, not long
My understanding (perhaps wrong) was that it was phonemically short, phonetically long.
Kantor’s dissertation on the secunda of the Hexapla posits a nine-vowel system, ī ē e ū ō o ε a ɔ̄ (p. 250).
β
Yes, my mistake. It became [v] only later, in Byzantine times.
A good witness to fricativized stops comes from Greek loanwords in the Mishna, especially as recorded by the Kaufmann and Parma codices and other witnesses to early pre-Masoretic Palestinian pronunciation. There’s a compendium of those in Shai Heijmans’ dissertation (in Hebrew; here). For example, στοά is borrowed as both אִיסְטְוָוה and אִסְטְבָֿהֿ, suggesting a distinctive fricative בֿ. Conversely, borrowings from words containing -μβ- are written with a בּ. It’s not a foolproof argument — there are exceptions, and as good as those Mishnaic manuscripts are, you can’t be sure that they are free from the effects of later reading traditions.
My point is not about the date of the spirantisation but about when it become contrastive.
I am not sure what you mean by that. How is it not contrastive if it is predictable (from the stress and other things?) And that brings up the point that stress was contrastive, so /haˈben/ and /haˈʔeβen/ or so would be less confusable if only because of the different stress position.
[ˈʔaβɪn]
How so? I’m looking at Kantor’s dissertation, and all I can find are segholates like מֶלֶךְ transcribed as μελεχ (and melech in Jerome).
How is it not contrastive if it is predictable
Compare the aspiration of /p t k/ in English: completely predictable, and thus subphonemic. The aspirated and unaspirated versions are allophones. But If Pāṇini were describing English, he’d make each of /p t k/ into two distinct consonants. Same with the stop/fricative distinction in Hebrew, until it became contrastive with the loss of “vocalic” schwa and loss of gemination in stops. That was certainly later than Josephus’ time.
Segolates in LXX transcriptions and the Hexapla are consistently written with the first vowel ε ο rather than η ω. Does Kantor suppose that ε ο represented long vowels here? Why?
I may be wrong about the date of fronting of the vowel in the absolute singular form of a-segolates. I think there is variation between different lexemes, but I may be hallucinating.
A good witness to fricativized stops comes from Greek loanwords in the Mishna, especially as recorded by the Kaufmann and Parma
Interesting! Though again, I’m not disputing that the /k t p g d b/ phonemes were (predictably) realised as fricatives in certain environments quite early (personally, I suspect this was due to Aramaic influence.) Neither the LXX nor the Hexapla transcription systems actually have any way of representing this, even if it was already established: the Greek χ φ θ are used for /k p t/ whether realised as fricatives or not, with κ τ reserved for ט ק. LXX-era Greek didn’t yet possess any of the relevant fricatives (Quintilian cites Cicero as mocking a Greek witness for being unable to pronounce Latin /f/), and Origen seems to have been operating with some sort of traditional transcription system at variance with contemporary Greek phonology.
@DE There do seem to be some languages that systematically distinguish long vowels from geminate vowels, but you always wonder about the analyses …
Would one of those languages be Māori/Polynesian languages in general? (Looking at you Hawaiian, which seems to have a superfluity of doubled vowels in its spelling.) The macron I put just there is thanks to modern typesetting, and is controversial as against spelling as Maaori. Of course in Colonial times, the phonemic difference with long vowels was mostly observed in the breach in spelling. Which means these days the difference is all but lost.
We can distinguish tūī/tuuii vs tui. The best I can gloss the enmacroned pronunciation is ‘Toohey’ — which is an Australian brewery/Irish name. IOW that’s not ‘-uu-‘ as in ‘continuum’.
The role of vowels in getting from the spelling to the pronunciation in te reo is a mess, not helped by the early Missionaries/linguists being trained on German Principles, so the vowels don’t have the values you’d expect in English spelling. You just have to know that ‘rua’, ‘tui’ (no macrons) are two syllables; ‘hoiho’ and ‘taupō’ are three vowels but two syllables, whereas ‘Kaikōura’s five vowels give four syllables [kaiˈkoːuɾa], as indicated by the macron (wot we usually do not ‘ave).
Not a mess, but maybe underspecified for that wot native speakers don’t need help with. We shall all be thankful that the spelling wasn’t devised with English conventions.
@DE: You’re right about the initial vowel. The Encyclopedia of Hebrew language and linguistics is now at archive.org, and there I am told that segholation had not yet taken effect at the time of the Hexapla: to wit, another aleph-initial qatl noun, אֶרֶץ, transcribed as αρς, as well as גֶּבֶר γαβρ. Also, the place name אֶבֶן הָעֵזֶר ʾeḇen hāʿēzer (Sam. I 7:12) is transcribed in the Hexapla ἀβενέζερ (same as the LXX).
AntC: “doubled vowels” in Hawaiian necessarily mean that there’s a glottal stop between them, which is marked in normative orthography, most everywhere nowadays in Hawai‘i itself, at least. Some older place names (Ni‘ihau, Kaho‘olawe) might still be spelled without it on the mainland.
I came across a paper on an American Northwest Coast language with a supposed V:/VV contrast, but I’ve forgotten the details. I remember that the VV sequences were in fact realised as VʔV or just as glottalised (a fairly common alternation, also found in Kusaal, though the ʔ is then not actually a stop, but an approximant.) At a suitable level of abstraction, you can say that all Kusaal glottalised vowels are long, but to get there, you have to undo the characteristic apocope process which shortens word-final long vowels in most contexts. Thus “buy” is “underlyingly” da’a, but in actual fact normally surfaces as da’ [da̰] [daʔ]. Agolle Kusaal also has some short glottal vowels before nasal consonants in closed syllables, but there the Toende Kusaal, Nabit, Talni and Farefare cognates have modal vowels, so that looks like a separate secondary development and might be handwaved away with a bit of imagination.
[Accounts other than my own of the Western Oti-Volta languages which have glottal vowels all take the VʔV/Vʔ realisations as simply involving a consonant /ʔ/, but this is unequivocally incorrect at a phonemic level, even though it sometimes works on a phonetic level.]
Even leaving aside LXX and Hexapla transliterations, certain segolate forms in TH actually make more sense on the basis that the first vowel was in fact read as short, incidentally.
On the one hand, you have a-segolates where the vowel was not fronted before /j/ like /jajin/ “wine” and /bajiθ/ “house.” Either you assume that pathach could be long (as Khan does, but he denies that vowel length was contrastive at all) or you have to say that some segolates had short vowels in TH. And Khan’s view makes it a mystery why the pausal forms have qametz, which is simple to account for if you just say that the vowel was lengthened in pause.
On the other hand, you have “death”, traditionally supposed to be /mɑ:wɛθ/, with the qametz read as long. But this makes no phonetic sense: why would /a/ undergo lengthening in such a context? Instead, it should be read as /mɔwɛθ/: all that’s happened is that /a/ has been rounded to /ɔ/ before /w/, which is phonologically entirely natural.
Tzere and kholem in TH segolates were short /e o/, not /e: o:/. The actual rule was not that /ɛ ɔ/ became /e: o:/ under stress, as David Qimchi’s system posits and learners of BH have been told ever since, but that under stress, /ɛ ɔ/ became short close /e o/. These vowels consistently pattern like pathach, not qametz, in TH morphophonemics e,g. in double-ayin segolates, and in the qal perfective of verbs. In fact, a good heuristic for identifying short tzere and kholem is noting where morphologically parallel forms have pathach.
This understanding of TH vowel length neatly matches the stress sandhi phenomena I was banging on about above; from a historical point of view, it implies that Original TH actually had vowel length virtually unchanged from the LXX/Hexapla, and an overall system of vowels which was nearly unchanged apart from the development that stressed /ɛ ɔ/ -> /e o/. The length system was never rearranged à la Qimchi; it remained much as in second-Temple Hebrew until vowel length was lost altogether (the Khan stage of TH.)
I really ought to write all this up properly …
“I wonder if Egyptian Greek was particularly conservative phonologically? Or even pronounced with a strong, distinctively Egyptian, accent, maybe …” It certainly looks like the whole newfangled fricativization simply never made it to Egypt (or at least not sooner than it unrelatedly appeared in German), notwithstanding the first FILIPPVS showing up in the 2nd c. AD elsewhere. The distinctively Egyptian accent, meanwhile, extended to lack of h-dropping in stressed syllables and to inconsistent or no voicing of /d g/.
…if you take word boundaries into account, as you do when you’re the speaker. Not so much when you’re the hearer; instead, in that case, you infer the word boundaries from aspiration (and word-final glottalization and pre-fortis clipping and other phonetic phenomena in addition to your knowledge of which words exist and which ones are likely to show up in the context in question) – so they’re actually contrastive. I think English needs to be described with two “phonemic” levels – unlike French or southern German for example.
The canonical example, which I learned here from John Cowan, is night rate : nitrate : Nye trait.
There already is an -aw- contraction, to a full holam, as in the construct case of מָוֶת māweṯ, viz. מוֹת môṯ, and likewise in nouns like שׁוֹר šôr ‘ox’ (< *ṯawr; pl. שְׁוָרִים šwārîm, for the fun of it).
Suchard (The Development of the Biblical Hebrew Vowels, §5.2) argues that the two words which exhibit this form (māweṯ ‘death’ and tāweḵ ‘midst’) avoided the contraction of the diphthong by virtue of ending with an unemphatic voiceless plosive. Later on, stressed /a/ assimilated to <å> before the bilabials /w/ or /m/ (§9.1.1 6.a.i.) All accented and word final vowels are lengthened (ibid. 8.b. and §4).
There is actually a surprisingly good analogy for that with the TH bgadkpat phenomenon: the fricativisation appears word-initially after vowels within phrases, but only phrases marked as more closely connected by the complex cantillation system (and only if certain other criteria are met, which are closely linked with the stress sandhi that I was talking about.)
But in both cases, the distribution is predictable if you take suprasegmental features into account: it’s all about the phonetic realisation of junctures (which are are much more carefully marked in TH than in English orthography, though not in the ways that modern linguists are familiar with …)
https://en.wikipedia.org/wiki/Juncture
I think the whole area of suprasegmenal influences on external sandhi is interesting cross-linguistically and also historically; the initial mutations of the Insular Celtic languages and of Mende started out as transparent contact phenomena, as did external tone sandhi in Kusaal, where processes which obviously arose as industry-standard tone-spreading and tone-dissimilation rules have been rendered opaque by the loss of underlying final vowels. (So much so that the usual “fixes” by inventing “floating tones” are about as helpful as inventing word-final “floating vowels” would be for describing Welsh soft mutation; it would just be confusing diachronic and synchronic analyses.)
Speaking of Suchard, I wonder what IEists make of his recent post, which would turn Verner’s Law upside down.
(My previous comment was a reply to DM, in case that’s not obvious.)
There already is an -aw- contraction, to a full holam
Sure, but that’s just the construct, it’s to holem not qametz, and is just the expected outcome of *aw not followed by a vowel. It doesn’t shed any light on the vocalisation of the absolute form.
“All accented vowels are lengthened” is exactly what I am denying. This didn’t happen until length itself altogether ceased to be contrastive in (later) TH. My main reason forr denying this is to do with stress sandhi, but note too that pathach appears frequently in stressed syllables. (Come to that, tzere and holem appear in stressed closed syllables, including non-word-finally.)
Why does stressed pathach become qametz in pause?
Briefly (to show – I hope – that I’m not just handwaving with this stuff about stress):
Final CV:C syllables never allow for stress to be retracted to a preceding syllable; occasionally, they lose stress altogether before a word with initial stress, and then the two are linked into a single stress unit by maqqeph; but stress shift within the word never happens.
But I’ve got ahead of myself by talking about “final CV:C syllables” as if vowel length distinctions in stressed syllable was an accepted thing. So backtracking:
There are are two sorts of polysyllabic words that don’t permit stress retraction before a following word with initial stress.
One is consonant-final absolute forms of (non-segolate) nouns. Now both LXX and Origen, which do show vowel length contrasts in stressed syllables) nevertheless always show η or ω, not ε or ο, in the final syllable of masculine absolute nouns with a single final consonant (not simplified from a geminate, not a cluster): this lengthening was therefore present already in actual BH nouns; it is reasonable to suppose that it was also present when the vowel is transcribed as α and Greek has no means to indicate length.
The other is verb forms ending in CVC where the vowel is hireq or shuruq, or more generally, a vowel usually written plene in the consonantal text: the vowels that in David Qimhi’s scheme are long by nature rather than position, and derive historically from proto-Semitic long vowels or diphthongs.
However, this group does not include verb forms were the vowel is pathach, or tzere or holem when they are not normally written plene: in these cases LXX and Origen have α ε ο: the vowels were certainly short at that stage, whatever happened later.
So the stress sandhi facts can be explained by the simple and phonologically plausible rule that ultraheavy CV:C syllables did not allow for the stress to be retracted: if you assume that the stress sandhi rule was being applied to a language with vowel length as it is represented in the LXX and the Hexapla.
The Masoretic pointing itself does not make this analysis impossible: it’s common ground that the signs mark vowel quality and not length. Nor does this seem at all unlikely to me on first principles. We know that Hebrew vowel length changed between the LXX and the Arabic transcriptions that Khan adduces to show that vowel length was no longer contrastive: the question is only, When?
Now it’s not impossible to explain the facts differently. You have to assume that nouns and verbs were prosodically different in TH, but in fact you need to do that anyway for other reasons, and there are plenty of examples of languages in which nominals and verbs show systematic prosodic differences (including, inevitably, Kusaal.)
So the usually-fixed final stress of non-segolate nouns is a prosodic feature of nouns, specifically. Could be … but then, why do some verbal forms behave like nouns, while others don’t? And how is it that this division among verb forms correlates exactly with whether the final syllable contained a long vowel/diphthong in BH or not? And is it just coincidental that nouns, which behave in this respect like verbs with a final syllable containing a long vowel/diphthong, are shown by the Greek transcriptions to have actually had long vowels in their final CVC syllables in BH?
One thing that isn’t explained by my theory (which is mine) is why BH nouns had this lengthening of originally short vowels in final syllables of absolute forms. It can’t have been just the effect of stress, because it (usually) fails to happen when the final consonant has been simplified from a geminate.
It also is orthogonal to the question of the length of original short vowels in open syllables immediately preceding the stressed syllable, which is a whole other can of worms. Hebrew morphophonemics would make more sense if they were still short, but transcriptions suggest otherwise. In practice, though, there aren’t all that many cases where Greek evidence helps, and with Aramaic you have to bear in mind that contemporary Aramaic itself didn’t actually allow short (non-schwa) vowels to stand in this position.
I haven’t commented on it yet because I also have things to say about Korean and Vedic…
But I think (nowadays) that Verner’s law is not reverse tonogenesis. Instead, it happened before (most of) Grimm’s and simply turned all voiceless consonants into their voiced counterparts except in “positions of strength”: word-initially* and right after stressed vowels. Later, Grimm’s struck and turned all the remaining aspirates into fricatives.
The beginning of words is a “strong position” in numerous undisputed sound changes, e.g. bgadkpat or the High German consonant shift or the gorgia toscana. The position right after the stress is the only one where southeastern German, which is stress-timed, allows consonant length (e.g., Satellit is spelled as if stressed on the middle syllable, but instead it has final stress and is therefore pronounced as if spelled satelitt).
It is important that consonant clusters were largely exempt from Verner’s law. Clusters are another “position of strength”, but shouldn’t prevent (forward or) reverse tonogenesis.
* Actually, there are a few words that suddenly make sense if Verner’s law happened word-initially. (Starting here on slide 34; note that slide 41 calls a regional Central German phenomenon “High German”, but that’s irrelevant here.) But they’re still outnumbered where this clearly did not happen. Maybe word boundaries originally played no role and Verner’s law was only blocked if the preceding word happened to end in a stressed syllable, giving lots of words a version with a voiced and a version with a voiceless initial, and then the less common form of each was eliminated… that wouldn’t be any stranger than the elimination of [English as she is] spoke.
Could that just be a matter of timing? Final CVC getting lengthened, final CVC: being long as a whole already, and then consonant length was eliminated word-finally?
@Y: There already is an -aw- contraction, to a full holam
@DE: Sure, but that’s just the construct, it’s to holem not qametz, and is just the expected outcome of *aw not followed by a vowel. It doesn’t shed any light on the vocalisation of the absolute form.
If I understood you, you are saying that -aw- assimilated to -åw- in the (absolute) måwet, for purely phonological reasons. Then why would the same phonological processes not produce å, but a full holam (ō) elsewhere?
An alternative explanation of stress sandhi in TH:
All non-segolate absolute forms of nouns which ended in a consonant had final stress in BH. (Things are less straightforward with constructs, and you do in fact still need a TH stress retraction rule for feminine absolutes of the pattern CVCV, but those are relatively minor details.)
On the other hand, verbs in BH were stressed on a final CVC syllable if the vowel was long; otherwise, they had stress on the penult.
Original TH inherited this as the underlying stress system, but the actual surface stress system prohibited penultimate stress in all words ending in a consonant, and also in all words ending in CVCV unless they were followed by pause. In the prohibited cases, underlying penultimate stress had to shift to the final syllable, but this “fix” was blocked in close juncture before a word with initial stress; in that case, a penult stress on CV- moved to the preceding open syllable if there was one, words of the form CVCVC kept their penult stress, but words ending in CVCCVC (oddly) lost stress altogether. (I may have got some details wrong here: I can’t locate my old notes on this at present.)
So anyway: this is a system in which many verb forms (and some noun forms with pronominal suffixes, too) have underlying shapes which can never stand unchanged in the text, but have to be transformed by context-dependent rules first. Although it would involve a lot of cleverness on the part of the Masoretes to implement all this so accurately, they were pretty good at it, and a piece of evidence in favour of it is that you actually do need something pretty much as elaborate as this to capture the very different morphophonology of TH nouns and verbs.
The other advantage is that you then don’t need my theories about vowel length. But it’s a more elaborate scenario in other respects, so I think you’d need some positive evidence that I was actually wrong in hypothesising that an LXX-style length system survived long enough to underlie the Masoretes’ invention of the Tiberian system at the beginning. As I said, that actually doesn’t strike me as implausible a priori; but there could easily be evidence I’m unaware of that simply refutes the idea.
Could that just be a matter of timing? Final CVC getting lengthened, final CVC: being long as a whole already, and then consonant length was eliminated word-finally?
Sure; something like that. But what it doesn’t explain is why this only happened in nominals, not verbs. My alternative scenario, where nominals and verbs had different stress patterns already in BH could help with that (although that only pushes back the difficulty: how did that stress difference come about initially anyway? But however you cut it, there is an odd difference here between nominals and verbs in BH already.)
Then why would the same phonological processes not produce å, but a full holam (ō) elsewhere?
Because the enviroments are different. In the absolute, the /w/ is followed by the epenthetic segol, and in the construct it’s not, so you’ve got *aw before a consonant, which regularly become /o:/, here, as elsewhere. In a case like שׁוֹר šôr ‘ox’ the absolute has been remodelled after the construct: that’s not an isolated phenomenon with segolates, but has happened with other words too (e.g. דְבַשׁ “honey.”)
Just noticed that my scenario B, which I intended as an alternative to my outlandish theories about LXX vowel length surviving until TH was first created, actually covertly assumes a contrast between underlying short and long vowels in penultimate open syllables. Stress may not be removed to the antepenultimate if the penult vowel is hireq or shuruq, but only in cases where the LXX/Hexapla have α ε or ο.
I suppose you could try to get round this by suggesting that those clever old Masoretes had a sort of standing protocol to the effect that stressed /i u/ could never lose stress unless they were word-final (assuming that the Masoretes treated them as /ij uw/ gives the wrong answer.) But it seems a lot easier just the assume that they were /i: u:/.
I do think that there may be something in supposing that verbs were underlyingly stressed on the penult unless the final syllable was CVCC or CV:C, though. The alternations between pausal forms, ordinary forms, and forms preceding an initial-stress word would all be easier to account for in that basis. (I suspect I may be reinventing the wheel here.)
שְׁוָרִים šwārîm, for the fun of it
indicating, apparently (if we merely posit a floating final vowel) that the post-resurrection feast will be sliced and slow-roasted on a spit?
Found my previous screedette on the bgadkpat phenomenon in Tiberian Hebrew, if anyone is interested.
https://languagehat.com/six-degrees-of-deuteronomy/#comment-3922056
(Kagi Search actually works for locating old LH comments. I may even end up paying for it.)