(Turkic) Vowel Harmony is not an analytically unifiable phenomenon

Jonathan North Washington
Swarthmore College
Tu+8, 2023-03-05

Overview

My talk in a nutshell

Turkic VH comprises
≥4 distinct phenomena,
which require
different analytical mechanisms.

These phenomena are related to one another as different levels of grammaticalisation.

The next ~18 minutes:

  • Previous approaches to VH
  • My proposal
    • Grammaticality cline
    • A close look at each type of VH
    • A lot of Turkic VH patterns
  • A clarification of my position

Background

Vowel harmony (VH), briefly

A vowel's quality is impacted by content which is not immediately adjacent

Claims of previous approahces

  • Inherent correlation of phonetics & phonology (Kaun 2004, Smith 2018)
    • E.g., rounding harmony involves continue rounding of the lips
  • Sensitive to phonological considerations [alone] (van der Hulst 2018)
    • I.e., it's all an abstract system—no need to consider phonetics
    • "[Lack of correspondence with phonetic considerations is not surprising]" (Nevins & Vaux 2006)
  • Non-iterative VH processes are unexpected (McCollum & Kavitskaya 2022)
    • Iterative VH processes are default (somehow?)

My claim

None of this are wrong, all of this happens

Vowel Harmony Phenomena

  • Attested at various points of grammaticalisation:
    Phonetics
    Phonology
    Morphology
    coarticulation assimilation underspecification lexicalised  trigger
  • Crucially:
    • VH originates in coarticulation
    • VH can (/does) become dissociated from phonetic basis
      • may phonologise as assimilation process
      • may further phonologise as underspecification
      • trigger features may become lexicalised
    • At least 4 different types of thing being called "vowel harmony"
  • Accounts for conflicting stories

Vowel Harmony Phenomena

  • grammaticalisation cline:
  • Phonetics
    Phonology
    Morphology
    coarticulation assimilation underspecification lexicalised  trigger

  • In more detail:
  • VH phenomenon area of grammar extent nature of phonetic target
    coarticulation phonetics* limited & variable target not hit
    assimilation phonology limited iterativity potentially non-canonical target created
    underspecification phonology unlimited iterativity canonical target created
    lexicalised trigger lexicon N/A N/A

  • Rest of talk: showing how these are different from one another
coarticulation
assimilation
underspecification
lexicalised trigger

Underspecification

  • Most broadly documented Turkic VH phenomenon
Tuvan anteriority harmony (Исхаков & Пальмбах, 1961)
/teve-dA/
[teveˈde]camel-loc
/xɑjɑ-dA/
[xɑjɑˈdɑ] cliff-loc                

Basic patterns of Turkic VH

  • Iterative, perseverative anteriority harmony
    Tuvan anteriority harmony (Исхаков & Пальмбах, 1961)
    URSRgloss
    /teve-dA/[teveˈde]camel-loc
    /xɑjɑ-dA/[xɑjɑˈdɑ]cliff-loc
  • Iterative/non-iterative, persaverative rounding harmony
    Spoken Turkmen rounding harmony (Clark, 1998; etc.)
    stemgloss-/(I)ŋ/ poss.2sg-/dA/ loc
    /ɡɯːð/daughter[ɡɯːðɯŋ][ɡɯːðdɑ]
    /pʊl/money[pʊlʊŋ][pʊldo]
  • Non-iterative anticipatory umlaut (raising, fronting)
    Uyghur rounding harmony (Clark, 1998; etc.)
    URSRgloss
    /bær-(ɨ)ʃ-GA/[beɾɨʃˈkæ]give-vn-dat
    /bɑr-(ɨ)ʃ-GA/[beɾɨʃˈqɑ]go-vn-dat

Involvement of consonants in VH

Kyrgyz: dorsal stops targets of anteriority harmony
URSRgloss
/it-GA/[ikʰe]dog-dat
/ɑt-GA/[ɑqʰɑ]horse-dat
em-GA/eɡe]feed-dat
/ɑsmɑn-GA/[ɑsmɑɴˈɢɑ]sky-dat
Kyrgyz: dorsal fricatives, dorsal nasals, and laterals targets of anteriority harmony
stemgloss-/GA/ dat-/lAr-(I)ŋ/ pl-poss.2sg
/ɑlmɑ/apple[ɑɫmɑˈʁɑ][ɑɫmɑɫɑˈɾɯɴ]
/tizme/list[tʰizmeˈɣe][tʰizmeleˈɾ]

Are these C pairs allophones then?
  • loanwords: [ɑrɑˈket] effort, [ɡɑɫstuk] tie
  • compounds: [bɑˈjɯm] Begayim (name), [bɑʁeˈlek] pant bottoms
How many consonants are affected?
All of them? All involving the tongue?
coarticulation
assimilation
underspecification
lexicalised trigger

Underspecification

  • Vast range of underspecified vowels
Kyrgyz (a-l), Chuvash (m-n): (backness)-underspecified vowels
URSRglossspecification
/kel-dI/[kelˈdi]come-pst.def[3]I = [+high]
/l-dI/[qʰɑɫˈdɯ]stay-pst.def[3]
/kel-GAn/[kelˈɣen]come-pst[3]A = [−high]
/l-GAn/[qʰɑɫˈʁɑn]stay-pst[3]
/kel-ʧU/[kelˈʧy]come-pst.impf[3]U = [+high,+round]
/l-ʧU/[qʰɑɫˈʧu]stay-pst.impf[3]
/kel-/[keˈlyː]come-gerUː = [+high,+round,+long]
/l-/[qʰɑˈɫuː]stay-ger
/kel-r/[keˈleːɾ]come-fut.idf[3]Aː = [−high,+long]
/l-r/[qʰɑˈɫɑːr]stay-fut.idf[3]
/beʃ-/[beˈʃøː]five-collOː = [−high,+round,+long]
/ɑltɯ-/[ɑɫˈt]six-coll
/pul-At-Əp/  [puˈɫɑdəp]be-npst-1sgƏ = [∅]?
/pyl-At-Əp/[pyˈlʲedɘp]interrupt-npst-1sg
  • (This is just a taste of the possibilities...)
coarticulation
assimilation
underspecification
lexicalised trigger

Underspecification

  • Straightforward way of explaining various phenomena
Underspecified Vs contrast with fully specified Vs
cf. "opaque" /i/ vs. harmonising /I/
(Turkish)
/jɑz-(j)Abil-(j)AʤAk-sIn/[jɑzɑbilɛˈʤɛkʲsin]write-abil-fut-2sg
/jɑz-dI-n/[jɑzdɯn]write-pst.def-2sg
Vowels that respond to rounding harmony in different ways
cf. /I/ (no rounding specification), /Y/ [-round], /U/ [+round]
(Southern Altay)
/bol-(I)p/[boɫup]be-vadv
/bus-(I)l/[buzul]mess up-pass
ol-nYŋ/oɫdɯɴ]road-gen
/kyj-dY/[kyjdi]burn-pst[3sg]
/ʧɯdɑɫ-LU/[ʧɯdɑɫdu]endurance-orn
/til-LU/[tildy]tongue-orn
coarticulation
assimilation
underspecification
lexicalised trigger

Underspecification

  • Vowel produced is otherwise a normal phonetic target
  • Realisation of harmonising vowel dependent on inventory;
    e.g., /A/ as [ɑ] ~ [æ]tat, [ɛ]tur, [e]kir, [ʲe]chv, [i͜ɘ]kaz, [i]ybe, [ə]fyg
  • Underspecified Vs participate in same phonological processes as non-harmonising Vs
    Standard Turkish lowering (a-b) and umlaut/raising (e-f)
    (Göksel and Kerslake 2005; Kornfilt 1997; own phonetic work [down])
    URSRgloss
    /jɛr/[jæɾ̞̊]place
    ɛp-lAr/ɛlæɾ̞̊]pocket-pl
    /jɛr-(j)A/[jɛˈɾɛ]place-dat
    ɛp-lAr-(j)A/ɛplɛˈɾɛ]pocket-pl-dat
    /jɛr-(s)I/[jeˈɾi]place-poss.3
    ɛp-lAr-(s)I/ɛpleˈɾi]pocket-pl-poss.3
  • ⇒ Underspecified V always realised as expected for V not involved in VH

Standard Turkish umlaut and lowering

Recordings of P09, see Washington (2016, 2019) for experimental details
coarticulation
assimilation
underspecification
lexicalised trigger

Underspecification

  • Realisation of underspecified Vs abstracted from phonetics
  • 8-vowel systems like Turkish (Lewis 1967, Kornfilt 1997, Göksel & Kerslake 2005)
    • Phonologically:
      [-bk][+bk]
      [-rd][+rd][-rd][+rd]
      [+hi]iyɯu
      [-hi]eøɑo
    • Phonetically:
      • 3 vowel heights
      • /ɯ/ phonetically non-high
coarticulation
assimilation
underspecification
lexicalised trigger

Underspecification

  • Iterativity is just a series of underspecified Vs occurring in sequence
Tuvan anteriority harmony (Исхаков & Пальмбах, 1961)
/teve-lAr-(I)m-dA/
[tevelerimˈde]camel-pl-poss.1sg-loc
coarticulation
assimilation
underspecification
lexicalised trigger

Phonological assimilation

  • Assimilation VH does not rely on underspecification
  • More limited capacity for iteration
Central Crimean Tatar: rounding harmony limited to first two syllables of word (McCollum & Kavitskaya, 2022)
 /qur–dI–ŋIz/
[qurduŋɯz](*[qurduŋuz])build-pst.def-2pl
Standard Turkish umlaut: limited to one syllable (own phonetic work [down])
 /jɛɾ-(s)I/
[jeɾi]                                            
  • New vowel target (as with underspecification)
  • New vowel target is non-canonical
  • Modeled as long-distance assimilation, no reference to underspecification
    • Intersection or delinking of features
  • Non-iterative by default!

Standard Turkish umlaut and lowering

Recordings of P09, see Washington (2016, 2019) for experimental details

More rounding harmony patterns at different levels of grammaticalisation

Qaraqalpaq: variation in iterativity of RH (Menges, 1947)
URSRgloss
/øt-kAr-Á-dI/[øtkøˈrødɪ]~ [øtkøˈredɪ]pass-caus-npst-3sg
Kazakh: variation in iterativity of RH, span associated with duration of C (McCollum, 2015a) — beginning of grammaticalisation of coarticulation to assimilation?
/qʊr-(I)l-(I)p/[qʰʊɾʊˈɫʊp]~[qʰʊɾʊˈɫəp]~[qʰʊɾəˈɫəp]build-pass-vadv
/qʊs-(I)p/[qʰʊˈsəp](*[qʰʊˈsʊp])vomit-vadv
Central Crimean Tatar: rounding harmony limited to first two syllables of word (more forms) (Kavitskaya, 2013; McCollum & Kavitskaya, 2022)
/dost-(I)m/[dostum]friend-poss.1sg
/tuzluq-(I)m/[tuzluʁɯm](*[tuzluʁum])salt shaker-poss.1sg
coarticulation
assimilation
underspecification
lexicalised trigger

Coarticulation

  • Phonetic* effect
  • Variable/inconsistent V(/C) qualities produced
  • Inconsistent extent of spread
  • No new V(/C) target—instead, articulation steered from existing target
Kyrgyz: intervening anterior Cs trigger fronted posterior Vs (own field work and pilot experiments)
URSRgloss
/toq-nIKI/[toqtuqu] one whose belly is full-poss.pred
/tok-nIKI/[toktük̠ü]~[toktüqu]electricity-poss.pred
/sok-(s)I/[soɣü](*[soʁu])juice-poss.3
/toqoj-LUː/ [toqojlüː]~[toqojlü͜u] (*[toqojɫuː]) forest-orn
coarticulation
assimilation
underspecification
lexicalised trigger

Coarticulation

  • Gets grammaticalised as assimilation
Turkish (Celements & Sezer, 1982)
/idrɑː-(s)I(n)-dA/[idrɑːkʲinˈdɛ](*[idrɑːɯndɑ])perception-poss.3-loc
/golf-(s)I(n)-dA/[golfynˈdɛ](*[ɡfundɑ])golf-poss.3-loc
Kazakh, with variation (own field work, corpus searches)
/bank-dA/[bɑŋkˈti͜ɘ](~[bɑŋkˈtɑ])bank-loc
/ru-nIŋ/[ruˈdɘŋ](~[ruˈdəɴ])steering wheel-gen
  • Coarticulation ultimate origin of all VH phenomena?
coarticulation
assimilation
underspecification
lexicalised trigger

Coarticulation

Kyrgyz: intervening anterior Cs trigger fronted posterior Vs (own field work and pilot experiments)
URSRgloss
/toq-nIKI/[toqtuqu] one whose belly is full-poss.pred
/tok-nIKI/[toktük̠ü]~[toktüqu]electricity-poss.pred
/sok-(s)I/[soɣü](*[soʁu])juice-poss.3
/toqoj-LUː/ [toqojlüː]~[toqojlü͜u] (*[toqojɫuː]) forest-orn
  • Not really spreading, how to implement?
  • * "It's just phonetics" ... but how?
    • "Phonology-like" patterns, but phonological targets already established
  • Second level of phonology?
  • lexical/tok-nIKI/
    PHONOLOGY
    targets|toktuqu|
    "PHONETICS" (or ..?)
    lexical[toktük̠ü]~[toktüqu]
coarticulation
assimilation
underspecification
lexicalised trigger

Extra-phonological patterns

  • Morpheme triggers VH in a way not derivable from phonological form
  • Implementable via floating feature
Western Yugur: sporadic anteriority changes, harmony per etymology (Roos, 2000)
URSRgloss
/un-GA/
[unki] (*[unqɑ])come out-vadv← *yn
/pɯl-GA/[pɯi](*[pɯɑ])know-vadv← *bil
/jy-GA/[jyɣɑ](*[jyɣi])wash-vadv← *ju
/jiɣ-GA/[jiqqɑ](*[jikki])gather-vadv← *jɯɣ
Kazakh & Kyrgyz: *j~n + *et- do how retains deleted V's anteriority (own field work)
stem-/GAn/ pst-/Á-sIŋ/ npst-2sg
kaz/jt/[qʰɑjtˈkʰi͜ɘn](*[qʰɑjtˈqʰɑn])[qʰɑjˈtʰi͜ɘsɘŋ](*[qʰɑjˈtʰɑsəɴ])
kir/nt/[qʰɑntˈkʰen](*[qʰɑntˈqʰɑn])[qʰɑnˈtʰes](*[qʰɑnˈtʰɑsɯɴ])

Summary

  • Grammaticalisation trajectory of VH
  • VH has origins in phonetics
  • Some VH patterns are still "phonetic", some are "phonological"
  • Different types of phonological mechanisms required
    • Underspecification
    • Unlinking, linking of multiple features
  • VH targets can find their way into the lexicon (a kind of "overspecification")
  • A "phonology-like phonetics", a second level of phonology

Thank you!


Thanks also to: one anonymous participant, two anonymous reviewers for Tu+8, various reviewers and the editors of the OUP Handbook on vowel harmony, IU University Graduate School Grant-in-Aid of Doctoral Research, Swarthmore Faculty Research Funding

Appendices

Historical fronting to [æ]

Kazakh: fronting between palatals (field work)
URSRgloss
ɑʃ-GA/æʃˈqʰɑ](*[ʃɑʃˈqʰɑ], *[ʃæʃˈkʰi͜ɘ])hair-dat
ɑʃ-(I)m/æˈʃəm](*[ʃɑˈʃəm], *[ʃæˈʃɘm])hair-poss.1sg
  • UR is VH trigger?
  • Alternatively: [æ] is not "front" in correct anteriority dimension
    cf. Nevins and Vaux (2006) on /j/
Tatar: grammaticalised fronting between palatals (field work, corpus searches)
æɕ-GA/æɕˈkʰæ](*[ɕæɕˈqʰɑ])hair-dat
æɕ-(I)m/æˈɕɘm](*[ɕæˈɕəm])hair-poss.1sg

Secondary fronting phenomena

  • Kazakh, e.g. /ʃɑʃ-NI/ → [ʃæʃtə] 'hair-ACC', /ʒɑj/ → [ʒæj] 'slow'
    • cf., Tatar /ɕæɕ-NI/ → [ɕæɕnɘ] 'hair-ACC'

Kazakh [æ]

  • Anterior root to vowel harmony when not stem-final:
    e.g. /æn-GA/ [æŋɡiɘ] 'song-DAT'
  • Posterior root to vowel harmony when not stem-final:
    e.g. /kʉnæ-GA/ [kʉnæʁɑ] 'guilt-DAT'
  • Tongue root posterior, tongue body anterior?