lecture 2: theoretical approaches to morphologically

Morphophonology 07/11/2017

1

Lecture 2: Theoretical approaches to morphologically conditioned phonology

1. QUESTIONS ARISING FROM LAST LECTURE:

How many types of morphologically conditioned phonological patterns can there be in a language? How different can morphologically conditioned patterns in the same language be from one another? What happens when the same word contains two affixes which trigger conflicting morphophonological patterns? Which pattern prevails?

2. TODAY: A SURVEY OF MODERN THEORIES DESIGNED TO ANSWER THESE QUESTIONS

(1) What these theories have in common: the indexation of phonological constraints or subgrammars to morphological stems or stem types

• Stratal Ordering (Lexical Morphology and Phonology; Kiparsky 1982ab, 1984, 1985;

Halle & Mohanan 1985; Mohanan 1982, 1986), converted in the age of Optimality Theory to Stratal OT (Kiparsky 2000, 2003, 2008; Bermúdez-Otero 2011)

o Key claims of Stratal theory: a universal small, fixed, number of morphologically conditioned phonological patterns (strata), each typically associated with an affix ordering ‘zone’ and often in a relationship of decreasing phonological ‘oomph’; no connection to process morphology

• Cophonology Theory (e.g., Orgun 1996, Inkelas, Orgun & Zoll 1997, Anttila 2002, Inkelas & Zoll 2007) (bears similarities to Indexed Constraint theory (e.g., Benua 1997ab; Alderete 1999, 2001; Itô & Mester 1999))

o Key claims of Cophonology Theory: potentially as many distinct morphologically conditioned phonological patterns as there are morphological constructions in a language; no built-in connection to affix ordering; same technology accomplishes process morphology

Cophonology theory is basically Stratal OT, stripped of the assumption that there’s an upper bound on the number of strata and stripped of the assumption that strata are extrinsically ordered.

3. COPHONOLOGY THEORY

• Associates each individual morphological construction with its own phonological grammar

• Is designed to capture language-specific detail, within a broader framework that allows cross-linguistic comparison

• Like LMP and Stratal OT, is inherently cyclic (interleaving follows from architecture)


2

4. STRATAL ORDERING

• Every morphological construction is assigned to one of a small number of distinct and strictly ordered levels/strata

• Each level/stratum is associated with its own phonological subgrammar • There is some small fixed number of these, between 2-4 (maybe 5) in original LMP; 2-2

(root, stem, word) in Stratal OT • Case studies in Stratal OT in particular often focus on opacity, which is problematic for

standard OT, and has been argued to follow from stratal ordering • Challenges arise from languages with more than the allowed # of strata, and/or languages

in which strata and affix order don’t correlate (2) “The key principles of Stratal Phonology are cyclicity and stratification” [B-O6]

4.1 THE ORIGINAL CASE STUDY IN STRATIFICATION: ENGLISH (e.g., Kiparsky (1982a)) (3) Level 1 Class 1 (“+-boundary) derivation, inflection Level 2 Compounding , Class 2 (“#-boundary”) derivation Level 3 Class 2 (“#-boundary”) inflection [Kiparsky 1982b moves this to Level 2] (4) Derivational affixes Class 1: -al, -ity, -ic, -ive, -ion, -ate, -ous, in-, con-, pre-, en-, de-, re-... [Latinate]

Class 2: -ness, -hood, -less, -ful, re- un-, non-, under-,... [Germanic]

(5) Semantic generalizations: Class 1: often irregular, less productive Class 2: regular, more productive

(e.g. curiosity vs. curiousness; *blandity, blandness, drill, driller, etc.)

(6) Ordering generalizations

root+I+I root+I#II root+II#II *root#II+I

act+iv+ate act+ion-less hope-less-ness *hope-less+ity

I+I+root II#I+root II#II#root *I+II#root

in+con+ceivable non#con+formity non#re#fillable *ir#re+fillable


3

(7) Phonological generalizations

isolation class 1 suffix class 2 suffix

i. STRESS (stress-changing) (stress-neutral)

párent parént+al párent#hood

orígin orígin-al, origin-ál-ity

próperty próperty#less(#ness)

ii. TRISYLLABIC LAXING opáque [ej] opác-ity [æ] opáque#ness [ej] divíne [aj] divín-ity [ɪ] divíne#er [aj] tóne [ow] tón-ic [a] tóne#bearing [ow] iiii. PALATALIZATION/SPIRANTIZATION (I) classic [k] classic-ism [s] lick-ing [k] classic-ist [s] clique-ish [k] cyclic [k] cylic-ity [s] pleas-er [z] please [z] pleas-ure [ʒ] pleas-ing [z] iv. NASAL PLACE ASSIMILATION ir-resolute un-wrap il-legal un-lawful im-mobile u{n,m}-moved (ditto) i{n,ŋ}-congruous u{n,ŋ}-clasp co{n,ɱ}-fused u{n,m}-fortunate v. CLUSTER PRESERVATION VS. SIMPLIFICATION

iamb [m] iamb+ic [mb] iamb#s [m]

long [ŋ] long+est [ŋg] long#ing [ŋ]

prolong [ŋ] prolong+ation [ŋg] prolong#ed [ŋ]

sign [n] sign+al [gn] sign#er [n]

malign [n] malign+ant [gn] malign#ing [n]

hymn [m] hymn+al [mn] hymn#s [m]

damn [m] damn-ation [mn] damn#er

(8) Neat, huh? (9) Problem: exceptions (Aronoff & Sridhar 1983, Halle & Mohanan 1985)

a. -able: • /n/-deletion (Str. 2): damn-able

• truncates -ate (Str. 1): navigate, navigable; extricate, extricable

b. -ize: • /n/-deletion: solemn-ize

• Velar softening (Str. 1): Satanicize

• Vowel deletion (Str. 1): summar-ize (cf. curry, *curr-ing)

• /g/ deletion (Str. 2): monophthong-ize

c. -ist • /n/-deletion (Str. 2): column-ist

• Velar softening (Str. 1): classic-ist

Gussenhoven 1986: maybe these suffixes should all be Level 1. Stress shift is possible but not necessary in Level 1.

Ø But: if within a level there can be morphologically conditioned phonology, what is the point of saying it’s a phonologically uniform level?

Ø We’ll come back to this question when we visit process morphology on Friday.


4

4.2 CYCLICITY AND INTERLEAVING

(10) Terms I personally prefer to keep distinct (e.g. Chapter 7 of Inkelas 2014): Ø Cyclicity: the same phonological pattern is (re-)applied on each of several steps of word-

formation (relevant for stratum-internal recursion of phonology) Ø Interleaving: phonological patterns are applied at each step of word formation, but the

the phonological mappings associated with the invidual morphological constructions in question are not identical (relevant to cross-stratal application of phonology)

Cyclicity illustrated (Turkish) (11) Turkish: epenthesis (driven by syllabification considerations) applies each time a suffix is

added, i.e. cyclically, not once to the whole word a. i. /ʧaj-m-E/ ʧajɯma *ʧajma çayıma ‘tea-1SG.POSS-DAT’ ii. /el-n-I/ elini *elni elini ‘hand-2SG.POSS-ACC’ iii. /konuʃ-r-m/ konuʃurum *konuʃrum konuşurum ‘converse-AOR-1SG’ b. i. /ʧaj-dE/ ʧajda çayda ‘village-LOC’ ii. /konuʃ-tI/ konuʃtu konuştu ‘converse-PAST’

Cyclic Noncyclic Input to Cycle 1 /ʧaj/ /ʧaj-m-E/ Syllabification, Epenthesis, Vowel Harmony [ʧaj]σ [ʧaj]σ[ma]σ Input to Cycle 2 [ʧaj]σ + /m/ Syllabification, Epenthesis, Vowel Harmony [ʧa]σ[jɯm]σ Input to Cycle 3 [ʧa]σ[jɯm]σ + /Ε/ Syllabification, Epenthesis, Vowel Harmony [ʧa]σ[jɯ]σ[ma]σ Output ʧajɯma

çayıma ʧajma çayma

ü M


5

Interleaving illustrated (Turkish)

(12) Stress-neutral suffixes: a word consisting of a root and only these suffixes receives (default) final stress. Word-stratum: default final stress assignment

Pre-stressing suffixes: place stress on the preceding stem-final syllable. Overrides default final stress. Stem-stratum: stress stem-final syllable

(13) Words with only neutral suffixes:

a. arabá ‘car’ b. araba-lár ‘car-PL’ c. araba-lar-á ‘car-PL-DAT’ (14) Words with only pre-stressing suffixes: a. arabá ‘car’ b. arabá-mı ‘car-INTERR’ c. arabá-yla ‘car-COM = with a car, by car’) (15) Word with both kinds of suffixes: gel-me-di ‘come- NEG-PAST = didn’t come’

Input to Stem Stratum /gel-me/ Stem-final stress assignment [gélme]S Input to Word Stratum [gélme]S + /dI/ (Default) final stress assignment [gélmedi] Output ü

Often cyclicity and interleaving are conflated, used interchangeably, and often that’s fine for the point being made. Still, it’s important to understand the difference. ? Stratal theory is both cyclic (in that any given stratum could be specified as cyclic) and has interleaving (interaction across strata). (16) “Stratal Phonology, however, adds two other important claims. First, cyclic domain

structure is sparse: relatively few morphosyntactic constituents trigger phonological cycles. Secondly, there are different P-functions for cyclic nodes of different rank.” [B-O]


6

(17) Illustration from Bermudez-Otero paper:

Student Questions: 1. Iamwonderingif

StratalPhonologyhasabandonedtraditionalconceptsofmorph/morpheme.However,in(2),firsttwopartsofthestemseemtobearnomeaningthemselves.sowhyshouldwesplititintotwoparts?

2. Howdoestheauthordecidethestem[foraccommodate]is"commod"insteadof"acommod”?

3. Is“-ion”SLorWL?

Student Questions: 4.Iamcuriousabouthowitisdetermined,inthecaseofmultipleWLorSLaffixes,which/howmanyaffixestriggeraphonologicalcycle.IftwosuccessiveWLaffixesresultinonlythesecondtriggeringthecycle.Howdoesthisgetdetermined(which,orboth,affixesmakethedetermination)andhowdoesitrelatetolook-aheadinphonology?

(18)

(19) àdbracadábra is the stress pattern expected for a 5-syllable word accòmmodátion has a stress pattern influenced by the inner cycle on accómmodàte

• The ability of cyclicity to make inroads into the opacity problem, if not solve it outright, is often cited as an argument in favor of cyclic models; see e.g. Kiparsky 2000

• Larry Hyman will present many more examples of cyclicity in Week 4!

Page 7 of 45

(2) Nword

©

Nstem SGaffix

Astem Naffix

Nstem

©

Aaffix

Vstem

©

Naffix

a. affix √ Vaffix

b. a- commod -ateSL -ionSL -lessWL -nessWL -∅WL

c. PWL( PSL( PSL( a-, commod-, -ate), -ion), -less, -ness, -∅)

d. PSL 1st

cycle accómmodàte

2nd

cycle accòmmodátion

PWL 3rd

cycle accòmmodátionlessness

The order of P-function application is thus intrinsically determined by morphosyntactic

constituency: the computation of the phonological form of the parts precedes and feeds the

computation of the phonological form of the whole. Stratal Phonology derives a great deal of

its empirical content from this simple notion. Notably, like all cyclic frameworks, stratal

theories predict that morphosyntactically induced opacity is subject to Cyclic Containment:

(3) Cyclic Containment

In cases of morphosyntactically induced phonological opacity, a linguistic

expression inherits its opaque phonological properties from a constituent

defining an immediate cyclic subdomain.

The stress profile of the English word accòmmodátionlessness, for example, is doubly

opaque: first, the word exhibits prefenestral primary stress (i.e. primary stress outside the final

trisyllabic window); secondly, pretonic secondary stress fails to fall on the initial syllable (cf.

monomorphemic items like àbracadábra). As shown in (2), this is because accòmmodátionlessness

inherits the metrical contour of the noun stem accòmmodátion-, which defines an immediate

cyclic subdomain: accòmmodátion- is a cyclic constituent, and there is no other cyclic node

between accòmmodátion- and accòmmodátionlessness. In turn, accòmmodátion- inherits the foot-

Page 7 of 45

(2) Nword

©

Nstem SGaffix

Astem Naffix

Nstem

©

Aaffix

Vstem

©

Naffix

a. affix √ Vaffix

b. a- commod -ateSL -ionSL -lessWL -nessWL -∅WL

c. PWL( PSL( PSL( a-, commod-, -ate), -ion), -less, -ness, -∅)

d. PSL 1st

cycle accómmodàte

2nd

cycle accòmmodátion

PWL 3rd

cycle accòmmodátionlessness

The order of P-function application is thus intrinsically determined by morphosyntactic

constituency: the computation of the phonological form of the parts precedes and feeds the

computation of the phonological form of the whole. Stratal Phonology derives a great deal of

its empirical content from this simple notion. Notably, like all cyclic frameworks, stratal

theories predict that morphosyntactically induced opacity is subject to Cyclic Containment:

(3) Cyclic Containment

In cases of morphosyntactically induced phonological opacity, a linguistic

expression inherits its opaque phonological properties from a constituent

defining an immediate cyclic subdomain.

The stress profile of the English word accòmmodátionlessness, for example, is doubly

opaque: first, the word exhibits prefenestral primary stress (i.e. primary stress outside the final

trisyllabic window); secondly, pretonic secondary stress fails to fall on the initial syllable (cf.

monomorphemic items like àbracadábra). As shown in (2), this is because accòmmodátionlessness

inherits the metrical contour of the noun stem accòmmodátion-, which defines an immediate

cyclic subdomain: accòmmodátion- is a cyclic constituent, and there is no other cyclic node

between accòmmodátion- and accòmmodátionlessness. In turn, accòmmodátion- inherits the foot-


7

4.3 B-O: THERE IS ONE COUNTEREXAMPLE TO CYCLIC CONTAINMENT IN ALBRIGHT 2015

(20)

(21)

(22)

(23) “OO-Faith demands that inflected forms be faithful to a privileged [inflected] base form.

Converging evidence from several distinct cases of over- and underapplication indicate that the base of Lakhota verb paradigms is a form that ends in the ablaut vowel [a], which should condition velar aspiration. This quality is then carried over to other inflected forms by BASE-IDENT(frication)”

There are other examples of this kind in the literature; see e.g., Steriade 2008

4.4 BUT BACK TO STRATAL ORDERING THEORY: HOW MANY STRATA?

(24) “Stratal Phonology, however, adds two other important claims. First, cyclic domain structure is sparse: relatively few morphosyntactic constituents trigger phonological cycles. Secondly, there are different P-functions for cyclic nodes of different rank.”

(25) Stratal OT offers Stem, Word, Phrase But there are many languages for which analysts have posited more lexical strata than just two (Stem, Word).

4.4.1 MALAYALAM (26) Mohanan’s factory metaphor (1986:47): “There is a conveyor belt that runs from the entry gate to the exit gate passing through

each of these rooms. This means that every word that leaves the factory came in through the entry gate and passed through every one of these rooms.”

UR → [ level 1 ] → [ level 2 ] → [ level 3 ] → [ level 4 ]

Faithfulness to non-contrastive phonetic properties in Lakhota Adam Albright, MIT ([email protected])

It is well known that members of inflection paradigms may show unexpected phonological similarity, creating exceptions to general phonological processes (Kenstowicz and Kisseberth 1977; Kiparsky 1978). Approaches based on Output-Output Faithfulness (Kenstowicz 1997; Steriade 2000; McCarthy 2005) derive this similarity using special faithfulness constraints that hold only among paradigmatically related surface forms. Stratal approaches, on the other hand (Bermúdez-Otero 2011), account for this similarity using input-output faithfulness to interme-diate steps in the derivation. Under the latter approach, what distinguishes paradigmatically re-lated forms from other strings is that faithfulness constraints may be promoted partway through the derivation, effectively “turning off” the process. In this paper, I present new data from a paradigm uniformity effect in Lakhota (Siouan; Dakotan sub-branch) that cannot be analyzed using faithfulness to intermediate forms. The reasons are simple: (1) the necessary intermediate form does not seem to exist, and (2) even if it did, later stages in the derivation would require faithfulness to be too low to have any lasting effect on the output. Furthermore, paradigm uni-formity targets a property that is never contrastive in any language, raising the question of whether any IO-Faith constraint should ever be able to preserve it. I argue instead that the facts follow naturally from an account that incorporates OO-Faithfulness for phonetic properties. In all Dakotan dialects, stops and affricates contrast for aspiration and ejection: ka ‘there’ vs. kʰa ‘to mean’ vs. k’a ‘to dig’. The phonetic quality of aspiration varies considerably be-tween and within dialects, and it is realized sometimes with velar frication (kˣa) and sometimes with glottal frication (kʰa) (Ullrich 2008, p. 6). In the Lakhota dialect, both are employed, with the distribution determined primarily by the following vowel. Ullrich (2008, pp. 697–8) states that “[t]he two types of aspirated stops occur in complementary distribution and never contrast in meaning.” Nevertheless, because the difference in frication noise is contrastive for fricatives (/x/ vs. /h/), he transcribes them differently. Inspection of dictionary entries involving aspirated stops reveal the distribution in (1), favoring velar frication before non-high vowels, especially when back. (The vowel transcribed as [ũ] is actually realized partway between [ũ] and [õ].)

(1) Velar vs. glottal aspiration in Lakhota

pʰi pʰĩ pʰu tʰi tʰĩ tʰu kʰi kʰĩ kʰu pˣe pˣo pˣũ tʰe/tˣe tˣo tˣũ kʰe kˣo kˣũ pˣa pˣã tˣa tˣã kˣa kˣã

This distribution is productively enforced, and can be illustrated by placing the same mor-pheme in different following contexts. Lakhota generally bans codas, but it is possible to vary the following context by placing morphemes in certain syntactic contexts, where the final vowels of certain morphemes are deleted, as in (2). When this happens, the quality of aspira-tion is determined by the quality of the following vowel, respecting the distribution in (1).

(2) Readjustment of aspiration quality depending on the following vowel

/makha/ ‘earth’ + /a-máni/ ‘APPL-walk’ makxámani ‘travel on foot’ + /itʃú/ ‘take’ makhítʃu ‘take up land, settle’ + /eglakĩjã/ makheɡlakĩjã ‘across the earth (adv.)’ /napha/ ‘flee’ + /ijája/ ‘leave’ [naphíjaja] ‘leave a place fleeing’

Certain stem-final vowels also show paradigmatic [a]∼ [e]∼ [ĩ] alternations, in a process known as ablaut: [sapa] ∼ [sape] ∼ [sapĩ] ‘black’. The distribution in (1) leads us to expect that aspi-rated stops have velar aspiration before [a], dorsal before [ĩ], and variable realization before [e] (depending on place of articulation). However, aspiration is always velar before ablaut vowels:

Faithfulness to non-contrastive phonetic properties in Lakhota Adam Albright, MIT ([email protected])

It is well known that members of inflection paradigms may show unexpected phonological similarity, creating exceptions to general phonological processes (Kenstowicz and Kisseberth 1977; Kiparsky 1978). Approaches based on Output-Output Faithfulness (Kenstowicz 1997; Steriade 2000; McCarthy 2005) derive this similarity using special faithfulness constraints that hold only among paradigmatically related surface forms. Stratal approaches, on the other hand (Bermúdez-Otero 2011), account for this similarity using input-output faithfulness to interme-diate steps in the derivation. Under the latter approach, what distinguishes paradigmatically re-lated forms from other strings is that faithfulness constraints may be promoted partway through the derivation, effectively “turning off” the process. In this paper, I present new data from a paradigm uniformity effect in Lakhota (Siouan; Dakotan sub-branch) that cannot be analyzed using faithfulness to intermediate forms. The reasons are simple: (1) the necessary intermediate form does not seem to exist, and (2) even if it did, later stages in the derivation would require faithfulness to be too low to have any lasting effect on the output. Furthermore, paradigm uni-formity targets a property that is never contrastive in any language, raising the question of whether any IO-Faith constraint should ever be able to preserve it. I argue instead that the facts follow naturally from an account that incorporates OO-Faithfulness for phonetic properties. In all Dakotan dialects, stops and affricates contrast for aspiration and ejection: ka ‘there’ vs. kʰa ‘to mean’ vs. k’a ‘to dig’. The phonetic quality of aspiration varies considerably be-tween and within dialects, and it is realized sometimes with velar frication (kˣa) and sometimes with glottal frication (kʰa) (Ullrich 2008, p. 6). In the Lakhota dialect, both are employed, with the distribution determined primarily by the following vowel. Ullrich (2008, pp. 697–8) states that “[t]he two types of aspirated stops occur in complementary distribution and never contrast in meaning.” Nevertheless, because the difference in frication noise is contrastive for fricatives (/x/ vs. /h/), he transcribes them differently. Inspection of dictionary entries involving aspirated stops reveal the distribution in (1), favoring velar frication before non-high vowels, especially when back. (The vowel transcribed as [ũ] is actually realized partway between [ũ] and [õ].)

(1) Velar vs. glottal aspiration in Lakhota

pʰi pʰĩ pʰu tʰi tʰĩ tʰu kʰi kʰĩ kʰu pˣe pˣo pˣũ tʰe/tˣe tˣo tˣũ kʰe kˣo kˣũ pˣa pˣã tˣa tˣã kˣa kˣã

This distribution is productively enforced, and can be illustrated by placing the same mor-pheme in different following contexts. Lakhota generally bans codas, but it is possible to vary the following context by placing morphemes in certain syntactic contexts, where the final vowels of certain morphemes are deleted, as in (2). When this happens, the quality of aspira-tion is determined by the quality of the following vowel, respecting the distribution in (1).

(2) Readjustment of aspiration quality depending on the following vowel

/makha/ ‘earth’ + /a-máni/ ‘APPL-walk’ makxámani ‘travel on foot’ + /itʃú/ ‘take’ makhítʃu ‘take up land, settle’ + /eglakĩjã/ makheɡlakĩjã ‘across the earth (adv.)’ /napha/ ‘flee’ + /ijája/ ‘leave’ [naphíjaja] ‘leave a place fleeing’

Certain stem-final vowels also show paradigmatic [a]∼ [e]∼ [ĩ] alternations, in a process known as ablaut: [sapa] ∼ [sape] ∼ [sapĩ] ‘black’. The distribution in (1) leads us to expect that aspi-rated stops have velar aspiration before [a], dorsal before [ĩ], and variable realization before [e] (depending on place of articulation). However, aspiration is always velar before ablaut vowels:

(3) Paradigm uniformity under ablaut

‘hit’ ‘try to do’ ‘tell the truth’ 3sg+PROG apˣa-hã ijutˣa-hã witʃakˣa-hã 3sg+NEG apˣe-ʃni ijutˣe-ʃni witʃakˣe-ʃni 3sg+EMPH apˣe-xtʃa ijutˣe-xtʃa witʃakˣe-xtʃa 3sg+FUT apˣĩ-kte ijutˣĩ-kte witʃakˣĩ-kte

The preservation of velar frication is problematic for a stratal account. First, it is not clear what intermediate stage of the derivation would be responsible for conditioning velar aspiration in forms like [apˣe-ʃni] or [apˣĩ-kte], since these forms do not contain the back vowel that would ordinarily condition it. One possibility is that these forms really do contain [apˣa] at some stage of the derivation; another possibility is that they simply contain [apˣ], and velar aspiration is the default realization when non-prevocalic. Although there is no direct support for either claim, let us assume for the sake of argument that the context is indeed present to derive inter-mediate [apˣ]. Then, an IO-IDENT constraint for aspiration quality must be promoted in order to preserve this quality. However, the forms in (2) show that this is unworkable: even if IO-IDENT(asp) is ranked highly at some point in the derivation, it must be demoted again at the phrasal level, so that the quality of aspiration can once again between determined by the fol-lowing vowel context. However, this later evaluation would obliterate intermediate [apˣ], and thus fails to predict paradigm uniformity. A second argument against the stratal approach comes from the fact that the putative IO- IDENT constraint must preserve the dorsal or glottal quality of aspiration. As far as I have been able to establish, this phonetic distinction has not been documented to be contrastive in any language, which raises the question why its effects are only seen in cases of paradigm uni-formity, and never basic contrasts. Both of these problems are circumvented in an approach in which OO-Faith demands that inflected forms be faithful to a privileged base form. Converging evidence from several dis-tinct cases of over- and underapplication indicate that the base of Lakhota verb paradigms is a form that ends in the ablaut vowel [a], which should condition velar aspiration. This quality is then carried over to other inflected forms by BASE-IDENT(frication), as in (4). Additionally, I argue that by distinguishing different dimensions of faithfulness (IO vs OO), we gain some insight into why aspiration quality can be preserved paradigmatically, but not as a lexical con-trast. Following Flemming (2008) and Steriade (2008), I argue that the these dimensions are driven by opposing forces: lexical contrasts should be perceptually maximally distinct, while allomorphs should be minimally distinct. This principle may preclude constraints such as IO- IDENT(frication quality) from CON altogether, or at least make it very unlikely that they will be satisfied. OO-IDENT, on the other hand, may freely targets such properties, since they contrib-ute to detailed similarity that helps to make allomorphs readily identifiable. This distinction is unavailable in an approach that eschews different dimensions of Faithfulness.

(4) Misapplication of aspiration quality

/apʰa-he/ Base-Ident(fric) *Cˣ/[+hi,…] *Cʰ a. apʰahe *! ☞ b. apˣahe

/apʰĩ-kte/ Base-Ident(fric) *Cˣ/[+hi,…] *Cʰ a. apʰĩkte *! * ☞ b. apˣĩkte *


8

(27) Mohanan (1982) proposes the following level ordering schema for Malayalam:

Stratum 1: Derivation Stratum 2: Subcompounding (head modifier semantics)

Stratum 3: Cocompounding (coordination semantics)

Stratum 4: Inflection

(28) Stratum n-

deletion Vowel sandhi

Nasal deletion

Stem-final vowel

lengthening

Gemination Tone and stress

1 ✓ ✓

2 ✓ ✓ ✓ ✓ 3 ✓ ✓ ✓ ✓ 4

(29) Gemination: affects obstruents in Dravidian stems at internal Subcompound (stratum 2)

juncture

a. meeşa + peṭṭi -kaḷ → meeşa+ppeṭṭi-kaḷə S ‘table’ ‘box’ -PL ‘table boxes = boxes made out of tables’ b. meeşa + peṭṭi -kaḷ → meeşa+peṭṭi-kaḷə C ‘tables and boxes’ c. kaaṭ + maram → kaaṭṭǝ+maram S [1982:37] ‘forest’ ‘tree’ ‘forest tree’ d. aaṭ + maaṭ -kaḷ → aaṭǝ+maaṭǝ-kaḷǝ C [1982:37] ‘goat’ ‘cow’ -PL ‘cattle’ (30) Stress assignment (stratum 3): stress a syllable with a long vowel, and the final syllable,

and the initial syllable unless followed by another stressed syllable. The leftmost stressed syllable is most prominent.

(31) Tone assignment (stratum 3): insert a LH melody. L links to syllable with main stress; H links to span containing final syllable and leftmost nonitial stressed syllable

Morphology Phonology

The lexicon

Level 1 Level 1

Level 2 Level 2

UR

... ...


9

(32) Tone and stress a. ṯii + waṇṭi + aappiis Subcompound fire cart office ‘fire station’

[ṯíiwaṇṭiyàappìisə] | | L H

b. yákṣan + kínnaran + gáṉḏharwwan Cocompound Yakshas Kinnaras Gandharwas ‘Yakshas, Kinnaras and Gandharwas’

[yákṣan] [kínnaran] [gáṉḏarwwan] | | | | | | L H L H L H Interim summary: Four (4) ordered lexical strata

4.4.2 TURKISH (33) Turkish: >2 several lexically assigned stress patterns alone

o Final stress: default o Stem-final stress: certain suffixes (overrides default) o ‘Sezer’ stress: imposed on place names

(34) Basque (Hualde 1988), Sekani (Hargus 1988), Kashaya (Buckley 1994)… up to 4-5 strata

have been motivated. Interim summary: More than two ordered lexical strata


10

4.5 STRATAL THEORY WEAKEND: LOOP OR NONORDERING EFFECTS

4.5.1 THE LOOP: MALAYALAM Stratal ordering predicts S inside C. Not predicted: C inside S. But: both occur (35) kaamuki ‘lady love’ bhaarya ‘wife’ sahooḏari ‘sister’

(36) Subcompound inside cocompound

[[ káamukii ] [[ bháaryàa ][ sahòoḏari ]]] -maarə ‘lady love and wife’s sister’ L H L H

(37) Cocompound inside subcompound

[[[ káamukii ][ bháaryàa]] [ sahóoḏari ] -maarə ‘sisters of lady love and wife’ L H L H L H (38) a. meeşa ‘table’ peṭṭi ‘box’ kasaala ‘chair’ b. meeşakasaalaappeṭṭikaḷǝ ‘boxes made from tables and chairs’ lit. [[ta-ch]C-bx]S-PL c. meeşapeṭṭikkasaalaa kaḷǝ ‘chairs made from tables and boxes’ lit. [[ta-bx]C-ch]S-PL (39) Mohanan: Loop

Stratum 1 → Stratum 2 ↔ Stratum 3 → Stratum 4

Loops have also been posited in Basque (Hualde 1988), Sekani (Hargus 1988) and Turkish (Inkelas & Orgun 1998).


11

4.5.2 TURKISH: LEVEL NONORDERING (INKELAS & ORGUN 1998) (40) a. Neutral suffixes inside place name: Sezer pattern imposed on outer stem [Kandil-li] (pace name) cf. kandil-li ‘oil lamp-ASSOC’ kanˈdilli kandilˈli [Ayran-cı] (place name) cf. ayran-cı ‘yogurt drink-AGT’ ajˈranʤɯ ajranˈʤɯ [Kuzgun-cuk] (place name) cf. kuzgun-cuk ‘raven-DIM’ kuzˈgunʤuk kuzgunˈʤuk b. Neutral suffixes outside Sezer stem: Sezer pattern prevails on inner stem [Menteşe]-cik ‘Menteşe-DIM cf. menteşe-cik ‘hinge-DIM’ ˈmenteʃeʤik menteʃeˈʤik [Kandil-li]-ye ‘Kandilli-DAT’ cf. kandil-li-ye ‘oil lamp-ASSOC-DAT’ kanˈdillije kandilliˈje c. Prestressing suffixes inside place name: Prestressing pattern prevails (on inner stem) [çam-lı-^ca] cf. çam-lı-ca ʧamˈlɯʤa ʧamˈlɯʤa (place name) ‘pine-ASSOC-MIT’ d. Prestressing suffixes outside Sezer stem: Sezer pattern prevails (on inner stem) [Ankara]-^mı cf. araba-^mı ˈankaramɯ araˈbamɯ ‘Ankara-INTERR’ ‘car-INTERR’ [Ankara]-lı-laş-^mı-dı cf. yaban-cɯ-laş-^mı-dı ˈankaralɯlaʃmɯdɯ jabanʤɯˈlaʃmɯdɯ ‘Ankara-ASSOC-VBL-NEG-PAST’ ‘foreign-AGT-ASSOC-VBL-NEG-PAST’


12

4.5.3 MOSES-COLUMBIAN SALISH (NXA'AMXCIN) Czaykowska-Higgins (1993): level ordering theory would require at least ten levels to account for morphologically conditioned stress patterns in Moses-Columbian Salish (Nxa'amxcin). (41) Suffixes in Nxa'amxcin are either dominant (stress-deleting) or recessive

(stress-preserving), in terms of their effect on base stress. o Root + recessive suffixes = root stress:

/ kʷuɫn-n-t-sa-xʷ-taʔ/ → kʷúɫncxʷtaʔ ‘lend-CTR-TR-1SGO-2SGS-IMP = lend it to me!’ [208]

o Root + dominant suffixes = stress on rightmost dominant: /kas- p iq-cin-cut-mix/ → kaspiqcncútǝxʷ ‘UNR-cook-FOOD-REFL-IMPF = he’s going to cook’ [209]

• There is no way to predict this phonological difference from the morphological properties of suffixes.

• Dominant and recessive suffixes are freely interspersed among each other in Nxa'amxcin words. The chart below illustrates in/transitiv(izer) suffixes:

• Either posit a loop or have a lot of levels with identical phonology.

Conclusion: a nexus between morphological zone and phonological subpattern cannot be taken for granted. If we apply the term ‘level’ to phonological subpattern, levels are not always ordered. Strata aren’t limited to 2…. they aren’t ordered….it’s starting to look like Cophonology Theory. But what about the generalizations?

4.6 AFFIX-SPECIFIC MORPHOLOGICALLY CONDITIONED PHONOLOGY

4.6.1 THE ORIGINAL CASE STUDY: ENGLISH (42) Level 1 Class 1 (“+-boundary) derivation, inflection Level 2 Compounding , Class 2 (“#-boundary”) derivation Level 3 Class 2 (“#-boundary”) inflection [Note: Kiparsky 1982b puts this on Level 2]

(43) Derivational suffixes Class 1: -al, -ity, -ic, -ive, -ion, -ate, -ous, -ism, -ist-... [Latinate]

Class 2: -ness, -hood, -less, -ful, -ish, -ing, -ize, -er… [Germanic]

CYCLICITY AND STRESS IN MOSES-COLUMBIA SALISH 269

of the in/transitiv(izer) category are either dominant or recessive. (73) lists suffixes of this category according to the order in which they occur relative to each other and indicates whether they are dominant or recessive. Suffixes of this category can cooccur (see Czaykowska-Higgins (1990b) for details).

(73) Order of Inltrans Suffixes

-xix- -min- -nun- -m- -tul- -t- -cut

-xax -xit- -stu- -waxw

-n- D R D R D R D

Within this category, there are no semantic/syntactic criteria by which one can predict whether a particular suffix will be dominant or recessive. Thus, for instance, while -stu 'causative' and -tul 'redirective' are both

grammatical-function-changing morphemes, one is recessive and the other is dominant. If one brings together the distributional and cyclic properties of the in/transitiv(izer)s with the properties of suffixes of the other categories, it turns out that it is necessary to postulate at least ten distinct strata for Cm. These strata are given in (74). I have specified for each stratum whether it is cyclic or noncyclic.

(74) Stratum 1: Primary Affixes, Lexical Suffixes Cyclic

Stratum 2: =mix (LS), =min (LS), =tn, Noncyclic

=xn, =lqst, =lqs

Stratum 3: -xix, -xax Cyclic

Stratum 4: -min, -xit Noncyclic

Stratum 5: -nun Cyclic

Stratum 6: -m, -stu, -1, -n Noncyclic

Stratum 7: -tul Cyclic

Stratum 8: -t, obj., subj. Noncyclic

Stratum 9: -cut, -waxw Cyclic

Stratum 10: -mix Noncyclic

The need to postulate ten strata for Cm undermines the predictive value


13

(44) Semantic generalizations: Class 1: often irregular, less productive Class 2: regular, more productive

(e.g. curios-ity vs. curious#ness; *bland-ity vs. bland#ness, drill vs. drill#er, etc.)

(45) Ordering generalizations: I closer to root than II

root+I+I root+I#II root+II#II *root#II+I

act+iv+ate act+ion-less hope-less-ness *hope-less+ity

(46) Phonological generalizations (sample)

isolation class 1 suffix class 2 suffix

i. STRESS (I) (stress-changing) (stress-neutral)

párent parént+al párent#hood

ínfinite infínit+ude ínfinite#ness

orígin orígin-al, origin-ál-ity

próperty próperty#less(#ness)

ii. TRISYLLABIC LAXING (I) opáque [ej] opác-ity [æ] opáque#ness [ej] divíne [aj] divín-ity [ɪ] divíne#er [aj] tóne [ow] tón-ic [a] tóne#bearing [ow] iiii. PALATALIZATION/SPIRANTIZATION (I) classic [k] classic-ism [s] lick-ing [k] classic-ist [s] clique-ish [k] cyclic [k] cylic-ity [s] pleas-er [z] please [z] pleas-ure [ʒ] pleas-ing [z] (47) Problem: exceptions (see e.g., Aronoff & Sridhar 1983, Halle & Mohanan 1985)

a. -ist • Velar softening (Str. 1): clássi[s]-ist • No Stress shift (Str. 2): hóspital, hóspital-ist (cf. hospitál-ity) b. -al • Stress shift (Str. 1): párent, parént-al • No Trisyllabic laxing (Str. 2): t[oː]n-al, vs. t[a]n-ic etc.

• Maybe -ist should be Level 1, but an exception to Stress shift (Gussenhoven 1986)? • Mark -al as extrametrical, therefore evades Trisyllabic Laxing (Hayes 1981)?

But: if within a level there can be morphologically conditioned phonology, what is the point of saying it’s a phonologically uniform level?

Stress Shift Trisyllabic Laxing Spir/Pal -ity: yes yes yes -ic: yes yes no -al: yes no ⎯ -ist: no no yes -ing: no no no

Five suffixes, five levels? Looks like Cophonology Theory.


14

4.7 BACK TO MALAYALAM

Mohanan and Mohanan (1984:588): Palatalization (of velars after a front vowel) is morphologically conditioned beyond what 4 levels can capture. It applies to the causative and verbalizing suffixes (which M&M locate in Stratum 2) but not at the internal compound juncture at Stratum 2; it applies to the dative suffix but not to the plural suffix (both stratum 4). It does not apply in Stratum 3 or postlexically. (48) Stratum Palatalization applies Stratum Palatalization does not apply 2 mara-k’k’-

‘cat-CAUS’ 2 kuṭṭi-kaḷi ‘child-game’

2 wira-k’k’- ‘tremble-VBLZER’

4 kuṭṭi-k’k’ǝ ‘daughter-DATIVE’

4 kuṭṭi-kaḷ ‘child-PLURAL’

4.8 INTERIM SUMMARY

• Stratal OT: 2 (“stem” and “word”) • Studies of highly affixing languages in the LMP framework: up to 5 (Malayalam, Turkish,

Sekani, Kashaya, Basque …) • Cophonology theory and Indexed Constraint theory: bounded only by the number of

morphological constructions in the language. In theory, each morphological construction could be associated with a distinct constraint ranking.

2, 3, 4 very large number this is a big difference!

isn’t this an empirical issue? why hasn’t this been resolved??

• Stratal OT is focused on high-level generalizations and on cross-linguistic tendencies, not on language-specific details.

• Once those are taken into account, Cophonology Theory is a more practical model. • But do we lose all generalization?

5. HOW DIFFERENT CAN MORPHOLOGICALLY CONDITIONED PATTERNS IN THE SAME LANGUAGE BE FROM ONE ANOTHER?

(49) Proposed means of constraining language-internal phonological variation • Strong Domain hypothesis (LMP; Kiparsky 1984)

o Rules all apply in Stratum 1, turn off at some later level • Stratum Domain hypothesis (LMP; Halle & Mohanan 1985)

o Each rule turns on at some stratum, turns off at some later stratum • Stratal OT (Kiparsky 2003)

o “I assume that the constraint system of level n+1 may differ in ranking from constraint system of level n by promotion of one or more constraints to undominated status”

These three are meaning-less if levels are not fully ordered


15

• Grammar Dependence (Alderete 1999, 2001a): formulated within Indexed Constraint theory. Strata differ only in the ranking of Faithfulness. Markedness constraint ranking, representing the default for the language, is invariant. See Inkelas & Zoll 2007.

• Grammar Lattice (Anttila 1997, 2002): any and all variation is possible; the ‘genius’ of a language is expressed via a master partial constraint ranking that all cophonologies must respect.

5.1 GRAMMAR LATTICE

Grammar Lattice: cophonologies of a language arranged in a hierarchy (lattice) (Anttila 1997, 2002, 2009) (50) Example (19) from Lecture 1: [Turkish] C-final root V-final root ‘do’ ‘come’ ‘understand’ ‘say’ jap gel anla søjle Facilitative /-Iver/: epenthesis jap-ɯver gel-iver anla-jɯver søyle-jiver Progressive /-Ijor/: deletion jap-ıjor gel-ijor anl-ɯjor søyl-yjor (51) Analyzed with a grammar lattice: Master Ranking *VV » {MAX-V, DEP-C} Cophonology A (-Iver) Cophonology B (-Ijor) *VV » DEP-C » MAX-V *VV » MAX-V » DEP-C

6. YES, BUT

• Are there any limits on variation? • What do we know about the actual range of variation in languages, anyway? • Could Cophonology theory describe a crazy language that has, say, Mandarin tone in

nouns and Carib stress on verbs?


16

1. Language-internally there is more variation between, say, ideophones and non-ideophones than there is between, say, verbs and nouns.

• In Yir-Yoront (Cape York Peninsula, Australia) ideophones allow certain onsets not otherwise permitted in the language, yet ban certain codas which are otherwise possible (Alpher 1994: 162).

• Any grammar that can describe the kinds of asymmetries holding between ideophones and non-ideophones in Yir-Yoront is clearly capable of describing comparable, or even more extreme, asymmetries between categories like noun and adjective, or passive vs. active verb, etc.

• Thus any formal principle that tolerates the kind of language-internal variation that exists between ideophones and non-ideophones will over- generate variation potential in other sectors of grammar.

• In this way, then, any broad-based constraints on grammatical variation are already doomed to miss a large part of the picture.

2. Language-internal variation and cross-linguistic variation are logically and materially related: language splits evolve from dialect splits, and dialect splits result from language-internal variation.

• Anttila (1997, 2002) has explicitly related subgrammatical variation, for which he posits cophonologies, to the kind of free variation giving rise to dialect splits, based on solid diachronic and synchronic data from Finnish (see also Anttila & Cho 1998on Korean).

• The fact that the degree of difference between two languages is typically greater than the degree of difference found among patterns within a language is surely related to the fact that distantly related languages differ from one another more than closely related languages do.

3. Under the right historical circumstances, languages can achieve amazing internal diversity.

• In Michif, a so-called ‘‘mixed language’’ in which the nominal systems is largely based on French and the verbal system largely based in Cree (see e.g., Bakker 1997). Michif has distinct co-existing phonological systems, one French-influenced and one Cree-influenced. It was the unique history of the language that gave rise to these unusually different cophonologies.

• In English, where the phonological differences between ‘‘level 1’’ and ‘‘level 2’’ in English are largely due to differences between English and French, most ‘‘level 1’’ morphophonemic alternations are due entirely to allomorphy in words borrowed from French (see e.g. Bermudez-Otero & McMahon 2006)


17

7. SUMMARY: THE KEY DIFFERENCES

Cophonology theory is basically Stratal OT, stripped of the assumption that there’s an upper bound on the number of strata and stripped of the assumption that strata are extrinsically ordered.

Stratal OT

Cophonology Theory

Level ordering (each morphological ‘zone’ of the word associated with a unique phonological subpattern)

yes no

Fixed, small # of levels yes no Differences between subpatterns strictly constrained

yes no

Can handle idiosyncratic morphologically conditioned phonology

no yes

Can handle process morphology no yes Cyclic or Parallel (±Interleaving)

Cyclic (interleaving)


18

References Alderete, John. 1999. Morphologically governed accent in Optimality Theory. University of Massachussets, Amherst, Ph.D.

dissertation. Alderete, John. 2001. Dominance effects as transderivational anti-faithfulness. Phonology 18(2). 201–253. Anttila, Arto. 2002. Morphologically conditioned phonological alternations. Natural Language and Linguistic Theory 20(1). 1–

42. Anttila, Arto & Young-mee Yu Cho. 1998. Variation and change in Optimality Theory. Lingua 104(1–2). 31–36. Benua, Laura. 1997a. Affix classes are defined by faithfulness. University of Maryland Working Papers in Linguistics 5. Benua, Laura. 1997b. Transderivational Identity: Phonological Relations Between Words. Amherst: University of Massachusetts

PhD dissertation. Bermúdez-Otero, Ricardo. 2011. Cyclicity. In Marc van Oostendorp, Colin Ewen, Elizabeth Hume & Keren Rice (eds.), The

Blackwell Companion to Phonology, vol. 4, 2019–2048. Malden, MA: Wiley-Blackwell. Bermudez-Otero, Ricardo & April McMahon. 2006. English phonology and morphology. In Bas Aarts & April McMahon (eds.),

The handbook of English linguistics. Oxford: Blackwell. Buckley, Eugene L. 1994. Theoretical aspects of Kashaya phonology and morphology. (Dissertation in Linguistics Series).

Stanford: CSLI Publications. Gussenhoven, Carlos. 1986. English plosive allophones and ambisyllabicity. Gramma 10(2). 119–141. Halle, Morris & K. P Mohanan. 1985. Segmental phonology of modern English. Linguistic Inquiry 16. 57–116. Hargus, Sharon. 1988. The lexical phonology of Sekani. (Outstanding Dissertations in Linguistics Series). New York: Garland

Publishing Co. Hayes, Bruce. 1981. A metrical theory of stress rules. Bloomington: Indiana University Linguistics Club. Hualde, José. 1988. A lexical phonology of Basque. Inkelas, Sharon & Cemil Orhan Orgun. 1998. Level (non)ordering in recursive morphology: evidence from Turkish. In Steven

Lapointe, Diane Brentari & Patrick Farrell (eds.), Morphology and its relation to phonology and syntax. Stanford, CA: CSLI Publications.

Inkelas, Sharon, Cemil Orhan Orgun & Cheryl Zoll. 1997. Implications of lexical exceptions for the nature of grammar. In Iggy Roca (ed.), Constraints and Derivations in Phonology, 393–418. Oxford: Clarendon Press.

Inkelas, Sharon & Cheryl Zoll. 2007. Is Grammar Dependence Real? A comparison between cophonological and indexed constraint approaches to morphologically conditioned phonology. Linguistics 45. 133–171.

Itô, Junko & Armin Mester. 1999. The structure of the phonological lexicon. In N Tsujimura (ed.), The handbook of Japanese linguistics, 62–100. Malden, MA: Blackwell.

Kiparsky, Paul. 1982a. Lexical morphology and phonology. In I. -S. Yang (ed.), Linguistics in the morning calm, 3–91. Seoul: Hanshin.

Kiparsky, Paul. 1982b. Word-formation and the lexicon. In Frances Ingemann (ed.), 1982 Mid-America linguistics conference papers, 3–32. Lawrence, Kan: University of Kansas.

Kiparsky, Paul. 1984. On the lexical phonology of Icelandic. In Claes-Christian Elert, Iréne Johansson & Eva Strangert (eds.), Nordic prosody II: papers from a symposium, 135–164. Ume\aa: University of Ume\aa.

Kiparsky, Paul. 1985. Some consequences of lexical phonology. Phonology Yearbook 2. 85–138. Kiparsky, Paul. 2000. Opacity and cyclicity. The Linguistic Review 17. 351–367. Kiparsky, Paul. 2003. Finnish noun inflection. In Diane Nelson & Satu Manninen (eds.), Generative approaches to Finnic and

Saami linguistics, 109–161. Stanford: CSLI Publications. Kiparsky, Paul. 2008. Fenno-Swedish Quantity: Contrast in Stratal OT. In Bert Vaux & Andrew Nevins (eds.), Rules, constraints

and phonological phenomena, 185–220. Oxford: Oxford University Press. Mohanan, K. P. 1982. Lexical phonology. Massachusetts Institute of Technology Ph.D. dissertation. Mohanan, K. P. 1986. Lexical phonology. Dordrecht: Kluwer. Mohanan, K. P. & Tara Mohanan. 1984. Lexical phonology of the consonant system in Malayalam. Linguistic Inquiry 15(4).

575–602. Orgun, Cemil Orhan. 1996. Sign-based morphology and phonology: with special attention to Optimality Theory. University of

California, Berkeley. Steriade, Donca. 2008. A pseudo-cyclic effect in Romanian morphophonology. In Asaf Bachrach & Andrew Nevins (eds.),

Inflectional identity, 313–360. Oxford: Oxford University Press.

lecture 2: theoretical approaches to morphologically

Documents