Voiceinduced vowel lengthening Scheer Tobias 2017 Voiceinduced vowel

Voice-induced vowel lengthening Scheer, Tobias. 2017. Voice-induced vowel lengthening. Papers in Historical Phonology 2: 116 -151. Fringe Workshop on Laryngeal Features in Historical Phonology at the Third Edinburgh Symposium on Historical Phonology Edinburgh, 29 November 2017 Tobias Scheer Université Côte d'Azur, CNRS

spontaneous vs. non-spontaneous voicing obstruents: non-spontaeous voicing – phonologically active sonorants, vowels: spontaneous voicing – phonologically inactive Chomsky & Halle (1968: 300 f) solid and consensual empirical record • for example, sonorants and vowels do not participate in final devoicing

spontaneous vs. non-spontaneous voicing documented transmission of voicing from vowels/sonorants to voiceless obstruents and vowels • intervocalic voicing • Cracow-Poznań voicing (external sandhi) ja[g] możesz ‘how can you’ • voice-induced vowel lengthening Very common phenomena external sandhi voicing Polish Bethin (1984, 1992), Gussmann (1992), Rubach (1996), Cyran (2011, 2012, 2014) Catalan Wheeler (1986), Bermúdez-Otero (2006) West Flemish De Schutter & Taeldeman (1986) Breton Krämer (2000) Durham English Gussenhoven & Jacobs (2011: 196)

how to go about voice transmission from sonorants / vowels option 1: • follow the surface • all alternations involving sound are phonological in kind. Sure, sonorants and vowels do not have phonologically active voicing, except when they do. [sonorant voice] Rice & Avery (1989), Piggott (1992), Rice (1993), Avery & Idsardi (2001) and Clements & Osu (2002: 338). Overview: Botma (2011). option 2: • believe in your theory • there is something beyond phonology Sonorants and vowels never bear a phonological specification for voicing, and their voicing is never taken into account by phonological computation. Transmission of laryngeal properties to obstruents occurs post-phonologically, i. e. in the phonetics. Cyran (2011, 2012, 2014), Scheer (2015 a, b)

voice-induced vowel lengthening • very common • English, French, German (strong verbs), Russian, Korean etc. • possibly a phonetic universal: Delattre (1962), Chen (1970) • mechanical consequence of speech production • sometimes grammaticalized (contrastive) Overview: Maddieson (1996, 164 ff, 1999). Belasco (1953, 1958), Hoffman (1958), Zimmermann & Sapon (1958), Delattre (1962), Hogan & Rozsypal (1980), Luce & Luce (1985), Kluender (1988), van Santen (1992).

Slavic "compensatory" vowel lengthening Origin • when word-final yers were lost, languages in the Western area of the Slavic territory lengthened the preceding vowel. • Western Slavic (save Lower Sorbian) plus westernmost languages of South Slavic (Serbo-Croatian, Slovenian). • geographic extension and possible restrictions to a subset of vowels Bethin (1998, 96 ff), Timberlake (1983 a, b), Kavitskaya (2002, 119 ff), Shevelov (1964, 447 f), Carlton (1991, 215 ff), Vondrák (1924, 309 -320), Rospond (1979, 65 ff), Stieber (1973, §§ 38 -43), Sanders (2003, 57 ff)

Slavic "compensatory" vowel lengthening Vowel lengthening before lost final yer • Serbo-Croatian lengthens before all consonants • Polish & Czech only after sonorants & voiced obstruents Nsg Gsg Nsg dol‑ъ dol-u bog‑ъ bog-a a. Old Polish dół doł-u b. Old Czech dól dol-u CS c. Serbo. Croatian "dale, valley" Gsg Nsg Gsg most‑ъ most-u bóh boh-a most-u bôg bȍg-a môst mȍst-a "God" "bridge"

Slavic "compensatory" vowel lengthening Properties • traditionally described as compensatory lengthening (compensating for the loss of the yer), but could as well be lengthening before word-final consonants. Carlton (1991: 217219) and Sanders (2003: 60 f) discuss reasons to doubt the compensatory causality for Western Slavic. • In all languages where it occurs lengthening is irregular, covering only an unpredictable subset of the words that offer the triggering context. • Sometimes also words that should not undergo the process do display lengthening. • This is true for all diachronic stages of the languages at hand, including the oldest record available. • Given its geographical extension, lengthening is assumed to have occurred in late Common Slavic (CS).

Western Slavic vowel lengthening before sonorants and voiced obstruents • in Western Slavic, lengthening only occurs before sonorants and voiced obstruents. • whatever the original trigger of the lengthening, Stieber (1973: § 41) demonstrates that in 15 th century Polish there was an active process that lengthened vowels before word-final voiced consonants: loans that were present then (but absent when yers were lost long before the 15 th century) appear with a lengthened vowel: Adaam staal captuur qhaan ‘Adam’ ‘steel’ ‘hood’ (modern spelling: chaan) ‘khan’.

Western Slavic vowel lengthening before sonorants and voiced obstruents • o > oo / __C+voice # where C+voice = sonorants and voiced obstruents here and below data are restricted to the vowel o, which of all vowels is most inclined to undergo lengthening. The situation of this vowel is also easier to assess since (like e) it was always short in CS. Hence long o in post-CS languages can only be the result of lengthening: there is no original long oo, which would have occurred before any consonant.

Western Slavic vowel lengthening before sonorants and voiced obstruents Old Czech lengthening before word-final sonorants and voiced obstruents alternating roots non-alternating roots Nsg masc. Gpl fem. /neutr. Nsg masc. Nsg Gsg Nsg Gpl Nsg Gsg __D bóh boh-a nóh roh-u nóž nož-ĕ vod-a vód stoh-u vóz voz-u slov-o slóv bob-u poss. ‑óv ‑ova lov-u Dpl ‑óm kov-u rod-u __R stól stol-u hor-a hór bor-u dvór dvor-a škol-a škól hrom-u mój moj‑ĕ strom-u dóm dom-u lom-u sól sol-i zvon-u vól vol-u hon-u pól pol-u boj‑ĕ hnój hnoj-ĕ __T rok-u brok krok tok brok-u krok-u tok-u

Western Slavic vowel lengthening before sonorants and voiced obstruents Old Czech, exceptions • environment met but no alternation: roh – roh-u (non-alternating roots) • overgeneralization of the long vowel to forms where it is not final: Nsg hróz‑a, stvór‑a, mór‑a, smól‑a Gpl hróz, stvór, mór, smól • environment not met but alternation occurs: póst - post-u "fasting Nsg, Gsg" (> mod. Cz. půst – půst-u).

Western Slavic vowel lengthening before sonorants and voiced obstruents Old Polish: more of the same Lengthening • before voiced obstruents: bóg, bób, róg, nóż, wóz, miód, wróg, chłód, gród, żłób • before sonorants dół, stół, sól, dwór, mój, król, tchórz • Old Czech roh - roh u, bob - bob-u should lengthen but do not, while in Old Polish they do: róg – rog-u, bób – bob-u. • Old Polish dom – dom-u should lengthen but does not, while Old Czech displays regular dóm – dom-u.

Western Slavic vowel lengthening before sonorants and voiced obstruents developments since Old Czech / Old Polish Czech Polish 1. final devoicing nóž [nɔɔʒ] > nóž [nɔɔʃ] nóż [nɔɔʒ] > noż [nɔɔʃ] 2. raising oo > uu nóž [nɔɔʃ] > nůž [nuuʃ] nóż [nɔɔʃ] > nóż [nuuʃ] 3. short ą > ę – ząb‑u > zęb‑u Plus analogical activity both ways • eliminating rightful alternations OCz vod‑a ‑ vód "water Nsg, Gpl" > MCz vod‑a - vod • creating illegitimate alternations sobot-a - sobót "Saturday Nsg, Gpl" powrót - powrot-u "return Nsg, Gsg"

Western Slavic vowel lengthening before sonorants and voiced obstruents • hence the modern situation is hopeless: alternations are entirely unpredictable and lexicalized The modern situation with all its ins and outs is described by Bethin (1979: 253 f, 259), Gussmann (1980: 53 f, 113 ff, 2007 : 261 ff), Szpyra (1989: 160 ff, 1992: 288 ff) and Grzegorczykowa (1999: 114 ff) for Polish and by Trávníček (1951: § 12) for Czech. • in addition, Polish has lost distinctive vowel length ==> alternations are only visible when associated to vowel quality: • nóż [nuʃ] - noż‑a [nɔʒa] • ząb [zɔmp] - zęb-u [zɛmbu]

Western Slavic vowel lengthening before sonorants and voiced obstruents Interpretation • in both languages final devoicing has obliterated the surface trigger: final voiced obstruents are no longer voiced on the surface. Hence a phonetic account based on surface voicing is out of the question. • It is also ruled out since in the modern languages the alternation involves a category change: o and u/uu are different phonemes (unlike the original short-long distinction). • Category changes in the modern languages can thus only be due to phonological patterning. • But even in Old Czech and Old Polish, the unpredictable lexical idiosyncrasy strongly suggests that the alternation was always lexicalized, from day one.

Western Slavic vowel lengthening before sonorants and voiced obstruents Analysis • a process like o > oo / __R, D # was never carried out by phonological computation at any stage of any Slavic language (or of Common Slavic). • it could not be a phonological process because sonorant voicing has no phonological existence. • Rather, phonetic length was phonologized through a modification of the lexical entry (restructuring): 1. /bog/ [boog] 2. /boog/ [boog] 3. /boog/ [book] 4. /boog/ [buuk] phonetic lengthening phonetic length lexicalized final devoicing raising

Western Slavic vowel lengthening before sonorants and voiced obstruents • Synchronic alternations in inflectional paradigms (in the old and modern languages) are managed by allomorphy: there is no common underlier: 1. /buug/ selected in Nsg (and Asg for non-animates) 2. /bog/ selected elsewhere • synchronic alternations involve a change of category (between two phonemes), hence could not be phonetic in kind.

Western Slavic vowel lengthening before sonorants and voiced obstruents • in order for this analysis to work, phonetics must be able to specifically lengthen vowels before word-final voiced obstruents and sonorants (but not anywhere else). • a process conditioned by a word boundary looks phonological. . . • in English, Vx in CVx. C# is significantly longer in duration than it is in CVx. CV# (Klatt 1973) • CVC beat, stoop, glide, room • vs. CVCV(C) beaten, stupid, gliding, rumour • for each given consonant Cx, the preceding vowel in CVCx# was significantly longer than in CVCx. V(C). • the difference in duration between vowels followed by wordfinal and internal consonants is significantly higher when the consonant is a sonorant or a voiced obstruent, as compared to voiceless obstruents.

Western Slavic vowel lengthening before sonorants and voiced obstruents We are thus facing a lexically specific phonologization of a phonetic property. How is it decided which lexical items are concerned? • maybe token frequency: lengthened items belong to basic vocabulary (Czech): lengthened: God, knife, table, house, ox, court not lengthened: angle, bell, hunt, metal, haystack

Zoom out: the cross-linguistic picture If voice-induced lengthening is phonetic in kind (and never phonological), the following prediction is made: • a pattern where this process is triggered by voiced obstruents, but not by sonorants, cannot exist. • because the distinction can only be made in the phonology. there does not seem to be any case on record (in the massive documentation of voice-induced lengthening) where • lengthening is induced by obstruents • but not by sonorants

phonetics or phonology? vowel length management done in the phonetics or in the phonology ?

phonetics or phonology? if phonological in kind • we are necessarily facing two distinct processes: lengthening in voice languages (only [voice] present), shortening in aspiration languages (only [spread gl. ] present) • sonorants must be phonologically specified for [voice] cross-linguistically • voice-induced vowel length modification cuts across voice(English) and aspiration languages (Russian, French). • there is no evidence that there are two distinct processes.

phonetics or phonology? if phonetic in kind • there is no need to duplicate the uniform pattern • sonorants don't need to bear phonological specifications for voicing • the abovementioned prediction is made: sonorants always side with voiced obstruents. There is no pattern where only the latter lengthen the preceding vowel. • if phonological in kind, the reverse prediction is made: vowel lengthening induced by voiced obstruents but not by sonorants should be the trivial, common pattern. ==> this is wrong for sure. there does not seem to be any case on record (in the massive documentation of voice-induced lengthening) where • lengthening is induced by obstruents • but not by sonorants

phonetics or phonology? Possible counter-example Winter's Law (Balto-Slavic) [Winter 1978] • IE short V = B-Sl short V iff under non-acute intonation and followed by a consonant other than an IE voiced obstruent (mediae) • that is, V > VV /__voiced obstruent or under acute intonation But • Winter's examples where non-acute V lengthen involve roots with a laryngeal, whose loss produces compensatory lengthening: CVHC > CVVC. His non-acute vowels that remain short are not followed by a laryngeal. • in glottalic theory, IE mediae (voiced obstruents) are ejectives. Hence there is no effect of voicing, but a "laryngeal" effet produced by ejectives and laryngeals. • Kortlandt (1978, 1981), Collinge (1985: 225 ff).

Zoom out: the cross-linguistic picture contrast with external sandhi voicing: trigger only phonetic phonological phonetic a. lengthening of a preceding vowel b. voicing of a preceding voiceless obstruent sonorants and only voiced obstruents yes no yes Cracow-Poznań yes Warsaw

lengthening or shortening? lengthening caused by sonorants and voiced obstruents or shortening caused by voiceless obstruents (pre-fortis clipping) ?

English phonetic vowel duration controlled by the voicing of the following C: vowels (both phonologically short and long) are • longer before sonorants and voiced obstruents • shorter before voiceless obstruents • bit vs. bid [short vs. short lengthened] • beat vs. bead [long vs. long lengthened] extensive study and description: Rositzke (1939), Peterson & Lehiste (1960), House (1961), Raphael (1972), Laeufer (1992), van Santen (1992), Gonet & Stadnicka (2006) etc. • classically interpreted as lengthening before voiced consonants. • House (1961) introduces the idea that it's the reverse: shortening before voiceless obstruents. • pre-fortis clipping coined by UCL phoneticians in the 1980 s

English (non-)argument: V# • extra length also occurs on word-final vowels, hence that cannot be due to a following voiced consonant • but: word-final lengthening is documented independently and hence has nothing to do with the voice-controlled pattern second vowel significantly longer in poppa pose than in pop oppose Oller (1973), Umeda (1975), Nakatani, O'Connor & Aston (1981), Beckman & Edwards (1990). But see issues discussed by Turk (2000).

English: phonetics or phonology? • the regular English situation bears all properties of a phonetic pattern: • there are not 4 different i's in the language • extra length/shortness is not contrastive • but Pöchtrager (2006), Pöchtrager & Kaye (2013) do have a purely phonological management whereby voicing is encoded as length: • voiced obstruents have 2 x-slots, voiced obstruents only one. • but their analysis cannot do voice assimilation pl. rock-/z/

English: phonetics or phonology? Bermúdez-Otero's (2014) • cases where the phonetic length is further exploited show that length management is phonological • ==> (Canadian) raising write [rʌit] vs. ride [rɑɪd] writer [rʌiɾɚ] vs. rider [rɑɪɾɚ] • argument: • vowel length distinctions show up as two distinct diphthongs • the process needs access to the voice value of the following C • this information is obliterated by flapping of both t and d • flapping is phonological in kind (independent evidence) • ==> hence diphthong selection must be done before flapping, that is before phonetics, i. e. in the phonology

English: phonetics or phonology? morphological situation (Bermúdez-Otero's 2014) word a. cíte cýcle nítrate b. citátion c. éye-ful d. tíe shop rai- pronunciation T belongs to sing? diphthong stressed? yes yes no no no yes yes no yes c[ʌi]t c[ʌi]cle n[ʌi]trate c[ɑɪ]tation [ɑɪ]e-ful t[ɑɪ]e shop the same morpheme a different morpheme

English: phonetics or phonology? Bermúdez-Otero's (2014) conclusion • Canadian raising is stem level (class 1): class 2 affixes don't do it • really? • we only see cases where the diphthong and the following C belong to the same morpheme • there is no clear case where raising is triggered by a stem level affix. Bermúez-Otero mentions Idsardi (2006: 25) reporting Canadian raising in • i-th "the i-th number", with -th being probably stem level.

English: phonetics or phonology? • if there is no morphological condition on Canadian raising, • i. e. if in all cases of raising the diphthong and the following consonant are tautomorphemic, • ==> then there is an analysis available along the Slavic pattern where the diphthong is modified in the phonetics and the result lexicalized: 1. phonetic modification of the vowel: offglide peripheralization (Thomas 2000, Moreton & Thomas 2007). write /rɑɪt/ [rɑit] vs. ride /rɑɪd/ [rɑɪd] 2. lexicalization of the result write /rɑit/ [rɑit] vs. ride /rɑɪd/ [rɑɪd] 3. raising (under stress) and flapping in the phonology: write /rɑit/ [rʌit] vs. ride /rɑɪd/ [rɑɪd]

that's it