Evidence for Stratal Phonology: Stem-level phonology and morphological structure

Presentation at the University of Manchester

1. Evidence for Stratal Phonology Stem-level phonology and morphological structure Pavel Iosad 23rd October 2018 1
2. Outline • Stem-level cyclicity in Stratal Phonology • Stem-level cyclicity and parts of speech: Welsh svarabhakti • Stem-level cyclicity and phonemic structure: Irish vowel separation • Converging evidence for stem-level cyclicity: Russian [e] ∼ [’o] revisited 2
3. Stratal Phonology and the stem level
4. Basic assumptions of Stratal Phonology • As defined by Bermúdez-Otero (2018), Stratal Phonology • respects cyclicity • respects stratification • builds on parallelist constraint-based theories 3
5. Roots, stems, and words • Roots are lexical items with no part-of-speech characterization • Roots are not cyclic domains • Stems are lexical items with POS characterization, but not inflectable words • Some stems define cyclic domains for stem-level phonological computation • Stem-level domains can be recursive • Words are autonomous lexical items with the full set of inflections • Words are cyclic domains for word-level phonological computation • Word-level domains are not recursive • Utterances are cyclic domains for phrase-level phonological computation • Phrase-level domains are not recursive 4
6. The lexical syndrome • In Lexical Phonology and Morphology, ‘lexical’ rules had a number of properties (Kaisse & McMahon 2011) • Cyclic reapplication • Non-derived environment blocking • Categorical application • Exceptionality • Structure Preservation 5
7. Good evidence for stratification • Some languages provide good evidence for stem-level constituency • Lexicon stratification: English (Giegerich 1999), Hebrew (Meir 2006) • Spanish: morphological constituency (Bermúdez-Otero 2013) (1) Spanish manos ‘hands’ word stem root man- stem vowel -o- inflection -s 6
8. Worse evidence for stratification • Not all languages offer such apparently clear evidence for the distinction between stem and word level • How do we distinguish between • Evidence for process ordering; and • Evidence for stratification? • Stratification is a middle ground between • Non-morphological process ordering • Morpheme-specific domain structure 7
9. Stems and parts of speech: Welsh svarabhakti
10. Welsh svarabhakti • Apparently well-behaved repair of sonority sequencing violations (Hannahs 2009) (2) Epenthesis in monosyllables a. [ˈoːχɔr] ochr ‘side’ b. [ˈɔχrɛ] ochrau ‘sides’ (3) Deletion in polysyllables a. [pɛːrɪɡ] perygl ‘danger’ b. [pɛˈrəɡlon] peryglon ‘dangers’ 8
11. Welsh svarabhakti and the stem level i • It turns out that svarabhakti-related phenomena suffer from the stem-level syndrome (Iosad 2017) • Part-of-speech specificity: √ llwfr ‘cowardly’ in Nantgarw (Thomas 1993) (4) Nouns: transparency a. [ˈɬʊvrɪn] llyfryn ‘coward’ b. [ˈɬʊvrɔd] llyfriaid ‘cowards’ 9
12. Welsh svarabhakti and the stem level ii (5) Adjectives and deadjectival derivations: overapplication a. [ˈɬuːvʊr] llwfr ‘cowardly’ b. [ɬʊˈvʊrdra] llyfrdra ‘cowardice’ c. [ɬʊvʊˈrai] llyfrháu ‘to become cowardly’ • Exceptionality 10
13. Welsh svarabhakti and the stem level iii (6) a. [ˈbaːrav] barf ‘beard’ b. [ˈfɪrv] ffurf ‘form’ c. [ˈsoːvɔl] sofl ‘stubble’ d. [ˈɡwɛvl] gwefl ‘lip’ • Cyclicity: less in Modern Welsh, but rife in Middle Welsh • am(y)l ‘plentiful’, but <amylach> ‘more plentiful’ • kened(y)l ‘nation’, but <kenedyloed> ‘nations’ 11
14. Where is the stem? • Phonologically, svarabhakti ‘looks like’ a stem-level pattern • But: morphological evidence for stems is much weaker • No obvious stratification • Little obvious stem-based morphology • Some verbalizing suffixes, but that is about it 12
15. Strata with weak morphological evidence • Crucially, patterns of cyclic misapplication • follow the derivational history • never straddle part-of-speech boundaries: no patterns like [ˈɬuːvʊr] Adj ∼ [ɬʊˈvuːrɪn] N ∼ [[ɬʊvˈr] Adj ai] V • Predicted by Stratal Phonology from first principles: stem-based storage 13
16. Stems and overapplication: Irish vowels
17. Irish vowel inventory • Long vowels: at least 5 [iː uː eː oː ɑː] • Most consonants can be phonemically ‘non-palatalized’ or ‘palatalized’ • Long vowels have a free distribution (7) a. [kʲuːnʲ] ciúin ‘quiet’ b. [bˠiːnˠ] buíon ‘band, company’ • Short vowels: more restricted distribution 14
18. Irish short vowels: distribution • See Ó Maolalaigh (1997) for the generalizations • All examples from Cois Fhairrge (De Bhaldraithe 1945, 1953) unless stated otherwise (8) a. [ˈtʲitʲimʲ] tuitim ‘I fall’ b. [ˈkur] cur ‘putting’ c. [ˈdinʲə] duine ‘man’ d. [ˈkudʲ] ∼ [kidʲ] cuid ‘share’ e. [ˈfʲis] fios ‘knowledge’ f. [ˈtʲuki] tiocfaidh ‘will come’ 15
19. Irish morphology: slenderization • Irish morphology makes extensive use of changes in the palatalization of final consonants (9) a. [bɑːd] bád ‘boat.nsg’ b. [bɑːdʲ] báid ‘boat.gsg’ (10) a. [kruːnʲ] coróin ‘crown.nsg’ b. [kruːNəx] corónach ‘crown.gsg’ • Since the realization of short vowels depends on the palatalization of surrounding consonants, we expect short vowels to alternate 16
20. Irish short vowels: alternations (11) a. [filʲ] fuil ‘blood.nsg’ b. [fulə] fola ‘blood.gsg’ (12) a. [trɛdʲ] troid ‘fight’ b. [trʌdə] troda ‘fight.gsg’ • But there are many vowel patterns (13) a. [tilʲ] toil ‘will’ b. [tʌləx] tola ‘will.gsg’ 17
21. Irish short vowels: analysis • Three underlying vowels [i̵ ə a] • Phonemic analysis with allophony (Krauss 1958, Skerrett 1967, Bliss 1972) • Rule-based phonology with ‘separation rules’ (Wigger 1970, Ó Siadhail & Wigger 1975, Ó Siadhail 1989) • Non-linear analysis with feature-filling spreading (Ní Chiosáin 1991, 1994) • Element Theory analyses (Cyran 1997, Bloch-Rozmej 1998, Anderson 2014, 2016) • Hence • /fi̵l + ʲ/ → /filʲ/ fuil • /fi̵l + ə/ → /fulə/ fola • Underlying ‘vertical’ system 18
22. Irish short vowels: problems • The most worked-out rule-based analysis is by Ó Siadhail (1989), which is problematic in many ways • Ó Sé (1982): complementary distribution cannot be sustained due to exceptions in derived forms • Ó Maolalaigh (1997): in underived forms, the vertical analysis can be sustained but for a few exceptions • mionna ‘oath’, brionglóid ‘dream’ with [i] 19
23. Separation rules are stem-level: interaction with morphology • Separation rules follow some morphology, notably slenderization • In some varieties, evidence that they precede other morphology • Corca Dhuibhne (Ó Sé 1982, 2000) (14) a. [ɡidʲ] goid ‘steal.imp.sg’ b. [ɡitər] goidtear ‘steal.impers.pres’ 20
24. Separation rules and opacity • Also in Corca Dhuibhne, word-final [xʲ] deletion counterbleeds vowel separation (15) a. [klʌx] cloch ‘stone.nsg’ b. [klɛxʲə] cloiche ‘stone.gsg’ c. [klɛ] cloich ‘stone.dsg’ d. *[klʌ] 21
25. Separation rules are stem-level • Pre-sonorant lengthening: vowels lengthen/diphthongize before coda ‘fortis’ sonorants (e.g. Hickey 1986, Ní Chiosáin 1991) (16) Case inflection a. [ɡʲlʲɑːN] gleann ‘valley.nsg’ b. [ɡʲlʲɑːNtə] gleannta ‘valley.npl’ c. [ɡʲlʲæNə] gleanna ‘valley.gsg’ • Backness separation transparently interacts with PSL (17) a. [tuːN] tonn ‘wave.nsg’ b. [tiːNʲ] toinn ‘wave.dsg’ c. [tiNʲə] toinne ‘wave.gsg’ 22
26. More interaction with morphology: diminutives i • The productive diminutive suffix -ín slenderizes the final consonant of the stem (18) a. [Lʲaur] leabhar ‘book’ b. [Lʲaurʲiːnʲ] leabhairín ‘book-dim’ • This often leads to the expected alternations (19) a. [kruk] cnoc ‘hill’ b. [krikʲiːnʲ] cnuicín ‘hillock’ 23
27. More interaction with morphology: diminutives ii (20) a. [sʌp] sop ‘wisp, bundle (of straw)’ b. [sɛpʲiːnʲ] soipín ‘id.-dim’ • But crucially, short /a/ behaves differently in inflection-driven slenderization and before -ín • In inflection, /a/ in a slender context raises to [e] or [i] (21) a. [lʲæk] leac ‘flagstone’ b. [lʲekʲə] leice ‘flagstone.gsg’ (22) a. [ɡlas] glas ‘lock’ b. [ɡlɛʃ] glais ‘lock.gsg’ 24
28. More interaction with morphology: diminutives iii (23) a. [fʲær] fear ‘man’ b. [fʲirʲ] fir ‘man.gsg’ • In the diminutive context, we get cyclic misapplication rather than raising (24) a. [ɡad] gad ‘withe.nsg’ b. [ɡadʲiːnʲ] gaidín ‘withe.dim’ • We even get /a/ in a Cʲ_Cʲ context, which is basically impossible in underived forms 25
29. More interaction with morphology: diminutives iv (25) a. [bʲæn] bean ‘woman.nsg’ b. [bʲænʲiːnʲ] beainín ‘woman.dim’ • However, many lexical items variably apply the ‘inflectional’ separation rules (26) a. [aLt] alt ‘joint.nsg’ b. [æLtʲiːnʲ] ailtín ‘joint.dim’ c. [ɛLtʲiːnʲ] ‘id.’ 26
30. The stratal affiliation of separation rules • Separation rules can overapply before verbal inflectional suffixes (word-level?) • Separation rules can overapply before the productive derivational diminutive -ín • Separation rules interact transparently with Pre-Sonorant Lengthening, which itself is counterbled by diminutive slenderization (27) a. [kaiLʲ] coill ‘forest.nsg’ b. [keLʲə] coille ‘forest.gsg’ c. [kaiLʲiːnʲ] coillín ‘forest.gsg’ 27
31. Separation are stem-level: semantic evidence • Variable application of separation rules: • [sɛpʲiːnʲ] soipín is [[ √ səp + ʲiːnʲ]ℒ ]ℒ • [bʲænʲiːnʲ] beainín is [[ √ bʲan]ℒ + ʲiːnʲ]ℒ • Where De Bhaldraithe (1953) reports a distinction in meaning between variants, it goes in the predicted direction • Stem attachment: cyclic misapplication, compositional meaning • raca [rakə] ‘comb’, raicín [rækʲiːnʲ] ‘wee comb’ • scead [ʃkʲæd] ‘small piece’, sceaidín [ʃkʲædʲiːnʲ] ‘diminutive of scead’ • Root attachment: transparent separation rules, idiomatic meaning • roicín [rekʲiːnʲ] ‘cogwheel’ • sceidín [ʃkʲedʲiːnʲ] ‘small load’ 28
32. Separation rules show the stem-level syndrome • Exceptions in underived forms: [mʲiNə] mionna • Failure to apply in some derived forms: [ærʲimʲ] airm, gsg of [arəm] ‘weapon’ • Overapplication before plausibly word-level suffixes • Verbal inflection • Productive, compositional diminutive 29
33. But isn’t it inflection? • It appears that vowel separation rules and Pre-Sonorant Lengthening both belong to the stem level, as they overapply in word-level contexts such as diminutives • These processes are particularly active in case and number inflection of nouns and adjectives • Is case and number inflection stem-level? • I would argue this is quite plausible 30
34. Stem structure in Irish • In nouns, stem structure is not easily observable morphologically: there are no ‘thematic’ elements or overarching patterns of syncretism • In verbs, stem structure is more visible: inflection combines a choice of ‘stem’ with a set of person-number suffixes to signal TAM features • Nouns • Very few patterns are productive (Carnie 2008): probably a good deal of lexical storage • See Acquaviva (2006) for a morphosyntactic/semantic argument in favour of decomposing case and number inflections • Verbs • Recent morphosyntactic work compatible with the idea that Irish verbal stems represent spans of morphosyntactic terminals, just as envisaged in stem-storage theories (Acquaviva 2014, Ostrove 2018) • Overapplication of PSL is at least possible in verbs: cailleann ‘loses’ [kaLʲəN] or [kɑːLʲəN] (De Bhaldraithe 1953) 31
35. Conclusion • ‘Vowel separation’ patterns in Irish show all signs of belonging to the stem level • Cyclicity • Exceptionality • Variable application • This is despite the direct evidence for internal stem constituency often being somewhere between ‘subtle’ and ‘non-existent’ • No obvious evidence for stratification, either • Nevertheless, Stratal Phonology makes the right predictions 32
36. Converging evidence for stem structure: Russian
37. The [e] ∼ [’o] alternation i • A classic problem in Russian phonology (Trubetzkoy 1934, Lightner 1969, Polivanova 1976, Itkin 1994, 2007) • In native vocabulary, surface [e] only follows palatalized consonants and [ʂ ʐ t͡s] • Before a following non-palatalized consonant, some stressed [e]’s alternate with [o] (28) a. [sʲelʲ-skʲ-ij] сельский ‘rural’ b. [sʲol-a] сёла ‘village-npl’ • In some morphemes, [e] never alternates: 33
38. The [e] ∼ [’o] alternation ii (29) a. [bʲel-i̵j] белый ‘white’ b. [bʲelʲ-inʲkʲ-ij] беленький ‘white-dim’ • Yet in others, [o] after a palatalized consonant never alternates (30) a. [tʲotʲ-a] тётя ‘aunt’ b. [tʲot-uʂk-a] тётушка ‘aunt-dim’ 34
39. The historical background and nature of the pattern • Non-alternating [e] goes back to Old Russian *ě (written <ѣ>) • Alternating [e] goes back to Old Russian *e (written <е>) • Old Russian *e, but not *ě, > o / Cʲ_C • Later, [o] spread to a number of items where it is not motivated historically • Lightner (1969): underlying /ě/ and /e/, a backing rule, plus extra machinery to explain overapplication 35
40. The morpheme-based analysis • Lightner’s analysis is beset with empirical difficulties (Itkin 2007), but its use of juncture and constituency to deal with some of them signals morphological entanglement • A better analysis: the presence of [’o] derives not from the Cʲ_C context but from the properties of the following morpheme • Polivanova (1976): suffixes can ‘allow’ or ‘require’ [’o] in the preceding morpheme • Itkin (1994, 2007): suffixes that palatalize a preceding consonant also block [’o] (to be revised) • Cubberley (2002) gives a similar description 36
41. Stem structure and palatalization i • A stratal analysis of Russian has been defended previously by Rubach (2000); Blumenfeld (2003); Gribanova (2008); (2009) • In many respects, it represents an attempt to rationalize earlier analyses with extrinsic ordering by positing strata • Classic analysis (Lightner 1972, Plapp 1996, Halle & Matushansky 2002) • Underlying /i/: palatalizes non-velars; coronalizes velars • Underlying /i̵/: does not affect non-velars; palatalizes velars (and fronts itself) 37
42. Stem structure and palatalization ii (31) Verbal /i/ a. [krʲik] крик ‘shout.nsg’ b. [krit͡ʃʲ-it] кричит ‘to shout-pres.3sg’ c. [svʲet] свет ‘light.nsg’ d. [svʲetʲ-it] светит ‘to light-pres.3sg’ (32) Nominative plural /i̵/ a. [krʲik] крик ‘shout.nsg’ b. [krikʲ-i] крики ‘shout-npl’ c. [kʲit] кит ‘whale.nsg’ d. [kʲit-i̵] киты ‘whale-npl’ 38
43. Stem structure and palatalization iii • The crucial stratal difference is • Stem-level /ki/ → [t͡ʃʲi] • Word-level (/ki̵/ →) /ki/ → [kʲi] • …and similarly /e/ • Gribanova (2008, 2009): evidence for a stratal distinction from yer behaviour, supported by morphosyntactic evidence • Problem: ample evidence that palatalization is not caused by the features of the vowel (Padgett 2011) • Cf. the ‘palatalizing morphophonemes’ of Itkin (2007) (33) a. [vor] вор ‘thief’ b. [varʲ-uɡʲa] ворюга ‘thief.pejor’ 39
44. Stem structure and palatalization iv (34) a. [krʲuk] крюк ‘hook.nsg’ b. [krʲut͡ʃʲ-ok] крючок ‘hook-dim-nsg’ c. [krʲut͡ʃʲ-k-a] крючка ‘hook-dim-gsg’ • Suggested solution (Iosad & Morén-Duolljá 2010): palatalization is caused by a floating feature • Stratal differences in the outcome of the floating feature docking? 40
45. The [e] ∼ [’o] alternation and suffixes i • As Itkin (2007) observes, all suffixes that require a preceding morpheme to have [e] also cause stem-level palatalization of preceding consonants (35) a. [ɡrʲop] грёб ‘row.past.sg.masc’ b. [ɡrʲebʲinʲ] гребень ‘comb’ (36) a. [lʲod] лёд ‘ice’ b. [ɡala-lʲedʲ-it͡s-a] гололедица ‘ice crust’ (37) a. [ɡrʲoza] грёза ‘dream-nsg’ b. [ɡrʲeʒ-u] грежу ‘I dream’ c. [ɡrʲezʲ-it] грезит ‘(s)he dreams’ 41
46. The [e] ∼ [’o] alternation and suffixes ii • And conversely, all suffixes that require [’o] do not palatalize a preceding consonant (38) a. [tvʲerdʲ] твердь ‘firmament’ b. [tvʲord-i̵j] твёрдый ‘solid’ (39) a. [pa-sʲelʲ-it] поселит ‘(s)he will settle’ b. [pa-sʲol-ak] посёлок ‘settlement’ • Generalization: if a suffix causes stem-level palatalization, it also requires a preceding morpheme to take [e] if that morpheme has an [e] allomorph • The fronting is caused by the presence of the palatalizing feature, and is active at the stem level 42
47. ‘Indifferent’ suffixes i • Some palatalizing suffixes do not require preceding morphemes to take [e] (Itkin 2007) (40) Case suffixes in /e/ a. [utʲos] утёс ‘cliff.nsg’ b. [utʲosʲi] утёсе ‘cliff.prep.sg’ (41) Past tense plural /i/ a. [mʲorz-nu-tʲ] мёрзнуть ‘be cold.inf’ b. [mʲorz-l-i] мёрзли ‘be cold.past.pl’ 43
48. ‘Indifferent’ suffixes ii (42) Diminutive /ik/ a. [t͡ʃʲort] чёрт ‘devil’ b. [t͡ʃʲortʲ-ik] чёртик ‘wee devil’ (43) Diminutive /et͡s/ (with a yer) a. [rʲiʂot] решёт ‘sieve.gen.pl’ b. [riʂot-t͡s-a] решётце ‘sieve.dim’ • Similarly, some non-palatalizing suffixes do not influence the [e] ∼ [’o] alternation 44
49. ‘Indifferent’ suffixes iii (44) Female /ok/ (with a yer) a. [t͡ʃʲuʐi̵-zʲemʲ-it͡s] чужеземец ‘foreigner’ b. [t͡ʃʲuʐi̵-zʲem-k-a] чужеземка ‘female foreigner’ c. [nava-sʲol] новосёл ‘new settler’ d. [nava-sʲol-k-a] новосёлка ‘female new settler’ • ‘Indifferent suffixes’ generalizations: • Inflection or highly productive derivation • Never trigger stem-level palatalization • Itkin (2007) notes the contrast between ‘indifferent’ diminutive /ik/, /et͡s/ and [e]-requiring non-diminutive, non-compositional homophonous suffixes: 45
50. ‘Indifferent’ suffixes iv (45) a. [varʲ-on-i̵j] варёный ‘boiled’ b. [varʲ-enʲ-ik] вареник ‘dumpling’ (46) a. [lʲiʂ-on-n-i̵j] лишённый ‘deprived’ b. [lʲiʂ-enʲ-its] лишенец ‘one deprived of civil rights’ 46
51. Summary analysis • The [e] ∼ [o] alternation is a stem-level pattern • In frameworks with stem storage, if a stem has an [e] allomorph, it is chosen before a palatalizing suffix • This explains why only stem-palatalizing suffixes trigger fronting • Instead of absolute neutralization with underlying /ě/, the applicability of [e] ∼ [’o] is a matter of lexical storage • Word-level suffixes can palatalize preceding consonants, but do not affect stem allomorphy: obey locality and cyclicity • Consilience of • Phonological evidence: palatalization • Phonological evidence: [e] ∼ [’o] alternation • Morphological and semantic evidence • …despite the apparent lack of obvious stratification or stem morphology 47
52. Conclusion
53. Summary • The three cases considered here all suggest that Stratal Phonology makes the right predictions in several areas • Welsh: relationship between the lexical syndrome and part-of-speech characterization • Irish: distinction between stem- and word-level domains in the absence of robust root-stem-word morphology • Russian: convergent evidence for cyclic domains from several phonological and morphological phenomena • Stratal Phonology envisions just the right cyclic domain structure 48
54. Thank you! Diolch yn fawr! Go raibh maith agaibh! Спасибо! 48