Two Japanese Questions.

1) In Jangfeldt’s Mayakovsky bio, he says “After his return from Berlin in May 1924, Mayakovsky met with the Japanese author Tamisi Naito, who was visiting Moscow.” (In the original: “Efter hemkomsten från Berlin i maj 1924 träffade Majakovskij den japanske författaren Tamisi Naito, som var på besök i Moskva.”) I can find no reference to such an author elsewhere, and I suspect the name may have gotten garbled; anybody know who this might be?

2) In Goncharov’s Фрегат “Паллада” [The Frigate Pallada] (see this post), he repeatedly refers to a cry of Japanese oarsmen, “оссильян” [ossil’yan], which Goetze renders “ossilian.” The -l- makes this impossible Japanese, of course, and I suspect there are other distortions; anybody know what the original word or phrase is?

(The hapless Goetze transliterates Хагивари, the name of one of the Japanese officers [the Russian word for ‘officer’ here, баниос, is borrowed from Dutch banjoost, but I have no idea what that’s from], three different ways: Kagivari, Chagivary, and best of all Charivari. He also omits large chunks of text — several pages on what Goncharov perceives as similarities between Chinese and Japanese, half a dozen pages on the need for Japan to open itself to the outside world, and every passage in which a fellow member of the expedition, Goshkevich, makes an anti-Semitic remark. For shame!)

Comments

  1. Regarding 1) – it seems that the name is Tamiji Naito/Тамидзи Найто: https://commons.wikimedia.org/wiki/File:Mayakovsky_Pasternak.jpg

  2. Thanks very much! At first I was suspicious because even with that search term results seemed to be limited to the meeting mentioned in the bio, but then Google found this WorldCat page, and that led me to his Japanese Wikipedia page. Mystery solved!

  3. I suspect Charivari to be the work of the hapless printer or copy editor. Lectio difficilior potior and all that (which I always write as “lectior difficilior potior” before I fix it).

  4. ossilian is probably some form of Japanese verb 押し寄せる oshiyoseru (push on, surge forward).

    Volitional form would be 押しよせよう Oshiyoseyou (let’s surge forward!)

  5. I suspect Charivari to be the work of the hapless printer or copy editor.

    ossilian is probably some form of Japanese verb 押し寄せる oshiyoseru (push on, surge forward).

    Both very plausible!

  6. Re: Goshkevich’s antisemitism.

    It is probably completely lost on Western readers, but Gorzkiewicz is a Polish surname.

    Goncharov is engaging here in anti-Polish stereotyping (Poles are rabid anti-Semites, unlike us, civilized Russian people)

  7. Banjoost does not appear in any Dutch dictionary I can find online, not even the WNT, the Dutch analogue of the OED, and the normal Dutch word for ‘officer’ is officier. But some googling finds this passage [Google Books link] in an English précis, published in1841, of a French translation of Nippon by the German doctor and botanist Philipp Franz von Siebold, one of the first European students of Japan (and a very interesting person):

    When we had turned the northern point of the isle of Swo-Sima [apparently in Nagasaki bay], several officers and Japanese interpreters came on board to receive our papers and some persons as hostages. They announced to us that an officer of high rank bearing the title of banjost, and some messengers from the factory [trading post], were coming to visit us.

    Since the celebrated adventure of the British ship of war Phaeton [which seized the Dutch representatives in order to extort Japanese naval supplies, and then fled before the Japanese navy could stop them], in 1808, the government takes the precaution, before the Japanese officers and the Dutch from the factory board the ship, to demand hostages, and conduct them to Dezima [Dejima island, site of the factory], so that since this epoch, the Dutch vessels can no longer enter immediately the port of Nangasaki [sic], but are obliged to remain at anchor during some time in the bay near the isle of Takaboko (Papenberg) under the eye of the imperial guards. Toward noon arrived the goban josi, or envoy of the imperial guard.

    So apparently banjoost is a Dutchly corrupt version of goban josi (possibly also corrupt, but at least plausible-looking Japanese), which passed briefly into Dutch in the sense ‘Japanese officer’ (as a google for it will show) and thence into Russian. I think a direct transmission to Russian is less likely, despite the geographical roundabout involved.

  8. I presume Хагивари is meant to represent the Japanese surname Hagiwari.

  9. “every passage in which a fellow member of the expedition, Goshkevich, makes an anti-Semitic remark”

    !!! !!!!!

    I’ve read Goetze, and enjoyed it – what he bothered to include.

  10. Great rowing-related mysteries in world literature:

    (1) What song the Sirens sang.

    Then all of a sudden it fell dead calm; there was not a breath of wind nor a ripple upon the water, so the men furled the sails and stowed them; then taking to their oars they whitened the water with the foam they raised in rowing. Meanwhile I look a large wheel of wax and cut it up small with my sword. Then I kneaded the wax in my strong hands till it became soft, which it soon did between the kneading and the rays of the sun-god son of Hyperion. Then I stopped the ears of all my men, and they bound me hands and feet to the mast as I stood upright on the cross piece; but they went on rowing themselves. When we had got within earshot of the land, and the ship was going at a good rate, the Sirens saw that we were getting in shore and began with their singing.

    (2) What Ahab told his men:

    [Stubb:] “… Merrily, merrily, hearts-alive. Pudding for supper, you know;—merry’s the word. Pull, babes—pull, sucklings—pull, all. But what the devil are you hurrying about? Softly, softly, and steadily, my men. Only pull, and keep pulling; nothing more. Crack all your backbones, and bite your knives in two—that’s all. Take it easy—why don’t ye take it easy, I say, and burst all your livers and lungs!”

    But what it was that inscrutable Ahab said to that tiger-yellow crew of his—these were words best omitted here; for you live under the blessed light of the evangelical land. Only the infidel sharks in the audacious seas may give ear to such words, when, with tornado brow, and eyes of red murder, and foam-glued lips, Ahab leaped after his prey.

    (3) What оссильян means.

  11. josi is clearly 上使 joushi (shogun’s envoy; emissary)

    Found the full Japanese title 御番上使 (o-ban joushi).

  12. David Marjanović says:

    Nangasaki [sic]

    The Japanese g, I hear, is pronounced [ŋ] in large parts of the country.

  13. When we had turned the northern point of the isle of Swo-Sima

    Has to be an error for Iwo-Sima (which is what Goncharov calls it; I was startled at first, thinking of Iwo Jima).

    The Japanese g, I hear, is pronounced [ŋ] in large parts of the country.

    Yes, Goncharov occasionally has forms in -ng- that clearly represent Japanese -g-.

  14. Swift uses “Nangasac.”

  15. (1) There’s the common surname Hagiwara 萩原(はぎわら), but would a Russian speaker have heard the final “a” as “i”? Seems doubtful, but my Russian phonotactics is not what it should be.

    (2) 御番上使 goban jōshi indeed seems to be it, as SFReader said, transcribed that way in a few Japanese sources talking about the Siebold escort retinue that pop up quickly on google. “Envoy” makes sense, too. Nagasaki was run directly by the Shogunate for most of the Edo period. A *very* cursory skim online gives me the sense that out of the character’s many possible meanings, in this particular 番 ban here (the go would be an honorific, and judging from banjost sometimes omitted) the operative one is probably “a duty post occupied by various responsible individuals in turn” (cf. modern “police boxes” 交番 kōban, lit. “crossroads duty-post”). If so, I suppose the post (office? garrison?) manned here could be one keeping an eye on foreigners, and the 上使 jōshi a supervisory envoy dispatched to that from Edo.

    (3) 押し寄せる seems unlikely to me, though it depends on what the boats are doing. Are they making a coordinated rush upon some enemy vessel all at once? The image that comes to mind for the verb is of an onrushing wave, or maybe a crowd–i.e. not a point in motion, but a front. If it’s just a repeated call to keep the rowers in rhythm, 押し寄せる would be odd. In this latter case it might well be something meaningless like “hey-ho!” In Japanese there are a lot of such (amazingly standardized) to-be-said-during-action-X phrases.

  16. In this latter case it might well be something meaningless like “hey-ho!” In Japanese there are a lot of such (amazingly standardized) to-be-said-during-action-X phrases.

    That was my uninformed guess, but I was curious what a more accurate transcription would be. Probably unknowable now.

  17. Found another description of this cry in “Russian Views of Japan, 1792-1913 An Anthology of Travel Writing”:

    “Sharp, wild cries reach us from the shore; now they quieten down, now
    they start up again with renewed frenzy. We listen carefully – and still do not understand. We only understand when an oar-propelled boat appears from behind a junk. On it are a crowd of oarsmen, who to encourage their labours, strike up not the resonant song of the Volga boatmen, but a disorderly mixture of monotonous, disjointed sounds on only two tones, both repeated endlessly, without variation. ‘I-yosso! yosso! yosso!’, cried the Japanese boatmen, and their whole song and all its meaning resided in this. This boat brought to us the Japanese official”

    That’s 寄せよう yoseyou, I think.

    “Let’s go, let’s go!”

  18. If the final vowel was nasalized (at least in Nagasaki harbor in the 1850s), that’s definitely a plausible candidate.

  19. Interesting. As I thought, then, it is a rhythm-keeping call.

    Still, 寄せよう yoseyō rubs my sprachgefühl the wrong way here. For one, 寄せる yoseru doesn’t mean “go” at all. The basic meaning is “shift A towards B.” And I’m not downplaying it by translating “shift” rather than “move,” etc. The sense of moving effort is weak, which is why (1) it’s used easily for abstract things like “思いを寄せる” omoi (thought) wo yoseru “think towards someone” (=think about them, usually in a romantic way), and (2) when compounded with other verbs it adds little more than directionality, as in hiku (pull) + yoseru = hikiyoseru (pull towards), or the above oshiyoseru (surge at/towards) from osu (push, force forwards) + yoseru.

    Another is the volitional form. The English “let’s go” is a lot more pithy and forceful, a softened command if said in the right tone. (Image: “Get moving! Let’s go! Let’s go!” being shouted at soldiers parachuting in quick sequence off the deck of a hovering helicopter.) The Japanese volitional is definitely further on the dial towards “suggestion”, and seems oddly soft for a shouted order or hortation as in this case.

    On the other hand, with regional dialects before the Meiji state–especially in a region that never had a prestige written form like Kyushu–all bets are off, and it’s a question of how much of a guide MSJ can be at all.

    And yet…the “yosso” above (I bet the “i” is just the “y” dragged-out) does reminds me of よいしょ yoi-sho (which can sound like よっしょ yossho), one of those stereotyped phrases I mentioned above specifically associated with moving effort, be it climbing stairs, picking up boxes, running, even (esp. for the elderly) standing up or sitting down. The image is one of huff-puff out-of-breath-ness, if you can imagine a world where native speakers of English actually (all) said out loud to themselves “huff! puff!”in such cases. Though I really wonder if such “words” would really be the same over such a distance of time and dialect.

  20. @Elessorn: Google does turn out with some suggestive examples, like in some kind of weird faux-rowing: 「ここは、つくばみらいの軽井沢です!」と会場の皆さんに自然の豊かさを強調したあと、「小貝川のライン下りをします。福岡堰から大平洋まで元気よく櫓をこいで下さい。」「よいしょ!よいしょ!」の掛け声とともに全員が力強く櫓をこぐ仕草。

  21. J.W. Brewer says:

    Re the potential relevance of “yoseru,” if you’re an oarsman, the way you make the boat “go” is by shifting/pulling the inboard end of your particular oar between point A and point B, over and over again.

    It is said to be an inaccurate cliche that coxswains urge on those actually doing the rowing by shouting “stroke! stroke!” at them. But the point is that in any language the standard verb for what-the-rowers-should-be-doing may be more specific than one just meaning “go” or “move.”

  22. More specific? My first guess is that a Russian team leader in this situation would just repeat “Ras! Ras! Ras!” (“Раз! Раз! Раз!”) which can be literally translated as “one! one! one!”, for as many times it takes. If the action consists of two somewhat temporarily distinct parts of ready-go type than it would be “and-one! and-one!” etc. (и-раз).

  23. marie-lucie says:

    Let’s go! : perhaps rather Let’s row!

    Nagasaki, “Nangasac”

    As a student I had a Japanese roommate and learned a little bit of Japanese. I heard a lot of the language as she spent much time talking to her friends, in person or on the phone. She definitely pronounced the particle ga as “nga”. I think she was from Yokohama,

    As for final “ac”, the vowel “i” in a word-final syllable “ki” is often barely audible, so that the word seems to end in a palatalized “k”.

  24. Nick Karayev says:

    From the Russian commentary:

    Гончаров передает одно из японских восклицаний типа «Хэссинсё, хэссинсё…», или «Эссинъёйса, эссинъё», или «Ёссириёйса, ёссириёйса».

    http://www.goncharov.spb.ru/fr_2_1_komm/

  25. David Eddyshaw says:

    From Shoichi Iwasaki “Japanese”, Benjamins 2006 (p35 fn):

    “There are three types of speaker with respect to the use and non-use of the word-medial, syllable-initial velar nasal [ŋ] … Consistent nasal speakers always pronounce shoo-gakkoo as [ɕo:ŋɑk::o:], while consistent stop speakers always pronounce it as [ɕo:gɑk::o:]. Non-consistent speakers could use either form. Many Tokyo dialect speakers used to be consistent nasal speakers, but younger speakers are mostly consistent stop speakers. This change is believed to have started with those who were born around 1930 because the speakers in this age group were relocated out of Tokyo during the Pacific War (1941-45) to the area where [g] rather than [ŋ] appears in the relevant phonological environment.”

    I believe it’s also partly a question of register, with [ŋ] being more highfalutin, but can’t locate a reference at the moment. Akita neru …

    High i u are devoiced between voiceless consonants and word finally after a voiceless consonant, and in Tokyoese are present in a more or less spiritual sense only; the palatalisation of k before i does remain.

    So yes, [nɑŋɑsɑkʲ] or thereabouts.

  26. Heinrich R. Blutvergieben says:

    About “Nangasaki”:
    I know very little about Japanese.
    Has the “use and non-use” of [ŋ] explained in David Eddyshaw’s 8:22 post developed from pre-nasalization?
    Wikipedia has
    https://en.wikipedia.org/wiki/Late_Middle_Japanese#Prenasalization
    and
    https://en.wikipedia.org/wiki/Early_Modern_Japanese#Prenasalization

    I’m sure dialects have always existed, I don’t know from which time and place Westerners first picked up their “Nangasaki” (and which European language’s orthographic convention this spelling is based on), I don’t know if it’s the prestige variety described in the Wikipedia articles or even local Nagasaki dialect, and I have no idea whether the “use and non-use” of [ŋ] in cases like this is older than Middle Japanese, but I’d guess that at some point in the past there was sort of a [ŋɡ] (from what earlier form?) which later became [ŋ] in some lects and [ɡ] in others, so at various points in time and space there may have been a Na[ŋɡ]asaki, a Na[ŋ]asaki and a Na[ɡ]asaki. Does that make any sense?

    I had only noticed that Sino-Japanese /g/ often corresponds to Sinitic initial /ŋ/, for example in 語 “language”, 五 “five”, 魚 “fish”, and 月 “moon”, so when I had read the original post but not the comments, I guessed that there had perhaps been a general shift of /ŋ/ to /ɡ/ which would have neatly explained both the Nangasaki spelling (earliest Westerners hear -ŋa-, their spelling sticks like the y in “yen”) and the three wolffive fish moon, but then comes Shoichi Iwasaki clearly specifying that the “shoo-gakkoo” split occurs word-medially, thank you for destroying my awesome hypothesis. I take away from this that language is easier understood if I ignore facts.

  27. David Marjanović says:

    High i u are devoiced between voiceless consonants and word finally after a voiceless consonant, and in Tokyoese are present in a more or less spiritual sense only

    Or in a spirant sense. I don’t know where in Japan the colleague was from who explained that the soy sauce brand Kikkoman is pronounced [kçkːoːman] with a loudly hissing, completely voiceless and vowelless first syllable.

    Has the “use and non-use” of [ŋ] explained in David Eddyshaw’s 8:22 post developed from pre-nasalization?

    That’s the common opinion. Old Japanese is reconstructed without [b d g] altogether, and instead with /np nt nk/.

  28. present in a more or less spiritual sense only

    I am totally stealing that.

    the soy sauce brand Kikkoman is pronounced [kçkːoːman] with a loudly hissing, completely voiceless and vowelless first syllable.

    I think your colleague is overselling it a bit, though as you said, it depends on where he’s from. Rather than a spirant it’s essentially a devoiced vowel. There are, for examples, (almost) minimal pairs like:

    食(く)った /kuꜜt.ta/ [kɯ̥ᵝt.ta] “ate”
    切(き)った /kiꜜt.ta/ [kʲi̥t.ta] “cut” (past)

    (I say almost because the initial consonant in the second example is palatalized.)

    From an English (and German?) native perspective, it often feels like these vowels are actually dropped, but they’re definitely still there, taking up spoken morae. It’s no obstacle to comprehension if you say “tomorrow” /asitaꜜ/ [aɕi̥ta] as [aɕta] (or even [aʃta]!), but the difference is felt by native Japanese speakers.

  29. David Eddyshaw says:

    I haven’t got any information on Nagasaki dialect, but in the Routledge “Languages of Japan and Korea” there’s a sketch of the Ei dialect of Kagoshima, which is at least also in Kyushu. It says that, as in Tokyo, i and u are devoiced between voiceless consonants, and may actually be deleted with compensatory lengthening of the preceding vowel in the case of the equivalents of Standard final -shi -ji -su -zu, e.g. [i:ʃ] for ishi “stone.” Resulting word-final fricatives are devoiced, so the forms corresponding to Standard suzu “bell” and susu “soot” are both [su:s]. Word-final i u are altogether deleted after obstruents, which are then all neutralised as /Q/, realised as unreleased [t̚], or assimilated to a following word-initial consonant, e.g. kaQ for kaki “persimmon.” Intervocalic -g- always appears as [ŋ]; this actually preserves the contrast with -k-, which is voiced to [g] between vowels: [hago] for hako “box”; this voicing happens to all intervocalic k t p unless preceded by unvoiced i u.

    Anyway, that should make the Ei pronunciation of Nagasaki [naŋasat̚]

  30. I should mention that Goncharov treats Nagasaki as a plural noun, which makes sense (-i being a Russian plural ending).

  31. Heinrich R. Blutvergieben says:

    @DM: Thanks. Looks similar to Greek -nt- becoming [d], but that’s something I know even less about.

  32. Heinrich R. Blutvergieben says:

    @DE: Thank you, very interesting. Perhaps the example “kaQ for kaki ‘persimmon.'” would illustrate the point better if it included something coming after it – I’d find it hard to wrap my head around Standard Japanese kaki no ki being something like [kan:ot̚] in Ei, and Standard kaki no ki no ne [kan:on:one].

  33. David Marjanović says:

    Rather than a spirant it’s essentially a devoiced vowel.

    That’s what I’ve often read, but it’s nowhere near how he casually pronounced it. (He wasn’t explaining the pronunciation but the etymology.)

    minimal pairs like:

    I suppose that’s a minimal pair of [ɸ] vs. [ç] in the most extreme accents.

    they’re definitely still there, taking up spoken morae

    Sure; voiceless fricatives can do that, even though very few languages outside the Pacific Northwest let them. I should have transcribed [ç̩], assuming that displays legibly on anyone’s screen…

  34. David Eddyshaw says:

    @der arme Heinrich:

    The sketch actually does give “kaN-no” as the Ei equivalent of “kaki no” where N is described as “a moraic homorganic nasal before a consonant and variously but typically [ʉ̃] elsewhere” (describing prepausal standard Japanese final -n.) He seems at any rate to mean a moraic nasal sound unspecified for position, corresponding to non-nasal /Q/.

    Interestingly, in view of the relationship between intervocalic voiced stops and nasalisation, the sketch says that Ei dialect differs from most Kagoshima dialects in that when final -i -u are deleted after the equivalents of standard g b z/j, the resulting final consonant may be /N/ instead of /Q/, e.g. usaN or usaQ for usagi “rabbit”, asoN or asoQ for asobi “play”, aN or aQ for aji “taste.”

    I presume monomoraic words like ki aren’t subject to the vowel deletion but I can’t see any explicit statement about this in the sketch.

  35. David Eddyshaw says:

    “Goncharov treats Nagasaki as a plural noun”

    Quite natural. One Nagasak, two Nagasaks.
    Athens, Thebes, Mycenae … I’m sure there must be non-Greek plural cities out there too, but I can’t think of any offhand.

  36. Heinrich R. Blutvergieben says:

    @DE: Thank you, very interesting as always.

    Plural cities in the Niemczech: München, Essen, Bremen; curiously, neither Frankfurt nor Berlin behave like this.

  37. > I think your colleague is overselling it a bit, though as you said, it depends on where he’s from.

    If he speaks Standard Japanese, he’s not overselling it. Any voicing in the first syllable of Kikkoman would sound weird in connected speech. I don’t know much about devoicing in other dialects, except that there are certain Kansai dialects where it appears less (think “desu” instead of “des”).

    > they’re definitely still there, taking up spoken morae. It’s no obstacle to comprehension if you say “tomorrow” /asitaꜜ/ [aɕi̥ta] as [aɕta] (or even [aʃta]!), but the difference is felt by native Japanese speakers.

    They’re there in a phonemic sense, but in phonetic sense, they’re gone. That is, in the surface form, the mora length is on the ɕ. Put a dot under the ɕ in [aɕta] and you’re fine. The difference is felt by native Japanese speakers because some learners skip the mora, not because they skip the vowel.

  38. David Eddyshaw says:

    More on nasalisation and voiced stops: the Routledge book also has a sketch of Tsuruoka dialect up in gently radioactive Tohoku, which has -d- -g- for Tokyo intervocalic -t- -k- versus -ŋ- [sic] -ᵐb- -ⁿd- -ⁿdz- for standard -g- -b- -d- -z-; so kaki “persimmon” is [kagï], and “key” (standard kagi) is [kaŋï].

    The words corresponding to standard chichi “father” and chiji “governor” thus end up contrasting in prenasalisation rather than voicing: [tʃïʒï] versus [tʃïⁿdʒï]

    The natural idea that this all just preserves the premodern system as is doesn’t work; just as with Ei dialect, the voicing of intervocalic -t- -k- doesn’t happen after devoiced i u, so hato “pigeon” turns up as hado, but hito “person” is still hito. However, it does fit a system where the emic distinction was nasalisation and voicing wasn’t contrastive.

  39. David Marjanović says:

    Plural cities in the Niemczech: München, Essen, Bremen

    Nope, they’re all grammatically singular, and I’m pretty sure they always have been. After so many centuries of syn- and apocope, German has so few distinct endings left that they all have 5 functions on average.

    (…That’s a guesstimate.)

  40. I should probably mention that of course, ɕ can be syllabic, but k can’t (it’s moraic when it’s geminated, though). So kitta would be something like [kʲçt:a] and kutta something like [kxt:a]. The [kɸt:a] that David Marjanović suggests is interesting. Both [x] and [ɸ] seem OK to me. Maybe there’s some coarticulation taking place.

    The devoicing in kitta in Standard Japanese connected speech seems mandatory (to this layman), but not the one in kutta. I have the impression that in general, devoicing in Japanese is an understudied phenomenon. This is all I’ve read about it:

    http://hasegawa.berkeley.edu/Accent/0099.pdf

  41. David Eddyshaw says:

    I should add maybe for the benefit of the two people reading LH who don’t already know it, that the large numbers of Japanese words *beginning* with voiced stops are overwhelmingly loans, mostly (as ever) from Chinese, with a few native Japanese recruits like deru “emerge” which have managed to lose an earlier initial vowel. As far as I can make out, even word-initially, what are now voiced stops are thought to have been prenasalised in premodern Japanese. In the case of the Kan-on layer of Sino-Japanese loans, this probably actually matched the source language, the weird dialect of Chang’an in Tang times, which had turned perfectly innocent initial nasals into prenasalised stops, presumably just to confuse the Turks (hence Sino-Japanese ba for ma “horse” via ᵐba); in other cases, it was presumably just the nearest match available. It wasn’t any easier for Japanese speakers to pronounce Middle Chinese than it is for their modern descendants to pronounce English.

  42. Is Los Angeles plural? Probably not, but how about Las Vegas?

  43. No, neither, as they are short for the City of Los Angeles and the City of Las Vegas respectively. The original Spanish name for the former is “El Pueblo de Nuestra Señora la Reina de los Ángeles de Porciúncula”, again singular. The latter was never a Spanish city, and is named for the meadows (las vegas) that originally grew on the spot where the original fort was built.

  44. I think your colleague is overselling it a bit, though as you said, it depends on where he’s from. Rather than a spirant it’s essentially a devoiced vowel.

    If he speaks Standard Japanese, he’s not overselling it. Any voicing in the first syllable of Kikkoman would sound weird in connected speech.

    ? Exactly. That’s what (I thought) I wrote? (“devoiced vowel”)

    But then again, I can see how I wasn’t that clear. It seems to me that the implication of notations like [kʲçt:a] or [kxt:a] or [kɸt:a] is that a devoiced (phonemic) vowel is phonetically just a syllabic spirant. But is that true? Is there no difference between, say, [sɯ̥] and [su̥]? More to the point, Is there no difference between [aɕi̥ta] and [aɕʊ̥ta]**? My perception is that there is, that (e.g.) the Native English speaker is more likely to produce the latter, and that the sound difference is perceivable (if not at all necessarily perceived). The assumption would be that even when devoiced, vowels are to some degree still actually being articulated by native speakers, and that it’s possible for the foreign learner to get the spirant and devoicing parts right while still misarticulating the vowel. In other words, [ɕi̥] and [ɕʊ̥] would be equally devoiced but differently vocalized.

    **If this isn’t displaying right, I’m trying to indicate the devoiced version of the FOOT vowel [ʊ]

    I grant the objection that such effects may not be relevant when describing native Japanese speech, since potentially contrastive vowels like [ʊ̥] are absent anyway. This would make the notation dainichi suggested–[aɕ̥ta] (dot under the ɕ) completely adequate. We could also try to express the difference as that between two kinds of sibilants. Or just define vowel as “voiced sound” to begin with. Or object on principle to the idea of making phonetic transcriptions too exact (not the first time we’ve debated this issue in reference to, e.g. IPA here).

    But to the extent that articulation is in fact affected by the original phonemic vowel, shoudn’t we consider that vowel as more than just moraically “there”?

  45. David Marjanović says:

    More to the point, Is there no difference between [aɕi̥ta] and [aɕʊ̥ta]**? My perception is that there is, that (e.g.) the Native English speaker is more likely to produce the latter

    That’s because the difference in rounding doesn’t go away – it remains on the syllabic consonant. The English [ʃ] is somewhat rounded, unlike [ɕ], so it’ll sound to Japanese more like shu than like shi when syllabic.

    But to the extent that articulation is in fact affected by the original phonemic vowel, shoudn’t we consider that vowel as more than just moraically “there”?

    It’s there on the phonemic level. What I’m saying is that, in some Japanese accents, it’s not there on the phonetic level.

    Voiceless vowels are not the same thing, phonetically, as fricatives. That’s why whispering is possible, and why… oh, it’s all on Wikipedia.

  46. January First-of-May says:

    Athens, Thebes, Mycenae … I’m sure there must be non-Greek plural cities out there too, but I can’t think of any offhand.

    Naberezhnye Chelny. Velikie Luki.
    Cannes too, I think, but I’m not sure whether it’s actually based on something in French (Wikipedia is confusing).

    Apparently the Greek city of Syracusae is plural, but the modern city of Syracuse is not (whether Italian or American).

  47. While we’re at it, neither is Cincinnati.

  48. My guess would be that in English the only plural place names (i.e., those that can take plural verbs) are phrases with “the”: the Philippines, the Netherlands (and even these can be treated as singular: “The Philippines is a big country”). Even historical plurals that are treated as plurals in pretty much every other language and that look like English plurals, e.g. Athens, are treated as singular in English.

  49. Graham Asher says:

    Liddell and Scott give two variants of Syracuse: the first plural (with some variants I won’t give here) and the second singular: ἁι Συράκουσαι, ἡ Συράκουσα.

  50. Heinrich R. Blutvergieben says:

    The French Wikipedia articles on the communes of Les Abymes and Les Mureaux use mostly the plural,[1] but sometimes the singular.[2]

    [1] Plural examples: “Les Abymes sont érigées en paroisse en 1726”, “Les Abymes appartient à …”, “Les îles de la Seine situées en face des Mureaux”, “Les Mureaux comprennent”, “jusqu’aux Mureaux”, “Les Mureaux font partie”

    [2] Singular examples: “Les Abymes (en créole Zabim ou Zabym) est une commune française”, “Les Abymes, bien que n’étant pas une station balnéaire, est le deuxième pôle économique”, “Les Mureaux est une commune”, “Les Mureaux accueille des épreuves de voile”

    @DM: Right. I can’t think of any German plural city name, why is that? After all, there are plenty of non-city toponyms in the plural, for example die Alpen, Königreich beider Sizilien, and probably quite a few Stadtteile.

  51. Voiceless vowels are not the same thing, phonetically, as fricatives.

    Obviously. And of course, people feeling sure they’re hearing a phoneme not objectively present in a particular phonetic realization is a familiar psycholinguistic phenomenon.

    I offered the contrast between [aɕi̥ta] and [aɕʊ̥ta] (note that I didn’t write [aʃʊ̥ta] or [aʃta], though both of those are of course frequently produced by English-native learners of Japanese) out of my perception that (1) it’s important to get both the sibilant and the articulatory position of the (devoiced) vowel correct, and that (2) it’s quite common to achieve the former without the latter.

    Are you skeptical that in the [kç̩k:o:maɴ] you heard, the fricative [ç̩] was being produced in the articulatory position for [i̥]? Or that, say, a native English speaker might successfully produce a [ç̩] that would nevertheless sound different because produced in a different articulation?

  52. David Marjanović says:

    I can’t think of any German plural city name, why is that?

    I think we’d need an article, and place names with articles are exceedingly rare in the first place. (Country names with articles are more common than in English.)

    Athen, final stress, is singular.

    Are you skeptical that in the [kç̩k:o:maɴ] you heard, the fricative [ç̩] was being produced in the articulatory position for [i̥]?

    No, but that’s pretty much what makes it [ç], as opposed to [x] or at least [çʷ].

    (BTW, I heard [n], not the uvular [ɴ]. I guess that should help place him in Japan, right?)

    Or that, say, a native English speaker might successfully produce a [ç̩] that would nevertheless sound different because produced in a different articulation?

    I’m indeed skeptical of that. More likely, it just wouldn’t be [ç] anymore.

  53. Baden Baden.

  54. David Marjanović says:

    Baden-Baden is singular, and so is Baden in Austria. They look like “to bathe”; “baths” would be Bäder.

  55. Denmark has several place names with unambiguously plural (definite) morphology, like Hedehusene and Kandestederne. But I’d still go with det gamle Hedehusene treating it as an opaque proper noun.

    On the other hand it can’t always be treated as singular; saying Hedehusene er grøn sits badly, unlike København er grøn for instance, because it isn’t really singular. Grønne is no better, because it isn’t really plural 🙂 But strangely enough the neuter form grønt works with both — I think it’s understood as an adverbial function to an implicit verb like ‘adopt’ or ‘work’.

    (Non-participial adjectives in predicate position is almost the only position where modern Danish has strong gender/number agreement).

    In the fossils department we have Husum which is an old plural dative in allative function, ‘to the houses’.

  56. Wikipedia told me Baden is the old plural noun, rather than the present-day verb.

  57. David Eddyshaw says:

    “El Pueblo de Nuestra Señora la Reina de los Ángeles de Porciúncula”

    Same principle (more or less) as the given name [María de los] Dolores.
    I once knew a [María de] Covadonga, who actually went in normal life by the name Cova.

    The Spanish do names well (as the thread on Khevenhüller is also currently demonstrating.)

    Must be the Arabic influence … just as every Spanish woman is called María until proven otherwise, thus leading to all this inventiveness, so the fact that every other Muslim bloke is called Muhammad has generated a great variety of by-names.

  58. David Eddyshaw says:

    Putting the acute accent on “María” made me wonder why the stress is on the i. French “Marie” too.
    It’s not like “Darius”, where the i was long and thus accented by the usual Latin rule. The original Miriam/Maryam has [j], where it’s not even syllabic, let alone long. The Wikipedia article suggests that the Latin form is actually a repurposing of the gentilic Maria, but that can’t be right; the famous general Marius was stressed on the antepenult as expected, like his descendant the plumber.
    Presumably it’s just taken over from Greek; the penult acute accent there follows the regular rule, with the final -a taken as long. That would mean that the position of the accent was determined in the period where the Classical pitch-accent based position constraints applied, but borrowed into Latin after the system had become one of stress. I suppose that’s logically possible.

  59. David Marjanović says:

    after the system had become one of stress

    Stress is stress. Classical Greek had two kinds of stress instead of just one like Latin, but I bet both were heard as stress by native Latin speakers.

  60. David Eddyshaw says:

    @DM:

    That can’t have been the case in the classical period, at least. Part of the artistry of the masters of Latin hexameter is how they play off the word stress against the metre, typically setting up clashes in the first part of the line which resolve into agreement in the second part. There’s nothing like that in classical Greek hexameter, at least as far as the pitch-accent goes; I’ve read attempts at trying to locate word stress in Classical Greek independent of the accent by looking for something analogous to the Latin practices in Greek verse.

    But: reading your post more carefully (perhaps I should get into the habit) I see you know that … even so, how would a poor Latin speaker go about hearing both pitch and dynamic stress as stress in the same word if they fell on different syllables?

    Anyway … in this context, of course, it’s the phonology of a later period that matters, for both languages. But what about another word that on first principles you’d have thought would be borrowed by similar people in a similar time frame: εκκλησία? You get French église, Italian chiesa, Spanish iglesia (Welsh eglwys, come to that.)

    Thinking about it, it’s not too easy to come up with counterexamples. There are lots of Greek words with the accent on the final syllable, but you can readily imagine that that would have been too alien to Latin habits to carry over. I’m trying to think of Greek words which are proparoxytone but have a long penult, and were borrowed into Vulgar Latin rather than as learned imports. Can’t come up with many, apart from these -ia words, where the more I think about it, the more something odd seems to be going on.

    Learned imports into the various Romance languages of Greek words ending in -ia seem to regularly carry over the penult stress: monarchie, monarchia, monarquía; philosophie, filosofia, filosofía. On the other hand, Latin infamia gives French infamie but Italian infamia, Spanish infamia with penult stress. There must have been a lot of remodelling and analogy going on. I expect entire theses have been written about this, but I just never noticed it before.

    I’ll bet numerous Hatters know all about it … I request enlightenment.

  61. French has no stress, only the relics of it.

  62. David Eddyshaw says:

    True; but the relics fall on what is now the final vowel other than e muet, and are still in their historical place, so I thought it worth adducing “monarchie” < monar'chia, in contrast to (say) "gloire" < 'gloria. You’re right that I was perpetrating some anachronisms, I admit. And “monarchie” must be a construct of monarch + ie rather than an inherited unit, because there’s no way that the ‘ch’ [ʃ] could be a regular historical development. OK, fair cop.

    I just looked at the index of suffixes in Paden's "Introduction to Old Occitan" which just says under -ia: "Classical Latin -ĭam, -īam, influenced by Greek -ía." Seems undeniable, but doesn't solve the problem of how the Greek form led to Latin forms with a long ī.

    The -ia words with penult stress all look like learned borrowings rather than straightforward developments via Vulgar Latin (apart from Maria herself, who first got me wondering.) So I suppose the interesting thing is that the borrowers in question used -īa for the Greek words but -ĭa for words borrowed from Classical Latin. Come to think of it, in quite a lot of the Latin words, the -ia is actually part of a recognisable larger suffix, like -ntia, so you could imagine an Italian borrower (say) happily automatically turning -ntia into -nza as a unit in words like pestilenza.

    And now I think of it, there actually are non-Greek suffixes ending in -ia which regularly end up with stress on the i: -aria, for example, in chevalerie, cavallaria, caballería, Obviously not directly inherited from Classical Latin, mind.

  63. David Eddyshaw says:

    Come to think of it, shouldn’t the reflex of -īam in French be -oie? But voie (for example) is in any case from Classical Latin vĭam, not vīam … I am clearly out of my depth here.

    Did French -ie actually owe its vocalism to being a learned loan import, rather than a reflex of a Latin by-form in -īam?

  64. David Eddyshaw says:

    C’est la vie .. from vītam. Got my reflexes muddled. It’s Latin ē that gives French oi, not ī.
    I shall now go to bed and hope that someone knowledgeable can explain it all to me tomorrow.

  65. > Voiceless vowels are not the same thing, phonetically, as fricatives. That’s why whispering is possible

    In theory, right. But in practise, I think the amount of air required to make the vocal chords vibrate is so small, that if you just turn off the voicing and change nothing else, you can hardly hear anything. I think people tend to add frication when whispering for this reason, and that’s probably also part of the reason why the devoiced Japanese vowels tend to turn into fricatives.

    Also, I think we’re moving into the area where it’s a matter of where you draw the distinction between the definition of different sounds. For example, in Japanese, the surface form of (voiced) /fu/ would usually be described as something like [ɸɯᵝ], i.e. a unvoiced bilabial fricative followed by a vowel, but when I produce it, I don’t move my lips/mouth at all when I turn on the voicing. So is it really [ɸβ̩]? So the question of whether, say, a devoiced /fu/ is [ɸɯ̥ᵝ] or [ɸ:] seems like a very technical one which depends on detailed definitions.

    > Non-participial adjectives in predicate position is almost the only position where modern Danish has strong gender/number agreement

    Can you elaborate on that? Not in indefinite attributive position?

    > Husum which is an old plural dative in allative function

    Interesting! I always thought it was the same kind as Virum/Farum, but here -rum actually turns out to mean something like “place”. Who’d have thunk.

  66. marie-lucie says:

    We need Etienne here.

  67. Not in indefinite attributive position? — That is the other one, but something strange happens with place names: with an indefinite attribute adjective they can always construe in the neuter singular. Et grønt Hedehusene/København/Kødbyen are totally unremarkable even though Kødbyen is singular common definite in form, so this gives no strong evidence about the plurality of Hedehusene.

    But you could say en grøn Kødby because the name is so transparent (‘meat packing district’), that you can treat it as a common noun, while ?grønne Hedehuse is right out. Also this may actually be what is bleeding through in Hedehusene er grønt from my previous post, instead of my adverbial theory.

  68. David Marjanović says:

    how would a poor Latin speaker go about hearing both pitch and dynamic stress as stress in the same word if they fell on different syllables?

    They didn’t, and that’s my point! Outside of tone languages, pitch and stress come as a package; stress languages don’t distinguish different kinds of pitch on stressed syllables, pitch-accent languages do.

    Like pitch in tone languages, so is length independent of stress in some languages (Czech, Hungarian, Old High German…) while others distinguish length only in stressed syllables (MHG and modern German generally) and yet others don’t distinguish length at all, keeping it either constant (Polish) or making all stressed syllables long by default (Russian, Central Bavarian).

    French has no stress, only the relics of it.

    French has no phonemic stress. Phonetically, stress is very much present – it’s always utterance-final.

    For example, in Japanese, the surface form of (voiced) /fu/ would usually be described as something like [ɸɯᵝ], i.e. a unvoiced bilabial fricative followed by a vowel, but when I produce it, I don’t move my lips/mouth at all when I turn on the voicing. So is it really [ɸβ̩]?

    Maybe. Do you move your tongue?

    An obvious possibility is that you’re in the middle between fricative and vowel and produce an approximant…

    Interesting! I always thought it was the same kind as Virum/Farum, but here -rum actually turns out to mean something like “place”. Who’d have thunk.

    I thought -um in North Sea place names was a reduced version of “home”… is that perhaps true for Frisian but not for Danish, or something?

  69. We have -um¹ (Smørum) and -um² (Husum), the first one is as you say and the second is the old plural dative-for-locative. Not to be confused with -rum (Virum). TFoAK agrees.

    (-um² is dated to the Viking age, but I would guess it had been in productive use since PIE and only became lexicalized as a place name element when noun declensions began to disappear. Singular dative wasn’t distinctive enough to be lexicalized, though I’m sure it must have been used too).

  70. Steenstrup 1908 (in Danish) has more; he is of the opinion that spellings with -hem in old sources are mostly spurious and the dative plural is more common than commonly thought (then). I haven’t been able to find newer works online.

    EDIT: Why can I link to heimskrlngla.no and escape moderation while da.wikipedia.org is stopped? Did someone list WiPe as a spam host? One of the mysteries of modern life…

  71. There are several Anglophone settlements named “The Meadows”. The one I’m (slightly) familiar with, a neighbourhood of Nottingham, has been treated as a grammatically singular opaque proper noun by everyone I’ve heard speak of it.

  72. January First-of-May says:

    Cannes too, I think, but I’m not sure whether it’s actually based on something in French (Wikipedia is confusing).

    Apparently the Russian language had somehow conflated Cannae, the site of the famous Second Punic War battle (in what is now south-eastern Italy), which does appear to actually be plural, with the completely unrelated city of Cannes in France (the site of the famous festival), which is apparently supposed to be singular.

    (Neither of those are to be confused with the entirely unrelated, but also similarly pronounced, Caen in Normandy.)

  73. Sort of similarly, but Caen has no /n/ (just a nasal vowel).

  74. David Eddyshaw says:

    “Maria”:

    Thinking about it a bit more, because of the loss of contrastive length, in Vulgar Latin the position of stress was not predictable, as it is in Classical Latin. However, penult stress was the unmarked default, partly because that was commoner to begin with and partly because many of the words with original antepenult stress had either lost the penult vowel, or the penult vowel had lost syllabicity, as in “gloria.”

    In inherited vocabulary, there would have been few or no words ending in -ia(m) by this point, and none with antepenult stress, because in that case the “i” would always have lost syllabicity.

    Greek words in -ia would have been borrowed, naturally, as [ia], contrasting with the [ea] of the reflex of Classical vĭam (French voie.) This would constitute a reintroduction of a sequence otherwise missing from the language, and it would be natural for such words to acquire the default, penult stress, even if this were not reinforced by hearing the Greek words as bearing penult stress. Of course it may have been,as DM suggests; but I reckon you can account for the phenomena without this assumption.

    This helps with “Maria”, where you only have to suppose that the “i” was tense, not that it was actually *lengthened* at some point. I have difficulty with this lengthening because it assumes a change in contrastive length in Latin at a time period when Greek had developed dynamic stress, which itself was surely tied up with loss of contrastive vowel length in Greek; AFAIK the loss of length occurred at more or less the same time in both languages. SimiIarly, I don’t think Paden can be right in setting up the suffix as -īa; that would be anachronistic.

  75. David Eddyshaw says:

    @January:

    I’d forgotten Cannae (no Roman, I.) Thinking (a bit late in the day) to consult my trusty Latin grammar under “pluralia tantum”, I find there’s a whole subsection on place names. “Falerii” is another one, and “Leuctra.” My Greek grammar reminds me of “Delphi.” Greek and Latin do both treat such names as grammatically plural, I think. Some of them are metonymous uses of the name of the inhabitants, which makes sense; not all though, by any means.

  76. J.W. Brewer says:

    Re “The Meadows,” wikipedia’s likely incomplete list gives no fewer than a dozen different places in the U.S. (half free-standing municipalities, the other half neighborhoods within larger cities) named “Forest Hills.” Anarthrous, and treated as singular for verb-agreement and pronoun purposes. I assume the real-estate developers and promoters of the Anglophone world have generated lots of other structurally-plural toponyms that have a similar “marketing” sort of feel. The suburbs of Cleveland include about a dozen municipalities named X Heights, of which Shaker Heights might be the best known, plus several named Y Hills, including the seemingly redundant Highland Hills. Obviously someone thought perceived altitude would be good for property values.

  77. David Eddyshaw says:

    And Syracuse.

  78. David Eddyshaw says:

    In my Maria-disquisition I see that I’ve needlessly overcomplicated the issue: if Vulgar Latin had default penult stress, and there were no words in the inherited vocabulary ending in [ia] by that point, there was no model for not just giving loanwords in -ia the default stress pattern. The “i” in “Maria” doesn’t need to be of any particular quality: it just needs to be syllabic.

  79. In California, Willows was once Willow, Needles is named after The Needles, and Palm Springs always has been that. I’m pretty sure Cayucos is not a shortened name, unlike many plural Spanish names which are short for “Rancho de los…”

    (And as was discussed here once, Adirondacks is not a plural.)

  80. David E: Another source of antepenult stress, the mutus cum liquida exception, was lost altogether, which is why tenebras ‘darkness.ACC’ > Sp. tinieblas, not *tiéneblas.

  81. David Marjanović says:

    Of course it may have been,as DM suggests; but I reckon you can account for the phenomena without this assumption.

    Point taken.

  82. Non-participial adjectives in predicate position is almost the only position where modern Danish has strong gender/number agreement
    Interesting. It’s the only position where German doesn’t have gender / number agreement.
    On plural city names in German: many of those in -en are etymologically plurals (sometimes Dative plurals, like the Danish examples), but they are all treated as singular.

  83. Yes, that’s interesting — do you know when it disappeared? And is it just German or all of WG?

    It’s showing no sign of disappearing in Danish or Swedish at least, except as I said for participles in Danish — like German, but unlike Swedish, the perfect construes with forms of ‘to be’ for certain verbs, and formerly (up to the 1933 bible in fact) the participle was treated as a predicate adjective agreeing with the subject: Gjæsterne ere komne ‘the guests have arrived’. But now we use the same (old neuter) form as with the ‘have’ perfect (where the participle would logically agree with the object if there is one, but I think that was gone by Old Norse times already): gæsterne er kommet.

    (Swedish generalized the ‘have’ perfect long ago, and for strong verbs developed a ‘supine’ from the neuter participle in unstressed position, so regular participles in predicate position are never part of a perfect and retain agreement: Jag har brytit armen/benet / min arm är bruten / mitt ben är brutet ‘I have broken my arm/leg / my arm/leg is broken’).

  84. Yes, that’s interesting — do you know when it disappeared? And is it just German or all of WG?
    I’d have to check, but my books on Old High German and Middle High German are at home. I can check when I go back Mid-March.
    But as far as I remember, Dutch is like German in that regard; while it distinguishes neuter-singular from non-neuter-singular in other positions, it doesn’t do so with predicative adjectives. I don’t remember whether Old English still had number or gender distinction in that position.

  85. marie-lucie says:

    LH: Caen has no /n/ (just a nasal vowel

    Absolutely. It is pronounced as if written Can. The inhabitants are les Caennais, pronounced as if Cannais. They cannot be confused with the inhabitants of Cannes, who are les Cannois.

  86. David Marjanović says:

    And is it just German or all of WG?

    All I know is that the Highest Alemannic dialects in (mostly) central Switzerland retain agreement.

  87. Greg Pandatshang says:

    OT: I understand the basic principle of Latin accentuation. I’ve never been sure how to handle short /i/ and /u/ that could potentially be treated as semivowels. Should “puerī” be /ˈpu.e.riː/ or /ˈpwe.riː/? Should “Iulianus” be or /ju.ˈli.a.nus/ (treating the “i” as the antepenultimate vowel) or /ˈjul.ja.nus/? I guess such a vowel could even be treated as part of a rising diphthong, so, e.g. /ˈae̯.mi.li̯a/???

  88. The answer to these questions is basically ‘it depends’.

    The spelling is ambiguous, you have to know or deduce from morphology when a written i or u is a separate syllable or a semivowel. Allen in Vox Latina spends 5 pages on the semivowels without once giving a rule, but does mention the the Romans themselves felt free to change the syllabification to fit a meter; Iulius can be /i.(i̯)u.li.(i̯)us/ instead of /i̯u.li.(i̯)us/, abiete can be /ab.i̯e.te/ instead of /a.bi.(i̯)e.te/. (An ‘automatic glide’ is assumed to appear between i/u and neighboring vowels).

    Also, in general intervocalic is geminated/ambisyllabic and makes the first syllable heavy even though the vowel is short, and because of a dislike for writing two identical letters in a row, there are cases like reicit /rei̯.i̯i:.kit/ where i stands for the sequence of semivowel and vowel. On the other hand, Gaius is /ga:.(i̯)i.(i̯)us/.

  89. Greg Pandatshang says:

    Oh, wow, I assumed there would be some kind of rule of allophony determining everything. Sounds more like they were starting to develop a phonemic underlying distinction between sequences like jV vs. ĭV vs. ĭjV (assuming a short unstressed /i/) that, in addition to not being represented in the orthography, never quite gelled into regularity.

  90. Presumably the Latin that Cicero used when he chatted with his buddies in the Senate had unambiguous phonemes for full and semi-vowels. Clearly they were felt as being similar enough that the system could be tweaked when speechifying or constructing poetry, but the ambiguity we perceive is because of the rather hit-and-miss orthography.

    (According to Allen, the reason why intervocalic i always denotes a geminate is that single i̯ had fallen in that position so there was no contrast to maintain in writing).

  91. All I know is that the Highest Alemannic dialects in (mostly) central Switzerland retain agreement.
    Good to know! That means the loss of the agreement is probaly not Common West Germanic.

  92. @Greg: In general, I don’t think forms like /ˈpwee.riː/, /ˈjul.ja.nus/ and ˈae̯.mi.li̯a/ are standard classically. In Proto-Romance unstressed short /i/ and /e/ became glides after a consonant, but in classical Latin they’re usually separate syllables (barring, as Lars mentioned, poetic alterations to syllabification). Wikipedia says iūliānus has a long a, so the stress would have to go on the penult anyway. I don’t think stressed /u/ or /i/, as in “pueri”, could be turned into a semivowel in poetry (and likewise, I don’t think a semivowel could be converted into a stressed syllable in poetry). Also, I don’t think poetic contraction would change where the stress is placed on a word, so /ˈae̯.mi.li̯a/ looks particularly unlikely to me. Other types of contraction in Latin generally don’t change stress: e.g. “Vergili” with short stressed i in the penult (from earlier “Vergilii”).

    The main cases where you have Cj are in compounds with iacio. Actually, many of these have derivatives in English spelled with “j” that can help with remembering this, like “adjective”, “abject”, “object”. In non-compound words, Cj doesn’t really occur as far as I know. Cv is more common, but it’s a bit harder to figure out from the phonological context when it occurs compared to Cu (I think that’s why the modern standard has settled on writing the semivowels in an unasymmetrical fashion, with “u” and “v” distinguished but not “i” and “j”). But in any case, you can usually rely on English spelling to help you out if you know of a related English word, so “INVOLVO” generally had three syllables while “INSINVO” generally had four syllables.

  93. I’m surprised no one yet has referred to Virgil’s Laviniaque venit / Litora, where the first word has to be scanned Lavinnjaque. And of course there are two other “irregularities” (oral pronunciations) in the same line.

  94. Eli Nelson says:

    Oh, I have no experience with Latin poetry. Is it definite that “Laviniaque” has to scan as ending with …HLL, though, and not with …LLLL with a substitution of two light syllables for one heavy one? I think I read somewhere that such substitutions were possible in some positions in a line.

    (Apparently not here, though. I just Googled it, and all sources say that this must be scanned “Lavin.jaque”.)

  95. Another interesting difference between Danish/German is that adjectives with possessive determiners are strong (or “mixed”, I think it is, technically) in German, but weak in Danish. German “mein guter man”, but Danish “min gode mand”. I asked why on Linguistics Stackexchange a while ago but didn’t get an answer. I’m sure you guys can help me figure out the distribution for other Germanic dialects.

    http://linguistics.stackexchange.com/questions/8561/definite-or-indefinite-adjectives-with-possessive-personal-determiners-in-german

    Another “fun fact”: attributive adjectives without determiners can be both strong and weak in Danish: “Kært barn har mange navne” (Dear child goes by many names), but (vocative) “Kære barn!” (Dear child). I guess this can be explained as an implicit “et” (a) in the former case and an implicit “mit” (my) in the latter.

    > “So is it really [ɸβ̩]?” Maybe. Do you move your tongue?

    No, I only turn on the voicing, and maybe adjust the airflow (I’m not good at estimating this latter one).

    > An obvious possibility is that you’re in the middle between fricative and vowel and produce an approximant…

    Yeah, but this is again an area where I’m confused about the technicalities… can approximants be syllabic, or are they consonants by definition? This is what Wikipedia says about ɰᵝ:

    https://en.wikipedia.org/wiki/Voiced_labio-velar_approximant

    > The type of approximant is glide or semivowel. The term glide emphasizes the characteristic of movement (or ‘glide’) of /w/ from the /u/ vowel position to a following vowel position. The term semivowel emphasizes that, although the sound is vocalic in nature, it is not ‘syllabic’ (it does not form the nucleus of a syllable).

  96. Syllabic approximants do exist, but we usually call them vowels. Per contra, non-syllabic vowels are glides, which are a subtype of approximants.

  97. David Marjanović says:

    Syllabic approximants, even not counting [l], are not the same thing as vowels, though I’m sure no language distinguishes them.

    The yi, wu, yu of Standard Mandarin are probably syllabic approximants; they’re not simply vowels or approximant + vowel sequences.

  98. My understanding is that they are simply vowels. This is how they are spelled in Gwoyeu Romatzyh, for example.

  99. Eli Nelson says:

    @David Marjanović: I’ve never heard that distinction made before. My understanding was that in the IPA, pairs like [i] and [j] or [u] and [w] are strictly speaking defined as identical except for syllabicity. Obviously a high vowel can have a greater or lesser degree of closure, and apparently it’s traditional in Bantu linguistics to refer to “superhigh” or “superclose” vowels, but I’ve never heard these vowels referred to as “syllabic approximants”. There is the analysis of PIE *i and *u as syllabic allophones of *y and *w but I have never seen this argued to be anything but purely phonological. I’d think that if a sound has enough closure that you can’t call it a vowel, you’re probably into the realm of fricatives.

  100. David Marjanović says:

    My understanding is that they are simply vowels.

    That’s not quite what they sound like. But they’re clearly not voiced fricatives either.

    are strictly speaking defined as identical except for

    Wouldn’t be the first time the “International Phonemic Alphabet” misses a phonetic distinction. Fortis/lenis comes to mind (phonemic in much of southeastern German, without any added voice or aspiration or glottalization).

  101. Eli, I don’t know that the IPA has any official statement about where [i] and [j] etc. differ, and phonology is not what it does.

    Phonetically, there’s more to glides vs. vowels than tongue position, in particular the dynamics of tongue position. Vowels tend to have a stable mid position mid-segment, but glides keep changing.
    There was a series of papers on the subject in Lingua, vol. 118 (2008).

    David, what do you mean by “The yi, wu, yu of Standard Mandarin are probably syllabic approximants”? How would you represent yi and yu phonetically?

  102. Lojban does have an opposition [i] ~ [ji], [u] ~ [wu], but only in the four words i, ii, u, uu and potentially in borrowings and proper names.

  103. Looks like I sparked an interesting discussion!

    > Fortis/lenis comes to mind (phonemic in much of southeastern German, without any added voice or aspiration or glottalization)

    … or length? That’s interesting. Examples/references? We’re not talking about the /p-b/ /t-d/ /k-ɡ/ /s-z/ /ʃ-ʒ/ distinctions in Standard German then, obviously.

    (Not being a linguist) I have never understood the fortis/lenis distinction on the phonetic level. Sure, phonemically it makes a lot of sense.

    > David, what do you mean by “The yi, wu, yu of Standard Mandarin are probably syllabic approximants”?

    Sorry, I know that’s not a question for me, but I would guess the point is that they’re too close/rounded to be categorized as pure vowels. So the question would be which of these 3 are the same or different [i] [j̩] [ʝ̩] (I hope the syllabicity dots show).

    > pairs like […] or [u] and [w] are strictly speaking defined as identical except for syllabicity

    One (possibly too obvious to mention) point is that when used broadly, they can differ for a particular language. For example, AFAIK there’s lip movement between the /w/ and the /u/ in English “woo”, so (ignoring any diphthongization of /u/) [wu:] might make more sense than the possibly obscure/misleading [u̯u:].

  104. Quite right. In particular, the lip shape for what’s called [w] and for what’s called [u] are not necessarily the same within a language or across languages, though the IPA defines them both, broadly, as ’rounded’.

  105. Correction, [w] is a ‘labial-velar approximant’, which is another way of saying “an even more rounded high back vocal”.

  106. January First-of-May says:

    Back when I studied French, I could never really figure out their two different kinds of /w/. I think they’re technically different things even in IPA, but they still sounded very similar.

    Pretty sure Russian has [ji] (usually spelled ьи), but I don’t think it can appear at the start of a syllable. Ukrainian has separate letters for [i] and [ji], but I don’t know enough to say if there’s a minimal pair.

  107. @05-01, do you mean the oui/huit difference? If you think they are semivowels, the second is fronted [ʊ̯i/ʏ̯i]; if approximants, palatal [wi/ɥi]. Don’t say the IPA isn’t flexible!

  108. @David Marjanović: the IPA doesn’t attempt to assign a different symbol to every possibly distinguishable vowel sound. The vowel space is continouous, after all. Diacritics such as raised [i̝] and lowered [i̞] can be used. I don’t understand how the [j̩] that you’re talking about would be distinguished from [i̝].

    @dainichi: I have also wondered if the “lenis” “fortis” distinction of German dialects without aspiration/voice/glottalization includes length differences. That seems to be somewhat common in other languages (e.g. from what I remember about Navajo the difference between /s/ and /z/ is partly length, since /z/ is often voiceless phonetically). If it’s not, I really can’t imagine what the difference sounds like: are the fortis consonants louder? Do they have a stronger effect on the articulation of adjacent vowels?

    @Y: Oh, certainly there are probably many phonetic differences like this between glides and vowels that I’m probably not aware of. But the one you mention only seems possible if the glide is a non-syllabic transitional segment between two other vowels. I don’t see how it would be possible to pronounce [j̩] as an isolated syllable with a gradiant change from the silence before to the silence after.

    @JohnCowan: certainly distinctions like [i] ~ [ji], [u] ~ [wu] are possible broadly speaking, but I think the nature of the realization will be language-specific (which is what “broad transcription” means, more-or-less). In some languages, such as English, high vowels may be diphthongized slightly, so from what I understand English [ji] would start high, dip and then end high again. In other languages, word-initial semivowels, or onset semivowels, or even just specifically semivowels before [i] may be realized with more closure than the average [j], possibly even some frication, so it would start very high and then go down to just high. As Y mentioned, [w] and [u] might have different kinds of rounding.

    @January First-of-May: French has /w/ as in “Louis” and /ɥ/ as in “lui”; are those what you are talking about? The difference is like the one between /u/ as in “tout” and /y/ as in “tu”, just made non-syllabic.

  109. I don’t see how it would be possible to pronounce [j̩] as an isolated syllable with a gradient change from the silence before to the silence after.
    That’s how I pronounce it! As you move the tongue into position and back, the formants move and move back, plus some fricative noise (thanks to which j̥ sounds different from i̥ ).

  110. Eli Nelson says:

    @Y: So [j̩] is the sound I would get if I try to pronounce just the start of the word “yes” and then extend it? It’s true this doesn’t sound like [i], but as you say, it seems to have some friction, so wouldn’t it be better to transcribe it as [ʝ̩]? I guess it’s possible to argue for either. This is like the previously-mentioned issue of whether to transcribe [i̥] or [ç̩] (although I get the impression that voiceless fricatives are harder to distinguish from the corresponding approximants than voiced fricatives are even when they are non-syllabic).

    @David: is that what you were talking about in Chinese? That there is some fricative-like quality to the sound of “yi” that isn’t in for example the first syllable of French “idée”? That seems plausible to me, since Chinese has syllabic fricatives or at least very fricated vowels in syllables like si ci shi chi, but I don’t remember heard that mentioned before.

  111. [j̩] is the sound of ‘yes’ without the ‘es’; no need to extend it, just say ‘yes’ slowly enough so you can stop after the y. It’s not as close as [ʝ], which has a noticeable pressure difference between the back and front of the tongue. [j] doesn’t produce any fricative noise, which [ʝ] does.

    (And I speak as an expert on the letter y.)

  112. Maybe the start of [jǝp̚], as in “yup” would be better, where the tongue position of the vowel is close to that of the [j].

  113. David Marjanović says:

    Whew, I’ve made a lot of work for myself here!

    I now remember an example of [j̩] in English. It’s from the background noise of my late childhood, a song by the band Kelly Family that everyone else* was apparently listening to:

    ||: So-ometimes I wish I were an a-an-ge-el,
    so-ometimes I wish I were yyyyy… :||

    – evidently “you”, but the vowel never came. (BTW, the “I” in “I wish” was extremely reduced, too – though the one in “I were” wasn’t reduced at all, but pronounced as long as “wish” or “were”.)

    * Apart from the author of the graffito that read:
    STOPPT DIE
    TIERVERSUCHE
    NEHMT DIE
    SCHEIß
    KELLIES
    – stop animal experimentation, take the motherfucking Kellies instead.

    While many languages disallow the sequences /ji/ and /wu/, they do occur. English is a bad example because FLEECE and GOOSE are [ɪi̯ ʊu̯] or thereabouts in most accents, but the Russian -ии and BCSM -ији/-iji endings (but not the cognate Polish -ii) are [iji]. I also agree about Russian -ьи. I’m pretty sure [ji], [wu] and [jy] ([jʷy]?) also occur in some Mandarin accents.

    David, what do you mean by “The yi, wu, yu of Standard Mandarin are probably syllabic approximants”? How would you represent yi and yu phonetically?

    [j̩ ɥ̩]; and wu is [w̩].

    the point is that they’re too close/rounded to be categorized as pure vowels

    Yes!

    do you mean the oui/huit difference? If you think they are semivowels, the second is fronted [ʊ̯i/ʏ̯i]

    Rather [u̯i y̑i]. And the -t of huit isn’t silent.

    the IPA doesn’t attempt to assign a different symbol to every possibly distinguishable vowel sound.

    Yep, that’s part of my point.

    Diacritics such as raised [i̝] and lowered [i̞] can be used. I don’t understand how the [j̩] that you’re talking about would be distinguished from [i̝].

    Not at all, actually. It’s just that [i] is by definition at the very edge of vowel space, so [i̝] isn’t a vowel anymore, it’s raised out of that space, and transcribing it that way could be considered self-contradictory or misleading.

    @David: is that what you were talking about in Chinese? That there is some fricative-like quality to the sound of “yi” that isn’t in for example the first syllable of French “idée”?

    There is something there that isn’t in idée, but I wouldn’t call it fricative-like. It’s less fricative-like than my native /j/, which seems to be [ʝ̃] or something.

  114. Eli Nelson says:

    @Y and @David:

    Thanks to both of you for the helpful explanations.

    [i] is by definition at the very edge of vowel space, so [i̝] isn’t a vowel anymore, it’s raised out of that space

    I hadn’t thought about that. It does make sense.

    [j̩] is the sound of ‘yes’ without the ‘es’; no need to extend it, just say ‘yes’ slowly enough so you can stop after the y. It’s not as close as [ʝ], which has a noticeable pressure difference between the back and front of the tongue. [j] doesn’t produce any fricative noise, which [ʝ] does.

    I don’t seem to be able to navigate between the Scylla of /ʝ/ and the Charybdis of /i/, or maybe I’m just not good at recognizing the presence or absence of friction. When I try to say “yes” slowly cutting off the sound after the /j/, I feel like the sounds I produce fall into three categories: [j] followed by a very short vowel, like [jə], [i], and a third sound that seems a bit buzzing, which I’ve been interpreting as the presence of friction.

  115. David Marjanović says:

    On fortis & lenis, here’s all I’m not too tired to say tonight.

    First read from here to the end. 🙂

    The terms were actually invented for Swiss dialects, where the distinction is indeed length, and length alone. Let’s pretend that never happened. 😐 Swiss dialects distinguish short lenes from long lenes, even word-initially, and there are no fortis plosives.

    Because there are so few idioms where fortis vs. lenis is phonemic without any additional feature (voice etc.), the topic is massively underresearched as far as I can tell. I think the phonetic difference lies in air pressure (from the lungs, without any glottal shenanigans). Beyond my own proprioception, this fits the following observations:
    – They’re extremes on a spectrum. The Spanish /k/ is clearly lenis, the Russian and Japanese /k/ are clearly fortis, and the French /k/ is intermediate (with some variation in both directions, I think). Also, I’m bad at hearing the distinction for fricatives, because my kinds of German have a length distinction but no fortis/lenis distinction for them: the long ones are always fortes, the short ones seem to have free variation across the whole spectrum (but no further – they never get voiced).
    – The difference is only audible in the release. When the clusters |t(ː)t|, |dt|, |t(ː)d| and |dd| arise across morpheme boundaries, they surface as [tː], [tː], [d̥ː] and [d̥ː], respectively, in my kinds of German.

    We’re not talking about the /p-b/ /t-d/ /k-ɡ/ /s-z/ /ʃ-ʒ/ distinctions in Standard German then, obviously.

    For the first three that depends on the kind of Standard German! In Austrian Standard German that’s very much what we’re talking about; native speakers of dialects that lack it (Viennese, Carinthian) try to use it when speaking Standard German. In northern Germany, no; the fortes are additionally aspirated or glottalized depending on position (like in English), and the lenes are additionally voiced (unreliably, like in English, though perhaps a bit more often). There’s a zone in north-central Germany where aspiration (and apparently glottalization) is absent; it’s continuous with Dutch, where the same holds, and distinguishes fortes from reliably voiced lenes like most languages in Europe. South-central Germany, roughly speaking, has abandoned the phonemicity of the distinction altogether, leading to voiceless lenes almost everywhere; because this can’t be mapped to the spelling, it’s not considered good enough to carry over into Standard German.

    For the last two, no. In northern and central Germany, it’s a voice + fortis/lenis distinction as you’d expect from most of the rest of Europe (to the extent that people have the loanword phoneme /ʒ/, anyway). South of that, it’s a length distinction (and its phonotactics are a bit different).

    Do they have a stronger effect on the articulation of adjacent vowels?

    Neither fortes nor lenes have any effect on adjacent vowels that I’ve noticed, or that anyone else seems to have noticed.

  116. David Marjanović says:

    I don’t seem to be able to navigate between the Scylla of /ʝ/ and the Charybdis of /i/, or maybe I’m just not good at recognizing the presence or absence of friction.

    I bet there’s no language with a minimal triplet for /ʝ/-/j/-/i/!

  117. For fun, a paper by Hualde et al. on glides in Spanish (from the issue of Lingua I mentioned) notes:
    de Emiliana [demiljána]
    de mi liana [demili.ana]

    Eli, try the experiment with “yup”, but silently or in whisper, so your voice doesn’t obscure the friction sound, if any.

    [jǝ̥̆] isn’t very far from [j̩], anyway.

  118. I bet there’s no language with a minimal triplet for /ʝ/-/j/-/i/!

    I’d look at Klamath for a possible exception. Supposedly it has /ʝ/ and has vowel-glide vs. vowel-vowel contrasts.

  119. dainichi says:

    > de mi liana [demili.ana]

    Could you summarize for people with no access to Lingua? According to Wiktionary and my knowledge of Spanish “liana” is phonemically /ˈlja.na/. So does the /lja/ surface as two syllables for some reason?

  120. He just calls it exceptional hiatus, in the sense that unstressed /i/ and /u/ before other vowels as a rule turn into glides (another example: pie [pje] ‘foot’ vs. pié [pi.e] ‘I chirped’.) There are a bunch of papers around on exceptional hiatus in Spanish, which I haven’t seen.

  121. There’s a good deal of variation between [iˈV] and [ˈjV] in Spanish, depending on the word, speaker and rate of speech. But the RAE’s 2010 spelling reform abolishes all remaining traces of this in writing (changing pié to pie, and guión to guion) – which I’m not a fan of.

  122. dainichi says:

    @Y, @Lazar,

    Thanks, very interesting! (My intuition about Spanish isn’t worth much, but) I can imagine pronouncing “pié” with two syllables. However, I can imagine the same for “piar”, did that ever have an accent? Was it ever “liána”? Or were the rules somehow different for a and e?

  123. No – as far as I know, an accent was never used in those words.

    I think the traditional orthographic logic is that where an “exceptional hiatus” occurs, the i gets treated as if it were one of the “strong vowels” a, e, o – thus the accents to show final stress in pié and guión. Whereas in piar (with the a followed by a stress-attracting r) and liana (with the a in penultimate position), even treating the i as a strong vowel wouldn’t change the placement of the stress.

    It’s a little convoluted perhaps, but it means that at least some of the exceptional hiatuses get represented in writing.

  124. By my unreliable, non-native intuition of Spanish, the secondary stress of de mi liana is enough to lengthen the li- and break up the diphthong.

    Paging Alon Lischinsky…

  125. @Y: at least in my native dialect, I don’t think that would break the diphthong. De mi liana would be [de.miˈlja.na] in most contexts I can imagine. Even if I assume explicitly contrastive stress (in something like Tarzán se colgó de mi liana, no de la tuya) that would be [ðe↗ˈmi ˈlja.na], without a hiatus.

    Then again, intuition is unreliable, and there are other cases in which things seem a lot more labile. According to the conventional wisdom, for example, [ea] never forms a diphthong in Spanish, but for me arrear is disyllabic [aˈreaɾ] more often than trisyllabic [a.re̞ˈaɾ], and while mear is mostly disyllabic [me̞ˈaɾ], meando is almost invariably [ˈmean̪.do]. (I’ve tried to indicate that the vowel in the diphtongised version feels a bit more front to me, but I have No Real Evidence™ of that, just a gut feeling.)

  126. Thanks! The paging actually worked.

  127. @Y: shall we then page Pancho for a Mexican perspective, and Jesús for a Peninsular one?

  128. Canepari has remarked on the non-syllabic realizations of /e/ and /o/ before another vowel in Spanish, and represents them respectively with a small capital j and a sort of turned small capital Latin omega.

  129. How about the Dr. Seuss-ish Se colgó de mi liana, no de la de Emiliana?

  130. Se cagó en la liana.

  131. @Lazar:

    Canepari has remarked on the non-syllabic realizations of /e/ and /o/ before another vowel in Spanish, and represents them respectively with a small capital j and a sort of turned small capital Latin omega.

    Thanks for this. I confess I’ve never been able to grasp ᶜᵃⁿIPA, and it’s taken me a while to wade through the various reference materials on Canepari’s site before I could make sense of it.

    As far as I can tell, Canepari uses ⟨ɩ⟩ for a near-close near-front unrounded vowel, which in ᵒᶠᶠIPA would be ⟨ɪ⟩. Is that the “small capital j” you mean? If so, it doesn’t seem to me the best possible description of the sound in question, but then again IANAP.

  132. David Marjanović says:

    non-syllabic realizations of /e/ and /o/

    Is there actually a reason not to interpret them as [e̯] and [o̯] (or perhaps the same with Canepari’s mid-vowel symbols)? While some diphthongs are a lot more common than others worldwide, anything can be in either position in a diphthong; there are diphthongs that begin with [a̯] out there.

  133. @Alon: No, it’s not [ɩ] but rather a (real) small capital J – representing a semi-palatal approximant in his terminology. (And there’s one for /a/ too – sort of like ɯ with the sides chopped off.) He’s remarked on the overlap between vowels and approximants, but follows the (AFAIK) conventional practice of using the former for falling diphthongs and the latter for rising ones. For a more extensive example of this approach, see his description of Romanian in the “phonosyntheses of European languages” chapter.

  134. David Marjanović says:

    follows the (AFAIK) conventional practice of using the former for falling diphthongs and the latter for rising ones

    And yet he complains about the “International Phonemic Alphabet”? This is an actual phonetic difference between languages. (And phonemic in the Maastricht dialect “of Dutch”.)

  135. Well, I think you might be underestimating how many distinct approximants are given symbols in CanIPA – see the chapter “Consonants and contoids (2)”, pages 213-214, where he gives a partial list of vocoid-approximant equivalences. Could you give an example from the Maastricht dialect?

  136. David Marjanović says:

    The link to the paper is in moderation. “[eːʲ]” in the paper should be written [eːi̯].

  137. Oh. Well in cases like that, I have seen him use [j] to show slightly higher tension in the falling elements of diphthongs. In his description of Arabic he mentions variation between [i] and [j] as the second element of the diphthong ay.

Speak Your Mind

*