There is no problem of /v/ in Russian phonology

26th Manchester Phonology Meeting, University of Manchester, Manchester, UK

1. There is no problem of /v/ in Russian phonology Pavel Iosad University of Edinburgh pavel.iosad@ed.ac.uk 26th Manchester Phonology Meeting University of Manchester 26th May 2018 1
2. Overview • The problem of /v/ and proposed solutions 2
3. Overview • The problem of /v/ and proposed solutions • Solution: /v/ (de)voicing is phonetic 2
4. Overview • The problem of /v/ and proposed solutions • Solution: /v/ (de)voicing is phonetic • Russian voicing assimilation is a scattered rule 2
5. Overview • The problem of /v/ and proposed solutions • Solution: /v/ (de)voicing is phonetic • Russian voicing assimilation is a scattered rule • Russian /v/ is not a problem for the Contrastivist Hypothesis 2
6. The problem of /v/
7. The basic problem ‘{v*} functions as a sonorant if followed by a sonorant, and as an obstruent if followed by an obstruent’ (Halle 1959, p. 63) (1) a. [zovə] зова ‘call-GEN.SG’ b. [zof] зов ‘call’ (2) a. [bʲitvə] битва ‘battle’ b. [podvʲik] подвиг ‘feat’ (3) a. [pʲɪvʲet͡s] певец ‘singer’ b. [pʲɪft͡sa] певца ‘singer-GEN.SG’ 3
8. Transparency of sonorants • Jakobson (1978): sonorants are transparent to voicing assimilation (4) a. [ɐt#ozʲɪrə] от озера ‘from a lake’ b. [ɐt#ɫoʃkʲɪ] от ложки ‘from a spoon’ c. [ɐd#ɫɡʊna] от лгуна ‘from a liar’ 4
9. Transparency of /v/ „Falls ein Geräuschlaut einem stimmhaften Geräusch- laut vorangeht, so wird auch der erste von den bei- den stimmhaft, gleichgültig ob die beiden unmittelbar nacheinanderfolgen oder zwischen ihnen ein einfaches oder langes, hartes oder weiches v auftritt.“ (Jakobson 1956) (5) a. [k#ozʲɪrʊ] к озеру ‘to a lake’ b. [k#vorənʊ] к ворону ‘to a raven’ c. [ɡ#vdɐˈvʲe] к вдове ‘to a widow’ ‘Everything transpires as if {v} or {v,} had been absent’ (Halle 1959, p. 64) 5
10. Devoicing of /v/ (6) a. [vˠi-xat] выход ‘exit’ b. [f-xot] вход ‘entrance’ c. [is#fxodə] из входа ‘from an entrance’ 6
11. Proposed solutions
12. The basic idea • The problem: 1. [v] behaves as if it lacked [+voice] when voicing assimilation applies 2. [v] can get [−voice] from voicing assimilation 3. The product of [v] → [f] behaves as if it has [−voice] by triggering assimilation • The solution: assign [±voice] to [v] during the derivation 7
13. What is the /v/ underlyingly? • Since the behaviour of /v/ resembles that of sonorants, it shares some featural property with them • Options: • /u̯/ (Lightner 1972) • /w/ (e. g. Coats & Harshenin 1971, Hayes 1984, Kiparsky 1985) • /ʋ̝/ ‘narrow approximant’, or other non-glide (e. g. Panov 1967, Padgett 2002) • /V/ underspecified for [±voice] (Hall 2004, Reiss 2017) 8
14. Derivational accounts • A representative account is by Kiparsky (1985) Rule /zoW/ /zob/ /K#WoroNu/ /iz#WKuSa/ /LɡuN/ Lexical phonology Final devoicing zop Default voicing k#WoroNu is#Wkusa Postlexical phonology Final devoicing zow ̥ Voicing assimilation — is#w ̥ kusa Default voicing k#woronu lɡun /w/ strengthening zof k#voronu is#fkusa 9
15. Other accounts • Padgett (2002): Russian /v/ is featurally [+sonorant −wide] Common ground • Russian [v] is not [+voice] underlyingly • The phonological grammar can make it [±voice] 10
16. Russian /v/ and contrast • Russian fricatives: /f(ʲ) v(ʲ)/, /s(ʲ) (zʲ)/, /ʃ ʒ/, /x/ • Minimal contrast for [±voice] (except /x/): why is the /f/–/v/ pair special? ‘[O]ur study of Russian /v/ […] shows a segment that does have a contrastive “twin” – there are surface [v]’s and [f]’s – and yet the segment under analysis still be- haves in a non-parallel fashion with respect to the fea- ture that determines the contrast […] Russian shows that lack of contrast with respect to [a feature] F is not a necessary condition to predict irregular behavior with respect to F’ (Reiss 2017, pp. 43–44). 11
17. A proposal • Russian /v/ and /f/ are not ‘contrastive twins’ anywhere in the phonological grammar 12
18. A proposal • Russian /v/ and /f/ are not ‘contrastive twins’ anywhere in the phonological grammar • /v/ is /V/ with no underlyingly laryngeal specification, and stays that way Any devoicing to ‘[f]’ is phonetic implementation 12
19. A proposal • Russian /v/ and /f/ are not ‘contrastive twins’ anywhere in the phonological grammar • /v/ is /V/ with no underlyingly laryngeal specification, and stays that way Any devoicing to ‘[f]’ is phonetic implementation • /f/ is underlyingly voiceless, and stays that way Any voicing to ‘[v]’ is phonetic implementation 12
20. A proposal • Russian /v/ and /f/ are not ‘contrastive twins’ anywhere in the phonological grammar • /v/ is /V/ with no underlyingly laryngeal specification, and stays that way Any devoicing to ‘[f]’ is phonetic implementation • /f/ is underlyingly voiceless, and stays that way Any voicing to ‘[v]’ is phonetic implementation ‘When […] an unpaired obstruent becomes voiced be- fore a voiced obstruent, it does not become identifiable with any other phoneme: just as a voiced realization of č remains a č, so a voiced f remains a f.’ (Andersen 1969, p. 126) 12
21. Russian /v/ (de)voicing is phonetic
22. The basic argument ‘I steer clear of the issues of syntax-phonology inter- face […] and also of the murky domain of gradient phonetic-y phenomena (including claims that sonor- ants “optionally and gradiently” devoice word-finally) […]’ (Reiss 2017, p. 30) • I do not 13
23. The basic argument ‘I steer clear of the issues of syntax-phonology inter- face […] and also of the murky domain of gradient phonetic-y phenomena (including claims that sonor- ants “optionally and gradiently” devoice word-finally) […]’ (Reiss 2017, p. 30) • I do not • The relationship between categorical and gradient (de)voicing in Russian deserves to be taken seriously • A modular approach to the division of labour ‘[P]honologists need to decide whether an alterna- tion falls into the realm of phonological computation before they propose a phonological analysis for it’ (Scheer 2015, p. 319) 13
24. Three propositions • The (de)voicing of sonorants in Russian is phonetic 14
25. Three propositions • The (de)voicing of sonorants in Russian is phonetic • The (de)voicing of /v/ in Russian is phonetic in much the same way 14
26. Three propositions • The (de)voicing of sonorants in Russian is phonetic • The (de)voicing of /v/ in Russian is phonetic in much the same way • There is every reason to suspect that obstruent (de)voicing in Russian is a scattered rule 14
27. Sonorant (de)voicing i • Sonorant devoicing is uncontroversially possible in Modern Standard Russian (e. g. Avanesov 1972) (7) Before voiceless (and devoiced) obstruents a. [sʲer̥p] серб ‘Serb’ b. [sʲer̥p] серп ‘sickle’ (8) Word-finally, especially after another consonant (Lyubimova 1975) a. [mˠi̵sl̥ʲ] мысль ‘thought’ b. [ʒˠizn̥ ʲ] жизнь ‘life’ 15
28. Sonorant (de)voicing ii • Shevoroshkin (1971): sonorants devoice in cases of transparency (9) a. [iz#mxa] из мха ‘from moss’ b. [is#m ̥ xa] • Kiparsky (1985) treats it as a variable postlexical rule • Long disputed • Disconfirmed experimentally 16
29. Sonorant (de)voicing iii ‘It seems likely that effects involving sonorants should be handled by the phonetic component. […] These con- clusions essentially remove from the phonology nearly all effects treated as postlexical phonology by Kiparsky (1985).’ (Padgett 2002, p. 6) We should treat this seriously! • If we interrogate the phonological status of sonorant (de)voicing, we should apply the same rigour to /v/ • No good evidence that /v/ (de)voicing is phonological • Plenty of suggestive evidence that it is phonetic 17
30. Sonorant and [v] transparency • Kulikov (2012): no phonological [v] transparency to voicing assimilation (10) a. [is fxodə] из входа ‘from an entrance’ b. *[iz vxodə] c. [ɐt vdɐvˠi] от вдовы ‘from a widow’ d. [ɐd vdɐvˠi] 18
31. Analysis of transparency • Sonorant devoicing – and triggering of devoicing by sonorants – is phonetic • Variable • Sensitive to prosodic boundary strength • Comparatively rare • Voiceless sonorants are not the default case, phonetically 19
32. Analysis of transparency • Sonorant devoicing – and triggering of devoicing by sonorants – is phonetic • Variable • Sensitive to prosodic boundary strength • Comparatively rare • Voiceless sonorants are not the default case, phonetically • Voicing of [v] – and triggering of voicing by [v] – is phonetic • Variable • Sensitive to speech rate • Comparatively rare • Voiced obstruents are not the default case, phonetically 19
33. Analysis of transparency • Sonorant devoicing – and triggering of devoicing by sonorants – is phonetic • Variable • Sensitive to prosodic boundary strength • Comparatively rare • Voiceless sonorants are not the default case, phonetically • Voicing of [v] – and triggering of voicing by [v] – is phonetic • Variable • Sensitive to speech rate • Comparatively rare • Voiced obstruents are not the default case, phonetically If sonorant (de)voicing can be phonetic, then so can [v] voicing 19
34. The problem of -Cv# i • A [f] ← /v/ by final devoicing is claimed to trigger assimilation variably (11) a. [trʲezvˠij] трезвый ‘sober-ATTRIB’ b. [trʲesf] трезв ‘sober-PRED’ c. [trʲezf] • Controversial experimentally • Analysis by Padgett (2002) • [trʲesf] is phonologically /trʲezʋ̝/ → [.trʲesf.] with regular cluster devoicing • [trʲezf] ∼ [trʲezv] is [.trʲe.zʋ̝.] with phonetic devoicing 20
35. The problem of -Cv# ii • Knyazev (2004): incomplete/variable devoicing in -CC# clusters is not specific to /v/ (12) a. [sɫuʒbə] служба ‘service’ b. [sɫuʃp] служб ‘service.GEN.PL’ c. [sɫuʒ̥p] • Nothing of the sort in morphologically underived -CC# sequences 21
36. The problem of -Cv# iii (13) a. [mozɡə] мозга ‘brain-GEN.SG’ b. [mosk] мозг ‘brain’ c. *[moz̥k] Analysis by Knyazev (2004) • Full devoicing in [mosk] is assimilation (phonological rule) • Partial devoicing in [trʲezf] and [sɫuʒ̥p] is coarticulation (phonetic rule) • Both [trʲezf] and [sɫuʒ̥p] escape assimilation via a derived environment effect 22
37. The problem of -Cv# iv Proposed analysis • No phonological evidence for /v/ → [f] in [trʲesf] ∼ [trʲezf] • Independent evidence for coarticulatory devoicing in derived clusters • Both [trʲesf] and [trʲezf] are phonologically [trʲezv] with phonetic devoicing 23
38. Behaviour specific to [v] i • Vorontsova (2007), Knyazev, Petrova & Vorontsova (2007) describe the behaviour of [v#v] clusters • Expected behaviour (14) a. [plof#vˠidəɫsʲə] плов выдался ‘pilaf turned out’ b. [plov#vˠidəɫsʲə] • Knyazev, Petrova & Vorontsova (2007), Vorontsova (2007) find a third option: [f#f] Or rather [f#v̥], with devoicing but not complete neutralization 24
39. Behaviour specific to [v] ii Analysis by Knyazev, Petrova & Vorontsova (2007) • [f#v] is found across a strong prosodic boundary • [v#v] is found in the absence of a boundary • [f#v̥] is found with weaker boundaries and is an instance of coarticulation Proposed analysis • As with [v] ‘transparency’, devoicing instantiates the phonetic tendency for obstruent voicelessness No evidence for phonological devoicing 25
40. Russian (de)voicing and rule scattering
41. Russian [v] and the life cycle • Are obstruent and [v] (de)voicing handled by the same rule? 26
42. Russian [v] and the life cycle • Are obstruent and [v] (de)voicing handled by the same rule? • Historically, they are different sound changes 26
43. Russian [v] and the life cycle • Are obstruent and [v] (de)voicing handled by the same rule? • Historically, they are different sound changes • Two predictions of the life cycle of phonological processes 26
44. Russian [v] and the life cycle • Are obstruent and [v] (de)voicing handled by the same rule? • Historically, they are different sound changes • Two predictions of the life cycle of phonological processes • Obstruent and [v] devoicing are different 26
45. Russian [v] and the life cycle • Are obstruent and [v] (de)voicing handled by the same rule? • Historically, they are different sound changes • Two predictions of the life cycle of phonological processes • Obstruent and [v] devoicing are different • (De)voicing can be a scattered rule 26
46. The history of devoicing Final devoicing of obstruents follows the fall of the yers • Earliest attestations in the 13th century • Does not occur in parts of the south-west and in some central dialects 27
47. The history of devoicing Final devoicing of obstruents follows the fall of the yers • Earliest attestations in the 13th century • Does not occur in parts of the south-west and in some central dialects Assimilative voicing: earliest attestations from late 12th century 27
48. The history of devoicing Final devoicing of obstruents follows the fall of the yers • Earliest attestations in the 13th century • Does not occur in parts of the south-west and in some central dialects Assimilative voicing: earliest attestations from late 12th century Assimilative devoicing: from late 13th century • Voicing and devoicing are not the same change 27
49. The devoicing of [w] • Historically, /v/ is *w • Often, w remains, possibly vocalizing before a consonant/word-finally • Earliest <f> for *w is from the early 17th century 28
50. The devoicing of [w] • Historically, /v/ is *w • Often, w remains, possibly vocalizing before a consonant/word-finally • Earliest <f> for *w is from the early 17th century • When [v] devoices, the outcome can be [x] (15) stolo[x] столов ‘table.GEN.PL’ 28
51. The devoicing of [w] • Historically, /v/ is *w • Often, w remains, possibly vocalizing before a consonant/word-finally • Earliest <f> for *w is from the early 17th century • When [v] devoices, the outcome can be [x] (15) stolo[x] столов ‘table.GEN.PL’ • Russian [f] should not be the contrastive twin of [v] (see already Andersen 1969) 28
52. Russian (de)voicing is scattered Coexisting rules • Final devoicing • Assimilative (de)voicing (may or may not be the same rule) • Sonorant and [v] (de)voicing • Reminiscent of rule scattering Not classical rule scattering • Is there evidence for phonetic devoicing of obstruents, too? 29
53. Evidence for scattering in obstruents • Voicing assimilation across phonological words is arguably phonetic • Variable, sensitive to boundary strength, not necessarily complete • Padgett (2012) endorses the position that postlexical obstruent assimilation is the same as sonorant (de)voicing 30
54. Further evidence: -Cv# revisited • Morphologically derived -CC# clusters have the same gradient voicing as -Cv# clusters • But not underived -CC# clusters (16) a. [sɫuʃp] служб ‘service.GEN.PL’ b. [sɫuʒ̥p] c. [trʲesf] трезв ‘sober.PRED’ d. [trʲezf] e. [mosk] мозг ‘brain’ f. *[moz̥k] 31
55. Conclusion on clusters • Gradient obstruent (de)voicing exists in Russian alongside phonological obstruent (de)voicing • Its effects are observed postlexically, and in the rare cases of unassimilated word-level clusters • Classic case of rule scattering 32
56. Incomplete neutralization? • If there is gradient obstruent (de)voicing, we need to think about incomplete neutralization • Final devoicing (e. g. Pye 1986, Dmitrieva, Jongman & Sereno 2010, Shrager 2012, Kharlamov 2014) • Voicing assimilation (e. g. Burton & Robblee 1997, Kulikov 2013) • Common thread: voicing can be neutralized, but other cues often aren’t Also noted for [v] devoicing by Knyazev (2004) • Ripe for an investigation acknowledging the possibility of rule scattering 33
57. Conclusion
58. What is the behaviour of Russian /v/? • Padgett (2002) is mostly right: Russian /v/ can be devoiced in the phonetics, just like sonorants • This being granted, we need no recourse to /v/ devoicing in the phonological grammar 34
59. Analysis Russian [f] has no [+voice] counterpart • Just like [x], [t͡s] and [t͡ʃʲ] 35
60. Analysis Russian [f] has no [+voice] counterpart • Just like [x], [t͡s] and [t͡ʃʲ] Russian [v] has no [−voice] counterpart • Just like [m], [n], [l]… 35
61. Analysis Russian [f] has no [+voice] counterpart • Just like [x], [t͡s] and [t͡ʃʲ] Russian [v] has no [−voice] counterpart • Just like [m], [n], [l]… • There is no evidence that [v] and [f] are distinguished by [±voice] in the phonology • To the extent they show parallel (de)voicing patterns, this is entirely due to phonetic processes 35
62. [v] in the contrastive hierarchy b p f v m … d t s z n t͡s … k ɡ x … [±son] b p f … d t s z t͡s … k ɡ x [±voi] p f … t s t͡s … k x … [±cont] p … t t͡s … k… f … s … x b … d z … ɡ [±cont] b … d … ɡ z v … m n … [±nas] v m n … − − − + + − + + − + 36
63. Explanation in phonology • Why is [v] like this? 37
64. Explanation in phonology • Why is [v] like this? It is underlyingly /w/ vel sim. • No evidence that it is • Complex account with various problems 37
65. Explanation in phonology • Why is [v] like this? It is underlyingly /w/ vel sim. • No evidence that it is • Complex account with various problems There is a sonorant [ʋ̝] → [f] • No evidence that this is a phonological process • Does not fit the phonetic evidence on the nature of Russian [v] (Bjorndahl 2015) 37
66. Explanation in phonology • Why is [v] like this? It is underlyingly /w/ vel sim. • No evidence that it is • Complex account with various problems There is a sonorant [ʋ̝] → [f] • No evidence that this is a phonological process • Does not fit the phonetic evidence on the nature of Russian [v] (Bjorndahl 2015) [v] is not specified for [±voice] (Hall 2004, Reiss 2017) • Fits the account here… • … but still requires phonological /v/ → [f], for which the evidence is slim 37
67. Conclusion • Russian /v/ is not affected by any phonological devoicing process 38
68. Conclusion • Russian /v/ is not affected by any phonological devoicing process • There is no evidence that [v] and [f] are ‘contrastive twins’ for [±voice] 38
69. Conclusion • Russian /v/ is not affected by any phonological devoicing process • There is no evidence that [v] and [f] are ‘contrastive twins’ for [±voice] • In a modular framework with language-specific realization of features, asymmetric behaviour follows from asymmetric specification 38
70. Conclusion • Russian /v/ is not affected by any phonological devoicing process • There is no evidence that [v] and [f] are ‘contrastive twins’ for [±voice] • In a modular framework with language-specific realization of features, asymmetric behaviour follows from asymmetric specification • The asymmetric specification follows from the contrastive hierarchy 38
71. Conclusion • Russian /v/ is not affected by any phonological devoicing process • There is no evidence that [v] and [f] are ‘contrastive twins’ for [±voice] • In a modular framework with language-specific realization of features, asymmetric behaviour follows from asymmetric specification • The asymmetric specification follows from the contrastive hierarchy • Russian [v] is not a problem for the Contrastivist Hypothesis 38
72. Thank you! Спасибо! 38